This document discusses using cash flow patterns as a proxy for firm life cycle stages. It proposes that cash flow patterns provide a parsimonious yet robust way to identify a firm's life cycle stage without distributional assumptions. The paper aims to validate cash flow patterns as a life cycle proxy and examine how life cycle stage affects profitability persistence over time, determinants of future profitability, and market valuation of firms. Key findings include that return on net operating assets does not fully mean-revert within 5 years when sorted by life cycle stage, and that mature firms appear undervalued by market participants, especially those with extreme book-to-market ratios.
Original Description:
Determine business maturity profile from its cashflow statement
This document discusses using cash flow patterns as a proxy for firm life cycle stages. It proposes that cash flow patterns provide a parsimonious yet robust way to identify a firm's life cycle stage without distributional assumptions. The paper aims to validate cash flow patterns as a life cycle proxy and examine how life cycle stage affects profitability persistence over time, determinants of future profitability, and market valuation of firms. Key findings include that return on net operating assets does not fully mean-revert within 5 years when sorted by life cycle stage, and that mature firms appear undervalued by market participants, especially those with extreme book-to-market ratios.
This document discusses using cash flow patterns as a proxy for firm life cycle stages. It proposes that cash flow patterns provide a parsimonious yet robust way to identify a firm's life cycle stage without distributional assumptions. The paper aims to validate cash flow patterns as a life cycle proxy and examine how life cycle stage affects profitability persistence over time, determinants of future profitability, and market valuation of firms. Key findings include that return on net operating assets does not fully mean-revert within 5 years when sorted by life cycle stage, and that mature firms appear undervalued by market participants, especially those with extreme book-to-market ratios.
Victoria Dickinson, PhD, CPA E. H. Patterson School of Accountancy University of Mississippi
Current Draft: November 2010
Correspondence: Victoria Dickinson, University of Mississippi E.H. Patterson School of Accountancy, Conner 204C, University, MS 38677; Phone: (662) 915-5448; Fax: (662) 915-7483; email: vdickins@olemiss.edu.
This paper is based on my dissertation at the University of Wisconsin Madison. I would like to thank the members of my committee: Don Hausch, Jed Frees, Holly Skaife, Terry Warfield, and John Wild (Chair), along with Mike Donohoe, Frank Heflin, Ryan LaFond, Matt Magilke, Bill Mayew, Brian Mayhew, Per Olsson, K. Ramesh, Greg Sommers, Jenny Tucker and Rodrigo Verdi for their helpful discussions and comments. This paper has also benefited from the comments of conference participants at the American Accounting Association and from workshop participants at the College of William and Mary, Florida State University, Miami University, University of Alabama at Birmingham, University of Florida, University of Mississippi, University of Utah, and University of Wisconsin - Madison. 2 Cash Flow Patterns as a Proxy for Firm Life Cycle
Abstract: This paper examines the validity of cash flow patterns as a proxy for firm life cycle. Cash flow patterns provide a parsimonious, but robust, indicator of firm life cycle stage that is free from distributional assumptions inherent when using a univariate or composite measure. Life cycle stage is an important determinant of the level and time series properties of profitability. Return on net operating assets (RNOA) does not mean-revert (spread of seven percent after five years) when sorted by life cycle stage which has implications for forecasting growth rates and for determining forecast horizons. Determinants of future profitability identified in prior literature such as asset turnover and profit margin are differentially affected by life cycle stage. Furthermore, market participants undervalue mature firms, specifically those with extreme book-to-market ratios. The cash flow pattern proxy has numerous applications in forecasting and analysis and is a useful control variable for future research.
3 1. Introduction Business firms are evolving entities and the path of evolution is determined by internal factors (e.g., strategy choice, financial resources, and managerial ability) and external factors (e.g., competitive environment, macroeconomic factors). Firm life cycles are distinct phases that result from changes in these factors, many of which arise from strategic activities undertaken by the firm. Lev and Zarowin (1999) document that: (1) the rate of business change has increased over time, and (2) the value-relevance of earnings has decreased over time. Taken together, these findings suggest that a non-earnings-based measure that captures firm life cycle stage would be useful to investors and creditors. In response, this paper develops and validates a parsimonious proxy for firm life cycle based on a firms cash flows patterns. Capturing life cycle at the firm level (rather than at the individual product or industry level) 1
is a difficult undertaking. The firm is an aggregation of multiple products, each with a distinct product life cycle stage. Additionally, the firm may compete in multiple industries such that its product offerings are quite diverse. As a result, firm-level life cycle stage is difficult to assess because it is a composite of many overlapping, but distinct product life cycle stages. The economics literature has addressed individual attributes of life cycle theory such as production behavior (Spence 1977, 1979, 1981; Wernerfelt 1985; Jovanovic and MacDonald 1994), learning/experience (Spence 1981), investment (Spence 1977, 1979; Jovanovic 1982; and Wernerfelt 1985), entry/exit patterns (Caves 1998), and market share (Wernerfelt 1985). I propose a life cycle proxy based on accounting information (specifically cash flow patterns) that can be linked to these constructs from the extant economics literature.
1 The following studies examine the validity of product life cycle for the following industries: German automobile manufacturers (Brockhoff 1967); pharmaceuticals (Cox 1967); tobacco, food, and personal care products (Polli and Cook 1969); household cleansers (Parsons 1975). 4 Anthony and Ramesh (1992) was one of the first studies to demonstrate the usefulness of firm life cycle in explaining market performance. However, their sample period ended before the Statement of Cash Flows was a required disclosure. Therefore, their life cycle measure relied on a composite of economic characteristics such as sales growth, dividend payout, capital expenditures, and firm age. By necessity, their life cycle proxy had to rely on portfolio sorts to draw distinctions between the life cycle stages. On the other hand, the classification scheme used in this paper is accomplished by using a firms operating, investing, and financing cash flows in combination to assign life cycle stage. The classification methodology is organic in that life cycle stage identification is the result of the firms performance and allocation of resources, as opposed to arbitrary assignment. I use the cash flow patterns proxy to document the economic characteristics and market behavior of firms within each life cycle stage. Specifically, I validate the cash flow proxy in relation to competing measures for firm life cycle and find that it is better aligned with the functional form of firm profitability. Economic theory predicts a nonlinear progression of variables such as earnings, return on net operating assets (RNOA) 2 , asset turnover (ATO), profit margin (PM), sales revenue, leverage, dividend payout, size and age across the life cycle continuum which is consistent with the distribution that results from using cash flow patterns as a life cycle proxy. Once the descriptive validity of the cash flow pattern proxy for life cycle is established, I examine which life cycle stages demonstrate persistence in profitability and investigate the inter- temporal convergence of profitability by life cycle stage. Previous research documents that profitability measures mean-revert over time (Brooks and Buckmaster 1976; Freeman, Ohlson and
2 RNOA removes the effect of financing from the profitability metric and is measured as operating income divided by net operating assets (NOA). NOA excludes financial assets from the denominator since they are already valued at fair value on the balance sheet. NOA also subtracts out operating liabilities from operating assets. This is because operating liabilities reflect a source of leverage that can increase profitability. 5 Penman 1982; Fairfield, Sweeney and Yohn 1996; Fama and French 2000; Nissim and Penman 2001) and understanding the evolution of profitability improves predictability. Conditioning an inter-temporal analysis by life cycle stage is expected to explain differences in convergence rates across firms. The results support that prediction. Specifically, RNOA maintains a differential spread of three to 10 percent between decline and mature firms even after five years. This difference is economically significant and suggests differences in firm life cycle are an impediment to the mean- reversion of profitability. Prior research has demonstrated that changes in future accounting returns (specifically RNOA) are explained by level and change of current profitability, growth in net operating assets, and by increases in the asset turnover (ATO) (Fairfield and Yohn 2001). Since firm life cycle stage differentially explains profitability, including the cash flow pattern proxy for life cycle also provides significant information about future change in RNOA, and the explanatory power of the cash flow pattern proxy outperforms alternative life cycle proxies. Furthermore, the explanatory power of changes in ATO for changes in RNOA is concentrated in the mature life cycle stage. This is consistent with economic theory that states competitive pressures will drive mature firms to focus on efficiency and cost containment. Also, Penman and Zhang (2006) find that increases in profit margin (PM) result in negative future RNOA because those increases are achieved through a reduction of operating expenses which are not sustainable. Again, this effect is concentrated in the mature life cycle stage, where product differentiation efforts (which manifest in the PM) have reached diminishing returns (Oster 1990, Shy 1995). Considering that life cycle affects non-convergence of profitability over time and that life cycle differences interact with the determinants of profitability, the next step is to investigate what role life cycle plays in understanding current firm value and predicting stock returns. The market does not fully capture information provided by the life cycle proxy such that mature firms earn 6 positive excess returns in the year following life cycle stage assignment. This indicates that investors underestimate the persistence of the elevated profitability of mature firms and instead expect their profitability to mean-revert to a normal level. Prior literature has found that profitable trading strategies can be developed based on extreme values of the book-to-market ratio (B/M) where low B/M firms are referred to as glamour firms and high B/M firms are considered value firms (Fama and French 1992, Lakonishok et. al. 1994). Two common B/M trading strategies (Piotroski 2000, Mohanram 2005) focus on using fundamental analyses to separate out firms that are undervalued versus those that overvalued among the glamour/value firms. The fundamental analyses used in this line of research are comprised of composite scores of numerous metrics (ranging from eight to nine signals for each fundamental analysis score). The cash flow pattern proxy for life cycle provides a simple and effective way to identify the winners among both B/M extremes using considerably less data. All market results are robust to consideration of delisting returns and execution issues regarding shorting illiquid securities. In summary, this paper presents and validates cash flows patterns as a parsimonious proxy for identifying firm life cycle stage. This classification is useful in the following contexts: (1) to better assess growth rates and forecast horizons in valuation models; (2) to better understand how economic fundamentals affect the level and convergence properties of future profitability; (3) to identify firms where potential unidentified risk factors and/or market mispricing exist based on differences in life cycle stage; (4) as a control variable for distinct economic characteristics of the firm related to firm life cycle that affect firm performance. These contributions benefit equity investors, creditors, auditors, analysts, regulators, and researchers alike. The remainder of the paper is organized as follows: Section 2 presents and validates the firm life cycle proxy, Section 3 7 demonstrates the usefulness of the proxy in explaining future profitability, Section 4 examines the market implications of life cycle information for future stock returns, and Section 5 concludes.
