Investment Efficiency with Imprecise Cost of Equity. Lee Biggerstaff Miami University. Brad Goldie Miami University. Haim Kassa Miami University

Transcription

1 Investment Efficiency with Imprecise Cost of Equity Lee Biggerstaff Miami University Brad Goldie Miami University Haim Kassa Miami University December 2017 Contact Information: Biggerstaff is an Assistant Professor of Finance, Farmer School of Business, Miami University, Oxford, OH 45056, Goldie is an Assistant Professor of Finance, Farmer School of Business, Miami University, Oxford, OH 45056, Kassa is an Assistant Professor of Finance, Farmer School of Business, Miami University, Oxford, OH 45056,

2 Investment Efficiency with Imprecise Cost of Equity Abstract Cost of equity is a direct input when a firm makes investment decisions. We find that firms with more imprecision in their cost of equity are more likely to deviate from optimal investment level. Our results are robust through time and are consistent when we sort on various firm characteristics. Imprecise estimates of cost of equity lead to both over and under investment. We also show a theoretical link between imprecision in cost of equity estimates and deviation from optimal investment. We further show that imprecise cost of equity and leads to poorer firm operating performance.

3 1. Introduction Efficient investment of capital is critical to shareholder wealth as a significant component of enterprise value stems from future growth opportunities. Most firms rely on discounted cash flow (DCF) analyses to make capital investment decisions, which requires estimating future cash flows and discounting them at the at the appropriate rate to reflect risk and timing. The efficacy of DCF models to direct investment is a function of both the accuracy of the forecasted cash flows and the identification of the correct discount rate. To date, most research has focused on factors that may impact the accuracy of forecasted cash flows (Biddle et al., 2009; Goodman et al., 2014) and little research has focused factors that impact the ability of managers to identify the correct discount rate. A firm s true cost of equity is unobservable and estimates for the cost of equity often lack precision (Fama and French, 1997). There are many factors that can impact the precision of a cost of equity estimate. A firm using the CAPM to estimate its cost of equity needs to identify the appropriate risk-free rate, the expected market risk premium, and their firm s beta factor. Uncertainty about any of these components leads to lower precision in the cost of equity estimate. In this paper, we study the effect of imprecise cost of equity estimates on corporate investment efficiency and show that firms with imprecise cost of equity estimates are more likely to invest at inefficient levels when compared to firms with more precise estimates. We measure the relative precision of a firm s cost of equity estimate with the standard error of beta from a CAPM regression, which we call CE Imprecision. 1 Higher 1 We use 36 months of returns and the CRSP value-weighted index in all tests. Results are robust to alternative estimation windows, return frequency, and benchmark return. 1

4 values of CE Imprecision indicate that the confidence interval around the estimated beta coefficient is larger and thus the firm is more likely to rely on an estimated cost of equity that is higher or lower than the firm s true cost of equity. We rely on CAPM based measure because survey evidence suggests that most firms use the CAPM to estimate their cost of equity (Graham and Harvey, 2001; AFP, 2013). Consistent with Fama and French (1997), we show that cost of equity estimates often lack precision. For example, the estimated CAPM beta coefficient for Honeywell International is 0.90 with a 95% confidence interval that ranges from to This translates to an estimated cost of equity between 6.76% and 9.84%. 3 Although this confidence interval appears economically large, it is the most precisely estimated beta coefficient from the S&P500 index. 4 We also document significant dispersion in the precision of firms cost of equity estimates. When we sort firms into quintiles based on CE Imprecision, the average firm in the lowest quintile has a beta estimate of 0.84 with a 95% confidence interval from 0.40 to 1.29, while the average firm in the highest quintile has a beta estimate of 1.55 and a 95% confidence interval from to Two conclusions can be made from looking at these distributions. First, firms with relatively precise cost of equity estimates still have significant uncertainty about their true cost of equity. Second, a large number of firms rely on cost of equity estimates that are basically uninformative in regards to their true cost of equity. We provide some validation of this measure using information from analysts. Implied cost of capital can be estimated using analysts earnings forecast and current 2 Using 36 months of returns ending 12/31/2016 with the CRSP VWRETD used to measure the market return. 3 This assumes a risk free rate of 2% and a market risk premium of 7%. 4 A table highlighting the most and least precise beta estimates is found in the appendix. 5 When using 7% as the expected market risk premium, this translates to a range of 6.23% for the average firm in the lowest quintile and a range of 36.82% for the average firm in the highest quintile. 2

5 stock price. Higher values of CE Imprecision should result in greater dispersion in analysts implied cost of equity as analysts also have a difficult time identifying the correct cost of equity. We do find that firms with higher values of CE Imprecision have greater dispersion in the analyst-implied cost of equity, which supports the notion that a firm with a less precise beta estimate is more likely to use an estimated cost of equity that deviates significantly from its true cost. Following the prior literature (see, for example, Goodman et al., 2014), we measure a firm s investment efficiency with Abnormal Investment, which is the absolute value of the residual from a regression of capital investment on Tobin s Q, cash flow, prior asset growth, and prior investment. 6 Larger values of Abnormal Investment indicate larger deviations from expected investment and thus lower investment efficiency (Bae et al. 2017). Using this basic framework, we study the relationship between CE Imprecision and Abnormal Investment and find that higher values of CE Imprecision are associated with larger values of Abnormal Investment. This is consistent with the view that imprecise cost of equity estimates lead to sub-optimal investment decisions. We find that this relation holds in the presence of firm and year fixed effects, which indicates that the relation is not driven by a time trend or unobserved firm characteristic. Additionally, we continue to find the positive relation in an instrumental variable analysis, which provides further evidence of a causal relation. 7 The intuition of this link is straight forward. First, most firms rely on variations of discounted cash flow analysis to make investment decisions (Graham and Harvey, 2001) 6 These regressions are run for each combination of industry-year with at least 30 observations. 7 We instrument for CE Imprecision with the distance between firm headquarters and NYC under the premise that longer distances increase information acquisition costs, which decreases the precision of the cost of equity estimation. 3

