The Roles of Performance Measures and Monitoring in Annual Governance Decisions in Entrepreneurial Firms* Ellen Engel University of Chicago

Transcription

1 The Roles of Performance Measures and Monitoring in Annual Governance Decisions in Entrepreneurial Firms* Ellen Engel University of Chicago Elizabeth A. Gordon Rutgers University Rachel M. Hayes University of Chicago October, 2001 First draft: January, 2001 * We appreciate the research assistance of Bruce Bower, Donald McLaren, Raluca Mihaila, Lee- Jean Tao, and Sandy Wu and the data assistance of Taorong Jiang. We received helpful comments from John Core, Greg Miller, Karen Nelson, Scott Schaefer, Abbie Smith, an anonymous referee, and workshop participants at Harvard Business School, the University of Utah Winter Accounting Conference, the 2001 Journal of Accounting Research Conference and the 12 th Annual Conference on Financial Economics and Accounting. This research was supported by a grant from The Kauffman Center for Entrepreneurial Leadership. Engel also acknowledges research support from the FMC Faculty Research Fund and Hayes from the William Ladany Faculty Research Fund at the University of Chicago Graduate School of Business.

2 The Roles of Performance Measures and Monitoring in Annual Governance Decisions in Entrepreneurial Firms ABSTRACT This paper analyzes annual corporate governance decisions at firms making initial public offerings (IPOs) of common stock between 1996 and Our objective is to examine relations between firms corporate governance decisions and the informativeness of available measures of managerial performance. We consider financial measures such as earnings and stock return, as well as direct monitoring. We collect a sample of IPO firms from the manufacturing, internet, and technology (non-internet) industries, and examine how the use of various performance measures in annual compensation grants and turnover decisions varies with the information environment of the firm and with the extent of venture capital influence. Consistent with prior research that finds earnings are of limited usefulness in firm valuation for internet firms, we find internet firms place less importance on earnings and greater importance on stock returns in determining compensation grants than do non-internet firms. We also find that compensation grants of firms with little or no venture capital influence display significantly stronger association with accounting and stock performance measures than those of firms with more intense monitoring by venture capitalists. This result is consistent with direct monitoring and the use of explicit performance measures acting as substitute governance mechanisms.

3 1. Introduction For most firms, the initial public offering (IPO) is a key event in separating ownership from control. Pre-IPO firms tend to feature extremely high levels of inside ownership, and any outside owners, such as venture capitalists, tend to be specialists in developing and monitoring new ventures. Consequently, agency issues of the type considered by Berle and Means [1932] and Jensen and Meckling [1976] are not of first-order concern in the pre-ipo period. The initial sale of shares to the public is often the first event in a firm s history that necessitates careful consideration of how to mitigate owner/manager agency conflicts. As Baker and Gompers [1999, 2000] observe, corporate governance decisions made at the time of the IPO are, therefore, crucially important. Recent research by Bushman and Smith [2001] explores the idea that governance structures are impacted by the firm s financial accounting system. If accounting- and marketbased measures of managerial performance allow owners to effectively assess how well managers are serving their interests, then we expect governance structures to place greater emphasis on formal, pay-for-performance contracting based on such measures. Similarly, if managers actions are easily observable to the firm s owners, then direct monitoring may play a greater role in governance. Thus, the usefulness of available measures of managerial performance is a key determinant of firms governance decisions. A developing literature now offers considerable evidence linking compensation and other governance decisions in wellestablished firms to properties of accounting- and market-based measures of managerial performance. Bushman, Indjejikian and Smith [1996] and Hayes and Schaefer [2000] show that firms substitute away from objective and towards subjective measures of managerial performance when objective measures are less precise, while Bushman, Chen, Engel and Smith 1

4 [2000] link ownership concentration and monitoring incentives to the ability of accounting measures to provide information about changes in shareholder value. In this paper, we apply insights from this literature to newly public firms. We test three hypotheses regarding the use of accounting performance measures in annual governance decisions. Our first hypothesis is that where the informativeness of current accounting measures with respect to the contribution of managerial actions to shareholder value is relatively low, the sensitivity of CEO pay to current accounting performance should be relatively low. Further, when the informativeness of current accounting measures is relatively low, we expect boards to substitute away from accounting-based measures and toward market-based measures; hence, our second hypothesis is that the sensitivity of CEO pay to stock returns will be relatively high when the informativeness of accounting measures about managerial contributions to shareholder value is low. Third, we hypothesize that direct monitoring and the use of compensation tied to objective performance measures are substitutes, so we expect that when levels of direct monitoring are relatively high, the sensitivity of CEO pay to accounting measures and to stock returns will be relatively low. We test similar hypotheses using CEO turnover as the firm s governance decision. We offer several reasons why testing these hypotheses in the IPO setting has the potential to yield new insights about the use of accounting information in governance. First, as noted above, firms undergoing IPOs face many crucial governance decisions; an understanding of factors affecting such choices would further enhance our knowledge about how firms transition from private to public ownership. Second, the IPO context may offer a sharper test of the hypothesis that reliance on performance-based contracts can act as a substitute for direct monitoring. As compared to more well-established firms, incentives for direct monitoring in 2