2. Is the proposed life cycle classification scheme descriptively more valid? Gort and Klepper (1982) define five life cycle stages: (1) an introductory stage, where an innovation is first produced; (2) a growth stage, where the number of producers increases dramatically; (3) a maturity stage, where the number of producers reaches a maximum; (4) a shake-out stage, where the number of producers begins to decline; and (5) a decline stage, where there is essentially zero net entry. I propose that cash flows capture the outcome of these distinct life cycle stages. Livnat and Zarowin (1990) document that the decomposition of cash flows into operating, investing and financing activities differentially affects stock returns. Therefore, cash flows capture differences in a firms profitability, growth, and risk; and the combination of those cash flows are mapped into life cycle theory to derive the life cycle classification used throughout the paper. A survey of economic theory related to life cycle and the predicted relation to cash flows is summarized in Table 1. The combination of cash flow patterns represents the firms resource allocation and operational capabilities interacted with the firms choice in strategy. Predictions about each cash flow component can be derived from economic theory which forms the basis for the cash flow patterns proxy for life cycle. For example, introduction firms lack established customers and suffer from knowledge deficits about potential revenues and costs, both of which result in negative operating cash flows (Jovanovic 1982). However, profit margins are maximized during increases in investment and efficiency (Spence 1977, 1979, 1981; Wernerfelt 1985) which means that operating cash flows are positive during the growth and maturity stages. Wernerfelt (1985) points out that declining growth rates will eventually lead to declining prices such that operating cash flows will decrease (and become negative) as firm enter the decline stage. 8 Managerial optimism (Jovanovic 1982) encourages firms to make early investments which ultimately serve to deter entry into the market by competitors (Spence 1977, 1979, 1981). Consequently, investing cash flows are negative for introduction and growth firms. While mature firms decrease investment relative to growth firms, they will invest to maintain capital (Jovanovic 1982, Wernerfelt 1985). If the cost of maintenance increases over time (i.e., rising prices), investing cash flows are negative for mature firms, although at a lesser magnitude than cash outflows for introduction and growth firms. Decline firms will liquidate assets in order to service existing debt and to support operations which results in positive cash flows from investing. Pecking order theory states that firms initially access bank debt followed by equity later in their life (Myers 1984, Diamond 1991). Barclay and Smith (1995) demonstrate that growth firms will access debt that is shorter in duration than mature firms. Taken together, financing cash flows are positive for introduction and growth firms. However, mature firms will begin to service debt and distribute cash to shareholders as they exhaust positive net present value investment opportunities, which results in negative financing cash flows. There is a void in the literature with respect to whether financing cash flows will be positive or negative for decline firms. 3
The combination of a firms patterns of operating, investing, and financing cash inflows and outflows provide a firm life cycle mapping at a given point in time. By using the sign (positive or negative) of the net operating, investing and financing cash flows, eight possible cash flow pattern combinations exist. 4 The eight classifications are collapsed into the five practical life cycle stages
3 Likewise, the literature is silent regarding cash flows for shake-out firms. As a result, shake-out firms classified by default if the cash flow patterns do not fall into one of the other theoretically-defined stages. 4 Incorporating the sign and magnitude of the cash flows would likely improve performance of the proxy. However, if positive (negative) cash flows were separated into low- and high-positive (negative) cash flows, the number of patterns would increase to 64, which is less straightforward to connect to economic theory. For that reason, only sign is considered. 9 (mentioned at the beginning of the section): introduction, growth, mature, shake-out, and decline based on expected cash flow behaviors from Table 1. 5
A benefit of the cash flow pattern proxy used in this paper, is that it uses the entire financial information set contained in operating, investing, and financing cash flows rather than a single metric to determine firm life cycle. As mentioned in the previous section, prior literature uses sorts on variables such as age, sales growth, capital expenditures, dividend payout or some composite of these variables to assess life cycle stage (Anthony and Ramesh 1992, Black 1998). The drawback of these methods, however, is that an ex ante assumption is required regarding the underlying distribution of life cycle membership. By forming portfolios sorted on a single variable (or a composite of those sorts), a uniform distribution of life cycle stages across firms is inherently assumed. Cash flow patterns, on the other hand, are the organic result of a firms operations and achieve better congruence with economic theory (i.e., a normal distribution). Both size and firm age are common proxies for life cycle. 6 When size and age are used as a life cycle proxy, an implicit assumption is that a firm moves monotonically through its life cycle. This assumption arises because product life cycles are characterized by forward progression from introduction to decline. However, a firm is a portfolio of multiple products, each at potentially a different product life cycle stage. Substantial product innovations, expansion into new markets or
5 There are three cash flow types (operating, investing, and financing) and each can take a positive or negative sign which results in 2 3 = 8 possible combinations. The 8 patterns are collapsed into five stages as follows:
6 Recent accounting research that relies on firm size or age to capture life cycle effects includes Bradshaw et al. (2010), Khan and Watts (2009), Caskey and Hanlon (2007), Doyle et al. (2007), Desai et al. (2006), Freeman et al. (2006), Klein and Marquardt (2006), Wasley and Wu (2006), Bhattacharya et al. (2004), and Chen et al. (2002). An abstract search for firm age in SSRN yielded 592 results while a search for firm size exceeded the maximum results of 1,000 papers. 1 2 3 4 5 6 7 8 Introduction Growth Mature Shake-Out Shake-Out Shake-Out Decline Decline Predicted sign Cash flows from operating activities + + + + Cash flows from investing activities + + + + Cash flows from financing activities + + + + 10 structural change can cause firms to move across life cycle stages non-sequentially. For that reason, firm life cycle can be cyclical in nature and the firms primary goal is to maintain its firm life cycle at the growth/mature stage where the reward-risk structure is optimized. Additionally, firms can enter the decline stage from any of the other stages. The management literature documents a liability of newness phenomenon (Stinchcombe 1965; Jovanovich 1982, Freeman, Carroll and Hannan 1983; Amit and Schoemaker 1993), which means that variation in the level of initial endowments (monetary resources, technological or managerial capability, etc.) interacts with mortality rates. 7 Therefore, firms in the decline stage are likely to include very young firms that succumb to this high initial mortality rate. Additionally, firm life cycle differs from firm age because firms of the same age can learn at different rates due to imperfections in their feedback mechanisms (i.e., accounting quality). All of these factors result in a misalignment between firm performance and firm age causing life cycle progression to be nonlinear in age. Taken together, size and age should be maximized during the mature stage due to nonlinearity. The first step of this paper is to validate whether the life cycle stages based on cash flow patterns are consistent with economic theory with empirical analysis. Firms listed on the NYSE, AMEX and NASDAQ exchanges (excluding ADRs) with necessary data on Compustat comprise the sample.
