Paul Brockman Xiumin Martin Emre Unlu
Objective and motivation Research question and hypothesis Research design Discussion of results Conclusion
The purpose of this paper is to examine the CEO s portfolio sensitivities on the maturity structure of debt We hypothesize and find supporting evidence that creditors limit their exposure to compensation induced managerial risk-seeking behavior by reducing the maturity structure of corporate debt
The use of equity-based executive compensation has dramatically increased during the past few decades Increased by a factor of 6 between 1980 and 2000 Existing empirical literature examines the two features of equity-based compensation on risk-taking behavior Delta Sensitivity of CEO s wealth to stock price Higher delta exposes risk-averse and underdiversified managers to more risk, DISCOURAGING risk taking Vega Sensitivity of CEO wealth to volatility of stock returns Higher vega provides managers a convex payoff, ENCOURAGING risk taking
Empirical studies on risk-taking and incentives Knopf, Nam, and Thornton, 2002 JF Examine the relation between managerial incentives and hedging activities Firms with HIGHER delta hedge MORE Firms with HIGHER vega hedge LESS Chava and Purnanadan, 2007 JFE Examine the relation between CFO s incentives and choice of floating versus fixed rate debt CFOs with HIGHER delta choose LESS RISKY structures CEOs with HIGHER vega choose RISKIER schedules
Tchistyi, Yermack, and Yun, 2007 Working paper Examine the relation between managerial incentives and pricing schedule of performance sensitive debt CEOs with HIGHER delta choose LESS RISKY schedules CEOs with HIGHER vega choose RISKIER schedules
Coles, Daniel and Naveen, 2006 JFE Examine the relation between managerial incentives and degree of risky-policy implementations Firms with HIGHER vega carry MORE debt have HIGHER R&D spending have LOWER capital expenditures tend to be LESS diversified Mixed results for delta Findings support the view that HIGH vega ENCOURAGES risk-taking
Rajgopal and Shevlin (2002) JAE Examine the relation between vega and riskiness of new investments Firms with HIGHER vega make RISKIER investments Cohen, Hall and Viceira (2000) and Guay (1999) JFE Examine the relation between vega and stock volatilty Firms with HIGHER vega have HIGHER volatility Overall, empirical evidence suggests that managers respond to compensation incentives and take actions that change the firm s risk profile
Risk-taking can aggravate the conflicts of interest between bondholder and stockholders Risk-shifting (Asset substitution) How do creditors deal with risk-taking incentives? Extend credit on a short-term basis (Primary inquiry) Short-term debt can be used to alleviate risk-shifting problem Frequent monitoring and repricing Increase cost of debt (Secondary inquiry) Rational creditors expect the consequences of risk-taking incentives and price debt accordingly
Maturity structure Creditors will use short-term debt to regulate excessive risk taking Proportion of short-term debt is NEGATIVELY related to delta Proportion of short-term debt is POSITIVELY related to vega Cost of debt Cost of debt is NEGATIVELY related to delta Cost of debt is POSITIVELY related to vega
Analyses based on balance sheet data Examines the relation between average maturity of overall debt and managerial incentives Cross-sectional Endogenize leverage Changes regression Endogenize leverage, investment and compensation Analyses based on new issues data Examines the relation between maturity of new debt and managerial incentives
Analyses based on credit spreads Examines the relation between Cost of debt and managerial incentives Cost of longer term borrowing and managerial incentives
Sample selection 1992-2005 period Industrial firms with SIC in 2000-5999 range CRSP-COMPUSTAT-EXECUCOMP 6,825 firm-year observations based on 1,312 unique firms