6 and they rely on the Capital Asset Pricing Model (CAPM) to estimate the cost of equity that enters in the discounted cash flow analysis (Graham and Harvey, 2001; AFP, 2013). Second, a firm with an imprecise cost of equity estimate is more likely to utilize a rate that significantly deviates from the true cost of equity when compared to a firm with precise estimate. 8 Firms that over-estimate their cost of equity are going to pass on good investment opportunities and firms that under-estimate their cost of equity are going to investment in projects that provide poor returns to shareholders, both of which are suboptimal. Fama and French (1997) reflect this notion as they conclude that two of the ubiquitous tools in capital budgeting are a wing and a prayer, and serendipity is an important force in outcomes. If our measure of investment efficiency is picking up on sub-optimal investment, then we should also see under performance. We measure operating performance with return on assets and asset turnover to determine if firms with less precise cost of equity estimates have weaker performance. We find a negative relation between CE Imprecision and firm operating performance, which suggests that the deviations from expected investment do reflect sub-optimal investment. This relation holds in the presence of firm and year fixed effects. We conduct additional analyses to demonstrate the robustness of this result. First, we divide our sample into quintiles based on the market value of equity and perform a separate analysis for each sub-sample and we find that the relation holds universally. Second, we divide our sample based on Tobin s Q and again we find that the relation persists in all sub-samples. These analyses suggest that the relation is not limited to firms of a specific size or with different growth expectations. Third, we sub-sample based on 8 We demonstrate this formally in a model found in Appendix 1. 4

7 time period and divide our 50-year sample into 5 sub-samples that correspond to sequential 10-year periods. We do not find a significant relation in the earliest sub-sample from 1965 to 1975, but we do find a significant relation in all subsequent time periods. This is consistent with the notion that CAPM was not likely to be widely used by firms during the period from 1965 to 1975 as it had only recently been developed. Finally, we construct alternative measures for CE Imprecision: 1) an indicator variable for betas in the top and bottom deciles, 2) deviations of estimated beta from 1, 3) deviations of estimated beta from the average beta in the industry, and 4) standard deviation of estimated betas at an industry level. We find similar results when we repeat our main analysis using these alternative measures. Overall, the results in this paper show that imprecise cost of equity estimates lead to suboptimal investment, and this has an adverse effect on shareholder wealth. This paper contributes to the literature in a number of ways. First, it provides further evidence of the imprecision of cost of equity estimates (Fama and French, 1997; Ferson and Locke, 1998; Pastor and Stambaugh, 1999). Second, it adds to the literature that links activity in financial markets to real investment (Baker et al., 2003; Chen et al., 2007; Bakke and Whited, 2010). Finally, it contributes to our understanding the determinants of investment efficiency (Jensen, 1986; Richardson, 2006, among others). The remainder of the paper is organized as follows. The next section describes the motivation and research expectations. Section 3 describes the sample selection and construction of the variables used in the paper. Section 4 outlines the research methodology and results. Section 5 describes additional untabulated analyses and Section 6 concludes. 5

8 2. Motivation and Research Expectations Fama and French (1997) explore the precision of industry level estimates for the cost of equity capital from return based models and conclude that the estimates are distressingly imprecise. The lack of precision in cost of equity is potentially problematic if firms make capital budgeting decisions using discounted cash flow analyses. Specifically, if a firm discounts expected cash flows using an estimated cost of equity that is too low (high), then the project will look more (less) attractive. This could potentially lead to accepting bad projects and failing to accept good projects. Survey evidence indicates that firms do use discounted cash flow analyses to make capital budgeting decisions. Graham and Harvey (2001) report that 74.9% of CFOs always or almost always use net present value and 75.7% always or almost always use internal rate of return. Both methods utilize the cost of equity as part of the decision criteria. Additionally, firms appear to rely on the CAPM to estimate their cost of equity. Graham and Harvey (2001) report that 73.5% of CFOs always or almost always use the CAPM to estimate their cost of equity. A newer survey by the Association for Financial Professionals (AFP, 2013) report that 87% of public firms us the CAPM to estimate their cost of equity. The lack of precision in market based models for the cost of equity is either not fully grasped by finance executives or they are able to use unobserved information to increase the precision of their estimates. Fama and French (1997) report that standard errors of more than 3% are typical for industry estimates for the cost of equity and firmspecific estimates are expected to be less precise. This contrast with a 2013 survey 6

