ebook img

Management Ownership and Investment PDF

58 Pages·2016·0.57 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Management Ownership and Investment

Management Ownership and Investment in the Business Cycle ∗ Brian S. Chen January 2016 (First version: August 2015) Working Paper Abstract Does risk aversion amplify business cycle downturns? This study considers the risk exposure of CEOs and its effect on firm investment in times of high macroeconomic uncertainty. Exploiting exogenous variation in CEO equity ownership, I document that firms with larger CEO stakes decrease investment significantly more in periods of high aggregate uncertainty. I consider different explanations and find evidence that risk aversion explains these results. Firms with high CEO stakes decrease risk-taking in times of high uncertainty and experience lower stock returns subsequent to periods of high uncertainty, suggesting that high managerial equity ownership may also pose costs to firms. ∗ Email: 1 Introduction What role, if any, does risk aversion play in amplifying business cycle downturns? Recessions tend to be periods of high macroeconomic uncertainty, and increased uncertainty may lead to declines in investment and output (e.g., Bernanke 1983, Bloom 2009). If key agents in the economy are effectively risk-averse – perhaps due to limited risk-sharing – this may have the potential to magnify the effects of heightened macroeconomic uncertainty. One set of important agents with significant discretionary control over levels of investment are the senior managers of firms. Many executives hold large, highly concentrated ownership stakes in their firms. Large managerial ownership stakes are a commonly suggested solution to conflicts of interest between owners and managers of the firm, as they incent managers to curtail excessive consumption of private benefits and to exert more effort to maximize firm value, leading to greater alignment of interest between managers and owners (e.g., Jensen and Meckling 1976). But these large managerial ownership stakes may have pitfalls as well. They lead managers to be highly ex- posed to firm risks, which may become particularly salient in times of high uncertainty. Risk-averse managers may then make investment decisions inconsistent with those desired by diversified outside shareholders. Separately, large stock ownership stakes may lead managers to become excessively aligned with equity holders and encourage them to choose risk-shifting investments, at the expense of firm creditors. This paper examines the relationship between managerial ownership stake and investment throughout the business cycle. Recessions and periods of high macroeconomic uncertainty are particularly relevant periods to study the impact of managerial ownership on investment. If large managerial ownership stakes lead to excessive risk-aversion or risk-shifting, periods of high uncer- tainty may highlight the role of these factors on firm investment. Moreover, aggregate investment is highly procyclical and its decline contributes significantly to total output declines in downturns, hence playing a key role in business cycle fluctuations. Studying the microeconomic forces affecting the investment decisions of firms in downturns may thus shed light on the causes of business cycle investment volatility. In this paper, I compare firms with different CEO equity ownership stakes and study their in- vestment behavior in downturns and periods of high macroeconomic uncertainty, where uncertainty is measured either as in Jurado, Ludvigson, and Ng (2015), aggregating information from hundreds of macroeconomic indicators, or by the implied volatility on S&P 500 index options (VIX). I find that firms with high CEO stakes decrease investment significantly more during periods of high uncertainty. Figure 1 shows this basic pattern: firms in the highest quintile of CEO ownership stake cut investment significantly more during recessions than firms in the lowest quintile. While firms with high and low levels of CEO ownership stake differ along other characteristics as well, 1 this differential investment behavior in periods of high uncertainty is robust to controlling for a multitude of covariates, various fixed effects (including industry, firm, time, and industry-time fixed effects), and different definitions of ownership stake. The magnitude of the effect is economically significant: for a one standard deviation increase in uncertainty, firms in the highest quintile of CEO ownership stake decrease investment by 0.08 percentage points more than those in the lowest quintile, in the baseline ordinary least squares (OLS) estimates. For comparison, in the same re- gression and sample, the marginal effect of a one standard deviation increase in cash flow – one of the most studied predictors of investment in corporate finance – increases investment by 0.17-0.19 1 percentage points. These results are not driven by differences in the underlying cyclicality of high and low CEO-stake firms, as there is no pattern of differential investment in response to changes in the output gap, a direct proxy for the macroeconomic cycle. To causally identify the effect of CEO ownership stake on investment during downturns, I use an instrumental variable estimation strategy and exploit exogenous variation in CEO ownership stake due to exogenous CEO turnover. Using two-sample instrumental variables (TSIV) estimation, I obtain estimates of the effect of CEO ownership stake on investment declines that are larger in 2 magnitude than those in the baseline OLS results. The TSIV estimates imply that for a one standard deviation increase in uncertainty, firms with CEOs in the highest quintile of ownership, relative to firms in the bottom four quintiles, decrease investment by 0.20-0.45 percentage points more. For comparison, the sample mean investment level is 1.62%. In addition, I test for possible threats to identification, such as direct tenure effects on investment, and do not find evidence supporting them. The IV estimates are larger in magnitude than the OLS estimates as they plausibly correct for biases such as that due to differences in CEO personal risk aversion. More risk-tolerant CEOs will ceteris paribus choose higher ownership stakes, biasing the OLS estimate toward zero. One potential explanation for the larger decline in investment among high CEO stake firms relative to low stake firms during periods of high macroeconomic uncertainty is managerial risk aversion. Managers who have larger equity ownership stakes are more risk-averse over firm outcomes due to concentrated, undiversified exposure to firm-specific risk. In a basic, stylized “real options” model with a risk-averse manager, more risk-averse managers cut investment more in response to 1 See Fazzari, Hubbard, and Petersen 1988, Hoshi, Kashyap, and Scharfstein 1991, Kaplan and Zingales 1997, Lamont 1997, Rauh 2006, Cummins, Hassett, and Oliner 2006 for a few prominent examples of the large literature on firm investment sensitivity to cash flow. 2 Two-sample instrumental variables (TSIV) is used to maximize statistical power from the limited set of exogenous CEO turnover events. For a detailed description of the TSIV estimation method, see Angrist and Krueger 1992 as well as Inoue and Solon 2010. Examples of papers that use TSIV estimation include Angrist 1990 and Dee and Evans 2003. I also present reduced-form estimates that do not use TSIV estimation but exploit the same underlying source of exogenous variation in CEO turnover events. 2 increases in uncertainty. This is driven by the “bad news principle” articulated by Bernanke (1983): increased uncertainty leads to less investment because states of the world with poor investment payoffs now occur with increased probability. More risk-averse managers are more sensitive to 3 poor outcomes in “bad” states of the world, and thus decrease investment more. More generally, managers may attempt to decrease firm riskiness via other channels as well, such as re-allocating funds to less risky investments, decreasing operating leverage, or reducing the transformation of safe internal funds (such as cash) to risky operating assets. Empirically, I find that firms with higher CEO ownership stakes decrease selling, general, and administrative (SG&A) expenses, which are an investment-like expense and contribute to higher operating leverage. Moreover, these firms increase equity payouts and asset sales relatively more. One important alternative explanation for these results is that managers with higher owner- ship stakes are better-aligned with shareholders, and that the larger investment declines during periods of high macroeconomic uncertainty are actually optimal – or, at minimum, better – pol- icy for shareholders. I conduct a number of additional empirical tests to distinguish between the alignment and risk aversion explanations. First, I compare firms with high and low institutional ownership, as firms with high institutional ownership are likely to have better shareholder moni- toring of management and hence a higher degree of manager-shareholder alignment. There is zero (or weakly positive) impact of institutional ownership on investment during downturns, suggesting that relative declines in investment in times of high uncertainty are not preferred by shareholders. Second, I find that excess stock returns in the year subsequent to a period of high macroeconomic uncertainty are lower for firms with larger CEO stakes, which is inconsistent with the explanation that it is optimal to decrease investment more in times of high uncertainty. Third, firms with high CEO stakes cut idiosyncratic firm risk-taking relatively more during periods