Investment-Cash Flow Sensitivities Are Really Investment-Investment Sensitivities

Investment-Cash Flow Sensitivities Are Really Investment-Investment Sensitivities Robert M Bushman Kenan-Flagler Business School University of North Carolina at Chapel Hill Abbie J Smith Graduate School of Business, University of Chicago X Frank Zhang School of Management, Yale University First Draft: October 2004 This Draft: March 2007 Abstract We provide strong evidence that documented patterns in investment-cash flow sensitivities across a-priori partitions based on dividend payouts, firm age and Cleary’s (1999) metric not reflect financing frictions, but rather reflect the direct connection between capital investment and corresponding investments in non-cash working capital We argue that the cash flow variable typically used in estimating investment-cash flow sensitivities, earnings before depreciation, really serves as a proxy for investment in non-cash working capital, rather than as a proxy for cash available to fund investment in the face of financing constraints We document that the observed pattern in investment-cash flow sensitivities is driven by the fact that signal-to-noise ratio in this cash flow variable with respect to working capital investment varies systematically across the typical a-priori partitions used This analysis complements and extends research focusing on measurement error in q (e.g., Erickson and Whited (2000), Alti (2003)) and on the measurement of financial constraints (e.g., Kaplan and Zingales (1997), Cleary (1999)) to explain investment-cash flow sensitivities Our analysis reveals that estimating reduced form investment equations which measure cash flows as earnings before depreciation essentially results in capital investment being regressed on a noisy measure of working capital investment and thus does not speak directly to financing constraints We appreciate the comments of Steve Fazzari, Steve Kaplan, Anil Kashyap, Bruce Petersen, and seminar participants at Chinese University of Hong Kong, Emory University and the University of Toronto Bushman thanks the Kenan-Flagler Business School, University of North Carolina at Chapel Hill for financial support, and Smith appreciates support from the William Ladany Research Fund at the Graduate School of Business at the University of Chicago 1 Introduction Following Fazzari, Hubbard, and Petersen (1988) (hereafter FHP), a large empirical literature posits that if external financing frictions create a wedge between the cost of internal and external funds, then the capital investment decisions of financially constrained firms will be sensitive to internally generated cash flows, after controlling for investment opportunities Consistent with this hypothesis, empirical studies indeed document that investment-cash flow sensitivities are higher for firms a priori classified as being more financially constrained based on criteria such as dividend payout and firm age (see Hubbard (1998) for a review of the literature) On the other hand, a number of recent papers raise serious concerns about whether investmentcash flow sensitivities reflect financing frictions, focusing on measurement error in q (e.g., Erickson and Whited (2000), Alti (2003)) and on the measurement of financial constraints (e.g., Kaplan and Zingales (1997, 2000), Cleary (1999)) to explain investment-cash flow sensitivities In this paper, we develop a novel argument that focuses on the underlying structure of the cash flow variable as the key to understanding investment-cash flow sensitivities In particular, we argue that the cash flow variable typically used in estimating investment-cash flow sensitivities really serves as a proxy for investment in non-cash working capital, rather than as a proxy for cash available to fund investment in the face of financing constraints We provide strong evidence that documented patterns in investment-cash flow sensitivities across a-priori partitions based on dividend payouts, firm age and Cleary’s (1999) measure of financial constraints not reflect financing frictions, but rather reflect the direct connection between a firm’s capital investment and corresponding investments in non-cash working capital The empirical specification generally used in the literature defines firm-level cash flow as accounting earnings before depreciation (EBD) However, EBD is actually composed of a true cash component, cash flow from operations (CFO), and a non-cash component (which we label working capital accruals or WCACC) which reflects net investment in non-cash working capital such as inventories and accounts receivable.2 This disaggregation of EBD is the key to more fully understanding the essence of investment-EBD sensitivities We show that the documented pattern in investment-EBD sensitivities is driven by the working capital investment component (WCACC), not the cash component (CFO) That is, the empirical patterns are really reflecting capital investment-working capital investment sensitivities, rather than investment-cash flow sensitivities! This result raises a serious challenge to the financing constraint interpretation of investment-EBD sensitivities To the extent that fixed capital investment represents an increase in the scale of the firm, it is natural to expect a corresponding increase in non-cash working capital items such as inventories and accounts receivable.3 This relation has nothing to with financing constraints but rather is a manifestation of increasing scale The main innovation in our paper lies in our explanation for why capital investment-EBD sensitivities vary monotonically across a-priori partitions based on characteristics such as dividend payouts, firm age and Cleary’s (1999) metric We argue that this occurs because the informativeness of the WCACC component of EBD with respect to contemporaneous investment in working capital varies systematically across these partitions The essence of our argument is that the WCACC component of EBD consists of two aspects: investments in non-cash working capital which are positively related to growth, and random fluctuations in working capital which The accounting relation is EBD = CFO + WCACC To emphasize the interpretation of WCACC as working capital investment, we replicate all the main results of the paper substituting just the change in inventory (one component of WCACC) in place of WCACC The idea is