The sample period extends from 1989 (the first year data from the Statement of Cash Flows was available for all firms) through 2005. Firms with average net operating assets (NOA), sales revenue, or market value of equity less than $1 million are excluded from the sample because small denominators skew the profitability metrics used throughout the paper. Similarly, firms with an absolute book value of equity less than $1 million are excluded from the sample. Firms in the financial industries are excluded due to the unique nature of their cash flows. Firms with SIC codes
7 For example, Jovanovic (1982) presents an analytical model where firms hazard rates (probability of failure) initially increase in the early life cycle stages. 11 greater than 9100 are omitted to ensure only for-profit firms are included in the sample. These constraints result in a final sample of 48,369 firm-year observations. If mature firms are stable, then the greatest frequency of observations will be present in this category, while the lowest frequency of observations will be in the decline stage. Table 2 Panel A confirms this prediction with 41 (five) percent of the firms classified as mature (decline). The effect of firm life cycle on several economic characteristics is tested by regressing each economic characteristic on indicator variables for life cycle membership, with the intercept capturing the mature stage. The coefficients on introduction, growth, shake-out and decline are the effect on the dependent variable relative to the mature stage. [Insert Table 2 about here] Economic theory predicts that profitability is maximized in the mature stage and this is confirmed by the regressions on Earnings per Share (EPS) and Return on Net Operating Assets (RNOA) (all coefficients except for mature are negative). The components of profitability, profit margin and asset turnover, are a function of strategy and the competitive environment. For example, Selling and Stickney (1989) point out that product-differentiating firms focus on research and development, advertising, and capacity enlargement. These expenditures should result in a higher profit margin (PM), which results confirm is maximized in the mature stage. Selling and Stickney also indicate that as firms mature, competition becomes more intense and the emphasis shifts to cost reduction and improved capacity utilization. This means that asset turnover ratios (ATO) should increase in the mature stage, which is confirmed by the regression results (the coefficient for growth firms is negative which indicates that asset turnover in the growth stage is lower relative to the mature stage). The high level of ATO in the introduction stage may be due to investments in uncapitalized assets such as research and development and/or operating leases. 12 Immediate expensing of these assets result in lower GAAP asset levels, which in turn depress ATO (and RNOA) relative to mature stage firms. 8
Growth in sales (GrSALES) and capital investment (GrNOA) should monotonically decrease across life cycle stages (Spence 1977, 1979, 1981), both of which are verified for introduction through shake-out. Market-to-book (MB) is thought to proxy for both expected future growth and risk. This suggests that mature firms should possess the lowest relative market-to-book as compared with the tail observations and the results support this prediction. With respect to other measures of risk, firms should make greater use of financial leverage (LEV) in the growth stage (Myers 1984, Diamond 1991) and asset beta (ASSET BETA) (an unlevered measure of business risk) should be minimized for mature firms. Dividends (DIVPAY) are more likely to be paid by mature firms due to decreased investment opportunities. The results confirm the above for LEV, ASSET BETA and DIVPAY. Advertising intensity (ADVINT) and research and development (INNOV) should be highest for early-stage firms as they build their initial technology. Both variables are highest in the introduction stage (although decline firms appear to increase their research and development, perhaps in a turnaround attempt). The number of segments (SEGMENTS) is expected to increase through maturity as growth is executed via product and/or geographical expansion, which the results confirm. Merger activity (MERGER) should be greatest for growth firms as they will likely be targets for acquisition and results indicate that introduction and growth stages are where the highest degree of activity takes place. Finally, size (SIZE) and age (AGE) are maximized for mature stage firms consistent with life cycle being nonlinear in both of these variables. Cash flow patterns
8 I thank one of this papers reviewers for providing this insight. 13 result in a proxy that is representative of the economic theory underlying the life cycle phenomenon. 9
The preceding analysis is repeated for life cycle classifications based on Anthony and Ramesh (1992) (Table 2, Panel B) and age quintiles (Table 2, Panel C). In Anthony and Ramesh paper, the median values of dividend payout, sales growth, and capital expenditure (scaled by market value) are computed over a five-year horizon. Those values, along with age are used to assign scores to each observation. 10 Composite scores are used to form five equal-sized portfolios which they label Growth, Growth/Mature, Mature, Mature/Stagnant, and Stagnant. Similarly, five equal-sized portfolios are formed on firm age, designated as Young, Mid-Young, Middle, Mid-Old and Old. In the last column of each panel, Vuong statistics are computed to compare the performance of the cash flow patterns proxy to the Anthony and Ramesh (hereafter, AR) in Panel B and the age quintile classifications in Panel C. 11 The cash flow patterns proxy significantly outperforms the AR proxy for EPS, RNOA, PM, GrSALES, GrNOA, INNOV, and MERGER. All results hold for cash flow patterns versus age with the exception of GrSALES. Recall that the AR classification is the composite of sorts on sales growth, capital expenditures, dividend payout, and age. As a result, AR provides better explanatory power for DIVPAY and AGE (other components of ARs proxy). ARs measure is also more descriptive of
9 It is possible that the results reported in Table 2 are due to industry effects. If life cycle is actually an industry phenomenon, then a simple industry control would capture the differences across firms. The economics literature suggests industry life cycle patterns occur because the rate of innovation and intensity of competition change over the industry life cycle. However, individual firms life cycle stages may differ within an industry because innovation is a continuing process with firms entering and exiting the market throughout the entire industry life cycle. Furthermore, the life cycle stages of individual firms within an industry vary due to differences in a firms knowledge acquisition, initial investment and re-investment of capital, and adaptability to the competitive environment. All tests throughout the paper were repeated on industry-adjusted samples based on the Fama French (1997) industry classifications (results untabulated). In each case, the results were consistent with those reported throughout the paper. Therefore, firm life cycle stage is distinct from industry life cycle. 10 Terciles of low to high dividend payout, high to low sales growth, high to low capital expenditure and young to old age were given scores of one to three, respectively. The composite scores ranged from three to nine as in Anthony and Ramesh (1992). 11 A positive (negative) Vuong statistic indicates that the cash flow patterns classification provides a better (worse) fit for explaining each economic characteristic than does the alternative proxy. 14 ATO, MB, SEGMENTS and SIZE. To reconcile the differences in performance between the cash flow patterns proxy and the AR classification, it should be pointed out that the cash flow patterns capture the nonlinearity of earnings, RNOA, PM, level of sales (as opposed to sales growth), ATO, age, etc. across the economically-defined life cycle stages. All of these variables are maximized in the mature stage (i.e., an inverted-U shaped distribution). If profitability drives market performance, then the cash flow proxy will be effective in predicting future performance from a profitability and market perspective. Next, age quintiles outperform the cash flow proxy for the same set of economic characteristics in which AR provided better explanatory power, which is expected since age was a component of ARs measure. Since young firm contain both introduction and decline firms, age provides less detail than the cash flow proxy with respect to low-performing firms. Having validated the cash flow patterns proxy, in the next section I use the proxy to examine the inter-temporal behavior of firms conditional on their initial life cycle stages. Survivorship bias is an issue inherent in this type of analyses. Table 3 Panel A examines the proportion of firms that survive for five years subsequent to the initial life cycle identification. To ensure a time span of five years subsequent to initial identification, a reduced sample from 1989 to 2000 is used (the sample period ends in 2005). 12 In the pooled sample, only 78 percent of firms survive five years ahead. Possible reasons for absence include merger activity, going private, or bankruptcy. The proportion of firms that delist for merger or performance-related issues are reported according to their life cycle stage in the year prior to delisting, and this breakdown is displayed in the last two columns. 13
Delisting proportions by life cycle stage are significantly different from a uniform distribution across life cycle stages (all z-statistics are significant). Over 65 percent of merger activity involves growth
12 Since one-year-ahead profitability is required for tests of explanatory power in Section 3, all firms survive through year t + 1 by construction. 13 Cash flow data is insufficient for computing life cycle stage in the year in which the delisting takes place. Delisting codes of 200-299 are categorized as Merger and codes of 500-599 are categorized as performance-related (Beaver, McNichols and Price 2007). 15 or mature firms and over 68 percent of performance-related delistings involve introduction or decline firms. [Insert Table 3 about here] The survivorship analysis is repeated by life cycle stage for the proportion of firms in existence relative to the year of life cycle identification and z-statistics are computed on differences in proportion between each life cycle stage and the pooled sample. The survival rates for mature (decline) stage firms are significantly higher (lower) than the survival rates for the pooled sample in all subsequent years. The introduction and growth firms are significantly lower than the pooled sample for years t + 3 through t + 5. Table 3 - Panel B examines the transition of firm-observations from one life cycle stage to another in subsequent periods, again using the reduced sample to ensure five subsequent years from the initial classification. The shaded portions of the table represent the proportion of firms that remain in their initial stage in later periods. For example, 60.13 percent of mature firms remain in the mature stage one year after initial classification. This proportion monotonically decreases to 55.97 percent by year t + 5. Several observations are worth noting from this analysis: (1) introduction firms are likely to stay in introduction or to move to the growth or mature stage (over 80 percent of observations at end of five years), (2) growth firms are fairly stable but a large proportion (ranging from 33.11 in year t + 1 and increasing to 43.23 percent by year t + 5) will move to mature, (3) growth firms are not likely to move to decline (less than four percent are in decline by year t + 5), (4) mature firms are stable and slightly less than 30 percent transition to growth over the next five years, (5) movements from mature to decline are unlikely (only two percent transition to decline during the five subsequent years), (6) a small proportion of decline firms remain in decline (only 18 percent after five years), but movement to introduction (25 percent by year 5) or to growth (23 percent), mature (20 percent) or shake-out (12 percent) is also fairly 16 common. Taken together, introduction firms tend to improve their position, growth and mature firms are relatively stable, and decline firms (that survive) have unpredictable outcomes in subsequent periods.