of high uncertainty. Finally, high stake CEOs who possess options whose values are highly sensitive to volatility (i.e., high “vega” options) cut investment less in periods of high uncertainty. Results of these additional empirical tests are consistent with the managerial risk aversion explanation. This paper makes several contributions. The first is to show that CEO ownership is an empiri- cally important factor affecting firm investment policy. Despite an extensive literature on manage- ment ownership, there remains limited evidence of the causal impact of management ownership on 4 firms. This study presents causal evidence that large managerial ownership stakes lead to larger declines in investment in times of high macroeconomic uncertainty, adding to the limited literature on the causal impacts of managerial ownership. 3 Greater exposure to firm risk via a larger ownership stake is analogous to being more risk-averse over firm investment outcomes. 4 Li and Sun (2015) use the 2003 dividend tax decrease as an exogenous shock to effective managerial ownership to study the impact of managerial ownership on firm value. 3 The second contribution of the paper is to show that these investment declines are due to managerial risk aversion and concentrated managerial exposure to firm risk. While there is an extensive literature on the positive effects of large managerial ownership stakes to align manager and shareholder interests, less evidence exists on the potential costs of large ownership stakes and 5 the consequences of exacerbating managerial risk aversion. High-powered incentives for CEOs come at the expense of risk-sharing, which is consistent with the canonical contract theory trade- off between incentive and insurance provision. Contract theory commonly links strong incentives to higher average compensation for the risk-averse manager, but this study shows that risk-sharing 6 costs also manifest themselves as distortions to firm policy. Regardless of whether the observed CEO stakes are the outcome of optimal or sub-optimal contracting between managers and owners, the results illustrate a real cost of large management ownership beyond the need to pay managers more. Third, the paper describes a novel channel through which managerial moral hazard affects 7 macroeconomic outcomes and cyclicality. Large CEO ownership stakes mitigate moral hazard concerns but force risk-averse managers to be exposed to large amounts of firm-specific risk. In times of high uncertainty, managers with large equity stakes cut firm investment to reduce their 8 personal risk exposure. This exacerbates the high volatility of investment in the business cycle. The study provides evidence that “animal spirits” – in the form of risk aversion – play a significant role in investment behavior and business cycle amplification (Keynes 1936). While U.S. public equity markets are well-developed and a large fraction of U.S. households own equity, concentrated equity ownership remains common among the managers of many publicly-traded firms, and likely among many non-public firms as well. Due to imperfect risk-sharing and concentrated ownership stakes, CEOs – a small number of agents in the aggregate economy, yet possessing significant discretion 5 Faccio, Marchica, and Mura (2011) and Lyandres, Marchica, Michaely, and Mura (2015) study the impact of owner portfolio diversification on firm risk-taking and investment. 6 Becker (2006) finds that Swedish CEOs with higher wealth levels (hence lower risk aversion) hold larger equity ownership stakes in their firm, which is evidence on the trade-off between large ownership stakes and risk-sharing. Other studies of the trade-off between incentives and risk-sharing in executive compensation include Aggarwal and Samwick 1999, Prendergast 1999, and Prendergast 2002. 7 Rampini (2004) presents a theory in which entrepreneurial activity varies throughout the business cycle due to fluctuations in entrepreneurial net worth. Entrepreneurs are risk-averse but must bear non-diversifiable risk to prevent moral hazard. Productivity shocks affect entrepreneurial net worth, amplifying the initial shock by decreasing entrepreneurship. Other work on managerial moral hazard and macroeconomic outcomes includes the role of entrepreneurial moral hazard in external financing frictions and financial intermediation amplifying net worth or credit supply shocks (e.g., Holmstrom and Tirole 1997), and the impact of managerial moral hazard on bank or financial sector risk-taking, due to deposit insurance or bailout policy (e.g., Diamond and Dybvig 1983, Grossman 1992, Demirgüç-Kunt and Detragiache 2002, Dam and Koetter 2012). 8 Although managerial hedging can offset this risk exposure, in practice managers are unlikely to engage in much hedging of their large aggregate risk exposure, as it is either explicitly prohibited (e.g., shorting own-firm stock) or very costly (e.g., purchasing many options to offset exposure to aggregate uncertainty). 