Internal Information Quality and the Sensitivity of Investments to Market Prices and Accounting Profits

Transcription

1 Internal Information Quality and the Sensitivity of Investments to Market Prices and Accounting Profits Shane Heitzman* Marshall School of Business University of Southern California Mengjie Huang Gabelli School of Business Fordham University Abstract: January 10, 2018 We ask whether the quality of internal information matters for investment decisions. We predict that investment is more sensitive to internal profit signals and less sensitive to external price signals when managers have higher quality internal information. Consistent with recent theoretical and empirical research, we proxy for internal information quality using observable information properties. We find that the sensitivity of investment to profitability is increasing, while the sensitivity of investment to market-to-book is decreasing in internal information quality. Our focus on internal information and decision making offers new and unique insights on the importance of information quality and complements the growing literature on the role of external reporting quality in reducing financing frictions. Keywords: Investment; Information Quality; Internal Information; Accounting JEL Classification: M41; G31; D81; D83 We appreciate comments from John Gallemore, Ed Maydew, Nemit Shroff, Toni Whited, Jerry Zimmerman, two anonymous reviewers, and workshop participants at the University at Buffalo, University of Minnesota, University of Arizona, and the European Accounting Association Conference. *Corresponding author. 701 Exposition Blvd. HOH 822, Los Angeles, CA phone:

2 Internal Information Quality and the Sensitivity of Investments to Market Prices and Accounting Profits 1. Introduction Expanding our understanding of financing constraints on investment behavior, a growing body of research examines the role of external reporting quality in relaxing these constraints. 1 The evidence to date suggests that higher quality financial reporting reduces asymmetric information between insiders and outsiders, controlling moral hazard and adverse selection problems that preclude efficient investment. In this paper, we pivot the focus on information quality toward a key determinant of efficient decision making within the firm: the quality of the manager s internal information (Kinney 1999). Specifically, we ask how the quality of internal information affects the sensitivity of investment decisions to signals from internal and external sources. We consider this question from the perspective of a manager who has imperfect information about the value of the firm s investment opportunities and updates the firm s capital budget with signals obtained from internal sources (profit forecasts) and external sources (market values). The quality of a given investment signal is determined by the timeliness and precision of the information it provides about the relevant investment decision. The manager that obtains timelier and more precise forecasts of investment profitability from internal sources should place more weight on that information and less weight on the market s belief reflected in price. 2 While both signals can be incrementally informative, their influence on the decision should depend on their relative timeliness and precision. 1 See, for example, Biddle and Hilary (2006), Biddle, Hilary and Verdi (2009), Beatty, Liao and Weber (2010a and b), Balakrishnan, Core and Verdi (2014), Balakrishnan, Watts and Zuo (2016). While this focus on capital allocation is crucial for understanding the economic consequences of financial reporting, Bushman and Smith (2001) argue that financing constraints is just one potential channel. Recent work examines the impact of external information environments on the identification of investment projects (Badertscher, Shroff and White 2013). 2 The attributes of efficient investment in internal information clearly vary across firms as a function of firm complexity and the diffusion of specific information within the firm (Fama and Jensen 1983). A firm s equilibrium 1

3 To motivate the relation between internal profitability signals and investment responses, consider the well-known positive empirical association between investment and operating profits (the typical empirical proxy for cash flow). While early studies attribute this result to a financing constraints mechanism (e.g., Fazzari, Hubbard and Petersen 1988), it also arises when operating profit is informative about economic performance and thus investment opportunities (Alti 2003; Cooper and Ejarque 2003). Because a manager s investment decisions depend on the expected profitability of those decisions, this opens the door for the sensitivity of investment to profitability to depend on the quality of the manager s internal signals. Thus, for a given shock to expected profitability, the manager with timelier and more precise internal information about expected profits can reallocate capital more efficiently. Therefore, we predict that as the quality of the manager s internal information improves, the sensitivity of investment to internal performance signals, such as forecasted profitability, becomes stronger. In contrast, a manager with lower quality internal information is more likely to look to sources outside the firm. For publicly traded firms, stock prices provide an observable signal of the market s information. Empirical evidence suggests that the manager s investment decisions do respond to these external market signals, especially when private information production by outsiders is better reflected in stock prices (Luo 2005; Chen, Goldstein and Jiang 2007; Bakke and Whited 2010). When the manager s comparative information advantage improves such as through upgrades to the internal information systems their response to price signals should weaken. Thus, we predict that as the quality of the manager s internal information improves, investment decisions should become less sensitive to market prices. investment in the internal information system equates the marginal costs (such as the cost of gathering and communicating information) with the marginal benefits (more efficient decisions). 2