5 many newly public firms appear to be especially strong due to the high levels of retained ownership by key investors, including venture capitalists (VCs). The literature on VCs suggests that one key role of such financiers is to monitor top management (see Barry et al. [1990], Kaplan and Strömberg [2000a]). If VCs have specialized expertise and strong incentives to monitor, then one may expect VC-backed firms to make lesser use of earnings- and returnsbased performance measures as compared to non-vc-backed firms. Testing these hypotheses in the IPO setting requires a measure of across-firm differences in the informativeness of available objective performance measures. Most prior research has used historical data to assess the usefulness of various measures for performance evaluation. 1 While we lack such historical data to compute similar proxies for our sample of newly public firms, we argue that the recent wave of so-called new economy IPOs offers a unique opportunity to test these hypotheses with an alternative proxy for the informativeness of accounting measures of managerial performance. We note that the sudden rise of internet-based commerce in the late 1990s was accompanied by a considerable degree of uncertainty about the future prospects for profitability in this industry. The new opportunities provided by electronic commerce had no analogue in business history; consequently, there was no track record of similar firms on which to base expectations of future growth. In addition, recent research has shown aggregate accounting-based performance measures such as net income to be poorly associated with market performance at internet firms (e.g., Trueman, Wong and Zhang [2000], Hand [2000], Demers and Lev [2000]). Together, these factors suggest that traditional accounting measures of firm performance may be of limited usefulness in mitigating potential 1 For example, Lambert and Larcker [1987] use time series information to capture performance measure variation and the correlation of stock return and earnings to proxy for performance measure precision and sensitivity, respectively, with respect to managers actions. These proxies, designed to capture properties of performance measures as they relate to managers actions, arguably capture this notion with noise, as they rely on results of overall performance rather than solely the results of managerial actions. 3

6 owner/manager agency conflicts at internet IPOs. Hence, we take industry to be a proxy for information environment, which we define as the informativeness of accounting measures of managerial performance, and analyze how the use of firm-level measures of performance in governance decisions varies with industry and with the extent of direct monitoring of top management. It is clear that firms in different industries may differ on dimensions other than information environment. Our cross-industry comparison of firms governance choices may therefore reflect these other differences. To mitigate this concern, we incorporate firm-level measures of uncertainty, growth opportunities, the extent of moral hazard problems and CEO characteristics from prior research on firms corporate governance choices. Thus, our analyses control for differences on these dimensions and ask whether industry has additional explanatory power for firms governance choices. To conduct our analyses, we collect a sample of firms making initial public offerings between 1996 and Our data set, constructed directly from firms proxy statements and prospectuses, consists of 464 firms drawn from three industries: general manufacturing, technology (non-internet-based) and internet. We examine how differences in the information environment and the presence of direct monitoring by VCs affect the use of earnings and information in stock returns in annual CEO compensation grants and turnover decisions. Although not focused on IPO firms, related issues have been addressed by Ittner, Lambert and Larcker [2000], who consider the determinants and performance consequences of equity grants in new economy firms, and Anderson, Banker and Ravindran [2000], who study the use of stock options in the information technology industry relative to other industries. Our approach 4

7 departs from other work by emphasizing both the information and monitoring environments as determinants of corporate governance decisions in IPO firms. Our results are largely consistent with our hypotheses. We find that internet and noninternet firms place differing importance on earnings and information in stock returns in determining annual executive rewards. Total compensation, which includes both cash and grants of equity-based instruments, is positively related to earnings for manufacturing firms and stock returns for internet firms, but not vice versa. Moreover, we find that non-internet firms use of accounting earnings is positively related to cash compensation decisions only, while internet firms use of information in stock returns is positively related to stock-based compensation only. Partitioning our sample by venture capital influence, we find that earnings relate positively to total and cash compensation for non-vc-backed, non-internet firms, while no such relation exists for the VC-backed non-internet firms. Similarly, for internet firms, returns are positively related to total and stock compensation for non-vc-backed firms, but not for the VC-backed firms. Finally, analysis of determinants of CEO turnover in the post-ipo period reveals that the probability of CEO turnover is inversely related to stock returns for manufacturing firms, but not for technology and internet firms. In general, we take our results to be supportive of the notion that entrepreneurial firms corporate governance decisions are sensitive to the information environment facing the firm at the time of the IPO. We document significant industry-level differences in how newly public firms address governance issues. Notable among these differences is the variation in reliance on performance measures in determining compensation grants. As we hypothesized, internet firms place lesser reliance on earnings as a measure of managerial performance as compared to manufacturing and non-internet technology firms. Further, use of these performance measures 5