9 conducted by the Association for Financial Professionals, which reports that 89% of publicly traded firms perceive their cost of capital estimates to be accurate within 100 bps. There is no clear definition of corporate investment efficiency. Ideally, executives identify and invest in projects that maximize shareholder wealth. A firm s investment opportunity set is not observable, but q theory indicates that investment should be a positive function of Tobin s Q (Modigliani and Miller, 1958; Tobin, 1969; Hayashi, 1982; Panageas, 2005). Following Goodman et al. (2014), we define investment efficiency as the deviation between a firm s actual capital investment and the amount expected as a function of the firm s valuation, cash flow, past growth, and prior investment. Larger deviations, which represent over or under investment, are consistent with less efficient investment. Our expectation is that firms deviate farther from their expected level of investment when their estimated cost of equity is less precise. We can t make a directional prediction regarding systematic under or over investment because we are not arguing a systematic bias in cost of equity estimates, but rather a lack of precision. Firms with a wide confidence interval for beta are more likely to be using a cost of equity estimate that is significantly below or above the true cost. The firms underestimating their cost of equity are likely to overinvest and the firms overestimating their cost of equity are likely to underinvest. Alternatively, if managers are able to precisely estimate their cost of equity estimates using additional information, then we would not expect that investment efficiency be related to our measure of the precision of cost of equity estimates. 7

10 Further, deviations from optimal investment should lead to weaker operating performance as firms either reject good projects or accept projects with poor expected returns. Fu (2010) shows that the underperformance following seasoned equity offerings is associated with overinvestment by these firms. We expect that firms with less precise cost of equity estimates are likely to experience lower operating performance because of under/over investment. 3. Sample Selection and Variables Construction Data Sources Our primary analyses require information regarding the capital investments made by firms and a measure of the precision of cost of equity estimates. To construct this sample we pull firm accounting data from Compustat North America and stock returns from CRSP. We begin our sample in 1965 because of data availability in Compustat and we conclude our sample with fiscal year The relatively light data requirements of this study allow us to analyze a large number of firms over a long period of time. Overall, our sample contains 147,542 firm year observations for 14,491 unique firms. The construction of the sample is detailed in Appendix Table 1. In order to estimate analyst implied cost of capital we obtain analyst forecast and stock price data from I/B/E/S from 1983 to Measuring CE Imprecision To measure the CE imprecision for each firm-year observation we estimate the following regression using monthly returns for the 36 months ending the month prior to the start of the fiscal year: Ret i,t = b 0 + b 1 (CRSP Value Weighted Market Return t )+e i,t 8

11 where Reti,t is the return of firm i in month t, CRSP Value Weighted Market Returnt is our estimate for the market return. β1 is the estimated beta for the firm-year and the standard error of β1 is our measure of CE Imprecision. We only include observations where there are at least 12 monthly returns available to perform the regression analysis. Measuring Investment Efficiency Biddle et al. (2009) and Bae et al. (2017) define efficient investment as a firm undertaking all projects with positive net present values under no market frictions. Under- and over-investment is harmful to shareholders as managers either take on negative NPV projects or fail to invest in positive NPV projects, and thus both under- and over-investment are viewed as inefficient (Bae et al., 2017). We can t directly observe the investment opportunity set, thus we rely on expected investment to represent the efficient investment level. We rely on existing research to guide the model of expected investment. Standard neoclassical q theory indicates that investment should be a positive function of Tobin s Q as firms invest more when prices are high (Modigliani and Miller, 1958; Tobin, 1969; Hayashi, 1982; Panageas, 2005). We include cash flow to control for variation in internal financing capability and asset growth as growth firms may invest at higher levels. Additionally, we include lagged investment, which helps control for a firm-specific component of investment (Goodman et al., 2014). Specifically, we estimate the following model for each combination of year and industry with at least 30 observations to determine the expected level of investment for each firm: CAPEX i,t Avg. Assets i,t = b 0 + b 1 TobinsQ i,t-1 + b 2 CFO i,t Avg. Assets i,t + b 3 AssetGrowth t-1 + b 4 CAPEX i,t-1 Avg. Assets i,t-1 +e i,t 9

12 where CAPEXi,t is the capital expenditure for firm i in year t, TobinsQi,t-1 is the beginning of year value of Tobin s Q for firm i. CFOi,t is the cash flow from operations for firm i in year t. AssetGrowthi,t-1 is the growth rate of assets from year t-2 to year t-1 for firm i. Complete variable definitions are found in the appendix. We rely on the residual from this model to measure deviations from expected investment with negative values indicating under-investment and positive values representing over-investment. The primary dependent variable in our analyses is Abnormal Investmenti,t, which is the absolute value of the residual investment from the estimated model. Higher values this variable represent larger deviations from expected investment. We use the absolute value because imprecise beta estimates should lead to overinvestment when the estimated beta is below the true beta and underinvestment when the estimated beta is above the true beta. 4. Empirical Results CE Imprecision and Implied Analyst Cost of Equity We use the dispersion of analyst-implied cost of equity in order to determine if CE Imprecision has an impact on the difficulty of estimation of a firm s cost of equity. We use four different models of analyst-implied cost of equity from Gebhardt, Lee and Swaminathan (2001), Claus and Thomas (2001), Gode and Mohanram (2003) and Easton (2004). The details for each estimation method are described in the Appendix Table A3. We estimate the models individually for each analyst so that we can measure the variation between analysts at a given point in time. Specifically, we take the standard deviation of implied cost of capital across analyst for each forecast period. Table 2 presents the results of regressions of the dispersion of analyst implied cost of equity 10

