WPS8534 Policy Research Working Paper 8534 Who are America’s Star Firms? Meghana Ayyagari Asli Demirguc-Kunt Vojislav Maksimovic Development Economics Development Research Group July 2018 Policy Research Working Paper 8534 Abstract There is wide spread concern about a growing gap between over time. While pricing power, as measured by markups, top-performing publicly listed firms and the rest of the predicts star firm status, a large fraction of star firms have economy and the implications of this for rising inequality low markups and there is no evidence that star firms are in the U.S. Using conventional return calculations, there cutting output or investment more than other firms for is indeed a widening gap between star firms (defined as the same markup. The effect of star status is persistent. those in top 10 percent of return on invested capital in any Five years later, star firms have higher growth, profits, and year) and the rest of the economy over time, especially in Tobin’s Q. A small subset of exceptional firms may pose industries that rely on a skilled labor force. However, once more pressing policy concerns with much higher returns measurement error in intangible capital is accounted for, and the potential to exercise market power in the future. this gap shrinks dramatically and has not been widening This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at ayyagari@gwu.edu, Ademirguckunt@worldbank.org, and vmaksimovic@rhsmith.umd.edu. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Who are America’s Star Firms? Meghana Ayyagari ∗, Asli Demirguc-Kunt † and Vojislav Maksimovic ‡ JEL Classification : E22, L1 ∗ School of Business, George Washington University, Ph: 202-994-1292; Email: ayyagari@gwu.edu † The World Bank, Ph: 202-473-7479; Email: Ademirguckunt@worldbank.org ‡ Robert H. Smith School of Business at the University of Maryland, Ph: 301-405-2125; Email: vmaksi- movic@rhsmith.umd.edu The authors thank Paulo Bastos, Miriam Bruhn, Mike Faulkender, Francisco Ferreira, David McKenzie, and Bob Rijkers for helpful comments and suggestions, and Elliot Oh for excellent research assistance. This paper’s findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. Introduction A great deal of attention has been paid to two trends: (1) the emergence of star firms that have pulled away from the rest of the economy (Furman and Orszag [2015], Koller, Goedhart, and Wessels [2017], Autor, Dorn, Katz, Patterson, and Van Reenen [2017]) and (2) introduction of new technologies together with a fundamental structural change towards a more intangible intensive economy (Corrado and Hulten [2010]) with corresponding implications for corporate investment and the overall economy.1 However we have little systematic evidence on the characteristics of the star firms. Questions of interest include: In which industries star firms occur, how long they retain their star status, and if these two trends are related - that is, is the rise of star firms related to the increased dependence on intangible capital? Determining whether the performance gap between star firms and other firms is the result of luck, market imperfections, or a reflection of successful idiosyncratic firm growth strategies2 is a public policy priority that will shape policies that promote or regulate high-value firms. The dominant concern is that these firms are gaining their distinction through market power, and relatedly, research has shown an apparent marked increase in concentration in U.S. industries in errez and Philippon [2017] and in markups and market power (De Loecker the last two decades (Guti´ and Eeckhout [2017]; Barkai [2016]).3 Autor et al. [2017] link concentrated winner-take-all markets with the fall in the labor share in the US economy as an explanation for the rise of star firms. Baker and Salop [2015] directly link a decline in the enforcement of anti-trust statutes to increases in errez concentration and rising inequality in the U.S.. Grullon, Larkin, and Michaely [2017] and Guti´ and Philippon [2017] argue that increased market power may be responsible for a combination of high profits and low investment. There is also concern that star firms may be creating systemic problems and disruptions through the economy. For instance, Van Reenen and Patterson [2017] suggest that the rise of star firms will lead to a fall in economic dynamism and productivity with declining pay and job opportunities for the average worker. 1 Several papers have explored the implications of the rise in intangible assets and knowledge capital on corporate investment (e.g. Peters and Taylor [2017]) and other macroeconomic impacts (e.g. Atkeson and Kehoe [2005], McGrattan and Prescott [2010], Eisfeldt and Papanikolaou [2014] and Perez-Orive, Caggese, et al. [2017]). 2 Successful idiosyncratic growth strategies may also be due to successful innovation or superior management practices (e.g. Bloom and Van Reenen [2007]). 3 For a strong dissenting view arguing that the observed changes in concentration computed at the national level are economically immaterial, see Shapiro [2018]. 1 In this paper, use a dataset of publicly listed firms in the United States from the Compustat database to identify star firms and the industries in which these firms are more likely to appear in. We examine the role of intangible capital, competition (Herfindahl-Hirschman industry concentra- tion index), and market power in influencing star firm status. We use three measures of market power - a firm-level measure of operating markups, firm-level market share, and a raw measure of firm size.4 Finally, we examine the persistence of the star firms’ performance and whether they differ in their investment and output per unit of capital compared to other firms in the economy. Following Furman and Orszag [2015], we define star firms as firms in the top 10% of Return on Invested Capital (ROIC) in the US in a particular year.5 Using conventional ROIC measures, in Figure 1, we find that there has indeed been a run-up in the ROIC of the top decile of large, non-financial sector, publicly-listed U.S. firms.6 Over the period 1965-2015, the ratio of the 90th percentile ROIC firm to the median ROIC firm has increased by over 69%. Importantly, we find that the star firms whose returns are diverging from the rest of the firms are in industries that require high cognitive skills and that in these industries average returns are higher. In industries where the tasks involve routine manual skills and which score low on non-routine cognitive and complex problem solving skills, we see lower returns and don’t see the star firms pulling away from the rest. However, conventional return metrics do not capitalize research and development, brand capital, or other forms of organizational capital. The consequences of not measuring intangible capital are far-reaching because they affect measures of firms’ earnings, identification of variable costs, capital investment and estimates of pricing power, outcomes which are subject to controversy. While De Loecker and Eeckhout [2017] show that there is a dramatic rise in firm market power in the 4 Our implementation of operating markups, which follow the work of De Loecker and Eeckhout [2017] and Traina [2018] is discussed below. 