3. How can life cycle classification help us better understand future profitability? Previous research documents that profitability measures mean-revert over time (Brooks and Buckmaster 1976; Freeman, Ohlson and Penman 1982; Fairfield, Sweeney and Yohn 1996; Fama and French 2000; Nissim and Penman 2001) and understanding the evolution of profitability improves predictability. Stigler (1963) reported that profitability displayed a strong central tendency over time, but that the convergence was incomplete. He stated that impediments to complete convergence stem from disturbances related to shifts in demand, advances in technology, and macroeconomic factors. Another potential impediment to convergence is differences in firm life cycle stage. Nissim and Penman (2001) suggest that truncated forecast horizons can be used if valuation attributes settle down to permanent levels within the forecast horizon. Therefore, understanding the convergence properties of profitability leads to better decisions with respect to growth rates and forecast horizons. Specifically, if convergence properties differ across life cycle stages, this information can be utilized to refine the valuation parameters for subsets of firms according to their current life cycle stage. [Insert Table 4 about here] To examine the convergence characteristics of profitability by life cycle stage, Table 4 Panel A reports the mean of annual median values of RNOA examined over a five-year period subsequent to the initial life cycle identification period. Pooled RNOA is relatively constant over time, ranging from 8.30 to 9.12 percent. Likewise, the mature stage is characterized by stable 17 profitability with RNOA ranging between 10 and 11 percent. However, the RNOA of both introduction and decline firms increase monotonically over time. [Insert Figure 1 about here] These medians are depicted graphically in Figure 1. Median RNOA by life cycle stage partially converges by year 3, but the difference in median RNOA between mature firms (10.41 percent) and decline firms (3.26 percent) is quite distinct five years later (difference of 7.15 percent which is substantial given that the median RNOA for the sample is 9.12 percent at time t). Growth and shake-out firms have relatively stable RNOA over the subsequent five years but the level is lower than that of mature firms. Indeed, mature firms maintain a sustainable advantage over the other life cycle stages while introduction firms earn considerably less even after five years (10.41 percent for mature compared to 5.31 percent for introduction). Therefore, convergence of RNOA remains incomplete at the end of five years, such that incorporating life cycle stage information will substantially impact forecasts of future RNOA. Next, I examine the frequency of RNOA decile membership by life cycle stage in Table 4 Panel B. Nissim and Penman (2001, Figure 4(b)) form deciles of RNOA and examine the convergence of the deciles over five subsequent years to document how RNOA evolves over time. They report a persistent difference in RNOA between the highest and lowest decile of RNOA after five years. If those differences were attributable to firm life cycle, we would expect to see an overrepresentation (underrepresentation) of introduction and decline (mature) firms in the lowest RNOA deciles. Consistent with these expectations, Panel B shows that nearly two-thirds of introduction firms (66 percent) and three-fourths of decline firms (78 percent) reside within the lowest three deciles of RNOA. Comparatively, only 17 percent of mature firms are contained within the lowest three RNOA deciles. Therefore, a partial explanation for the non-convergence in RNOA in Nissim and Penmans study is attributable to differences in firm life cycle stage as measured by 18 cash flow patterns. Specifically, the mean-reversion of Nissim and Penmans lowest deciles of firms is largely driven by the improvements in performance experienced by introduction firms and the decline firms that survive in the future. However, the highest deciles (those shown to decline over- time in Nissim and Penman) are not overpopulated by any one life cycle stage. Therefore, life cycle theory better explains the mean-reversion properties of low-profitability firms than of high- profitability firms. Given the differences in RNOA by firm life cycle demonstrated in the previous analyses, information about firm life cycle stage should be useful in explaining future profitability. I adapt Fairfield and Yohns (2001) model of future profitability to test for the incremental effect of life cycle stage in explaining one-year-ahead change in RNOA. 14 Current profitability (both level and change in current RNOA) is included in the model because it is known to be serially correlated with future profitability. The coefficients on current RNOA and RNOA are expected to be negative since profitability is mean-reverting (Brooks and Buckmaster 1976; Freeman, Ohlson and Penman 1982; Fairfield and Yohn 2001). Future changes in profitability can also occur due to a denominator effect, or growth in NOA. This growth, captured by GrNOA, is controlled for to ensure that the changes in future profitability are not driven solely by changes in investment. Prior research has shown the coefficient on GrNOA to be negative since investment in NOA is subject to diminishing returns. RNOA is decomposed into two components: asset turnover (ATO) and profit margin (PM). Asset turnover indicates the amount of assets needed to generate sales whereas profit margin indicates a firms ability to convert sales into profit. A cost leadership strategy is aimed at improving the asset turnover, while a product differentiation strategy is oriented toward improving profit
14 Forecasts of future profitability are improved when based upon operating income rather than GAAP net income (Fairfield, Sweeney and Yohn 1996; Nissim and Penman 2001). Therefore excluding the financing portion of the firm focuses the analysis on the sustainability of operating profitability. 19 margin. Fairfield and Yohn (2001) examine the effect of both levels and changes in ATO and PM on future change in RNOA. 15 Changes in ATO are indicative of increased efficiency in production and should represent permanent sources of profitability such that a positive relation with future profitability is expected (Penman and Zhang 2006, Fairfield and Yohn 2001). Penman and Zhang (2006) find a negative relation between change in profit margin and future profitability. They suggest that an increase in PM is derived from a current reduction in operating expenses, which is not sustainable and thus has negative consequences for future profitability. Next, I add alternative life cycle proxies (i.e., cash flow patterns, Anthony and Rameshs classification, and age quintiles) to determine which classification scheme best explains future profitability. Thus Model 1 is: 1 4 1 5 4 3 2 1 1 + = + + + A + A + + A + + = A t k t k t t t t t t LC D PM ATO GrNOA RNOA RNOA RNOA c | | | | | o (1) Because life cycle stages are captured by indicator variables set to one if the firm-observation is a member in that stage, the intercept captures either mature firms (for cash flow patterns and the AR classification) or middle-aged firms (for age quintiles). 16 The results of estimating Model 1 are reported in Table 5. All standard errors reported throughout the analyses are robust with respect to firm and year clustering. All of the profitability coefficients are significant (with the exception of change in PM) and are in the predicted direction. For ease of comparison, all of the life cycle coefficients are reported as the total effect (intercept + D k ) but the t-statistics pertain to the incremental difference between the each life cycle stage and the mature/mature/old stage. [Insert Table 5 about here]
15 They report that current levels of ATO and PM are not informative in forecasting one-year-ahead change in RNOA but that change in ATO is positively related to one-year-ahead change in RNOA. 16 The following alignments were made across classification schemes (cash flow patterns/AR/age quintiles): Introduction/Growth/Young, Growth/Growth-Mature/Mid-Young, Mature/Mature/Middle, Shake-Out/Mature- Stagnant/Mid-Old, and Decline/Stagnant/Old. 20 All cash flow pattern life cycle stage coefficients are significant with the mature stage representing positive future change in profitability (t = 14.89). For brevity, only the cash flow pattern life cycle stage names are used in Table 5, but I will include the alternative name in parentheses when discussing the results for the AR and age quintile classifications. In the AR classification, only the introduction (growth) stage was significantly and negatively related to future change in profitability. Significant positive associations with future profitability are found for the shake-out (mature/stagnant) and decline (stagnant) stages, which is counter to theory and intuition. All age quintile life cycle stages, with the exception of mature (middle-aged), have a significant association with one-year-ahead change in profitability, but again, the sign of the effect is counter to economic theory. Therefore, life cycle stage as measured by cash flow patterns is most consistent with theoretical expectations of future profitability. As a final check, Vuong statistics test the validity of pairwise comparisons of the models for explaining RNOA t+1 and the cash flow pattern classification outperforms both the AR classification (z = 4.19) and the age quintile classification (z = 3.89). Recalling that Fairfield and Yohn (2001) found ATO to be informative for explaining future profitability, economic theory suggests that mature firms should benefit the most from improvements in efficiency (Spence 1977, 1977, 1981; Wernerfelt 1985). This occurs because mature firms generate higher-than-normal profits, which attracts competition from other firms. Thus, to continue to maintain profitability, mature firms need to refocus their operations on cost containment and production efficiency. Selling and Stickney (1989) suggest that operational gains in efficiency are reflected in improvements in ATO. Therefore, I predict that the explanatory power of ATO for future profitability will be concentrated in mature firms (i.e., a positive coefficient on an interaction term, Mature ATO). 21 Product differentiation efforts will be reflected in higher profit margins (Selling and Stickney 1989), and growth firms are likely to exert the greatest effort to establish their brand identity and market share (Spence 1977, 1979, 1981). This suggests that growth firms will have the greatest benefit in the future from current expenditures on product differentiation and thus I expect a positive coefficient on an interaction between growth firms and change in PM, Growth PM. However, given the results in Penman and Zhang (2006) that increases in profitability due to increases in profit margin are not sustainable, I expect the incremental benefit of the product differentiation strategy to be mitigated by the time a firm reaches maturity. This would indicate that an interaction, Mature PM, would be negatively correlated with future changes in profitability. To test these assertions, Model 2 is:
1 4 1 5 4 1 4 4 1 5 4 3 2 1 1 ) ( ) ( + = = = + + A + A + + A + A + + A + + = A
t k t k k k t k k k k t t t t t t LC PM LC ATO LC D PM ATO GrNOA RNOA RNOA RNOA c o o | | | | | o (2) Results from estimating Model 2 are reported in Table 5. As in Model 1, life cycle indicator variables are included for the introduction, growth, shake-out, and decline stages whereas the mature firms are captured through the intercept. The coefficients presented in Table 5 (Model 2) represent the total effect (the main effect plus the incremental effect for each life cycle stage) on future change in RNOA. However, reported tests of significance (t-statistics) pertain to whether the mature stage coefficients are different from zero or in the case of the other life cycle stages, whether coefficients are statistically different from those of mature stage firms. As predicted, increases in ATO lead to significant increases in profitability primarily for mature firms, but this effect is only captured by the cash flow patterns and age quintile classifications. This result is consistent with economic theory stating that improving efficiency is more important for mature firms once the market is saturated. Contrary to predictions, increases in 22 PM are not associated with increases in future RNOA for growth firms in any life cycle classification. 17 However, consistent with predictions, the negative relation between changes in PM and future profitability documented in Penman and Zhang (2006) is concentrated in the mature stage for both the cash flow patterns and AR classifications. Finally, Vuong tests of pairwise performance comparison indicate that the cash flow pattern proxy outperforms the AR classification (z = 2.04) and slightly outperforms the age quintile classification (z = 1.67) once the ATO and PM interactions are considered. Given the fact that both the AR and age quintile classifications outperformed cash flow patterns when explaining ATO (Table 2), it follows that the performance of cash flow patterns would decline once the ATO interaction was included. Given the results of this analysis and the parsimony in computing the cash flow pattern proxy relative to the AR classification, the cash flow patterns proxy better provides a more tractable firm life cycle measure. Age is a parsimonious proxy, but young firms can be distressed (Stinchcombe 1965; Jovanovich 1982; Freeman, Carroll and Hannan 1983; Amit and Schoemaker 1993). While the Old classification may capture maturity, age does not provide a distinct separation of introduction versus decline firms among the Young classification. Cash outflows for research and development are included in the Statement of Cash Flows as operating activities, but they are more representative of investing cash flows. For this reason, in an untabulated analysis, the cash flow patterns specification for all analyses throughout the paper was re-estimated with research and development reclassified as investing cash flows with no substantial change in results from the results presented in the paper.
17 Some economics research claims that the benefits of product differentiation techniques such as advertising and marketing are difficult to capture at the individual firm level because advertising partly serves to increase demand for the entire market (Oster 1990, Shy 1995), i.e. there is a free-rider effect. Additionally, gains in innovation are difficult to capture due to the mobility of labor among competing firms (Porter 1980, Jovanovic and Nyarko 1995). These factors could mitigate the ability of product differentiation efforts to result in persistent increases in profitability. 23 4. What is the role of life cycle classification in understanding current firm value and predicting future stock returns?
In previous sections, I have demonstrated that cash flow patterns provide a reliable and parsimonious proxy for firm life cycle. Given the superior profitability of mature stage firms, this life cycle attribute should be associated with higher buy-and-hold stock returns in future periods conditional on a mature life cycle signal in the current period. Since the majority of firms are in the mature stage at any given point (41.18 percent of the sample), investing upon seeing a mature life cycle signal is easily implementable by all types of investors. To examine whether investors recognize the persistent profitability of mature stage firms, I examine 12-month buy-and-hold size-adjusted returns 18 computed by subtracting the appropriate market capitalization decile return (from the CRSP database) from a firms raw return. Size-adjusted returns are accumulated from the beginning of the fifth month of year t + 1 through the fourth month of the second year following the life cycle stage signal (t + 2). This allows market participants to assess life cycle stage based on the published financial statement data prior to portfolio formation. Additionally, delisting returns are used when they are included in the CRSP database and delisted firms without a corresponding delisting return are assumed to have a return of zero (Piotroski 2000) unless the delisting occurred due to performance-related reasons in which case the missing delisting return was set to -100.00 percent. 19
The independent variables used in the returns regression are earnings per share scaled by stock price t-1 (X/P) and change in earnings (X/P) because both earnings levels and changes have significant explanatory power for annual stock returns (Easton and Harris 1991). Prior research has
18 The results are invariant in all analyses for replacing size-adjusted returns with market-adjusted returns. 19 Prior research has used -30.00 percent for NYSE and AMEX firms (Shumway 1997, Mohanram 2004) and -55.00 percent for NASDAQ firms (Shumway and Warther 1999) that are delisted for performance. Sloan (1996) uses -100.00 percent for performance-related delistings. Since missing delisting returns for mergers are likely to be understated when setting to zero, I make no upward correction for merger-related missing returns to bias against finding results. 24 demonstrated a differential response to losses versus positive earnings (Hayn 1995, Basu 1997) so an indicator variable (Loss) is set to one if the company incurs a loss for the current year and this variable is included as a main effect and as an interaction with earnings. Common risk factors such as the book-to-market ratio (B/M), size (Size) as measured by the natural log of the market value of equity, and risk as captured by market model beta (Beta) (Collins and Kothari 1989, Fama and French 1992) are included in the model, all scaled by beginning of the year stock price. Therefore, Model 1 is estimated as: 1 , , 7 , 6 , 5 , 4 , 3 , 2 , 1 0 1 , / ) / ( / / + + + + + + + + A + + = t i t i t i t i t i t i t i t i t i Beta Size M B Loss P X Loss P X P X RET c | | | | | | | o (3) Results from estimating Model 1 are reported in Table 6 and all coefficients are significant and consistent with prior research. Standard errors used to compute significance are robust to clustering by firm and year. [Insert Table 6 about here] In Model 2, an indicator variable capturing membership in the mature category is added to the specification. The results from sections two and three indicate that the highest level of profitability is attained and persists for mature firms (Figure 1). Investors may undervalue these mature firms if they do not fully recognize the implications of the life cycle proxy (i.e., recognize the signal of the cash flow patterns at the financial statement date) on firm value. If this potential undervaluation is true, then the Mature variable should be positively associated with future size- adjusted returns, which is demonstrated in Model 2 (t = 6.60) of Table 6. 20
Prior research has found that the book/market (B/M) ratio of a firm is strongly positively correlated to future stock returns (Fama and French 1992, Lakonishok et al. 1994). Fama and
20 In untabulated results, Shake-Out firms were bundled with Mature firms since the level and persistence of profitability of these two stages are fairly similar. This strengthened the magnitude and significance of the indicator variable. Grouping growth firms with mature firms substantially weakened the results, which indicates that the returns-earnings relation for growth firms is structurally different than that of mature and shake-out firms. 