Cross-sectional estimation framework Maturity structure = α 0 + α 1 Delta+ α 2 Vega + Controls Dependent variable: Maturity structure Two proxies ST3 : Proportion of total debt maturing in 3 years or less ST5 : Proportion of total debt maturing in 5 years or less Expected signs α 1 < 0 α 2 > 0
Treatment variables: Managerial Incentives Delta LPRCSEN : Change in CEO s wealth for 1% increase in stock price (logarithmic transformation) Vega LVOLSEN : Change in CEO s wealth for 1% change stock volatility (logarithmic transformation) How are managerial incentives computed? For option portfolio we use Core and Guay s (2002) approximation method Requires annual data Very high explanatory power
How are managerial incentives computed? Proxy statements report CEO s portfolio in four parts: Options from new grants Options from previous grants (exercisable) Options from previous grants (unexercisable) Stock holdings Core and Guay s (2002) approximation method yields delta and vega for the option portfolio using Black-Scholes equation Exercise price and time-to-maturity are approximated for previous grants For the stock portfolio BS-Delta STOCK =1 BS-Vega STOCK 0 (Guay, 1999)
Control variables Signaling Credit quality (size, credit rating dummy, Z-score) Earnings (abnormal earnings) Premature liquidation risk (leverage) Agency costs Bondholder-shareholder (asset maturity, growth opportunities) Shareholder-manager (stock ownership) Cash-flow volatility (volatility) Regulation (utility dummy) Tax (term structure)
Dependent Variables Independent Predicted Signs ST3 ST5 Variables Coefficient estimate p-value Coefficient estimate p-value Intercept 1.4281*** 0.000 1.2059*** 0.000 LPRCSEN - -0.0385*** 0.000-0.0179** 0.012 LVOLSEN + 0.0309*** 0.001 0.0277*** 0.003 LSIZE - -0.1558*** 0.000-0.0777*** 0.000 LSIZE2 + 0.0090*** 0.000 0.0041*** 0.001 LEVERAGE - -1.1068*** 0.000-0.6251*** 0.000 ASSET_MAT - -0.0026*** 0.000-0.0036*** 0.000 OWN + 0.4562*** 0.000 0.1567* 0.093 M/B + -0.0021 0.714-0.0044 0.382 TERM - 0.0014 0.699-0.0049 0.209 REG_DUM - 0.0108 0.429-0.0501*** 0.001 ABNEARN + -0.0171 0.648 0.0253 0.500 STD_RET + 0.0163 0.889 0.2232** 0.036 RATE_DUM - -0.0945*** 0.000-0.1413*** 0.000 ZSCORE_DUM - -0.1194*** 0.000-0.047*** 0.000 R 2 adj 0.251 0.226 N 6,825 6,825
Endogenizing leverage Maturity = α 0 + α 1 Delta+ α 2 Vega + α 3 Leverage + Controls Leverage = β 0 + β 1 Delta+ β 2 Vega + β 3 Maturity + Controls Estimation method 2 SLS 3 SLS (untabulated) GMM (untabulated)
Dependent Variables Independent Predicted Signs ST3 ST5 Variables Coefficient estimate p-value Coefficient estimate p-value Intercept 1.5525*** 0.000 1.2777*** 0.000 LPRCSEN - -0.0423*** 0.000-0.0200*** 0.003 LVOLSEN + 0.0311*** 0.001 0.0279*** 0.002 LSIZE - -0.1465*** 0.000-0.0723*** 0.001 LSIZE2 + 0.0082*** 0.000 0.0037*** 0.004 LEVERAGE - -1.5295*** 0.000-0.8693*** 0.000 ASSET_MAT - -0.0025*** 0.000-0.0035*** 0.000 OWN + 0.5014*** 0.000 0.1828** 0.045 M/B + -0.0139** 0.016-0.0112* 0.053 TERM - 0.0007 0.801-0.0053* 0.075 REG_DUM - 0.0049 0.739-0.0535*** 0.000 ABNEARN + -0.0091 0.784 0.0298 0.367 STD_RET + -0.1845 0.117 0.1072 0.363 RATE_DUM - -0.0723*** 0.000-0.1285*** 0.000 ZSCORE_DUM - -0.1840*** 0.000-0.0843*** 0.002 R 2 adj 0.182 0.204 N 6,825 6,825
Change-regressions ΔMaturity = α 0 + α 1 ΔDelta+ α 2 ΔVega +ΔControls Change is computed from t-1 to t (if t-1 is unavailable t-2 is used)
Dependent Variables Independent Predicted Signs ΔST3 ΔST5 Variables Coefficient estimate p-value Coefficient estimate p-value Intercept 0.0058* 0.093 0.0014 0.530 ΔLPRCSEN - -0.0440*** 0.003-0.0259** 0.025 ΔLVOLSEN + 0.0483** 0.012 0.0247* 0.090 ΔLSIZE - -0.0831 0.176 0.0647 0.291 ΔLSIZE2 + 0.0018 0.650-0.0029 0.475 ΔLEVERAGE - -1.1325*** 0.000-0.6638*** 0.000 ΔASSET_MAT - -0.0007 0.449-0.0005 0.614 ΔOWN + 0.7247*** 0.007 0.5065** 0.015 ΔM/B + -0.0009 0.940-0.0289** 0.017 ΔTERM - -0.0040 0.339-0.0073** 0.020 ΔABNEARN + -0.0226 0.377 0.0553** 0.026 ΔSTD_RET + 0.0376 0.798 0.1341 0.302 ΔRATE_DUM - -0.1201*** 0.000-0.1929*** 0.000 ΔZSCORE_DUM - -0.1242*** 0.000-0.0571*** 0.000 R 2 adj 0.095 0.063 N 5,513 5,513