4 We use the sum of capital expenditures and research and development to proxy for investment, market-to-book to proxy for the market s valuation of investment opportunities, and earnings before depreciation and R&D to proxy for profitability. This is consistent with the empirical definitions of investment, Q and cash flow used in much of the prior literature. Our empirical tests focus on the sensitivity of investment to market-to-book and profitability conditional on proxies for internal information quality. We predict that as internal information quality improves, the correlation between investment and internal signals (profits) becomes stronger, while the correlation between investment and external signals (market values) becomes weaker. We follow Gallemore and Labro (2015) and Goodman et al. (2014) and base our primary proxies for internal information quality on the speed of earnings news, managerial earnings guidance, internal control weaknesses and unintentional financial statement errors. The use of observable instruments for unobservable internal information constructs is unavoidable. Fortunately, the correspondence between the quality of the manager s internal information and the quality of what they report externally (and hence what we can observe) is arguably high. This assertion is supported theoretically by Hemmer and Labro (2008) who show that the decision usefulness of external financial reporting is inherently tied to the quality of information for managerial decision making, and empirically by Dichev et al. (2013) who find that over 80% of CFOs rank internal use of externally reported earnings as very important and emphasize the use of one number for internal and external communications. Complementing this direct evidence, several other studies document consistent links between the properties of internal information and external reporting. For example, firms with weaknesses in their internal controls also appear to have lower quality GAAP accruals (Ashbaugh-Skaife et al. 2008; Doyle, Ge and McVay 2007), face higher borrowing costs (Costello and Wittenberg- 3

5 Moerman 2011), are slower to release their financial statements (Ettredge, Li and Sun 2006) and provide less accurate guidance (Feng et al. 2009). 3 Brazel and Dang (2008) find that firms implementing an ERP system release their audited financial statements with less delay, while Dorantes et al. (2013) find that ERP firms issue external earnings guidance more often and with higher quality. Using novel settings to identify investments in internal information, both Samuels (2016) and Ittner and Michels (2017) provide evidence that internal information quality is positively reflected in external reporting quality. Shroff (2017) and Cheng, Cho and Yang (2017) find that changes in external reporting requirements (i.e. GAAP) can alter the internal information environments managers rely for decision making. While it is conceivable that managers could obscure high quality internal information when reporting to shareholders, they cannot generate high quality financial reports and disclosures from low quality internal information. We are not the first to recognize that a sort on the observable attributes of external information is also a sort on the unobservable attributes of internal information. Thus, we rely on prior literature to select our internal information proxies and take steps to ensure that our interpretations are not confounded by external reporting incentives. Our main result is consistent with our prediction: investments by managers with higher quality internal information are more sensitive to profits and less sensitive to market prices. This finding is robust across a diverse set of information quality proxies including the speed of earnings release and managerial guidance and is consistent with the interpretation that managers with high quality internal information are less likely to defer to the market s opinion of investment opportunities. Instead, they place more weight on timely internal information about shocks to profit opportunities. 3 See also Cheng, Dhaliwal and Zhang (2013) who argue that investments in internal controls (following disclosure of a material weakness) reduce financing constraints by improving the quality of information reported to capital markets. 4

6 To provide additional support for our conclusions, we first conduct tests that exploit timeseries shocks to internal information quality. First, we examine how firms respond to external and internal investment signals after they remedy internal control weaknesses. We predict and find that when internal control problems are fixed, internal information quality increases and subsequent investment becomes more sensitive to accounting profits and less sensitive to market prices. Second, we use the adoption of SFAS 142 as an exogenous shock to firms internal information environment. SFAS 142 compliance likely requires managers to acquire additional information, thus enriching the internal information environment (Cheng, Cho and Yang 2017) and by extension, internal decision making and control. Our results suggest that affected (goodwill) firms rely more on profit signals and less on price signals when investing in the post-sfas 142 period. Although our results suggest that internal information quality affects investment responses in predictable ways, they are subject to a number caveats. First, we acknowledge the implications from prior research that agency conflicts cause investment to be sensitive to internal funding and control for cash holdings in the regression. 4 The inclusion of cash follows from our assumption that the stock of internal funds offers a cleaner proxy for the cash-based agency conflicts described by prior research (Jensen 1986; Biddle, Hilary and Verdi 2009; Nikolov and Whited 2014) and is consistent with research on external financing-based motives for cash holdings (Opler et al. 1999; Bates, Kahle and Stulz 2006). Consistent with Biddle, Hilary and Verdi (2009) and others, we find that an increase in information quality reduces the sensitivity of investment to cash holdings. Second, we recognize that the quality of the external signal (price) is also relevant and leads to 4 For example, under an adverse selection hypothesis, improving information quality reduces financing frictions when firms need external capital and thus reduces the sensitivity of investment to internal funding proxies. Additionally, moral hazard problems can lead managers to spend cash or exploit overpriced equity to undertake projects that generate private benefits. Under a moral hazard hypothesis, improving information quality enhances monitoring and reduces the sensitivity of investment to both market valuations and internal funds. 5