13 forecasts on CE Imprecision and other firm characteristics. We find that firms with greater CE Imprecision have significantly greater dispersion in analyst-implied cost of equity. This provides evidences that firms with greater imprecision in the estimates of their betas have noisier estimates of their cost of equity. CE Imprecision and Abnormal Investment Our primary analyses investigate the link between the imprecision of a firm s cost of equity and its deviation from its expected level of capital investment. To conduct these analyses, we rely on pooled OLS regressions and our dependent variable is Abnormal Investment. Higher values of this variable indicate a larger deviation from the predicted level of investment and are likely to indicate a level of corporate investment that does not maximize shareholder wealth. Our variable of interest is CE Imprecision, which is constructed such that higher values indicate a cost of equity estimate that is less precise. We control firm size (Ln(Assets)) as larger firms may have higher ability managers and more sophisticated systems that allows them to invest more efficiently. We control for growth expectations (Tobin s Q and Asset Growth) to allow for differences in investment efficiency between firms that are investing in new projects versus firms that are likely to be maintaining and replacing existing assets. We included cash flows as firms generally rely on internally generated capital to fund capital investments. Finally, we include estimated beta to control for differences in systematic risk at each firm. Table 3 presents coefficient estimates from our primary analyses. In column 1, we demonstrate a simple univariate relation between CE Imprecision and Abnormal Investment. The coefficient estimate is (t-stat=19.35), which indicates that firms with more imprecise betas tend to have larger deviations from their expected level of 11

14 capital investment. We show that this relationship continues to hold in the presence of control variables along with indicator variables for industry (2-digit SIC) and year. Finally, in the last specification, we include firm and year fixed effects and the coefficient on CE Imprecision remains positive and significant. This specification provides evidence that a unobservable, time invariant firm characteristics are not driving the observed relation between CE Imprecision and Abnormal Investment. The link between CE Imprecision and Abnormal Investment is also economically meaningful. The average firm in our sample has capital expenditures that average about 7 percent of assets and abnormal investment averages about 3 percent of assets. Our coefficient estimate from the univariate regression indicates that a 1 standard deviation increase in CE Imprecision would increase abnormal investment by 0.54 percent of assets ( x 0.474). This represents an increase of 18% of the sample mean of abnormal investment. We note that the firms with the least precise estimates of cost of equity are dramatically different from those firms with the most precise cost of equity estimates. Specifically, they are much smaller firms with much higher values for Tobin s Q. Although we control for both size and Tobin s Q in our regression analysis, it is a reasonable to question if our primary result holds for different sub-samples of firms. We first address this by running our analyses on firms sub-sampled based on size. In each year, we sort firms into quintiles based on size and then run our pooled OLS regression within each of these quintiles. Coefficient estimates from these regressions are presented in Panel A of Table 4. We find that the coefficient on CE Imprecision remains positive and significant in each sub-sample. Panel B of Table 4 presents coefficient estimates 12

15 where firms are sorted based on Tobin s Q and again, we continue to find a positive and significant coefficient on CE Imprecision. Finally we look at different periods of our sample to determine if the relation between CE Imprecision and Abnormal Investment is consistent over time. Specifically, we create subsamples using ten sequential years of data. Panel C of Table 4 presents the coefficient estimates generated by each sub-sample and the estimated coefficient on CE Imprecision is consistently positive and significant with the exception of the subsample from 1965 to The lack of a significant coefficient estimate during this early sample period is consistent with the notion that CAPM was not widely used by companies because it had just recently been developed. We argue that an imprecise cost of equity estimate should lead firms to deviate from their expected level of investment through both under- and over-investment. As such, we don t expect to a consistent positive or negative deviation from expected investment for firms with an imprecise cost of capital estimate. We provide evidence supporting this notion in Table 5. In this table our dependent variable is Residual Investment, which is the signed residual from our model of investment. 9 Negative values of Residual Investment indicate under-investment and positive values indicate overinvestment. In the first specification we use the full sample of observations and do not find a significant coefficient estimate on CE Imprecision. This indicates that a noisy cost of equity estimate does not lead exclusively to over- or under-investment. In the second specification, we limit observations to those with negative values of Residual Investment (i.e. firms that are under investing). We observe a negative and significant coefficient on CE Imprecision, which indicates that among firms that are under-investing, those with a 9 We take the absolute value of Residual Investment to calculate Abnormal Investment. 13

16 less precise cost of equity estimate exhibit a greater degree of under-investment. In the final specification we only look at firms that are over-investing and find a similar pattern. The over-investing firms with a noisy cost of equity estimate over-invest to a greater degree than firms with a more precise cost of equity estimate. Instrumental Variable Analysis Establishing causality in empirical studies is a difficult task. In this study, we implement a 2-stage least squares analysis using an instrumental variable to help establish a causal relationship between CE Imprecision and Abnormal Investment. The instrument that we employ is the distance between a firm s headquarter location and New York City, NY. The usefulness of our instrument hinges on two criteria: 1) relevance and 2) excludability. The existing literature looking the geographic locations of firms has documented patterns where the cost of information acquisition for investors increases as the distance between the investor and the firm increases. Isolated firms are likely to experience stock prices that are less informative because they are farther from investors. Consistent with existing literature, we find that firm s farther from NYC have significantly higher CE Imprecision, which establishes the relevance criteria. The exclusion criterion is more difficult to establish, because there is no clear-cut empirical test to establish exclusion. To meet the exclusion criterion, the instrument can t capture an omitted economic factor that is directly correlated with a firm s investment efficiency. There are a multitude of factors that can impact the ability of managers to efficiently allocate capital. Goodman et al. (2014) indicates that high-ability managers are able to more efficiently allocate capital because they produce better quality cash flow forecasts, but it seems unlikely that manager quality is directly related to distance to investors. 14