Endogenizing leverage, investment and compensation Maturity = α 0 + α 1 D+ α 2 V + α 3 L +α 4 RD +α 5 CAPEX+ Controls D = β 0 + β 1 V+ β 2 L + β 3 RD + β 4 CAPEX+ β 5 Maturity+ Controls V = γ 0 + γ 1 D+ γ 2 L + γ 3 RD + γ 4 CAPEX+ γ 5 Maturity+ Controls L = δ 0 + δ 1 D+ δ 2 V+ δ 3 RD + δ 4 CAPEX+ δ 5 Maturity+ Controls RD = ε 0 + ε 1 D+ ε 2 V+ ε 3 L + ε 4 Maturity+ Controls CAPEX = ζ 0 + ζ 1 D+ ζ 2 V+ ζ 3 L + ζ 4 Maturity+ Controls Estimation method 2 SLS 3 SLS (untabulated) GMM (untabulated)
Estimation summary Consistent results with prior analyses Positive (negative) relation between proportion of short-term debt and vega (delta) Consistent results with recent literature on managerial incentives Coles et al. (2006) High (low) vega (delta) results in risky firm policies Higher leverage Higher RD and lower CAPEX
Evaluation of economic significance Delta A change from the 50 th percentile to the 95 th percentile reduces short-term debt (ST3) by 8.7% (OLS) 9.6% (2-eqn 2SLS) 10.0% (Changes) 27.7% (6-eqn 2SLS) 30.2% (6-eqn 3SLS) 35.3% (6-eqn GMM) Median value of short-term debt =32%
Evaluation of economic significance Vega A change from the 50 th percentile to the 95 th percentile increases short-term debt (ST3) by 4.4% (OLS) 4.4% (2-eqn 2SLS) 6.9% (Changes) 20.1% (6-eqn 2SLS) 9.9% (6-eqn 3SLS) 27.0% (6-eqn GMM) Median value of short-term debt =32%
Sample selection 1992-2005 period Industrial firms with SIC in 2000-5999 range CRSP-COMPUSTAT-EXECUCOMP New debt issues are drawn from SDC 355 public issues 642 Rule 144A issues 2,368 private issues 4,343 syndicated issues 7,388 total debt issues representing 873 unique firms Two samples Unconsolidated sample (issue-year) Consolidated sample (firm-year)
Estimation framework Maturity of new issue t = α 0 + α 1 Delta t-1 + α 2 Vega t-1 + Controls Dependent variable: Maturity Unconsolidated sample (issue-year) LMAT: Maturity of new issue (logarithmic transformation) Consolidated sample (firm-year) LWEIGHT_AVG_MAT: Weighted-average of maturity of all new issues during the fiscal year (logarithmic transformation) LAVG_MAT: Arithmetic average of maturity of all new issues during the fiscal year (logarithmic transformation) Expected signs α 1 > 0 α 2 < 0
Unconsolidated sample (Transaction level) Consolidated sample (Firm-year level) Dependent variable LMAT Dependent variable LWEIGHT_AVG_MAT Dependent variable LAVG_MAT Independent Coefficient Coefficient Coefficient p-value p-value Variables estimate estimate estimate p-value Intercept 2.9590*** 0.000 1.8058*** 0.000 1.9432*** 0.000 LPRCSEN 0.1114*** 0.000 0.1132*** 0.000 0.1069*** 0.000 LVOLSEN -0.0770*** 0.006-0.1300*** 0.000-0.1369*** 0.000 LSIZE -0.0644 0.482 0.1407* 0.082 0.0682 0.414 LSIZE2-0.0003 0.947-0.0082* 0.068-0.0021 0.661 LEVERAGE 0.0154 0.919-0.2871** 0.029-0.3062** 0.019 ASSET_MAT 0.0045*** 0.002 0.0058*** 0.000 0.0066*** 0.000 OWN -1.3134*** 0.005-1.5970*** 0.000-1.4757*** 0.000 M/B -0.0113 0.485-0.0520*** 0.001-0.0566*** 0.000 TERM 0.0201 0.376-0.0002 0.989-0.0023 0.875 REG_DUM -0.1943** 0.040-0.0332 0.702-0.0374 0.688 ABNEARN 0.0052 0.961 0.0300 0.799 0.0443 0.71 STD_RET -2.4487*** 0.000-2.6250*** 0.000-2.7106*** 0.000 RATE_DUM -0.0913*** 0.003-0.0803** 0.011-0.0678** 0.032 ZSCORE_DUM 0.0139 0.768 0.0627 0.175 0.0803* 0.097 R 2 adj 0.376 0.343 0.317 N 7,388 3,122 3,122
Sample selection 1994-2005 period Industrial firms with SIC in 2000-5999 range CRSP-COMPUSTAT-EXECUCOMP Credit spreads for traded debt are from Datastream Indenture details and credit ratings are from Mergent-FISD Bonds with special features (i.e. call, convertible etc.) are excluded 268,400 bond-day observations based on 266 bond issue and 114 unique firms