7 reverse predictions on the sensitivity of investment to external and internal signals. Employing proxies for stock price informativeness used in prior research, our evidence suggests that more informative price signals increase the sensitivity of investment to market prices and decrease its sensitivity to profits. Third, the moral hazard hypothesis predicts that high quality information reduces the sensitivity of investment to prices because managers in those firms are less likely to respond opportunistically to mispriced equity. To address this, we include proxies for mispricing and their interactions with market-to-book. Our results are unchanged. Finally, to address the evidence that shows that measurement error in market valuations biases the coefficient on marketto-book toward zero (Erickson and Whited 2000), we follow Shroff (2017) and utilize techniques developed in Erickson and Whited (2002) and Erickson, Jiang and Whited (2014) to correct for measurement error in the proxy for investment opportunities. Our results are robust. Our study contributes to the literature in several ways. First, we build on the descriptive theory that links attributes of the firm s information environment to its managers investment decisions, extending prior research by focusing on the equilibrium relation between the design of the internal information system and decision making within the firm. Second, we empirically examine the role of internal information quality by exploiting its intrinsic link to observable information constructs. This builds on a promising stream of research that expands the boundaries of inquiry on the internal information environment and leverages the coordination of information demands between users inside and outside the firm to construct empirical proxies for internal information quality. Third, our empirical strategies focus on the identification of internal information effects while controlling for the influence of asymmetric information problems between managers and capital providers. In doing so, our focus on internal information and decision making complements the growing 6

8 literature on the importance of external reporting quality in reducing financing frictions that impede efficient investment. 2. The Investment Framework and Hypothesis Development 2.1. The empirical investment framework We rely on structure extended from neoclassical investment theory to motivate the link between internal information quality, investment, and internal and external investment signals. Under the q theory developed in Hayashi (1982) and Summers (1981), in perfect markets without financial frictions, investment is determined solely by marginal q, which should capture all the factors relevant to the investment decision. Because theoretical q is unobservable, the basic empirical model is implemented with a market-based proxy for q and is usually adapted to include other factors predicted to affect the investment decision, i.e.: I it M it 1 EBD it Cash it 1 = α A 0 + α 1 + α it 1 A 2 + α it 1 A 3 + α it 1 A 4 Leverage it 1 it 1 (1) + α 5 ln(a it 1 ) + e it I/A is investment scaled by beginning total assets, where investment is capital expenditures plus research and development. M/A is the beginning market-to-book asset ratio, the most commonly used empirical proxy for q. Evidence from the investment literature shows that the coefficient on market-to-book is decreasing in adjustment costs and measurement error in market-based proxies for q (Erickson and Whited 2000). However, building on the interpretation of market-to-book as the market s valuation of investment opportunities at the beginning of the period, there is growing evidence that the coefficient on market-to-book increases when managers incorporate feedback from external signals in market prices into their investment decisions (Chen et al. 2007; Bakke and 7

9 Whited 2010). There is mixed evidence that the market-to-book coefficient picks up opportunistic responses to market mispricing (Polk and Sapienza 2009; Bakke and Whited 2010). 5 EBD is operating profits or earnings before depreciation (and R&D). We employ operating profit (EBD) because both the cash flow and accrual components should reflect a manager s internal information about expected profitability. However, the economics literature first employed EBD as an empirical proxy for liquidity. Fazzari et al. (1988) show that investment is positively correlated with EBD, concluding from that the existence of costly external finance (and perhaps motivating its early use in the accounting and finance literature on financing constraints). Subsequent work by Kaplan and Zingales (1997) raises important questions about the reliance on investment-cash flow associations to identify the existence of financing constraints. The emerging consensus in the literature is that the loading on EBD is likely not a story of financing constraints. 6 Operating earnings naturally reflect information about the profitability of investment opportunities, paving the way for alternative explanations for the positive correlation between operating earnings and investment (Hennessy, Levy and Whited 2007). First, when firms have market power, the proxy for average q (market-to-book) will diverge from true investment opportunities (marginal q), allowing current earnings to explain more of the variation in current period capital allocation (Cooper and Ejarque 2003; Moyen 2004). Second, the significance of operating earnings in the investment regression depends on measurement error in the market s valuation of current and future investments. This can induce positive bias on the coefficient on 5 Market-to-book is also correlated with the conservatism in firm s accounting policies. However, the impact on our interpretations is not clear unless conservatism is correlated with internal information quality such that the conditional investment sensitivities are driven by bias in market-to-book triggered by conservatism. In a recent study on investment and firm-level conservatism, Lara, Osma and Penalva (2016) report a negative correlation between various conservatism measures and market-to-book, which mitigates our concerns. 6 In the accounting, Bushman, Smith and Zhang (2012) provide evidence that the loading on accrual-based proxies for cash flows in the economics literature can be partially explained by co-movement in capital expenditures and working capital investment suggesting that investment sensitivity to EBD is not a story about financing constraints. 8