8 appears to vary with the extent of direct monitoring, as measured by venture capital influence, in a manner supporting our prediction that the presence of intense monitoring by venture capitalists acts as a substitute for compensation contracts explicitly linked with financial performance measures. The remainder of the paper proceeds as follows: We describe our hypotheses and control variables in section 2. We discuss our sample selection procedures and present descriptive analyses of the data in section 3. In section 4, we present our empirical design and tests of hypotheses, and we offer conclusions in section Development of hypotheses and description of variables Our interest is in assessing how differences in the information environment and the presence of direct monitoring affect the performance measures used in annual governance decisions. Research into the use of performance measures for contracting has emphasized how the properties of performance measures can impact the relative extent of their use in contracts. This literature suggests that when multiple performance measures are available, firms will substitute away from noisy or imprecise measures of managerial actions (see, for example, Banker and Datar [1989]). This insight also implies that in firms where owners engage in considerable direct monitoring of top management (and hence are able to obtain reasonably precise measures of managerial performance), less reliance will be placed on objective performance measures. Prior research has found both accounting- and market-based performance measures to be related to the corporate governance decisions we examine, compensation and turnover (see, for example, Lambert and Larcker [1987], Murphy [1999], and Weisbach [1988]). 6

9 As noted earlier, Generally Accepted Accounting Principles (i.e., GAAP) has been criticized for not adequately capturing the performance of new economy firms. Valuation studies of internet companies support this criticism. Trueman, Wong and Zhang [2000] do not find a significant association between net income and market value. Hand [2000] finds a positive relation between both R&D and marketing costs and market values, suggesting unrecorded assets. The limited usefulness of accounting in capturing value-enhancing activities of internet firms and our earlier discussion of the uncertainties in the internet industry suggest that the properties of managerial performance measures for internet firms may differ from those used in other industries. In particular, all else equal, the governance choices of internet IPOs should reflect less reliance on accounting-based measures of performance and greater use of other information as captured by stock returns relative to non-internet IPO firms. Thus, we expect to observe systematic differences in the use of performance measures between internet and non-internet firms. In developing our hypotheses about industry differences in the use of accounting measures in governance decisions, we draw from evidence concerning the relatively low usefulness of traditional accounting measures in valuation for internet firms. However, we do not suggest that the use of accounting measures should be identical in settings of valuation and contracting with executives. 2 As Paul [1992] observes, in a valuation setting, information is aggregated to measure the value of the firm, while information used in contracting is aggregated to measure managers contribution to firm value. The latter notion is more difficult for boards of directors to observe; thus, in practice, we observe boards of directors frequently employing 2 Gjesdal [1981] shows that the ranking of information systems in valuation settings may differ from the rankings of information systems in contracting with managers. Bushman, Engel, Milliron and Smith [2000] develop a theoretical link between the use of earnings in valuation and in contracting with managers, with the compensation weight on earnings associated with the weight on earnings in valuation. 7

10 accounting-based measures in annual contracting with executives. Bushman, Engel, Milliron and Smith [2000] find evidence that the weight placed on earnings in managers cash compensation contracts is positively, though not perfectly, associated with the weight on earnings in valuation over time. Our research design assumes that it is plausible to expect a link between the two uses of accounting information, and we exploit this link in developing a proxy for the informativeness of accounting measures in the process of evaluating executives. We also expect to observe differences in the use of performance measures in governance decisions depending on the extent to which owners are able to engage in direct monitoring of top management. While monitoring is generally difficult to observe directly, the varying extent to which venture capitalists (VCs) hold large ownership stakes in our sample firms offers a proxy for direct monitoring. As the recent literature on the role of VCs documents, these financiers specialize in the formation and development of new enterprises, and it is common for them to play a significant supervisory role (see, for example, Barry et al. [1990], Gompers [1995], and Kaplan and Strömberg [2000a, 2000b]). We therefore expect these owners to have specialized expertise in monitoring the actions of management. Hence, firms with greater VC ownership will rely more heavily on information gathered through monitoring in assessing the performance of top management and will substitute away from other available performance measures. Our analysis here is related to several previous streams of literature. First, prior research on the association between executive compensation grants and performance measures has documented differences in the use of performance measures based on industry-specific or firmspecific characteristics. Ely [1991] conducts a multi-industry analysis of executive compensation over the period 1972 through 1982 and documents that the relation between CEO cash compensation and a set of performance measures differs across the industries studied and 8