4 over large firm investment choices – may end up amplifying business cycles due to personal risk aversion. This paper is related to a few prior strands of literature. First, there is an extensive literature on the misalignment of managers with owners of the firm, and the curative effect of managerial ownership (e.g., Jensen and Meckling 1976, Leland and Pyle 1977, Jensen 1986, McConnell and Servaes 1990, Himmelberg, Hubbard, and Palia 1999, Bertrand and Mullainathan 2003, Shue and Townsend 2014). While the benefits of managerial ownership are well understood, there are fewer 9 studies on the costs of high managerial ownership. Panousi and Papanikolaou (2012) argue that managerial risk aversion explains their finding that firms with high managerial ownership decrease investment more in face of higher idiosyncratic firm volatility. While their results are consistent with this study, there are important differences: i) this study focuses on macroeconomic uncertainty 10 and investment in the business cycle; ii) this study exploits exogenous variation in managerial ownership; and iii) I use a different measure of ownership stake that is more directly linked to the manager’s risk exposure and risk aversion. Second, this paper relates to studies on the relationship between managerial characteristics and firm investment. Bertrand and Schoar (2003) argue that managerial fixed effects – or “styles” – have explanatory power for firm investment, but Fee, Hadlock, and Pierce (2013) show that the empirical estimation of managerial style effects must differentiate between endogenous and exogenous managerial turnover events. In this paper, I exploit the exogenous CEO turnover data 11 from Fee, Hadlock, and Pierce (2013). Other studies show that specific CEO characteristics, ranging from overconfidence to prior work or life experience, affect firm investment policy (e.g., Malmendier and Tate 2005, Malmendier, Tate, and Yan 2011, Schoar and Zuo 2011, Pan, Wang, and Weisbach 2013, Benmelech and Frydman 2014). Third, previous studies on real options models of investment and the importance of uncertainty shocks in investment are related (e.g., Bernanke 1983, Dixit and Pindyck 1994, Bloom, Bond, and Van Reenen 2007, Bloom 2009). Finally, there are related studies that examine other factors that lead to differential behavior across firms in response to the business cycle (e.g., Philippon 2006, Eisfeldt and Rampini 2008). The remainder of the paper proceeds as follows. Section 2 describes the data, the baseline empirical specification, and the TSIV identification strategy. Section 3 presents the basic empirical results. Section 4 discusses the interpretation and potential explanations for the results, presents 9 Examples of papers on the costs of high management ownership include Morck, Shleifer, and Vishny (1988) on managerial entrenchment at high levels of ownership, and Friend and Lang (1988) on high management-ownership firms choosing sub-optimally low levels of firm debt. 10 In a part of the analysis, this study uses idiosyncratic firm volatility as an outcome, because firm volatility can be endogenously affected by managers’ decisions. In Panousi and Papanikolaou (2012), idiosyncratic firm volatility is an explanatory variable. 11 I am grateful to Charles Hadlock and Ted Fee for generously providing this data. 5 evidence from further tests to distinguish between these explanations, and contains robustness checks. Section 5 discusses the macroeconomic implications of the results. Section 6 concludes. 2 Data and empirical strategy 2.1 Data Quarterly firm data from Compustat is matched to annual firm data on executive compensation and shareholdings from Execucomp, covering the years 1992-2013. Data from Execucomp does not cover all firms in the Compustat database, but only a subset that is approximately the universe of 12 firms in – or formerly in – the S&P 1500. I exclude financial firms (SIC codes 6000-6799) and regulated utilities (SIC codes 4900-4949), as is customary in studies of firm investment. 