that, when a company expands its manufacturing capacity and sells more products, its inventory levels naturally increase 3 are largely independent of growth.4,5 Recognizing that partitions based on dividends, age and Cleary’s metric to a large extent partition firms based on growth, the idea is that for growing, capacity expanding firms, working capital investment is also expanding As a result, the informativeness of WCACC with respect to working capital investment dominants the effect of the random component for such firms, and the WCACC component of EBD correlates highly with contemporaneous capital investment On the other hand, for slow growth or steady state firms, random fluctuations dominant the working capital investment aspect of WCACC, resulting in a lower correlation between investment and EBD for these firms We provide evidence consistent with this argument It is important to put our results into the context of the existing literature that argues against a financing friction interpretation of investment-cash flow sensitivities Beginning with Poterba (1988), a number of studies have noted that potential measurement error in q complicates the interpretation of documented investment-cash flow sensitivities These papers include Gilchrist and Himmelberg (1995), Erickson and Whited (2000), and Alti (2003) It is most instructive contrast our paper with Alti (2003) Alti (2003) argues that measurement error in q relative to capturing near-term investment opportunities varies systematically across firms partitioned by dividend payout, and that cash flow reflects near-term investment opportunities As a result, cash flow should be relatively more important in explaining current investment for growth firms (relative to mature firms) because q has more measurement error relative to near-term opportunities for growth firms Consistent Zhang (2007) shows that working capital accruals, which by definition measure growth in operating non-cash working capital, are highly positively related to other growth attributes, such as growth in number of employees, growth in sales, growth in fixed assets, and growth in financing activities Such random fluctuations follow from simple timing issues such as inventory replenishment timing and cash collection timing on accounts receivable For example, consider a steady state firm that replenishes inventory to upper threshold, S, whenever inventory levels hit lower threshold, s This would generate timing bounces between s and S that are unrelated to growth with Alti (2003), we show that not only does cash flow reflect near-term investment opportunities, it directly embeds near-term investment in working capital (WCACC)! However, unlike Alti (2003), our explanation does not require that measurement error in q vary systematically across partitions Instead, we argue that extent to which EBD (WCACC) reflects working capital investment relative to random fluctuations in working capital varies systematically across the common a-priori partitions We only require that measurement error in q allows space for a direct measure of near-term working capital investment (WCACC) to load Finally, it is also instructive to compare our results with those of Kaplan and Zingales (1997, 2000) and Cleary (1999).6 While using different approaches, these papers attempt to directly measure financing constraints, and using these measures, find that investment-cash flow sensitivities are actually higher for financially unconstrained firms than for financially constrained firms.7 In our paper, we replicate Cleary’s result using the ZFC measure he developed.8 As we show, high ZFC represents growing firms while low ZFC represents low growth firms The connection between FHP and Cleary then lies in the fact that both empirical designs partition firms on the extent of capacity expansion Thus, investment-EBD sensitivities are higher for high ZFC firms (“unconstrained”) relative to low ZFC firms (“constrained”) because for growing firms, EBD (WCACC) is a good measure of such working capital investment and correlates highly with capital investment, while for slow growth firms, the opposite holds The rest of the paper is organized as follows Section lays our the empirical design Section presents our main empirical analyses of investment-cash flow sensitivities In section Fazzari, Hubbard, and Petersen (2000) vigorously dispute both the theoretical claims of Kaplan and Zingales (1997 and 2000) and their empirical conclusions See also Moyen (2004) for a related approach The premise of ZFC as a proxy for financial constraint in Cleary (1999) is that firms who cut dividends are more likely to be financial constrained Cleary (1999), estimates ZFC using discriminant analysis, classifying firms into dividend cut, no change, and dividend increase groups based on beginning-of-period variables that are chosen to proxy for firm liquidity, leverage, profitability, and growth we verify the robustness of our results to alternative empirical specifications, while section concludes Empirical Framework Our point of departure is the basic panel regression equation used in the literature 9: I t / K t    1 qt    EBDt / K t   et , (1) where It is capital investment in period t, Kt-1 is capital stock at the beginning of the period, qt-1 is average q, and EBDt is the cash flow variable commonly used in the literature, measured as earnings before extraordinary items plus depreciation The innovation in our design follows from the recognition that EBD is not a pure measure of current, internally generated cash flows, as it also reflects a range of non-cash elements dictated by fundamental accounting principles Using basic accounting identities, EBD can be disaggregated as EBD = E + DEPEXP = (ACCRUALS + CFO) + DEPEXP (2) = (WCACC – DEPEXP +CFO) + DEPEXP = WCACC + CFO where E is earnings before extraordinary items, ACCRUALS is total accruals (the difference between accounting net income and cash flow from operations), WCACC is working capital accruals (primarily changes in both inventory and accounts receivable minus changes in accounts payable), CFO is cash flow from operations, and DEPEXP is depreciation expense Our main empirical analysis uses a Fama-MacBeth fixed effects approach Specifically, we first subtract the time series average for each variable and each firm, and then estimate a series of cross-sectional regressions with the demeaned