11 that the actual performance measures that are related to compensation differ across the industries. Clinch [1991] finds that weights on accounting- and stock-based performance measures in compensation of key employees are increasing in the level of research and development activity of firms over the period Our analysis builds on this prior work first by using industry as a proxy for the usefulness of accounting performance measures and then by considering the joint effect of the information and monitoring environments. Our work is also related to studies of the determinants of managerial ownership. Himmelberg, Hubbard and Palia [1999] find that various firm and industry characteristics are associated with managerial ownership during the period They implicitly consider the properties of the information system to be a potential unobserved effect associated with managerial ownership in their firm fixed effects model. Demsetz and Lehn [1985] investigate determinants of ownership concentration, hypothesizing that control potential from more effective monitoring is positively associated with the extent of concentration of ownership. They find that ownership concentration is positively associated with the volatility of stock market and accounting rates of returns as proxies for control potential. Using industry dummies, Demsetz and Lehn also find that ownership concentration is lower in regulated firms in which subsidized monitoring occurs and is higher in media firms where amenity potential is greater. Our work differs from these analyses by explicitly focusing on capturing differences in the information and monitoring environments and probing how firms consider these differences in their annual governance decisions. Since our interest is primarily in understanding how accounting information is used to mitigate owner/manager agency problems, we emphasize annual compensation grants and CEO turnover decisions in our empirical analysis. While strong effort incentives are provided by 9

12 managers holdings of equity-based instruments, an examination of incentives provided by these holdings would necessarily be silent on the question of whether, and if so how, boards of directors of newly public companies make use of accounting information in governance decisions. 3 We note that boards invest considerable time and resources in assessing managerial performance, both for purposes of making annual performance-based compensation grants and in decisions involving the CEO s continued employment. Hence, an analysis of the information used in making such assessments helps to provide a more complete understanding of firms corporate governance practices. The balance of this section describes our empirical approach and the control variables used in our analysis Compensation grants We first study annual grants of cash and equity-based compensation to CEOs. We begin by describing how the level of annual compensation grants in IPO firms differs across sample industries. Using industry as a proxy for a firm s information environment and venture capital involvement as a proxy for direct monitoring, we perform regression analyses of the relation between accounting- and market-based performance measures and the levels of annual grants of cash and stock-based compensation. In these analyses, it is important to control for various other firm- and executive-level characteristics that may affect grants of CEO compensation, and we note that our control variables here play two key roles. First, prior research suggests these control variables are related to compensation grants to key executives. Second, a number of our 3 Examples of studies examining incentive aspects of equity-based instruments include Himmelberg, Hubbard and Palia [1999] and Demsetz and Lehn [1985]) discussed above. A recent study by Baker and Gompers [1999] conducts an analysis of the determinants of executive incentives at the time of the IPO for a sample of firms undergoing IPOs during the period

13 control variables, including capital intensity, R&D intensity, firm age, and firm risk, have been shown by Baker and Gompers [1999] to be useful predictors of the presence of VC backing. Hence, our regression estimates of the effect of VC backing on compensation hold constant these other factors documented to be related to VC presence. We draw from prior literature to consider control variables capturing firm- and CEOspecific characteristics. Firm characteristics such as size, growth prospects, and age are likely to impact grants of compensation to CEOs. Prior studies have documented a link between firm size and the level of compensation granted to CEOs (see, for example, Murphy [1999]). We control for firm size using the log of the market value of the firms equity. Annual performance-based or incentive pay may depend on growth opportunities if managers in growing firms have greater scope for taking actions that impact firm value. Most prior literature suggests that firms with higher book-to-market ratio, higher capital intensity (as an inverse measure of intangible intensity), and lower research and development (R&D) intensity are likely to have lesser growth opportunities. Also, if younger firms have greater growth opportunities or face tighter cash constraints, then they may elect to place a greater reliance on stock-based pay. We also consider firm-level uncertainty, based on recent work by Prendergast [2000] suggesting that pay-for-performance is more likely to be observed in situations where principals do not have a strong a priori sense for what actions an agent should be pursuing. This argument suggests that incentive pay may be more likely in cases where the overall environment is more uncertain. We proxy for uncertainty using a measure of the stock price volatility for the firm. If stock price volatility reflects overall uncertainty, then incentive pay may be in greater use when volatility is high. 11

14 Models of managerial reputation or career concerns (see Gibbons and Murphy [1992]), Jensen and Meckling [1976] offer various predictions as to how CEO characteristics such as age or tenure may relate to compensation. Older managers, for example, may have shorter horizons than investors, requiring greater reliance on formal incentive contracts. Alternatively, boards of directors may be better able to evaluate CEOs who have been in the job for a longer time, which may suggest a lesser reliance on formal incentives and greater use of implicit arrangements than with younger, newer CEOs. Compensation grants may also be affected by whether the CEO is new (i.e., it is the executive s first year as CEO) and whether the CEO founded the firm. The hiring of a new CEO around the time of an IPO may be indicative of a need for certain skills and expertise useful in managing a public company. A new CEO of an IPO firm, then, may be expected to have a larger impact on firm value (i.e., higher marginal productivity of effort), so higher-powered incentives may be appropriate. In addition, the initial compensation/ownership package provided to a new CEO may reflect labor market conditions or the new CEO s opportunity wage. Similarly, the need for annual performance-based pay may be expected to differ markedly for founder and non-founder CEOs due to possible alternative incentives in place for founders, including implicit incentives or larger equity ownership interests in the firm. We also consider the extent of existing stock holdings in the firm by the CEO. Greater CEO ownership may align the CEO s incentives with the interests of shareholders, reducing the need for additional performance-sensitive pay (see Benston [1985], Murphy [1985]). Alternatively, high ownership can lead to entrenchment. An entrenched executive may be able to consume more firm resources and demand higher compensation because the actions of the executive are less subject to market discipline (see Jensen and Meckling [1976]). Recent empirical work has produced mixed evidence on whether grants of stock-based compensation are 12