10 operating earnings as current profitability must play a larger role in explaining investment when market-to-book is a poor proxy for investment opportunities (Erickson and Whited 2000). Third, following Jensen (1986), operating earnings can also load if managers view excess cash flows as a source of capital for empire building. In this paper, we assume that managers have rational expectations and form unbiased forecasts about profitability. This implies that profits realized during the period (EBD) are a viable proxy for forecasted profits at an earlier time when the investment decision is made. Thus, we view operating profits primarily as an internal signal of productivity shocks and opportunity costs relevant for investment, and not as a proxy for internal funds. That said, we recognize the important role that internal funding plays in mitigating financing constraints when external financing is costly and structure our analysis to include the most direct proxy for the availability of internal funding cash and short term investments ( cash ) held at the beginning of the period. The finance literature finds that firms build cash balances when they anticipate financing constraints (Opler et al. 1999; Almeida, Campello, and Weisbach 2004; Bates, Kahle and Stulz 2009). Biddle et al. (2009) incorporate cash holdings and leverage into their measure of external financing frictions. We include leverage in the model following Hennessy s (2004) finding that debt overhang distorts investment and results in under-investment and because investment and leverage could be endogenously correlated if a positive shock to investment opportunities leads to both an increase in investment and an increase in debt issuance Empirical predictions: internal information quality and investment responses 7 In place of cash holdings, we explored other financial constraint measures as robustness checks, including the KZ index based on Kaplan and Zingales (1997), the Whited and Wu (2006) index, the SA index from Hadlock and Pierce (2010), and the combined cash and leverage measure as in Biddle et al. (2009). The main inferences are unchanged under all these alternative specifications. To address the concern that multicollinearity arising from correlation between cash holding and cash flow might bias our results, we exclude cash holding from the regressions in robustness tests and obtain similar inferences. 9

11 We are interested in whether an increase in the quality of internal information will cause investment to become more sensitive to internal investment signals and less sensitive to external signals. 8 M/A and EBD serve as proxies for a) an external signal of investment opportunities provided by the firm s market price, and b) an internal signal of expected profitability, respectively. The manager s investment responses to each of these signals depend on the signals relative quality. All signals that are informative should be used. However, the decision maker should place more weight on the higher quality signal, causing investment to become more sensitive to that signal. An investment in high quality internal information provides the manager with timelier and more precise internal feedback about the firm s productivity and opportunity costs. Such information enhances the quality of profit forecasts for both recent and proposed investments. This improves the manger s ability to reallocate capital to the most valuable projects in a timelier fashion. Under the internal information quality hypothesis, the sensitivity of investment to internal profit signals is increasing in the quality of internal information. 9 In contrast, as internal information quality declines, external sources of information such as market prices can become more important. The potential for market prices to guide these capital allocation decisions at the firm was recognized early on by Hayek (1945) with recent work including Dow and Gorton (1997), Dye and Sridhar (2002), Luo (2005), Gao and Liang (2013) and Zuo (2016). This literature starts with the assumption that managers are well-informed about 8 To be clear, in this paper, internal and external information refers to the source of the investment signal, i.e. whether the signal is derived inside the firm (from the internal management system) or outside the firm (from market prices). This is different from the internal and external economic forces that affect investment: internal factors are created by the firm, such as firm-specific technologies, resource and investment opportunities; external factors are created outside the firm, such as political and competitive factors. Managers and the market incorporate both internal and external factors in developing their investment signals, but with different comparative advantage. 9 An investment in higher quality internal information could potentially overlap with the high ability manager ; whether high ability managers are more likely to make those investments, or the high ability managers become that way because of the greater investment in internal information quality, the implication would be that managers with higher quality information are better able to synthesize information from forecasts and translate those into actionable and profitable investments. 10