2.1.1 Measuring CEO stake A commonly used measure of managerial ownership is the fraction of the firm’s equity owned by the CEO or by top executives of the firm (e.g., Panousi and Papanikolaou 2012). However, this measure is not ideal to test the role of CEO risk aversion, because it does not directly measure the CEO’s exposure to firm or aggregate risk relative to her entire wealth portfolio. A CEO can own a large fraction of the firm’s equity but have this stake represent a small fraction of their total wealth. CEO firm equity wealth A better measure is the fraction of the CEO’s total wealth in firm equity, i.e. , CEO total wealth where CEO total wealth includes non-firm equity wealth along with the present value of human 13 capital. In this study, I use as a proxy for CEO stake: CEO firm equity wealthit CEOstakeit = . CEO cash compensationit The numerator, CEO firm equity wealth, is calculated as the product of CEO shares owned, in- cluding restricted stock, and the share price. The denominator is calculated as the annual salary and bonus of the CEO. CEOstakeit serves as a proxy for fraction of CEO total wealth in the firm, where the denominator is a flow value (rather than present value) approximation for the value of 12 According to S&P, firms in the S&P 1500 collectively make up around 90% of US equity market capitalization. The S&P 1500 includes the S&P 500 which covers large-cap firms, but also includes the S&P MidCap 400 and S&P SmallCap 600. 13 CEOs are typically prohibited from taking measures to hedge against own-firm risk in their personal portfolio (e.g., such as shorting or buying put options on firm stock), but could hedge their exposure to aggregate risks (e.g., their market beta exposure). Any unobserved CEO hedging of aggregate risks in their personal portfolio will bias against finding any impacts of CEO risk exposure in the empirical tests. 6 CEO human capital, as data on CEO financial wealth is generally unavailable in the U.S. Alter- natively, one can interpret CEOstakeit as a normalized measure of the CEO’s firm equity stake. If CEOs of larger firms have larger dollar stakes and higher human capital as well as non-firm equity wealth, this normalization makes ownership stakes more comparable across firm sizes, types, and time periods (as it is inflation-invariant). I divide CEOstakeit into quintiles in my empirical 14 analysis to prevent outliers from driving the results. In robustness checks, I adjust the numerator of CEOstakeit, CEO firm equity wealth, for CEO option ownership by adding the estimated dollar “delta” exposure that CEOs face via their options. Delta is a standard measure of the change in value of the option with respect to a dollar change in the underlying stock price. 2.1.2 Summary statistics Table 1 displays firm characteristics by quintile of CEO ownership stake. Firms with high CEO stake tend to be more profitable. They also have higher average Tobin’s Q, lower levels of leverage, and faster growth. Firms with high CEO stake do not appear to differ in terms of stock return volatility. These statistics suggest that investment opportunities and financial constraints may differ across firms with differing CEO stakes. As many of these factors affect investment directly, 15 I control for them in the empirical analysis. 2.2 Baseline empirical specification While Figure 1 shows larger investment declines in recessions for high CEO stake firms, the data are not adjusted for any potential confounds. In an ideal experiment, one would assign otherwise identical firms different CEO ownership stakes and observe their investment patterns across the business cycle and in times of differing macroeconomic volatility. Without an experiment, I estimate the baseline regression J Invit ′ ∑ = β0+β1CEOstakeit−4+β2CEOstakeit−4 ×Mt +β3Mt+X itΓ+ δj1 {Industryi = j}+ϵit, Ait−1 ︸ ︷︷ ︸ j=1 Effect of interest (1) Invit for firm i in quarter t. Here, Ait−1 is capital expenditures scaled by lagged assets, CEOstakeit−4 is the lagged CEO stake, Mt is the time-series measure of macroeconomic uncertainty or business 14 Firms whose CEOs have very low cash compensation are thus included in the top quintile of the CEOstakeit variable. 15 All variables are winsorized at the 1% level, except total assets. 7 16 cycle, Xit is a matrix of controls, and δj are industry fixed effects. Controls include potential confounds and other determinants of investment: Tobin’s Q, cash flow, size, leverage, sales growth, cash-on-hand, stock return volatility, and CEO age. I exclude firm-years in which there is CEO turnover, to avoid matching the previous CEO’s ownership stake with firm policies implemented by the new CEO. In other specifications, firm fixed effects are used, exploiting within-firm variation in CEO stake, or the macroeconomic indicator Mt main effect is replaced with time fixed effects, leaving the effect of interest, CEOstakeit−4 ×Mt, identified by differential cross-sectional impacts of CEO stake in periods of high versus low Mt. I use a few different proxies for the macroeconomic state Mt. To measure macroeconomic uncertainty, I use a measure of macroeconomic uncertainty calculated by Jurado, Ludvigson, and Ng (2015) (henceforth “JLN uncertainty”), which aggregates information from the time-series of hundreds of economic variables and removes the predictable or “expected” component of these 17 time-series. Second, the implied volatility on S&P 500 index options (VIX) is used as a measure of macroeconomic uncertainty. Third, output gap is used to proxy for business cycle. The time- series of these three macroeconomic indicators is shown in Figure 2. While the three variables are correlated, the correlation between the uncertainty measures and output gap is relatively low (the 2 R from a quarterly regression of JLN uncertainty on the output gap is 0.1; that of a regression of VIX on output gap is 0.03). I also use an indicator variable for the acute period of financial crisis from 2008Q3 through to 2009Q2 along with time fixed effects to document the effect of CEO stake during the recent financial crisis. 