15 related to existing managerial ownership (see, for example, Mehran [1995], Yermack [1995], Kole [1997], Core and Guay [1999] and Anderson, Banker and Ravindran [2000]). We control for the CEO s existing holdings in the firm by including a measure of the CEO s percentage ownership of the firm s equity and also conduct sensitivity tests with an alternative measure of CEO dollar wealth in the firm Turnover Next, we consider the relation between firm-level performance measures and post-ipo CEO turnover. Prior literature has documented an inverse relation between the probability of a management change and market and accounting performance of the firm (see Coughlan and Schmidt [1985], Warner, Watts and Wruck [1988], Weisbach [1988]). As with compensation grants, we consider management turnover to be a corporate governance decision made by the board of directors and hypothesize that the directors consider the information environment of the firm in effecting management changes. We therefore expect to observe systematic differences between internet and non-internet firms in the association of performance measures with CEO turnover. 4 We also consider the impact of venture capital involvement on the turnover decision and include controls for other hypothesized CEO and firm-level determinants of turnover, including CEO tenure, age, ownership in the firm s stock, founder status, firm size and firm age. 4 In related work, Engel, Hayes and Wang [2001] probe the notion that the association between available performance measures and board-initiated CEO turnovers is linked with the relative usefulness of the performance measures. They conduct their analyses with a sample of firms included in the annual Forbes executive compensation surveys from and use historical information to develop proxies for the usefulness of performance measures. 13

16 3. Sample selection and descriptive analyses 3.1. Sample Our sample is drawn from the set of firms whose initial public offerings took place between May of 1996 and December of We identified our sample firms from lists of IPO firms produced by ipomaven.com. Our primary data requirements are financial data for the firm, and compensation and stock ownership data for the firm s chief executive. Financial information is taken from CRSP and Compustat. The compensation and stock ownership information is collected from SEC filings using the SEC s EDGAR database. These data are included in the prospectus filed at the time of the firm s IPO and in the proxy statements thereafter. Because firms were required to file financial documents with the SEC electronically starting in May 1996, we are able to obtain the IPO prospectus data from the SEC s EDGAR database for most firms in our initial list of IPOs. We selected three industries, internet, manufacturing, and technology (i.e., non-internet computer), for analysis. These three industries were those with the largest number of IPOs over the sample period. These industries allow for an attractive research design due to expected differences in their information environments, with internet and manufacturing firms at opposite ends of the spectrum and technology firms displaying characteristics of both. Since internet companies have no unique SIC code, we initially identified internet companies from a listing in Morgan Stanley Dean Witter s Internet Company Handbook v.2, including actively traded firms at June 1, We further identified internet companies in our sample using listings of sample firms in prior studies of internet companies (Demers and Lev, 2000; Hand, 2000; Trueman, Wong and Zhang, 2000). For the remaining sample IPO companies not classified as internetrelated, we determined if the firm is internet-related by examining the list of internet firms 14

17 produced by internet.com and by reading descriptions of sample companies in NASDAQ s weekly new companies report. The industry membership of other sample IPO firms is determined using SIC codes, with non-internet technology firms having 3-digit SIC codes of 357, 367 and 737 and manufacturing firms having 3-digit SIC codes between 300 and 399 (excluding 357 and 367). 5 Within these three industries, we identified 545 IPOs for which CRSP and Compustat data were available for at least some part of the sample period. We omitted ADRs and foreign firms, as well as small businesses that were not required to file online until later in 1996, leaving 475 firms. After the removal of spin-off companies, the sample consisted of 464 firms. Missing CRSP or Compustat data in the IPO year further reduced the sample for the regression analyses to 433 firms. Table 1 reports the industry composition and calendar year of the IPO transaction of our sample firms. Our sample consists of 216 (46.6%) firms in the internet industry, 91 (19.6%) in manufacturing and 157 (33.8%) in technology. The number of IPOs from the years 1996 to 1998 ranges from 67 to 80 per year but increases to 239 in 1999, primarily due to a surge in internet IPOs. It is interesting to note that while the number of internet IPOs has dramatically increased over the sample period, the number of manufacturing IPOs has declined. Approximately 16.0% of our sample firms have delisted subsequent to the IPO, due primarily (90%) to acquisition by another firm. 5 3-digit SIC codes of sample internet firms include 300, 357, 366, 367, 481 and 737, with over 90% of internet firms in the category 737, which includes computer programming, data processing and computer services firms. 15