17 Access to capital could potentially impact investment efficiency and is likely a function of distance to investors, which is potentially problematic. We recognize this concern, but most firms finance investment through internally generated cash flows, which means that differential access to outside capital is not likely to impact their investment efficiency. The coefficient estimates from our instrumental variable analysis are presented in Table 6. The first stage regression coefficients are found in column 1 and we find a positive and significant coefficient on Distance to NYC, which is consistent with the notion that firm prices are less informative when information acquisition is more costly. The second stage coefficients are presented in column 2. The coefficient estimate on CE Imprecision (IV) is positive and significant, which provides evidence of a causal relation between CE Imprecision and Abnormal Investment. Overall, the instrumental variable approach supports a causal interpretation of the positive relationship between CE Imprecision and Abnormal Investment. CE Imprecision and Firm Operating Performance We provide further implications by analyzing firm operating performance. We have shown evidence that CE Imprecision is associated with deviations from predicted levels of investments, but we have not established that these deviations are sub-optimal from an investor perspective. To measure firm performance, we utilize Return on Assets and Asset Turnover and we analyze the relationship between CE Imprecision and Return on Assets using pooled OLS regressions. The coefficient estimates from these analyses are presented in Table 7. The first specification utilizes industry and year fixed effects and the second specifications uses firm and year fixed effects. Across both specifications we find a negative and significant 15

18 coefficient estimate on CE Imprecision, which indicates that firm performance suffers when the appropriate cost of equity capital is hard to determine. The third and forth specifications measure firm performance with Asset Turnover and we continue to find a negative and significant coefficient on CE Imprecision. Overall, these results are consistent with Goodman et al. (2014), who document that less efficient investment is correlated with lower future operating performance. Alternative Measures of Imprecise CE Estimates Fama and French (1997) indicate that extreme estimates of beta are likely to have more error. We construct several alternative measures to capture the higher error associated with extreme betas. First, we create an indicator variable (Extreme beta) set to one for observations where the estimated beta is in the top or bottom decile of firms in the sample during the year. Second, we measure the deviation of the estimated beta from 1. Larger deviations from 1 should be associated with more error in the estimation of beta. Additionally, we measure the deviation of the estimated beta from the average estimated beta of firms in the same industry that year. Finally, we use a measure created at the industry-year level. We take the standard deviation of the estimated betas of firms in the industry during the year. We use these alternative measures in place of CE Imprecision and the coefficients from these alternative measures are reported in Table 7. We find positive and significant coefficient estimates using all 4 alternative measures of cost of equity imprecision. 5. Untabulated Analyses Manager Quality 16

19 Goodman et al. (2014) show that the accuracy of manager provided earnings guidance predicts less abnormal investment. The intuition underlying this result is that managers that produce more accurate earnings guidance are also likely to produce more accurate projections when making capital expenditure decisions. A potential critique of our analysis is that low quality managers make poor investment decisions through bad projections and result in noisy stock price (thereby producing imprecision when estimating beta). Following Goodman et al. (2014), we include the average managerial forecast error from the prior three fiscal years and find that the coefficient on CE Imprecision remains positive and significant when analyzing Abnormal Investment. This indicates that manager forecast quality is not driving our results. CEO Incentives Incentives from managerial stock and option holdings have been shown in influence corporate investment choices (Coles, Daniel, and Naveen, 2006). To account for the possibility that CEO incentives are driving both CE Imprecision and Abnormal Investment, we include CEO Delta and CEO Vega in our analysis of Abnormal Investment. CEO Delta measure the change in the value of a CEO s stock and option holdings for a change in firm value and CEO Vega measures the sensitivity of a CEOs wealth to stock return volatility. In untabulated results, we continue to find a positive and significant coefficient on CE Imprecision when controlling for CEO incentives. Private Information in Stock Prices 17

20 Stock prices reflect public information and private information and the prior literature has shown the amount of private information in stock prices influences the investment decisions of corporate managers (Chen et al., 2007; Bakke and Whited, 2010). Managers can learn from the private information that is incorporated in their stock price to make better investment decisions. CE Imprecision is positively correlated with the amount of private information in stock prices and thus its possible that this correlation is driving our results. Following Chen et al. (2007) we measure the amount of private information in stock prices for a firm-year as One minus R 2, where the R 2 is from a firmyear regression of daily returns on the market returns and industry returns. In untabulated results, we continue to find a positive and significant coefficient on CE Imprecision when controlling for the amount of private information incorporated into firms stock prices. Financial Reporting Quality Prior research has shown that firms financial reporting quality has an impact on Abnormal Investment as higher quality financial reporting reduces the information asymmetry between firms and outside investors (Biddle and Hillary, 2006; Biddle et al. 2009). This reduction in information asymmetry potentially makes it easier for capital constrained firms to raise outside capital and it reduces managerial ability to make value destroying investments in firms with sufficient internally generated capital. Beta estimates for firms with poor financial quality are likely to be less precise as more mispricing is likely to occur. In untabulated robustness tests we control for financial reporting quality using the Dechow and Dichev (2002) accruals quality measure (as implemented in Biddle et al. 2009). We continue to find a positive and significant 18