variables Finally, we use the time series standard deviation of the estimated coefficients to compute fixed-effect Fama-MacBeth t-statistics (see Rajan et al 2000) In section 6, we show that all results in the paper are robust to instead running specifications that include firm and year dummies in equation (1), as well as to a battery of other robustness tests Our empirical strategy then is to substitute CFO and WCACC for EBD in (1) above, and examine the pattern in both investment-CFO sensitivities and investment-WCACC sensitivities across partitions based on dividend payouts, firm age, and Cleary’s ZFC measure of financial constraints To illustrate the structure of EBD, the Appendix includes a statement of cash flows for Columbia Sportswear Co This statement is typical for U.S public companies The first section of the statement reconciles accounting net income (E) to cash flow from operations (CFO) The statement details the necessary adjustments to subtract items that increase net income but not provide cash in the current period, to add back items that decrease net income but not use cash in the current period, and to adjust for items that impact cash flow but not net income Thus, the process adds back depreciation (a non-cash expense) to earnings and adjusts for changes in working capital accounts as well as other non-cash revenues and expenses Changes in non-cash working capital accounts are the main difference between cash flows from operations and earnings plus depreciation (EBD) As illustrated in the Appendix, working capital changes typically include changes in accounts receivables, changes in liabilities, changes in inventories, and changes in other operating activities.10 Accrual accounting by its nature smoothes earnings, recognizing higher (lower) earnings than cash flows during periods of growth (decline) As a result, the difference between earnings and cash flows (ACCRUALS) is a function of firms’ business stage Analogous to changes in the number of employees and other growth attributes, WCACC is sensitive to firms’ growth stage 10 As a second illustration, consider a firm that buys inventory in response to an impending increase in demand, pays cash for it, but does not sell it by the end of the period Note that CFO is lowered because the firm paid out cash and WCACC is higher because inventory increased In this example, while actual cash is reduced, profits are not because unsold inventory is considered an asset, not an expense Inventory only impacts profits in the period when it is sold Summing the two components creates accrual accounting net income, and involves adding together cash flows (CFO) with investment in inventory (WCACC) A similar argument goes through when we allow for other working capital items like receivables and accounts payable During expansions, investment in fixed assets is naturally accompanied by investments in working capital to support growth.11 What we clarify in our paper is that these investments in inventory and receivables are included in the cash measure, EBD! Because capital expenditures and changes in working capital are two sides of the same growth phenomenon, the investmentEBD sensitivity is driven by the positive relationship between capital investment and WCACC In the introduction we argued that capital investment-EBD sensitivities vary monotonically across a-priori partitions based on characteristics such as dividend payouts, firm age and Cleary’s (1999) metric because the informativeness of the WCACC component of EBD with respect to contemporaneous investment in working capital varies systematically across these partitions To clearly illustrate our argument for why the correlation between investment and WCACC varies across a priori partitions, consider the following simple example Assume: Working capital investment, WCI = a*I, where I is capital investment That is, working capital investment is proportional to capital investment Let VAR ( I )  I2 ; and WCACC = WCI +  , where  is independent random fluctuation in WCACC due to timing issues Let VAR ( )   and assume that   is the same for all firms Given these assumptions, it is straightforward to show that ( I , WCACC )  a Cov( I , a * I )  2 I 1/ 2 1/  I * (a  I    ) (a  I    ) (3) As we document below (table 3), dividend payout, firm age and Cleary’s (1999) metric are all significantly correlated with firm growth In terms of our example then, our goal is to explain 11 Fazzari et al (2000) in their reply to Kaplan and Zingales (1997 and 2000), construct a measure of “total investment” by adding the changes in both inventory and accounts receivable to capital expenditures why ( I , WCACC )  Allowing  I to potentially be a function of growth we write this Growth partial derivative as  ( I , WCACC )  ( I , WCACC ) d I  Growth  I dGrowth Using (3), it is easy to show that the first term in (4), (4)  ( I , WCACC ) , is positive Thus, the  I overall sign of the total derivative is determined by the sign of d I As we show dGrowth empirically below (table 4), the standard deviations of investment, EBD and WCACC all increase across the growth partitions, implying that d I  In essence, as growth increases, dGrowth WCACC becomes more informative with respect to working capital investment as  I increases while   is constant and, therefore, is more positively correlated to capital expenditure (I) It is our contention that this argument explains observed investment-EBD sensitivities We turn now to our empirical analysis Empirical analysis of investment-cash flow sensitivities In section 3.1 we describe our sample and present descriptive statistics In section 3.2 we implement our main analyses of investment-cash flow sensitivities In table 2, we split the cash flow variable used in the literature, EBD, into its two components, CFO and WCACC, and analyze these two components separately In table 3, & 5, we analyze the relation between capital investment and EBD, CFO, WCACC, and INV across three commonly used partitioning variables: dividend payout ratio, firm age, and ZFC, the financial constraint index from Cleary (1999) Finally in table 6, we attempt to distinguish our argument from Alti (2003) 3.1 Sample and descriptive statistics Our sample selection is similar to that of Gilchrist and Himmelberg (1995), Almeida et al (2004), and Almeida and Campello (2004) We consider the universe of manufacturing firms (2000