18 3.2. Descriptive analyses We begin by examining descriptive information about compensation and firm characteristics, both at the time of the IPO and in subsequent years. We include mean and median information about these characteristics for the overall sample and separately for each of the three industries in Tables 2 and 3. Table 2 presents information about CEO compensation arrangements and several CEO characteristics, including the CEO s ownership percentage in the firm, CEO age and CEO tenure with the firm. 6 Table 2 reveals that the average value of total annual compensation grants over the sample period is quite high, with $8,641,950 granted in the IPO year and $3,597,380 granted in the post-ipo period, on average across all firms. There are substantial differences in the average value of total compensation grants across firms in the three industries, with average total compensation in the post-ipo period ranging from $1,200,940 in manufacturing firms to $7,477,600 in internet firms. While total cash compensation of CEOs in manufacturing and technology firms is, on average, greater than total cash compensation to internet CEOs, the average value of stock-based awards to internet CEOs causes the average value of internet CEO total compensation to be substantially higher than that of CEOs in manufacturing and technology firms. Further, the mean (median) value of the stock-based component of compensation for all firms is $8,306,340 ($521,200) and $3,177,260 ($0) in the IPO year and subsequent years, respectively. This suggests that stock-based awards are significant in value when awarded; however, some firms do not grant stock-based awards annually or at all. An examination of the sample finds that three-quarters of the firms issue options at least once during the sample period. 6 The appendix provides detailed descriptions and sources of data for each variable in the descriptive tables for both the IPO year and the post-ipo period. 16

19 This finding, combined with the median stock-based compensation value of zero in the post-ipo period, suggests that option grants are common in our sample firms but are, in some cases, lumpy in nature. Table 2 also reports that CEOs of internet firms are younger and have been in their role as CEO for a shorter time than their counterparts in manufacturing and technology IPO firms, with the differences most prominent between internet and manufacturing firms. Overall, Table 2 suggests there are considerable differences in the composition and level of compensation across the industries. Table 3 presents descriptive information about firm performance and other characteristics of sample firms at the IPO date and in the post-ipo period. Internet firms have, on average, substantially larger market capitalizations than manufacturing and technology, while non-internet firms have, on average, larger levels of sales than internet firms. Striking differences exist in both accounting and market return performance of IPO firms in the three industries. Internet firms display lower levels of net income, on average, in the IPO year and subsequent periods than non-internet firms, with internet firms reporting, on average, more overall losses (85.1%) than non-internet firms. In contrast, internet firms experienced substantially higher average market returns in the post-ipo period (121.48%) than manufacturing and technology firms (6.02% and 54.29%, respectively). These relations are consistent with the low associations between aggregate accounting measures of performance and market returns of internet firms documented in prior research and discussed earlier. Table 3 also reveals that the average volatility of monthly market returns of non-internet IPO firms (17.57% and 25.33%, respectively for manufacturing and technology firms) is lower than that of internet firms (33.19%). The differences in return volatility are consistent with the differences across industries in book-to-market ratio the inverse of market-to-book ratio, a 17

20 proxy for growth and investment opportunities which averages for internet firms, compared to and for manufacturing and technology firms, respectively, in the post- IPO period. Overall, the results of the descriptive analyses in Table 3 suggest differences exist in life cycle and in the level of uncertainty about firm value across firms in the three industries, validating our premise of substantial differences across firms in the sample industries. We also present descriptive information about sample firms by the extent of venture capital involvement, partitioning the sample by whether the influence of venture capital is significant, measured as equity ownership by VCs at a level of 20% or greater of the firm. Table 4 reports descriptive information by venture influence and reveals systematic differences in firmand CEO-related characteristics between firms with and without significant venture influence. Venture-influenced firms are, on average, younger and are less likely to be managed by the founder of the firm. Venture-influenced firms also display greater average market capitalizations, lower (and quite negative) net income and lower book-to-market ratios. CEOs of firms with more significant venture influence owned a smaller share of the firm both at the IPO date and in subsequent years. This smaller CEO ownership percentage may reflect either a crowding out of CEO ownership by VCs or the larger market capitalization of ventureinfluenced firms, which would lead to a smaller ownership percentage for a given dollar value of equity holdings by the CEO. 4. Empirical design and results In this section, we present our empirical analyses of the link between available performance measures and annual governance decisions relating to CEO compensation grants and turnover in the years immediately subsequent to the IPO. We analyze the use of 18