12 inside factors such as the firm s technological capabilities and specific investment opportunities. The value of those investment opportunities is also affected by outside forces like competition, product demand, political and geographical considerations and other factors about which the manager may be relatively less informed. This opens the door for outside investors to have a comparative information advantage and an economic incentive to reveal their information through trading. Their trades make prices more efficient and communicate more information to managers. 10 Consistent with this, Chen et al. (2007) and Bakke and Whited (2010) find that investment is more sensitive to market values when more private information production is reflected in the firm s stock price. Since managers that obtain higher quality internal information for decision making should depend less on external signals for investment decisions, the internal information quality hypothesis predicts that the sensitivity of investment to market valuations is decreasing in the quality of internal information Alternative explanation: External financial reporting quality effects In this paper, we focus on the role of internal information quality in the capital allocation process. Although we view this as distinct from research emphasizing the role of external financial reporting quality in mitigating agency conflicts that affect investment, it is related conceptually through the emphasis on the attributes of the information environment and empirically through the 10 For example, as Dye and Sridhar (2002) theorize and Luo (2005) demonstrates empirically, managers can make an announcement of a proposed transaction, observe the market reaction to the announced transaction, and decide whether to proceed based on the reaction. Unfortunately, such overt instances are relatively uncommon. In general, the researcher s ability to infer the precise context of a price movements is tricky without a clear measure of the underlying news. Instead, we opt to rely on the more general assumption that managers have some ability (even if noisy) to infer market information from price movements in context. That context might be the latest unemployment figures, an announcement of tax reform in the EU, a competitors announcement of a successful innovation. The manager can use price movements of its own stock and its peers stock in this process. 11 The market obtains at least some of its information from financial reports and disclosures made by managers. As the quality of internal information improves, the quality of reporting and disclosure also improves, and market prices can become more efficient and more highly correlated with actual investment decisions. In this world, higher information quality should increase the sensitivity of investment to both the internal profit signal (α 1 ) and the external market price signal (α 2 ). We do not find this prediction empirically. 11

13 basic investment model and the choice of information proxies. In this section, we briefly discuss the prior research on external financial reporting effects and describe how our methodology accounts for these potentially confounding effects in order to enhance our inferences about internal information effects. Theory. Following Bushman and Smith (2001) an important way for financial reporting attributes to influence investment by controlling agency problems: shareholders delegate investment decisions to a manager with more precise information about the firm s investment opportunities. The self-interested manager has an incentive to choose projects that generate private benefits. In response, shareholders control moral hazard and adverse selection costs through incentive alignment and monitoring. The assignment of decision rights, incentive structures and monitoring mechanisms are therefore chosen to maximize firm value. Information plays a key role. A reduction in information quality can affect the firm s marginal investments if it improves the manager s ability to implement projects that generate private benefits or increases marginal financing costs sufficient to turn a project s NPV negative. Motivated by the literature documenting that investment appears sensitive to internal funds, recent accounting studies use this agency-based framework to test predictions about the role of external reporting quality for investment. These studies generally argue that reporting high quality information to external users reduces financing frictions and thus increases investment efficiency. Under an adverse selection narrative, an increase in the quality of information reported to the market reduces the sensitivity of investment to internal funds and increases the sensitivity of investment to investment opportunities. The basic idea is that when external funds are costlier than internal funds, investment will be constrained by the availability of internal funds. An improvement in the quality of information reported to capital providers reduces the adverse 12

14 selection costs of external financing and thus reduces the reliance on internal sources of project funding. 12 Under a moral hazard narrative, managers respond opportunistically to market mispricing and internal funding shocks that facilitate empire building. With a positive internal funding shock, the manager would rather take on a negative NPV project that provides private benefits than distribute excess funds to investors. Reporting high quality information provides directors and investors with timely and precise information about managerial decisions ex post, reducing ex ante incentives for opportunistic investment when the firm holds excess cash or its equity is overpriced. 13 Evidence. Biddle and Hilary (2006) find that cash flows explain investment better in countries with less transparency. 14 Biddle et al. (2009) argue that firm-level reporting quality reduces deviations from predicted investment, the impact of incentives to over- or under-invest and the sensitivity of investment to cash holdings. Beatty, Liao and Weber (2010a) show that the impact of reporting quality on investment-cash flow sensitivity is weaker when firms can resolve asymmetric information through private debt markets, while Beatty, Liao and Weber (2010b) find 12 We recognize the possibility that the impact of information quality could be priced. However, the evidence is mixed on whether financial reporting quality is a priced risk factor (Francis et al. 2005, Core et.al 2008). Prior studies have relied on the adverse selection channel to justify investment-q sensitivity as a proxy for investment efficiency (e.g., Shroff et al. 2014). 13 For example, accounting information can provide contracting parties with more information to monitor managerial performance. Financial information that is more informative about investment opportunities and managerial actions allows superiors to write incentive contracts based on reported performance. This also reduces the ability and incentives to mislead superiors about potential project payoffs and thus reduces the likelihood the manager will overinvest in self-serving projects. 14 Biddle and Hilary (2006) is not directly comparable to our study. They focus primarily on a cross-country comparison of financial constraints and reporting quality. They derive country-level measures of financial constraints based on within-country regressions of investment on market-to-book and operating cash flow, and compare those to country-level measures of earnings quality (smoothness, timeliness, loss avoidance and aggressiveness). The idea is that firms in countries with higher quality earnings should not be as sensitive to cash flow for investment. The results, based only on 33 observations at the country level, support that prediction. In their firm-specific tests within the US, they rely on investment-cash flow sensitivities to test for the presence of financing constraints and reporting quality s ability to mitigate them. In addition to the findings in Bushman et al. (2012) who cast doubt on using investment-cash flow sensitivities as a measure of financial constraints, Biddle and Hilary (2006) use an uncommon measure of investment-cash flow sensitivity that lacks controls for investment opportunities embedded in market valuations or other controls such as size and leverage. To our knowledge, this approach has not been adopted by the literature. 13