21 coefficient estimate on CE Imprecision when controlling for financial reporting quality, indicating that CE Imprecision is not simply proxying for the financial reporting quality of the firm. 6. Conclusion Fama and French (1997) concluded that estimates of the cost of equity are imprecise, but they didn t explore how this imprecision impacted firms. We argue that a difficult to estimate cost of equity increases the likelihood that a firm is evaluating investment opportunities using a discount rate that is too high or too low. We find evidence supporting with this argument as the imprecision associated with estimating the cost of equity is positively correlated with deviations from expected investment. Consistent with the premise that these deviations from expected investment are sub-optimal, we also find that firms with imprecise cost of equity estimates have lower operating performance. Overall, our results suggest another channel through which financial markets impact real economic activity. 19

22 References Association for Financial Professionals AFP Estimating and Applying Cost of Capital Report of Survey Results. Bethesda, MD (2013). Bae, Gil Soo, Seung Uk Choi, Dan S. Dhaliwal, and Phillip T. Lamoreaux. Auditors and Client Investment Efficiency. The Accounting Review 92.2 (2017): Baker, Malcolm, Jeremy C. Stein, and Jeffrey Wurgler. "When does the market matter? Stock prices and the investment of equity-dependent firms." The Quarterly Journal of Economics (2003): Bakke, Tor-Erik, and Toni Whited. Which Firms Follow the Market? An Analysis of Corporate Investment Decisions. Review of Financial Studies 23.5 (2010): Biddle, Gary C., and Gilles Hilary. "Accounting quality and firm-level capital investment." The Accounting Review 81.5 (2006): Biddle, Gary C., Gilles Hilary, and Rodrigo S. Verdi. "How does financial reporting quality relate to investment efficiency?" Journal of Accounting and Economics 48.2 (2009): Chen, Qi, Itay Goldstein, and Wei Jiang. "Price informativeness and investment sensitivity to stock price." Review of Financial Studies 20.3 (2007): Cheng, Mei, Dan Dhaliwal, and Yuan Zhang. "Does investment efficiency improve after the disclosure of material weaknesses in internal control over financial reporting?" Journal of Accounting and Economics 56.1 (2013): Coles, Jeffrey L., Naveen D. Daniel, and Lalitha Naveen. "Managerial incentives and risk-taking." Journal of Financial Economics 79.2 (2006): Claus, J., and J. Thomas, 2001, Equity Premia as Low as Three Percent? Evidence from Analysts' Earnings Forecasts for Domestic and International Stocks, Journal of Finance, 56(5), Dechow, Patricia M., and Ilia D. Dichev. "The quality of accruals and earnings: The role of accrual estimation errors." The Accounting Review 77.1 (2002): Easton, P. D., 2004, PE Ratios, PEG Ratios, and Estimating the Implied Expected Rate of Return on Equity Capital, Accounting Review, 79(1), Fama, Eugene, and Kenneth French. Industry costs of equity. Journal of Financial Economics 43 (1997):

23 Ferson, Wayne E., and Dennis H. Locke. "Estimating the cost of capital through time: an analysis of the sources of error." Management Science 44.4 (1998): Fu, Fangjian. "Overinvestment and the operating performance of SEO firms." Financial Management 39.1 (2010): Gebhardt, W. R., C. M. C. Lee, and B. Swaminathan, 2001, Towards an Implied Cost of Capital, Journal of Accounting Research, 39(1), Gode, D., and P. Mohanram, 2003, Inferring the Cost of Capital Using the Ohlson Juettner Model, Review of Accounting Studies, 8(4), Goodman, Theodore H., et al. "Management forecast quality and capital investment decisions." The Accounting Review 89.1 (2013): Graham, John R., and Campbell R. Harvey. "The theory and practice of corporate finance: Evidence from the field." Journal of Financial Economics 60.2 (2001): Hayashi, Fumio. "Tobin's marginal q and average q: A neoclassical interpretation." Econometrica: Journal of the Econometric Society (1982): Jensen, Michael C. "Agency costs of free cash flow, corporate finance, and takeovers." The American Economic Review 76.2 (1986): Krüger, Philipp, Augustin Landier, and David Thesmar. "The WACC fallacy: The real effects of using a unique discount rate." Journal of Finance (2011). McNichols, Maureen F., and Stephen R. Stubben. "Does earnings management affect firms' investment decisions?." The Accounting Review 83.6 (2008): Modigliani, Franco, and Merton H. Miller. "The cost of capital, corporation finance and the theory of investment." American Economic Review 48.3 (1958): Pástor, Ľuboš, and Robert F. Stambaugh. "Costs of equity capital and model mispricing." Journal of Finance 54.1 (1999): Panageas, S. The Neoclassical q Theory of Investment in Speculative Markets. Mimeo, The Wharton School, University of Pennsylvania (2005) Richardson, Scott. "Over-investment of free cash flow." Review of Accounting Studies (2006): Tobin, James. "A general equilibrium approach to monetary theory." Journal of Money, Credit and Banking 1.1 (1969):