25 French attribute this result to risk due to the fact that high B/M firms may be closer to default. By contrast, Lakonishok et al. attribute the B/M effect to mispricing such that investors are overly optimistic about low B/M stocks (glamour firms) and pessimistic about value stocks (high B/M). Piotroski (2000) and Mohanram (2005) investigate whether fundamental analysis can be used to separate the winners from the losers at the two extremes of the B/M spectrum. Piotroski develops an FSCORE based on nine binary variables (i.e., earnings versus loss, increasing ROA, positive cash flows, cash flows greater than net income, increasing operating margin, increasing asset turnover, decreasing leverage, increasing current ratio, and non-issuance of equity) for value firms. Mohanram develops a GSCORE based on eight signals (i.e., ROA and cash flows greater than median for other low B/M firms in the same industry, cash flows greater than net income, lower earnings and sales growth variability than for other low B/M firms in industry, and research and development, advertising intensity and capital expenditures all greater than other low B/M in the same industry) for glamour firms. In both cases, partitioning on the respective composite scores can identify firms where significant excess returns can be earned based on the ex ante signal (FSCORE or GSCORE, along with the identification of low or high B/M). In Piotroskis study, high B/M firms with the highest FSCORE earned mean market-adjusted returns of 13.4 percent while high B/M firms with the lowest FSCORE earned -9.6 percent. Mohanram found that low B/M firms with the highest GSCORE earned a mean size-adjusted return of 3.1 percent while the lowest GSCORE among low book/market firms earned -17.5 percent in the year after portfolio formation. It is possible that life cycle information (measured parsimoniously by cash flow patterns) can effectively detect the upside of both the value and glamour trading strategies demonstrated in Piotroskis and Mohanrams studies. Since most of the variables that comprised the FSCORE and GSCORE are related to life cycle, especially as measured by cash flow patterns, the proxy used in 26 this paper may capture the essence of what makes the composite scores effective in predicting future returns. Combining the B/M screen with the mature life cycle stage screen may identify firms that comprised the long portion of the two B/M strategies. Furthermore, by restricting the B/M screen to only mature firms, the necessity to short some firms is eliminated which makes the strategy highly implementable for all types of investors. Since market participants have access to more information than what is captured by the accounting system, a high market value relative to book value (resulting in low B/M) may be reflective of future product development. In other words, mature firms that have low B/M ratios (mature-glamour firms) may be on the cusp of recycling back to the growth life cycle stage in future periods due to new product or market expansions. If the new development is successful, those firms should enjoy increasing profitability as the new development matures. Conversely, a low market value relative to book value (high B/M) is indicative of market inattention or neglect by information intermediaries such as analysts. When high B/M firms are also mature (mature-value firms), the risk of distress is much lower than stocks that are avoided due to anticipation of poor performance. It has been demonstrated that these types of value stocks are undervalued by the market which may result in a positive reaction when the market realizes their profitability did not mean-revert to lower expectations. Therefore, Model 3 examines whether interacting the mature life cycle stage with alternatively the lowest and highest quintiles of B/M results in positive size-adjusted returns in the year following the mature and B/M signal: 1 , , 10 , 9 , 8 , 7 , 6 , 5 , 4 , 3 , 2 , 1 0 1 , ) / ( ) / ( / ) / ( / / + + + + + + + + + + + A + + = t i t i t i t i t i t i t i t i t i t i t i t i M HighB Mature M LowB Mature Mature Beta Size M B Loss P X Loss P X P X RET c | | | | | | | | | | o (4) Results show that the interactions between Mature and both the lowest and highest quintiles of B/M generate positive size-adjusted returns in the year after portfolio formation (t = 3.05 and t = 2.16, respectively). Overall, the mispricing based on extreme values of B/M demonstrated in prior 27 research is likely to be more severe among mature firms, consistent with the market not recognizing the level and persistence of RNOA in these firms. Furthermore, there are ample observations in the Mature-Low B/M quintile portfolio (ranging from 104 firms in 2000 to 256 firms in 2003) and in the Mature-High B/M quintile portfolio (ranging from 180 firms in 1989 to 295 firms in 2002) to comprise an economically meaningful investment portfolio to exploit this mispricing in any given year.
5. Conclusion This paper develops and validates the use of cash flow patterns as an effective and parsimonious proxy for firm life cycle. Several performance measures and firm characteristics such as profitability, market performance, size, and age are nonlinearly related to firm life cycle. For that reason, univariate sorts on these performance measures result in an inappropriate assignment of life cycle stages such that the lowest portfolios are populated with both introduction and decline firms, which have differential implications for future performance. More importantly, a simple sort on univariate measures makes a distributional assumption of uniformity that is not supported by economic theory. The paper first validates the life cycle proxy according to economic theory and then examines how cash flow patterns capture the economic concept of life cycle relative to proxies used in past research (i.e., Anthony and Rameshs metric and firm age). Next, the persistence of profitability, measured by return on net operating assets (RNOA) is examined by life cycle stage. Convergence rates and the patterns of mean-reversion of future profitability are shown to differ by life cycle stage. Specifically, the spread in RNOA is three to 10 percent between mature and decline firms five years subsequent to portfolio formation (based on life cycle identification). This spread of seven percent is economically significant given that the median RNOA for the sample is 8.61 28 percent. The valuation and forecasting implications are that growth rates and forecast horizons should be conditioned on information pertaining to the firms current life cycle stage. Additionally, past research on the decomposition of RNOA has shown that change in asset turnover is an important driver of future changes in RNOA (Fairfield and Yohn 2001) but that improvements in future profitability due to increases in profit margin are not sustainable (Penman and Zhang 2006). Both of these results are primarily concentrated in mature firms. Increases in operational efficiency are critical for mature firms due to increased competition that is attracted to the superior profits earned by mature firms. At the same time, diminishing returns to product differentiation efforts appear to be evident among those same firms. Finally, the market valuation consequences of life cycle (as captured by cash flow patterns) are investigated and it is demonstrated that positive future excess returns can be earned for firms that fall into the mature category. Furthermore, life cycle adds crucial fundamental information to separate out the winners from the losers at both extremes of the book/market spectrum (i.e., glamour and value firms). The cash flow pattern proxy captures information priced by the market in a more efficient manner than the fundamental signals used in previous trading strategies which require the computation of composite scores based on numerous metrics. In summary, this paper uses basic accounting information to capture the construct of firm life cycle which embodies differences in resources, rates of investment, obsolescence rates, learning and experience curves, adaptation, product-differentiation, and production efficiencies. The cash flow pattern proxy for life cycle stage outperforms other proxies used in extant research including age, and better explains future profitability (both in rates of return and stock returns) given its foundation in economic theory.
29 REFERENCE LIST
Amit, R. and P. Schoemaker. 1993. Strategic assets and organizational rent. Strategic Management Journal 14, 33-46.
Anthony, J. and K. Ramesh. 1992. Association between accounting performance measures and stock prices. Journal of Accounting and Economics 15, 203-227.
Barclay, M. J. and C. W. Smith, Jr. 1995. The maturity structure of corporate debt. Journal of Finance 50, 609-631.
Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting and Economics 24, 3-37.
Beaver, W., M. McNichols and R. Price. 2007. Delisting returns and their effect on accounting-based market anomalies. Journal of Accounting and Economics 43, 341-368.
Bhattacharya, N. N., E. L. Black, T. Christensen, and R. Mergenthaler. 2004. "Empirical evidence on recent trends in pro forma reporting." Accounting Horizons 18, 27-44.
Black, E. L. 1998. Life-cycle impacts on the incremental value-relevance of earnings and cash flow measures. Journal of Financial Statement Analysis 4, 40-56.
Bradshaw, M. T., M. S. Drake, J. Myers and L. Myers. 2010. "A re-examination of analysts superiority over time-series forecasts." Working paper, Boston College, Ohio State University and the University of Arkansas.
Brockhoff, K. 1967. A test for the product life cycle. Econometrica 35, 472-484.
Brooks, L. and D. Buckmaster. 1976. Further evidence on the time series properties of accounting income. Journal of Finance 31, 1359-1373.
Caskey, J. A. and M. L. Hanlon. 2007. "Do dividends indicate honesty? The relation between dividends and the quality of earnings." Working paper, University of California at Los Angeles and Massachusetts Institute of Technology.
Caves, R. 1998. Industrial organization and new findings on the turnover and mobility of firms. Journal of Economic Literature 36, 1947-82.
Chen, S., M. L. DeFond, and C. Park. 2002. "Voluntary disclosure of balance sheet information in quarterly earnings announcements." Journal of Accounting and Economics 33, 229-251.
Collins, D. and S. P. Kothari. 1989. An analysis of intertemporal and cross-sectional determinants of earnings response coefficients. Journal of Accounting and Economics 11, 143-182.
Cox, W. E. J. 1967. Product life cycles as marketing models. Journal of Business 40, 375-384.
Desai, H., C. E. Hogan, and M. Wilkins. 2006. "The reputational penalty for aggressive accounting: Earnings restatements and management turnover." The Accounting Review 81, 83-112.
30 Diamond, D. 1991. Monitoring and reputation: The choice between bank loans and directly placed debt. Journal of Political Economy 99, 689-721.
Doyle, J. T., W. Ge, and S. McVay. 2007. "Determinants of weaknesses in internal control over financial reporting." Journal of Accounting and Economics 44, 193-223.
Easton, P. and T. Harris. 1991. Earnings as an explanatory variable for returns. Journal of Accounting Research 29, 19-36.
Fairfield, P., R. Sweeney, and T. Yohn. 1996. Accounting classification and the predictive content of earnings. The Accounting Review 71, 337-355.
__________ and T. Yohn. 2001. Using asset turnover and profit margin to forecast changes in profitability. Review of Accounting Studies 6, 371-385.
Fama, E. F. and K. R. French 1992. "The cross-section of expected stock returns." Journal of Finance 47, 427- 465.
________ and ________. 1997. Industry costs of equity. Journal of Financial Economics 43, 153-193.
________ and ________. 2000. Forecasting profitability and earnings. Journal of Business 73, 161-175.
Freeman, J., G. Carroll, and M. Hannan. 1983. The liability of newness: Age dependence in organization death rates. American Sociological Review 48, 692-710.