15 that firms with low reporting quality finance investment through leases rather than purchases. 15 Other studies that rely on financing frictions as the main channel for reporting quality include Balakrishnan, Core and Verdi (2014) who show that a negative shock to collateral values has a stronger effect on investment for firms with lower reporting quality, and Balakrishnan, Watts and Zuo (2016) who argue that a negative shock to the credit market has a stronger impact on investment for less conservative firms. Shroff, Verdi and Yu (2014) tackle the agency problem within the firm, showing that investment decisions by subsidiaries in high quality local information environments are more responsive to local investment opportunities. This aforementioned research on external financial reporting quality is relevant to our study insofar as we rely on a tight link between internal and external information to derive our empirical proxies for internal information quality. Thus, we must account for the potential that our empirical proxies for internal information quality could pick up attributes of external financial reporting quality that affect adverse selection and moral hazard problems that cause managerial investment decisions to depend on the availability of internal funds the dominant mechanism in prior research. While this alternative explanation generally works to bias against finding evidence consistent with the internal information hypothesis, we further control for these effects by interacting our information quality proxies with the level of cash holdings. Because our proxies 15 Unlike our paper, Beatty, Liao and Weber (2010a) find that higher accounting quality leads to a drop in the investment-cash flow association. But the two papers are very different and caution should be exercised in comparing the results. Beatty et al. are interested in the impact of accounting quality when firms are issuing debt. Thus, their sample is restricted to firms that issued public or private debt in the year of investment. This narrows the sample considerably. Additionally, unlike our focus on internal information, they focus on earnings-based measures of accounting quality, including: accrual quality, earnings persistence, the explanatory power of an earnings persistence model, and the earnings-cash flow correlation. Moreover, these proxies are measured in the year of investment (as opposed to lagged as in our study). Given these large and fundamental differences, our results do not have obvious implications for their stud and vice versa and we do not attempt to reconcile the difference. 14

16 are potentially sorts on external reporting quality, we predict the interaction will be negative: an increase in information quality should reduce investment sensitivity to cash holdings Empirical Methodology and Results 3.1. Research design To test the impact of internal information quality on investment sensitivity to external price and internal profit signals, we interact our proxies for internal information quality with market-tobook and operating earnings. Moreover, we include an interaction between internal information quality and cash holdings as a control for the impact of information quality on adverse selection costs and moral hazard problems. Specifically, we run the following regression: I it M it 1 EBD it Cash it 1 M it 1 = δ A j + τ t + α 1 + α it 1 A 2 + α it 1 A 3 + α it 1 A 4 IIQ + α 5 it 1 A it 1 EBD it Cash it 1 IIQ + α 6 IIQ + α A 7 IIQ + α it 1 A 8 Leverage it 1 it 1 (2) + α 9 ln(a it 1 ) + e it. IIQ is a proxy for internal information quality. M/A serves as the proxy for the external signal based on market prices and EBD serves as the proxy for the internal profit signal. δ j and 16 Conceivably, the same managers that invest more in high quality internal information are also creating an asymmetric information problem that increases adverse selection costs assuming that this information is not shared with market participants. Thus, what we label high quality internal information is potentially a proxy for low quality external information and thus our empirical evidence is observationally equivalent to prior studies on external reporting quality and financing constraints. We think this explanation is unlikely to be at work in our study for at least four reasons: 1) the empirical association between internal information quality measures (filing speed, guidance, etc.) and external reporting quality measures (analyst coverage and forecast accuracy) is positive, supporting our assertion that a sort on high internal information quality is a sort on high external reporting quality even if some firms make active attempts to muddy their information, 2) the association between internal information quality proxies and proxies for adverse selection in equity markets is generally negative, supporting the view that high quality internal information reduces asymmetric information with the market; 3) the logic implies that the increase in internal information quality leads to a higher cost of debt and possibly equity, running contrary to a large body of evidence, and 4) to the extent this mechanism is at work, it should actually play out on cash holdings as that best represents (in the model) the availability of internal funding and thus the potential for adverse selection problems to crop up. This does not appear to be the case. 15