2.3 Instrumental variable strategy Even with fixed effects and extensive controls for confounds, firm-time varying unobserved omitted variables may bias the empirical results. Firm fixed effects will address unobserved firm omitted variables by exploiting only within-firm variation in CEO stake and macroeconomic uncertainty, but cannot rule out the possibility that unobserved correlates of CEO stake that vary within firm and over time could drive the results. To address this, I exploit plausibly exogenous variation in CEO stake due to exogenous turnover in CEOs in an instrumental variable (IV) strategy. CEO turnover leads to changes in CEO stake because executives accumulate ownership stakes over time from stock grants, option grants, and equity payouts associated with long-term incentive plans. New CEOs will thus tend to have lower ownership stakes than pre-existing incumbent CEOs. However, CEO turnover in and of itself is problematic as an instrument as it can coincide with the desire for a firm to change their business 16 Execucomp contains annual data on CEO share ownership, which is merged to Compustat based on fiscal year. Hence CEOstakeit−4 uses data on CEO share ownership from the previous fiscal year. 17 These data are obtained from Sydney Ludvigson’s website. 8 strategy and their investment policy. Therefore I exploit only exogenous CEO turnover events, as documented by Fee, Hadlock, and Pierce (2013). Using news search, Fee et al. (2013) classify exogenous CEO turnover events as those caused by death, health, or “natural” retirement. Natu- rally retiring CEOs are restricted to those whose age falls in the retirement time window, whose resignation is not otherwise found to be forced in the news search, and whose firms are not under- 18 performing based on observables. The data on exogenous CEO turnover events runs from 1990 to 2007 and covers all firms in Compustat with book assets exceeding $10 million (1990 dollars). In IV regressions, I define the instrument, recentCEOit, as the indicator for recent new CEO subsequent to exogenous CEO departure, for the five years subsequent to the year of exogenous 19 turnover. As there are two endogenous variables – CEOstakeit and CEOstakeit × Mt – two instruments are needed: recentCEOit and recentCEOit ×Mt. The first-stage equations are CEOstakeit−4 = α10 + α11recentCEOit−4 + α12recentCEOit−4 ×Mt + ′ α13Mt +X itΛ1 + FEi + η1it, CEOstakeit−4 ×Mt = α20 + α21recentCEOit−4 + α22recentCEOit−4 ×Mt + (2) ′ α23Mt +X itΛ2 + FEi + η2it, and the second stage regression is Invit ′ ̂ ̂ = δ0 + δ1CEOstakeit−4 + δ2CEOstakeit−4 ×Mt + δ3Mt +X itΓ + FEi + νit, (3) Ait−1 ̂ ̂ where CEOstakeit−4 and CEOstakeit−4 ×Mt denote the two fitted values estimated in the first stage, and FEi denotes either the industry or firm-level fixed effect used in the regression. One issue with IV estimation in the sample is that statistical power is limited due to a limited 20 number of exogenous CEO turnover events. While the Execucomp data on the endogenous variable CEOstakeit is only available for a subset of the Compustat database, the CEO turnover data is matched to a larger sample of Compustat firms. To fully exploit the variation in the instrument, I use a two-sample instrumental variable (TSIV) estimator (e.g., Angrist 1990, Angrist and Krueger 1992, Dee and Evans 2003). The TSIV method combines moments from two datasets 18 See Fee et al. (2013) for additional details. Most CEO turnover events in their sample are not exogenous, as only 824 of a total of 7,179 CEO turnover events are classified as exogenous. Moreover, Fee et al. (2013) show that firm policies do not appear to change significantly around exogenous CEO turnover events, which supports the claim that these events are quasi-random and are not otherwise associated with desired underlying changes in firm policy. 19 To be consistent with the OLS specifications, the year of CEO turnover is excluded. If firm investment is particularly anomalous in the first year of a CEO’s tenure, excluding the year of CEO turnover also avoids this potential concern. 20 Variation from the instruments is needed to identify not only the effect of a higher CEO stake, but also the interaction effect of CEO stake in times of high macroeconomic uncertainty. 9

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.