21 performance measures in annual compensation grants to CEOs in years after the IPO. We examine grants of total CEO compensation and its two primary components cash and grants of equity-based instruments. We first analyze determinants of CEO compensation, including performance measures and CEO and firm characteristics, and then consider the impact of direct monitoring by venture capitalists on the use of performance measures. Finally, we examine factors associated with CEO turnover in the post-ipo period Economic determinants of annual grants of CEO compensation We first examine the impact of the firm s information environment on the relation between CEO compensation and firm performance for the post-ipo period. We hypothesize that firms with low quality information from accounting measures will substitute away from accounting measures toward other measures of performance. Our primary empirical tests focus on earningsbased measures as summary measures of accounting performance. Earnings-based measures are arguably the most prominent summary accounting measures of firm performance and are cited as the most frequently used performance measures in surveys of executive compensation contracts (see, for example, Murphy [2000]). We also conduct analyses using alternative accounting measures of performance based on components of earnings (e.g., sales, gross margin and operating profit), in light of the recent work discussed above that finds a weak relation between market valuation and summary earnings measures. We use stock returns as a proxy for a measure of value-relevant information other than earnings that we expect to receive more attention in contracting when accounting quality is low. If contracts make use of public, valuerelevant information, then stock returns will help explain compensation even if returns are not explicitly contracted upon. 19

22 We estimate the following model of the determinants of annual post-ipo grants of compensation and its components: log( CEO Compensation) = α + α Mfg + α Tech + α Accounting Performance 3 + α Stock Returns + α MV + α CEO Tenure + α Founder + α CEO Age j= 1 4 j + α Firm Age + α New Ceo + α Zero Cash Policy + α CEO Ownership% 9 + α Stdev Ret + α PPE _ TA + α RD _ TA + α Book to Market + ε j= 1 3 j (1) where CEO Compensation represents the log of total compensation (TCOMP) or one of its components, cash (TCASH) or stock-based compensation (TSTOCK) 7, Mfg (Tech) = 1 if the firm is a general manufacturing (technology) firm; 0 otherwise, j represents the industry category (1=internet, 2=manufacturing, 3=technology), Accounting Performance represents one of two aggregate accounting measures of performance for the fiscal year: core earnings or loss (EARN) or return (core) on assets (ROA), with core earnings measured as net income (loss) before extraordinary items, discontinued operations and special items, Stock Returns = total annual stock return of the firm for the fiscal year, MV = the natural log of the market value of common equity of the firm at the end of the fiscal year, CEO Tenure = CEO s tenure with the firm at the end of the fiscal year, Founder = 1 if the CEO is a founder; 0 otherwise, CEO Age = age of CEO at the end of the fiscal year, Firm Age = the number of years since the inception of the firm at the end of the fiscal year, New CEO = 1 if the CEO is new during the IPO year; 0 otherwise, Zero Cash Policy = 1 if the firm discloses an explicit policy not to pay cash compensation to CEO; 0 otherwise, CEO Ownership% = the percentage of the firm s equity owned by the CEO at the end of the fiscal year, 7 TCASH equals the sum of annual cash compensation, including salary, bonus and other annual cash as reported in the annual proxy statement. TSTOCK equals the sum of the value of annual grants of stock options and restricted stock. Value of restricted stock grants is as reported in the annual proxy statement or prospectus. Stock options grants are valued using Black-Scholes according to the terms disclosed in the proxy statement or prospectus, with volatility measured as industry median standard deviation of monthly stock returns in the year prior to the IPO, a risk-free rate of 6% and an assumption of no dividends over the option term. The risk free rate (as measured by the 12-month treasury bill rate) over the sample period ranged from 4.2% to 6.1%, averaging 5.4%. All results are similar if option holdings are valued using a risk free rate of 5%. 20