17 τ t represent industry and year fixed effects and are included to control for differences in industry organization and practices and time trends. 17 Standard errors are clustered at the firm and year level Sample and variable construction We begin with a sample of 83,645 firm years drawn from the sample period 1988 through We require the firm to be publicly traded on NYSE, NASDAQ or AMEX and have sufficient information to calculate at least one of our information quality measures. We exclude firms with SIC codes between and Firm-years with asset growth exceeding 100% are deleted to avoid the effects of large M&A transactions and seasoned equity offerings. We measure internal information quality using five proxies that draw on intuitive connections to the internal information environment. 18 In doing so, we acknowledge the concerns expressed by Leuz and Wysocki (2016) who note that measures of external reporting quality based on the quality of a firm s accruals are problematic for examining real effects like investment. Because they are inherently connected to the firm s underlying economics, interactions between earnings and accruals quality can explain variation in observed investment for reasons that have nothing to do with the quality of the accounting-based performance signal. To mitigate the direct influence of economic performance on our choice of proxies, we adopt a set of proxies that do not rely on 17 Although there does not appear to be a uniform consensus, a review of the related literature suggests that a typical study of investment and information employs some combination of industry, firm or year fixed effects. In our main tests, we employ industry and year fixed effects and cluster standard errors at the firm and year level. To facilitate better interpretation of the results, we rank the continuous IIQ measures into deciles that lie between 0 and 1. This procedure essentially removes much of the within-firm variation, and attempts at using firm fixed effects with our ranking variable is highly sensitive when firms do not vary in their ranking over the sample period. In a robustness test described in Section 4.6 we used the raw IIQ measure and estimate the model with firm and year fixed effects. Our results are robust. 18 An alternative approach is to look at macro-level measures of the information environment, where increasing the amount of information available about peer firms improves identification of investment opportunities. Shroff, Verdi and Yu (2014) and Badertscher et al. (2013) find that firms invest more in response to their investment opportunities when they operate in environments where other firms are providing more public disclosure. Of course, the proprietary costs of disclosure can increase the manager s incentive to reduce the quality of externally reported information, but in general, firms appear to benefit from operating in more transparent environments. 16

18 accruals and is arguably less directly afflicted by this concern. In addition, the regressions include profitability and market-to-book as independent variables, effectively controlling for the covariation in performance and the information quality measures. The first proxy, FilingSpeed, is the length of time it takes the firm to release earnings once the fiscal period closes. It is estimated as the number of days between the end of the year and the release of the annual earnings figure following Jennings, Seo and Tanlu (2014) and multiplied by -1. Because of the tight connection between information used internally and that reported externally, managers with low quality internal information will need longer to prepare external financial statements (Dorantes et al. 2013), delaying the auditor s ability to provide an opinion (Ashton, Willingham and Elliott 1987). The second proxy, Guidance, is an indicator variable equal to 1 if managers make at least one quarterly or annual earnings forecast during the previous year. If providing forward-looking information to the market increases managers exposure to litigation risk, managers with the highest quality information will be most likely to provide guidance. Thus, as suggested by Goodman et al. (2014), earning guidance behavior is a plausible instrument for internal information quality. Our third proxy is GuidanceAccuracy, the average accuracy of annual management earnings guidance issued in the three years prior to the investment year. Among managers that provide forecasts, higher quality information systems facilitate information gathering and enable the manager to make more accurate internal forecasts. Dorantes et al. (2013) show that the accuracy of firms earnings guidance increases following improvements to information systems. The fourth proxy, NoICW, equals 1 if the firm does not disclose any internal control weakness in the year prior to investment. According to Feng et al. (2009), internal control problems could 17