24 Table 1 Sample Summary Statistics Panel A presents summary statistics of our overall sample of firms. Panel B presents sample means for subsamples of the overall sample based on CEO Imprecision. Panel A: Summary Statistics Variable Mean Std. Dev. P25 Median P75 N Investment ,542 Residual Investment (0.00) 0.05 (0.02) (0.00) ,542 Abnormal Investment ,542 Assets 3,239 38, ,542 MVE 1,794 12, ,542 Tobin's Q ,366 Cash Flow / Average Assets ,542 Return on Assets ,542 Asset Growth Rate (0.02) ,542 Price (FYE) ,542 Beta ,542 CE Imprecision ,542 Idiosyncratic Volatility ,542 R ,542 Panel B: Sub-samples selected by CE Imprecision Quartile 1 (Lowest Quartile 5 (Highest CE Imprecision) CE Imprecision) Quartile 2 Quartile 3 Quartile 4 Investment Residual Investment (0.00) (0.00) (0.00) Abnormal Investment Assets 9,155 4,154 1, MVE 5,030 2,137 1, Tobin's Q Cash Flow / Average Assets (0.09) Return on Assets (0.13) Asset Growth Rate Price (FYE) Beta CE Imprecision Idiosyncratic Volatility R

25 Table 2 Analysts Cost of Equity Dispersion This table presents coefficient estimates from linear regressions of Analysts CE Dispersion on CE Imprecision. GLS, CT, GM, and Easton, and Price Target are five different methods of measuring the implied cost of capital following from the literature. Analysts CE Dispersion is the standard deviation of the implied cost of equity measured across analysts at the end of the year. CE Imprecision is the standard error of the estimated beta coefficient of the firm from a regression of the firm s monthly returns on an index return using the 36 monthly observations prior to the beginning of the fiscal year. All other variables are defined in the appendix. T-stats based on standard errors clustered at the firm level are included in parentheses. Analysts CE Dispersion VARIABLES GLS CT GM Easton Price Target CE Imprecision t *** *** *** *** *** ** *** *** *** *** (9.52) (3.53) (10.43) (3.45) (9.24) (2.10) (9.52) (2.88) (9.49) (2.80) Ln(Assets t-1) * *** ** *** *** *** *** *** *** (-1.81) (4.91) (-2.52) (6.20) (-3.01) (8.44) (-0.72) (6.02) (-7.73) (7.43) Tobins Q t ** *** *** *** *** ** *** *** ** ** (-2.19) (2.89) (-5.05) (2.88) (-4.57) (2.48) (-10.58) (-3.40) (-2.50) (2.18) Cash Flow t / Average Assets t *** *** *** *** *** *** *** *** *** *** (-11.19) (-11.93) (-14.05) (-15.11) (-7.96) (-8.91) (-6.44) (-4.00) (-22.19) (-10.88) Asset Growth t *** *** *** *** * *** ** *** (-4.96) (-4.04) (-4.98) (-3.53) (-1.71) (0.09) (-3.83) (-2.49) (2.89) (1.13) Investment t *** *** *** *** *** *** ** *** *** (5.03) (5.32) (6.01) (4.83) (3.63) (2.83) (2.07) (-0.07) (7.75) (3.98) Estimated Beta t *** *** *** (1.18) (-0.67) (1.14) (-1.37) (1.51) (-0.63) (3.23) (-0.88) (11.81) (5.19) Cash Flow Volatility t-4 to t *** *** *** *** *** *** *** *** ** (6.33) (4.30) (6.09) (4.20) (3.18) (1.41) (4.26) (4.85) (3.41) (2.30) Industry FE Yes No Yes No Yes No Yes No Yes No Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE No Yes No Yes No Yes No Yes No Yes Clustered Standard Errors Firm Firm Firm Firm Firm Firm Firm Firm Firm Firm Observations 19,794 19,794 19,794 19,794 15,182 15,182 10,865 10,865 28,573 28,573 R-squared

26 Table 3 Abnormal Investment This table presents coefficient estimates from linear regressions of Abnormal Investment on CE Imprecision. Abnormal Investment is the absolute value of the residual from industry-year regressions of investment on Tobin s Q, cash flows, asset growth, and lagged investment. Larger values variable of this indicate larger deviations from the predicated level of investment. CE Imprecision is the standard error of the estimated beta coefficient of the firm from a regression of the firm s monthly returns on an index return using the 36 monthly observations prior to the beginning of the fiscal year. All other variables are defined in the appendix. T-stats based on standard errors clustered at the firm level are included in parentheses. VARIABLES Abnormal Investment CE Imprecision t *** *** *** *** ** (19.35) (10.78) (8.69) (8.27) (2.09) Ln(Assets t-1) *** *** *** *** (-43.34) (-34.77) (-34.04) (-17.77) Tobin s Q t *** *** *** *** (3.34) (3.50) (3.51) (2.71) Cash Flow t / Average Assets t *** *** *** ** (3.72) (3.21) (3.17) (2.42) Asset Growth t ** (-2.22) (-1.47) (-1.47) (-0.64) Investment t *** *** *** *** (63.94) (42.57) (42.61) (15.86) Estimated Beta t (-1.25) (-0.41) Industry FE No No Yes Yes No Year FE No No Yes Yes Yes Firm FE No No No No Yes Clustered Standard Errors Firm Firm Firm Firm Firm Observations 147, , , , ,542 R-squared