Estimation framework System 1 SPREAD= α 0 + α 1 Delta+ α 2 Vega + α 3 Maturity + Controls Maturity = β 0 + β 1 Delta+ β 2 Vega + β 3 SPREAD+ Controls Examines creditors reluctance to lend low-delta/high-vega CEOs Expected Signs α 1 < 0 (lower the delta, higher the borrowing cost) α 2 > 0 (higher the vega, higher the borrowing cost)
Estimation framework System 2 SPREAD= α 0 + α 1 Delta+ α 2 Vega + α 3 Maturity + α 4 Delta*Maturity + α 5 Vega*Maturity +Controls Maturity = β 0 + β 1 Delta+ β 2 Vega + β 3 SPREAD+ Controls Examines how maturity affects creditors reluctance to lend lowdelta/high-vega CEOs Expected Signs α 3 > 0 (establishes positive maturity premium) α 4 < 0 (longer maturity exacerbates the delta effect) α 5 > 0 (longer maturity exacerbates the vega effect)
Estimation framework Dependent Variable: SPREAD SPREAD= Yield to maturity Interpolated Treasury yield with corresponding maturity Control variables for the spread equation Firm-specific Equity risk and return(campbell and Taksler, 2003; Kwan, 1996) Profitability, leverage, interest-coverage (Campbell and Taksler, 2003) Issue-specific Credit rating (Campbell and Taksler, 2003) Illiquidity (Chen et al., 2007) Issue size (Campbell and Taksler, 2003) Coupon rate (Elton et al., 2004) Benchmark Treasury (Longstaff and Schwartz, 1995) Economy-wide Slope of the yield curve (Fama and French, 1989; Collin-Dufresne et al.,2001) Eurodollar-Treasury spread (longstaff, 2004)
Average corporate bond yield spread (%) broken down by maturity Maturity AAA AA A BBB BB B CCC Short 0.359 0.533 0.644 1.016 2.923 2.826 5.671 Medium 0.534 0.730 1.006 1.546 3.111 4.483 7.212 Long 0.612 0.829 1.125 1.725 3.258 4.810 8.659
System 1 Independent Variables Dependent Variables SPREAD LMAT Estimate p-value Estimate p-value Intercept 5.4310*** 0.000 4.9098*** 0.000 LMAT 1.4524*** 0.000 SPREAD 0.0811*** 0.000 LPRCSEN -0.1258*** 0.000 0.1205*** 0.000 LVOLSEN 0.1406*** 0.000-0.1771*** 0.000 CONTROLS Not reported Not reported
System 2 Independent Variables Dependent Variables SPREAD LMAT Estimate p-value Estimate p-value Intercept 4.3976*** 0.000 4.9098*** 0.000 LMAT 1.7376*** 0.000 SPREAD 0.0811*** 0.000 LPRCSEN 2.6642*** 0.000 0.1205*** 0.000 LVOLSEN -3.2306*** 0.000-0.1771*** 0.000 LPRCSEN x LMAT -1.1236*** 0.000 LVOLSEN x LMAT 1.3582*** 0.000 CONTROLS Not reported Not reported
Evaluation of economic significance Benchmark case: Maturity, delta and vega are assumed to be at sample medians Sensitivity of spread to delta = -0.39% Sensitivity of spread to vega = 0.47% Comparison case: Maturity is evaluated at the 95% and delta/vega remain unchanged Sensitivity of spread to delta = -1.00% (factor of 2.56) Sensitivity of spread to vega = 1.19% (factor of 2.53) Interpretation Long-term debt exacerbates the agency costs of delta and vega related incentives
Alternative evaluation of economic significance Benchmark case: Maturity, delta and vega are assumed to be at sample medians Sensitivity of spread to maturity= 1.43% Comparison case 1: Delta is evaluated at the 5% and maturity/vega remain unchanged Sensitivity of spread to maturity= 2.82% (factor of 1.97) Comparison case 2: Vega is evaluated at the 95% and maturity/delta remain unchanged Sensitivity of spread to maturity= 3.09% (factor of 2.16) Interpretation Maturity premium is high when delta is low and vega is high
Creditors are aware of risk-taking incentives of executive compensation When managers are encouraged to take risk Creditors use short-term debt to protect themselves. Creditors also increase cost of borrowing (especially long-term borrowing) Managerial incentives influence maturity structure and cost of debt.