Freeman, R., J. Ohlson, and S. Penman. 1982. Book rate-of-return and prediction of earnings changes: An empirical investigation. Journal of Accounting Research 20, 639-653.
____________, A. S. Koch, and H. Li. 2006. "Can firm-specific models predict price responses to earnings news?" Working paper, University of Texas at Austin, University of Virginia and Santa Clara University.
Gort, M. and S. Klepper. 1982. Time paths in the diffusion of product innovation. Economic Journal 92, 630- 653.
Hannan, M. and J. Freeman. 1984. Structural inertia and organizational change. American Sociological Review 49, 149-164.
Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20, 125-153.
Jegadeesh, N. and S. Titman. 2001. Profitability of momentum strategies: an evaluation of alternative explanations. Journal of Finance 56, 699-720.
Jovanovic, B. 1982. Selection and the evolution of industry. Econometrica 50, 649-70.
___________ and G. MacDonald. 1994. The life cycle of a competitive industry. Journal of Political Economy 102, 322-347.
___________ and Y. Nyarko 1995. A Bayesian learning model fitted to a variety of empirical learning curves. Brookings Papers on Economic Activity: Microeconomics: 247-305.
31 Khan, M. and R. L. Watts. 2009. "Estimation and empirical properties of a firm-year measure of accounting conservatism." Journal of Accounting and Economics 48, 132-150.
Klein, A. and C. A. Marquardt 2006. "Fundamentals of accounting losses." The Accounting Review 81, 179-206.
Lakonishok, J., A. Shleifer, and R. Vishny. 1994. "Contrarian investment, extrapolation, and risk." Journal of Finance 44,1541-1578.
Lev, B. and P. Zarowin. 1999. The boundaries of financial reporting and how to extend them. Journal of Accounting Research 37, 353-385.
Livnat, J. and P. Zarowin. 1990. The incremental information content of cash-flow components. Journal of Accounting and Economics 13, 25-46.
Mohanram, P. 2005. Separating winners from losers among low book-to-market stocks using financial statement analysis. Review of Accounting Studies 10, 133-170.
Myers, S. 1984. The capital structure puzzle. Journal of Finance 39, 575-592.
Nissim, D. and S. Penman. 2001. Ratio analysis and equity valuation: From research to practice. Review of Accounting Studies 6, 109-154.
Oster, S. 1990. Modern Competitive Analysis. New York, NY: Oxford University Press.
Parsons, L. 1975. The product life cycle and time-varying advertising elasticities. Journal of Marketing Research 12, 476-480.
Penman, S. and X. Zhang. 2006. Modeling sustainable earnings and P/E Ratios with financial statement analysis. Working paper, Columbia University and University of California, Berkeley.
Piotroski, J. D. 2000. "Value investing: The use of historical financial statement information to separate winners from losers." Journal of Accounting Research 38, 1-41.
Polli, R. and V. Cook. 1969. Validity of the product life cycle. Journal of Business 42, 385-400.
Porter, M. E. 1980. Competitive Strategy: Techniques for Analyzing Industries and Competitors. New York, NY: The Free Press.
Selling, T. and C. Stickney. 1989. The effects of business environment and strategy on a firm's rate of return on assets. Financial Analysts Journal 45, 43-52.
Shumway, T. 1997. The delisting bias in CRSP data. Journal of Finance 52, 327-340.
__________ and V. Warther. 1999. The delisting bias in CRSPs Nasdaq data and its implications for the size effect. Journal of Finance 54, 2361-2389.
Shy, O. 1995. Industrial Organization: Theory and Application. Cambridge, MA: The MIT Press.
Sloan, R. 1996. Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71, 289-315
Spence, M. 1977. Entry, capacity, investment, and oligopolistic pricing. Bell Journal of Economics 8, 534-544. 32
_________. 1979. Investment strategy and growth in a new market. Bell Journal of Economics 10, 1-19.
_________. 1981. The learning curve and competition. Bell Journal of Economics 12, 49-70.
Stigler, G. 1963. Capital and Rates of Return in Manufacturing Industries. Princeton, NJ: Princeton University Press.
Stinchcombe, A. 1965. Organizations and Social Structure. Handbook of Organizations. Chicago, IL: Rand-McNally.
Wasley, C. E. and J. S. Wu. 2006. "Why do managers voluntarily issue cash flow forecasts?" Journal of Accounting Research 44, 389-429.
Wernerfelt, B. 1985. The dynamics of prices and market shares over the product life cycle. Management Science 31, 928-939.
33 Table 1 Economic Links to Cash Flow Patterns Cash Flow Type Introduction Growth Mature Shake-Out Decline Operating Firms enter market with knowledge deficit about potential revenues and costs (Jovanovic 1982)
() Cash Flows Profit margins are maximized during period of greatest investment (Spence 1977, 1979, 1981)
(+) Cash Flows Efficiency maximized through increased knowledge of operations (Spence 1977, 1979, 1981; Wernerfelt 1985)
(+) Cash Flows
Declining growth rates lead to declining prices (Wernerfelt 1985) Routines of established firms hinder competitive flexibility (Hannan and Freeman 1984)
(+/) Cash Flows Declining growth rates lead to declining prices (Wernerfelt 1985)
() Cash Flows Investing
Managerial optimism drives investment (Jovanovic 1982) Firms make early large investments to deter entry (Spence 1977, 1979, 1981)
() Cash Flows Firms make early large investments to deter entry (Spence 1977, 1979, 1981)
() Cash Flows Obsolescence increases relative to new investment as firms mature (Jovanovic 1982, Wernerfelt 1985)
() Cash Flows Void in theory
(+/) Cash Flows Liquidation of assets to service debt
(+) Cash Flows Financing
Pecking order theory states firms will access bank debt then equity (Myers 1984, Diamond 1991); Growth firms will issue short-term debt (Barclay and Smith 1995)
(+) Cash Flows Pecking order theory states firms will access bank debt then equity (Myers 1984, Diamond 1991); Growth firms will issue short-term debt (Barclay and Smith 1995)
(+) Cash Flows Focus shifts from acquiring financing to servicing debt and distributing excess funds to shareholders; Mature firms will issue longer term debt (Barclay and Smith 1995) () Cash Flows Void in theory
(+/) Cash Flows Focus on debt repayment and/or renegotiation of debt
(+/) Cash Flows 34 Table 2 Effect of Life Cycle Classification on Economic Characteristics
Panel A Life Cycle Stages defined by Cash Flow Patterns Pooled Introduction Growth Mature Shake- Out Decline
N 48,369 5,752 16,423 19,920 3,861 2,413 % of total N 100.00% 11.89% 33.95% 41.18% 7.98% 4.99% Dependent Variable Mean Coefficients (relative to Mature) (t-statistic) Adj. R 2
35 Table 2 - Continued Effect of Life Cycle Classification on Economic Characteristics
Panel B Life Cycle Stages defined by Anthony and Ramesh Classification Growth Gr/Mat Mature Mat/Stag Stagnant N 8,169 7,135 7,814 7,640 11,527 CF Patt. % of total N 19.32% 16.87% 18.48% 18.07% 27.26% over AR Dependent Variable Coefficients (relative to Mature) (t-statistic) Adj. R 2
36 Table 2 - Continued Effect of Life Cycle Classification on Economic Characteristics
Panel C Life Cycle Stages defined by Age Quintiles Young Mid/Young Middle Mid/Old Old N 9,705 9,647 9,641 9,725 9,651 CF Patt. % of total N 20.06% 19.94% 19.93% 20.11% 19.95% over Age Dependent Variable Coefficients (relative to Mature) (t-statistic) Adj. R 2
37 Table 2 - Continued Effect of Life Cycle Classification on Economic Characteristics
For the sample period 1989 to 2005. Means presented in Panel A are the means of annual medians except for total number of observations, dividend payout ratio, advertising, innovation, number of segments, and mergers which are annual means. The explanatory power of each life cycle classification is tested by regression each dependent variable on indicator variables for each life cycle stage except for Mature (Middle) which is captured in the intercept. This means that the regression coefficients capture the effect of each life cycle stage relative to maturity. Robust standard errors are clustered by firm and year. Coefficients significant at the 0.05 level or better are designated in bold. Vuong statistics report the explanatory power of the Cash Flow Pattern classification over Anthony and Rameshs (Age Quintiles) classification in Panel B (Panel C). Dependent variables are measured as follows: Earnings per share (EPS) is measured before extraordinary items (#58). Return on net operating assets (RNOA) = Operating Income (OIt)/Average Net Operating Assets (NOA). Profit margin (PM) = Operating Income (OI)/Net sales (#12). Asset Turnover (ATO) = Net Sales)/Average Net Operating Assets (NOA). Growth in Sales (GrSALES) is defined as (Net Sales/Lagged Net Salest) 1. Growth in NOA (GrNOA) is defined as (NOA/NOA) 1. Market-to-Book (MB) = Market Value of Equity/Book Value of Equity (#60). Leverage (LEV) = Net Financial Obligation/Common Equity (#60). ASSET BETA is the mean market model beta from a regression of daily raw returns on the value-weighted market return over the prior 250 days adjusted for leverage. Dividend Payout Ratio (DIVPAY) = Common dividends (#21)/Net Income (#172). Advertising Intensity (ADVINT) is Advertising Expense (#15) /Net Sales (#12), Innovation (INNOV) is [R&D (#46) plus Amortization Expense (#65)/Net Sales (#12)]. SEGMENT is the number of segments reported in the Compustat segment files. MERGER is the percentage of firms that have AA codes in Compustat (AFTNT1). SIZE is the log of market value of equity. AGE is defined as the log of the number of years since the firms first appearance in the CRSP database. All variables are winsorized at the 1st and 99 th percentiles to mitigate the influence of extreme values.