27 Table 4 Subsample Analysis This table presents coefficient estimates from linear regressions of Abnormal Investment on CE Imprecision. Abnormal Investment is the absolute value of the residual from industry-year regressions of investment on Tobin s Q, cash flows, asset growth, and lagged investment. Larger values variable of this indicate larger deviations from the predicated level of investment. CE Imprecision is the standard error of the estimated beta coefficient of the firm from a regression of the firm s monthly returns on an index return using the 36 monthly observations prior to the beginning of the fiscal year. All other variables are defined in the appendix. In Panel A, we sub-sample based on firm size; in Panel B, we sub-sample based on lagged Tobin s Q; and in Panel C, we analyze sub-samples based on time period. T-stats based on standard errors clustered at the firm level are included in parentheses. Panel A: Sub-sample analysis Firm Size VARIABLES Abnormal Investment Sample Smallest Firms Quintile 2 Quintile 3 Quintile 4 Largest Firms CE Imprecision t *** *** *** *** *** (2.75) (3.90) (4.16) (4.24) (6.26) Ln(Assets t-1) *** *** *** *** *** (-7.04) (-4.52) (-3.68) (-5.20) (-8.93) Tobin s Q t * *** *** *** *** (1.78) (3.95) (5.17) (4.41) (3.23) Cash Flow t / Average Assets t * * (1.78) (1.75) (-0.06) (-0.20) (-1.53) Asset Growth t ** (-1.23) (-0.52) (-2.27) (0.27) (-0.79) Investment t *** *** *** *** *** (21.94) (17.88) (21.13) (21.57) (20.17) Estimated Beta t (-0.39) (-1.34) (-1.51) (0.67) (1.19) Industry and Year FE Yes Yes Yes Yes Yes Clustered Standard Errors Firm Firm Firm Firm Firm Observations 29,528 29,507 29,513 29,507 29,487 R-squared

28 Panel B: Sub-sample analysis Tobin s Q VARIABLES Abnormal Investment Sample Lowest Q Quintile 2 Quintile 3 Quintile 4 Highest Q CE Imprecision t *** *** * *** *** (4.85) (4.73) (1.96) (3.40) (3.53) Ln(Assets t-1) *** *** *** *** *** (-15.27) (-18.68) (-16.35) (-17.40) (-18.52) Tobin s Q t (-0.15) (0.51) (0.49) (1.39) (1.00) Cash Flow t / Average Assets t * *** (-1.15) (-0.01) (-0.29) (1.85) (3.66) Asset Growth t *** * (-2.63) (-1.66) (-0.48) (-1.25) (-1.07) Investment t *** *** *** *** *** (19.31) (18.98) (20.34) (21.93) (17.37) Estimated Beta t ** * (-2.40) (-1.49) (-1.12) (0.16) (-1.74) Industry and Year FE Yes Yes Yes Yes Yes Clustered Standard Errors Firm Firm Firm Firm Firm Observations 29,528 29,507 29,513 29,507 29,487 R-squared Panel C: Sub-sample analysis Time Period VARIABLES Abnormal Investment Sample 1965 to to to to to 2015 CE Imprecision t *** *** *** *** (0.34) (5.40) (4.67) (4.50) (4.43) Ln(Assets t-1) *** *** *** *** *** (-15.70) (-16.88) (-22.02) (-16.91) (-15.26) Tobin s Q t *** *** ** (1.29) (3.44) (1.03) (5.12) (2.02) Cash Flow t / Avg. Assets t * *** (1.82) (4.39) (0.36) (-0.38) (-0.38) Asset Growth t *** (-2.74) (-0.34) (-1.04) (-0.06) (-1.22) Investment t *** *** *** *** *** (16.13) (20.30) (23.71) (19.81) (18.04) Estimated Beta t ** (1.07) (-0.84) (1.08) (-2.08) (-0.06) Industry and Year FE Yes Yes Yes Yes Yes Clustered Standard Errors Firm Firm Firm Firm Firm Observations 13,087 28,499 36,594 40,576 28,786 R-squared

29 Table 5 Residual Investment This table presents coefficient estimates from linear regressions of Residual Investment on CE Imprecision. Residual Investment is the residual from industry-year regressions of investment on Tobin s Q, cash flows, asset growth, and lagged investment. Larger values variable of this indicate larger deviations from the predicated level of investment. CE Imprecision is the standard error of the estimated beta coefficient of the firm from a regression of the firm s monthly returns on an index return using the 36 monthly observations prior to the beginning of the fiscal year. All other variables are defined in the appendix. T-stats based on standard errors clustered at the firm level are included in parentheses. VARIABLES Residual Investment Full Residual Invt. < 0 Residual Invt. >= 0 CE Imprecision t *** *** (-1.24) (-8.95) (5.17) Ln(Assets t-1) *** *** *** (-4.10) (27.70) (-28.07) Tobins Q t * *** (1.84) (-5.99) (1.47) Cash Flow t / Average Assets t ** ** *** (2.31) (-2.38) (3.36) Asset Growth t ** (-0.31) (-0.60) (-2.12) Investment t *** *** (1.06) (-40.63) (21.56) Estimated Beta t (-0.35) (0.18) (-1.59) Industry and Year FE Yes Yes Yes Clustered Standard Errors Firm Firm Firm Observations 147, , ,542 R-squared