19 lead to erroneous internal management reports, reducing the quality of the manager s internal information set. Our final proxy, NoError, is an indicator variable for the absence of restatements due to unintentional errors. Managers rely on accurate internal records and well-designed information systems for decision making (Gallemore and Labro 2015). Unintentional errors potentially spill over to investment decisions by distorting internal management reports. To reduce measurement error caused by outliers and to improve comparability across the marginal effects, we rank FilingSpeed and GuidanceAccuracy into deciles and scale them to range between 0 and 1 prior to running the regressions, with high values corresponding to higher information quality. In Panel A of Table 1, we report descriptive statistics for the main variables in our model. In Panel B, we report correlations. Our information quality measures are positively correlated with each other, but no correlation is greater than 0.27, consistent with our intent to identify independent constructs for information quality. Notably, our measures are correlated with firm size. Larger firms file annual reports faster, are more likely to provide guidance, have higher guidance accuracy, are less likely to disclose internal control weaknesses and are more likely to have restatements resulting from unintentional errors. To mitigate size-driven interpretations, we control for size in the investment regressions Main results The main results are reported in Table 2. The benchmark investment regression is reported in column (1) and the coefficients on M/A, EBD and Cash are all positive and significant consistent with prior research. In columns (2) through (6) we report the results from interacting each of our five information quality proxies with our proxies for market valuations, operating earnings and cash holdings. Across all five information quality measures, the coefficient on the interaction between M/A and IIQ is significantly negative. The coefficient on the interaction between M/A and 18

20 FilingSpeed of (t-stat = ) in column (2) implies that increasing FilingSpeed by one decile reduces the sensitivity of investment to market valuations by (0.019 / 9). These effects imply that investment-market-to-book sensitivities fall from for firms with the slowest filing speed to for firms with the fastest filing speed. Estimates based on management guidance yield similar inferences. At firms that issue guidance (column 3) the sensitivity of investment to market-to-book is 45% lower than at firms that do not issue guidance ( / 0.022, t-stat = ). We find similar effects when we focus on the accuracy of the guidance in column (4). In columns (5) and (6), we find that investment in firms with higher internal control quality or no restatements is also significantly less sensitive to market values. 19 Thus, across all measures, investment becomes significantly less sensitive to market valuations as internal information quality increases. These results support the internal information quality hypothesis which predicts that as internal information quality improves, managers place less weight on external market signals. Turning to the impact of internal information quality on investment sensitivity to internal profit signals, the interactions between EBD and the information quality proxies (IIQ) are positive and generally statistically and economically significant. In column (2), for example, the coefficient on the interaction between EBD and FilingSpeed of (t-stat = 9.44) implies that a one-decile decrease in the time to report earnings (roughly one week), leads to a (0.147 / 9) stronger sensitivity of investment to internal accounting signals. In column (3) we find that investment by firms that issue guidance is more than twice as sensitive to earnings (( ) / 0.100) as investment by firms that do not (t =8.64). Based on this smaller sample of guidance firms with sufficient data to calculate GuidanceAccuracy, we find some evidence that improvements in 19 For NoICW and NoError, we also conducted propensity-score matching analyses to control for differences in multiple dimensions between ICW and No ICW firms, as well as firms with and without unintentional financial statement errors. Similar inferences can be made from this analysis. 19

21 guidance accuracy increase the sensitivity of investment to earnings, although the effect is only marginally significant (t = 1.72). Compared to firms with internal control weaknesses, firms without weaknesses (NoICW) are about two-thirds more responsive to internal accounting signals (( ) / 0.052). However, regressions using the restatement-based proxy NoError (col. 6) show no evidence that firms without unintentional errors are significantly more responsive to internal signals. Similar to the results on market valuations, these results generally support the internal information quality hypothesis: investments are more sensitive to internal profit signals when managers have higher quality internal information. 20 Prior research suggests that adverse selection and moral hazard problems matter when information quality reflects asymmetric information between insiders and outsiders. To address this, we return to the interaction between IIQ and Cash that serves as our control for the agencybased asymmetric information explanations explored in Biddle et al. (2009) among others. Consistent with those studies, we find that investment becomes less sensitive to cash holdings when FilingSpeed, Guidance and GuidanceAccuracy increase, suggesting that higher information quality mitigates the propensity of managers to overinvest or reduces adverse selection costs of external financing One possibility is that information quality has a differential impact on investment sensitivities to the cash flow and accrual portions of earnings. For example, managers might discount the signal in forecasted accruals more than forecasted cash flows when information quality is low. Following Bushman et al. (2012), we decompose EBD into its cash flows and accruals components and find results consistent with internal information quality having a differential impact on the strength of the relation between investment and cash flows relative to accruals. The effective is positive for both, consistent with our main results, but stronger for interactions with cash flows. Because we lack strong ex ante expectations about the impact of information quality on the components of EBD, and because the results are not consistent across these components, we are hesitant to draw strong inferences from these tests and leave further analysis to future research. 21 The negative coefficient on the interaction between IIQ and M/A is also consistent with the moral hazard argument. To the extent low information quality provides managers with incentives to exploit mispricing by overinvesting when prices are high, or underinvesting when prices are low, the variation in investment sensitivity to market prices explained by information quality could be attributable to moral hazard problems and not to variation in the scope for learning from prices. We examine this explicitly in Section 4.4. and find that our results are not driven by this mispricing explanation. 20