23 Stdev Ret = the standard deviation of monthly annual stock returns for the firm for the fiscal year, PPE_TA = the ratio of property, plant and equipment to total assets at the end of the fiscal year, RD_TA = the ratio of research and development expenses to total assets at the end of the fiscal year, Book to market = the ratio of common equity to the market value of equity at the end of the fiscal year. The appendix further details descriptions of variables and sources for data. We estimate the model using data for all firms for all available years subsequent to the IPO date through The model is estimated for the log of total compensation (TCOMP) along with its two key components, cash (TCASH) and stock-based compensation (TSTOCK). 8 We also consider two aggregate measures of accounting performance earnings, EARN, and return on assets, ROA, both computed using core earnings in the numerator as described above. 9 The model includes proxies for various CEO- and firm-specific factors discussed in section 2 that may impact the level of annual compensation grants by firms to CEOs. 10 The results of the estimations of equation (1) are presented in Table 5. Panel A presents the results of the estimations involving total compensation, while Panel B includes the results 8 We perform a log transformation of the compensation measure before estimating the model due to skewness in the distribution resulting from a large number of firms not granting stock compensation each year. 9 We do not predict differences in the use of the two summary accounting measures of performance (EARN and ROA) by firms in the different industries, but rather include two earnings-based measures because firms differ greatly in the specific performance measure(s) used in their contracts. Potential heteroscedasticity and coefficient bias issues relating to EARN are addressed by the inclusion of market value as a scale proxy, as suggested by Barth and Kallapur [1996]. We also obtain qualitatively similar results when all estimations in this section are conducted with bottom-line earnings (i.e., net income) rather than core earnings and using two alternative measures of earnings: basic and diluted earnings per share. 10 An alternative specification used in some prior studies of the pay-performance relation would be a model of changes in compensation grants as a function of changes in accounting performance. Such a model mitigates the need to consider many of the CEO- and firm-specific variables we include as potential factors impacting grants of compensation to CEOs. We use a levels specification with appropriate controls for two primary reasons: 1) we feel a levels approach is more appropriate for the analysis of stock grants, which are an important compensation component for our sample firms and 2) the levels analyses with appropriate CEO- and firm-specific controls allow for more degrees of freedom, since a change specification would result in a large reduction in the number of sample firms due to the timing of the IPO for many sample firms. As a check of our results relating to the cash component, we regress the log of bonus (i.e., the more variable cash component) on returns, size, and changes in both of our earnings measures. We find significant positive coefficients on technology firms earnings changes in both the EARN and the ROA model. 21

24 using cash compensation and stock-based grants of compensation. 11 The results suggest that total compensation for internet firms is significantly positively associated with stock returns. This is in contrast with the results for non-internet firms, which display no association between total compensation and stock returns. Panel B suggests that the positive association between compensation and returns for internet firms is attributed to the stock-based component of compensation. Total compensation of manufacturing firms displays a significant positive association with earnings. Further, cash compensation of non-internet (technology) firms is significantly positively linked with both EARN and ROA. In contrast, internet firms display no association between compensation and EARN and a negative association between total compensation and ROA. 12 While we offer no conclusive explanation for this negative association with ROA, we observe that it is consistent with the findings in recent research of a negative association between earnings and market valuation in internet firms (e.g., Hand [2000]). We also consider the possibility that boards of directors of internet firms find accounting information other than core earnings to be useful performance measures in determining CEO compensation grants. We re-estimate equation (1) for total compensation and its components with measures based on components of earnings, including sales, gross profit and operating income. 13 We find similar qualitative results in these analyses, with cash compensation of non- 11 Panel B reports the results of the model with the log of stock compensation as the dependent variable using White s adjusted t-statistics, due to White s test suggesting the potential for heteroscedasticity. White s tests for all other models do not suggest the presence of heteroscedasticity. The stock regressions were also run using a tobit specification, due to the many grants of zero stock in any given year. The inferences in the tobit specification are unchanged from those presented in the table. 12 Tests of equality of coefficients on earnings measures between internet and non-internet firms reject the null hypothesis in the total compensation and total cash compensation models, except for ROA in the cash model. Similar tests on stock return measures between internet and non-internet firms do not reject the null hypothesis of equality in the total compensation and stock compensation models. 13 Each of the component measures is significantly correlated with EARN or ROA in the manufacturing and technology industries, while component measures generally display no or negative association with EARN or ROA in the internet industries. We separately include each of the alternative accounting measures in the estimation of equation (1) both in undeflated form (analogous to EARN) and deflated (analogous to ROA). Specifically, the 22

25 internet firms displaying significant positive association with the accounting-based measures. In contrast, cash compensation of internet firms show no association with accounting measures, while total and stock compensation of internet firms display significant association with stock returns. Overall, the results in Table 5 suggest systematic differences in the use of accountingbased measures and the information in stock returns for sample internet and non-internet IPO firms. In particular, the use of accounting information is associated with cash compensation decisions for non-internet sample firms, while internet firms appear to rely instead on stock market returns in compensation decisions in granting stock-based compensation. The results are consistent with our hypotheses that when the informativeness of accounting information is relatively low, the association between CEO grants of compensation and accounting performance will be relatively low and that the association between grants of compensation and other performance information will be relatively high. The estimations of equation (1) control for factors found to be related to compensation levels in prior research. All forms of compensation are positively associated with firm size as measured by the natural log of the market value of equity. The CEO s overall ownership percentage of the firm (CEO Ownership %) is negatively associated with total compensation and its cash and stock-based component. Total and stock compensation are also significantly positively associated with the New CEO dummy variable. The links between stock compensation and both CEO Ownership % and New CEO suggest that boards of directors consider existing executive holdings in the firm s stock in decisions concerning annual grants. More specifically, these links suggest that lower annual incentives through compensation are deflated measures we use are sales-to-total assets, gross margin percentage (gross margin divided by sales) and operating income-to-total assets. 23