38 Table 3 Survival Rate and Transition Matrix Analyses
Panel A Proportion of firms that survive beyond portfolio formation period N = 33,088
Base years range from 1989 to 2000 so that five subsequent years are available for each observation (sample period extends to 2005). Z-statistics (in italics below the proportion) from a test of equal proportions is computed for each life cycle stage relative to the pooled sample for the years subsequent to life cycle identification; and for each stage relative to a uniform distribution for the delisting categories. Z-Statistics in bold indicate a significant difference in proportions at 0.05 significance level or better. Delisting data was extracted from the CRSP Event database and was computed for all observations with adequate cash flow data to compute the life cycle stage in the year prior to delisting. CRSP categorizes delistings as follows: 200-299 are mergers and 500-599 are dropped securities due to performance. The proportion of delistings due to mergers (inadequate performance) is computed by life cycle stage and reported in the second to the last (last) column.
39 Table 3 Continued Survival Rate and Transition Matrix Analyses
Panel B Transition Matrix: Proportion of observations in each life cycle stage in years subsequent to portfolio formation N = 33,088
Stage at Portfolio Formation Stage in Future Period t+1 t+2 t+3 t+4 t+5
Base years range from 1989 to 2000 so that five subsequent years are available for each observation (sample period extends to 2005). For each life cycle stage at the time of portfolio formation (year t), this table reports the proportion of surviving firms by life cycle stage for each year subsequent to life cycle identification.
40 Table 4 Analysis of Return on Net Operating Assets (RNOA) by Life Cycle Stage
Panel A Inter-temporal Analysis of Median RNOA by Life Cycle Stage Pooled Introduction Growth Mature Shake- Out Decline N 33,088 4,121 11,742 13,424 2,409 1,392 % of total N 100.00% 12.45% 35.49% 40.57% 7.28% 4.21% Year Relative to Formation
t 9.12% -4.18% 9.54% 11.01% 8.26% -16.30% t + 1 8.53% -1.75% 8.38% 10.74% 7.85% -9.08% t + 2 8.30% 1.30% 7.86% 10.34% 7.96% -4.86% t + 3 8.43% 3.15% 7.84% 10.27% 8.37% 0.30% t + 4 8.71% 4.56% 8.03% 10.50% 8.60% 0.59% t + 5 8.93% 5.31% 8.19% 10.41% 9.45% 3.26%
Panel B Proportion of Life Cycle Stage by RNOA Decile Membership Pooled Introduction Growth Mature Shake- Out Decline N 48,369 5,752 16,423 19,920 3,861 2,413 % of total N 100.00% 11.89% 33.95% 41.18% 7.98% 4.99% RNOA Decile Lowest 35.95% 3.57% 2.44% 12.04% 50.73% 2 19.66% 8.32% 6.57% 14.69% 19.31% 3 10.57% 10.97% 8.92% 12.04% 7.67% 4 7.09% 11.54% 10.48% 8.70% 4.68% 5 5.49% 11.70% 11.31% 7.10% 2.86% 6 5.09% 11.37% 11.71% 7.46% 2.49% 7 5.18% 10.91% 12.08% 7.80% 1.82% 8 4.73% 9.89% 12.86% 7.93% 2.98% 9 3.76% 10.13% 12.73% 9.35% 2.69% Highest 2.47% 11.59% 10.90% 12.90% 4.77% Total 100.00% 100.00% 100.00% 100.00% 100.00% Likelihood Ratio Chi-Square: 11,311.17 (<.0001) 41 Table 4 - Continued Analysis of Return on Net Operating Assets (RNOA) by Life Cycle Stage
The sample period is 1989 to 2000 for Panel A and from 1989 to 2005 for Panel B. Panel A requires five subsequent years to portfolio formation to ensure that the results are not affected by a truncated time period. Return on net operating assets (RNOA) is the mean of the annual medians and is measured as Operating Income (OIt)/Average Net Operating Assets (NOA). In Panel B, life cycle stage membership is analyzed as a proportion of RNOA decile membership. A test of equal proportions across RNOA deciles by life cycle stage was rejected with a Likelihood Ratio Chi-Square of 11,311.17 (<0.0001).
42 Table 5 Explanatory Power of Life Cycle Stages for Future Change in RNOA
n = 48,369 Variable Predicted Sign CF Patt. A&R Age Quintiles Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 RNOA
Vuong test comparing: Model 1 Model 2 Z-Statistic p-value Z-Statistic p-value CF Patterns over A&R: 4.19 <0.0001 2.04 0.0417 CF Patterns over Age: 3.89 <0.0001 1.67 0.0947
43 Table 5 Continued Explanatory Power of Life Cycle Stages for Future Change in RNOA
For the sample period 1989 to 2005. Robust standard errors are clustered by firm and year. The dependent variable is RNOAt+1. The coefficients are the total effect (the main effect plus the incremental effect for each life cycle stage). The t-statistics pertain to whether the mature stage coefficients are different from zero or whether the other life cycle coefficients are statistically different from the mature stage coefficient. The following alignments were made across classification schemes (cash flow patterns/AR/age quintiles): Introduction/Growth/Young, Growth/Growth-Mature/Mid-Young, Mature/Mature/Middle, Shake-Out/Mature-Stagnant/Mid- Old, and Decline/Stagnant/Old. Return on net operating assets (RNOA) = Operating Income (OIt)/Average Net Operating Assets (NOA). Growth in NOA (GrNOA) is defined as (NOA/Lagged NOA) 1. Asset turnover (ATO) = Net sales (Compustat #12)/Average Net Operating Assets (NOA). Profit margin (PM) = Operating Income (OI)/Net sales (#12). All variables are winsorized at the 1st and 99 th
percentiles to mitigate the influence of extreme values.
44 Table 6 Buy-and-Hold Annual Size-Adjusted Returns to Life Cycle Strategy
n = 46,226 Variable Predicted Sign
Model 1 Model 2 Model 3 Intercept +/ 0.020 (4.30) 0.005 (0.86) 0.005 (0.94) X/P + 0.219 (4.13) 0.202 (3.80) 0.208 (3.90) X/P + 0.045 (5.25) 0.042 (4.94) 0.043 (4.97) Loss +/ 0.068 (6.77) 0.075 (7.40) 0.075 (7.31) X/P Loss -0.239 (-4.31) -0.222 (-4.01) -0.227 (-4.09) B/M
For the sample period 1989 to 2005. Robust standard errors are clustered by firm and year. This table presents 12-month buy-and- hold size-adjusted returns accumulated from the beginning of the fifth month of year t + 1 through the fourth month of year t + 2. Portfolio formation is based on life cycle stage or book-to-market ratio at the end of year t. Delisting returns are used when included in the CRSP database. Delisted firms without a corresponding delisting return are assumed to have a return of -100 percent in the delisting period when delisted for performance-related reasons (Delisting codes between 200 and 299); zero percent otherwise. Earning/Price (X/P) are earnings per share before extraordinary items scaled by beginning of the year stock price. Loss is an indicator variable set to one if earnings are negative in the current year. Book-to-Market (B/M) = [Book Value of Equity (#60) / Market Value of Equity] scaled by beginning of the year stock price. SIZE is the log of market value of equity scaled by beginning of the year stock price. BETA is the mean market model beta from a regression of daily raw returns on the value-weighted market return over the prior 250 days scaled by beginning of the year stock price. MATURE is an indicator variable set to one if the observation is in the mature category (based on cash flow patterns) at the end of the year. Low (High) B/M is the lowest (highest) quintile of the B/M ratio at the end of the year. All variables are winsorized at the 1st and 99 th percentiles to mitigate the influence of extreme values.
45 Figure 1 Convergence Analysis - Future RNOA by Life Cycle Stage
N=33,088
For the sample period 1989 to 2000. Life cycle stage is determined at time zero and median RNOA for each stage is computed for the five subsequent years.
-20.00% -15.00% -10.00% -5.00% 0.00% 5.00% 10.00% 15.00% 0 1 2 3 4 5 M e d i a n
R N O A
Year Relative to Portfolio Formation Year Introduction Growth Mature Shake Out Decline