5 ROIC is an important profitability metric in corporate finance measuring how efficiently a company can allocate its capital to profitable investment and has been widely used in the literature (e.g. Ben-David, Graham, and Harvey [2013]) and by practitioners (e.g. Koller [1994], Koller et al. [2017]). For instance, David Benoit writing for the Wall Street Journal argued that General Motors placated activist investors with the help of higher return on invested capital (ROIC). See The Hottest Metric in Finance: ROIC, Wall Street Journal (2016). 6 We restrict our sample to large firms (defined as firms with assets more than $200 Million in 2009 dollars, adjusted for inflation) to replicate the equivalent figure in the previous studies. However, the evidence in Council of Economic Advisors [2016], Furman and Orszag [2015], and Koller et al. [2017] is based on a proprietary dataset of US firms from McKinsey & Co. whereas Figure 1 is based on publicly available Compustat data. If we were to use the full sample of Compustat firms without restricting to large firms, we get much higher increases in ROIC for the top decile of firms. 2 US using Cost of Goods Sold (COGS) as a measure of variable cost, Traina [2018] argues that once we include Selling, General, and Administrative Expenses (SGA) which are an increasingly vital share of variable costs for firms, there is no rise in markups. To address these issues, we adjust the conventional ROIC measures, measures of capital stock, as well as variable costs and measures of pricing power from De Loecker and Eeckhout [2017] and Traina [2018] to take into account investment in intangible capital. Once we re-compute the ROIC calculations to factor in estimates of intangible capital from the finance literature (see Peters and Taylor [2017] and the references therein), we find that both the run-up by top decile of firms and the much higher mean returns in the cognitively skilled industries disappear. Thus, the differences that we found earlier in firm differentiation between industries are likely attributable, in great part, to not accounting for intangible capital consistently. Industries that rely heavily on complex cognitive skills are likely to have higher amounts of intellectual and organizational capital, which is not measured by ROIC prepared according to generally accepted accounting principles. Next, we show that once we adjust the markups based on operating expenses for intangible capital, there is indeed a rise in markups over time. We also find that markups are positively related to high profits and greater probability of being a star, especially in industries that rely on low routine manual skills. In the overall sample, higher profits and restrictions of output are positively related to pricing power. However, ROIC stars have higher Output (sales/invested capital) Capex, and R&D investment compared to other firms. In addition, a large fraction of star firms have relatively low markups. We find no evidence that star firms are cutting output or any type of investment compared to other firms, even at high levels of markups. Our results are robust to a number of checks and alternate specifications. We find all our conclusions above to hold even when we tighten the requirement for star status down to the top 100 or 150 firms (when ranked by ROIC) each year. There is no run-up over time of the top 100 or 150 firms once we correct for intangible capital. We do find that the effects of star status are persistent. Five year later, star firms have higher ROIC, sales growth, and Tobin’s Q suggesting that our results are not driven by firms that have randomly realized high returns in specific years. 3 We find similar results when we use an alternative definition of star status which categorizes star firms as those in the top decile of market value (Tobin’s Q), taking into account the adjustment for the value of intangible capital. We find that large firms and firms with high markups in industries with high complex problem and analytical skills are more likely to be stars using this alternative definition. To account for the fact that cash holdings at some of the technology companies are substantial, we use yet another definition of star status where we consider only non-cash working capital in our definition of ROIC.7 In addition, in sensitivity tests we also find that our results are robust to varying the fraction of intangible capital that is used to correct the ROIC measures. All our results hold with these alternative definitions. A policy implication of our analysis is that while high pricing power is associated with high profits and star firm status, we also see that star firms are not cutting output less than other firms with same markups, and that a significant proportion of star firms do not have high markups. There is also little evidence that extraordinary returns are being realized as a result of high industry concentration or high market share. To look at possible disruptive and system wide effects of star firms, we need to focus our search on a very small number of firms. The analysis of these firms is not straightforward, both because of their small numbers and their adoption of pricing policies that reduce current returns in expectation of higher subsequent returns. A very small number of firms are often cited in the press as disrupting conventional business models, Amazon, Facebook, Google, Apple, and Microsoft (AFGAM), and we do see that these firms (especially Apple) have supernormal returns to capital. However, their markups are not necessarily much larger than those of the 90th percentile firm over the sample period. As we discuss below, these firms may have more market power than is even evidenced by their markups. In particular, they may be following strategies that emphasize holding markups and profits below their short run optimal values and growing quickly as a means of dominating their industries in the long run. Such strategies pose complex public policy challenges. 7 It is not clear how we should treat firms’ holdings of cash and near-cash securities. At one extreme, they are required precautionary balances, part of the firm’s invested capital. At the other extreme, they mostly consist of excess cash retained by the firm’s managers and should not be used in evaluating the economic value of the firm’s business. 4 Our paper contributes to the recent literature on star firms. Several researchers have used alternative definitions of star firms, and have focused on the consequences of the rise in industry concentration. Autor et al. [2017] use a definition of star firms based on productivity and market share and argue that star firms contribute to inequality within the US because they are more profitable and have lower wage to sales ratios, despite paying higher unit wages. Hall [2018] studies mega firms, defined as firms with more than 10,000 workers and finds no evidence that industries with high proportion of these firms have high markups but does find that markups increase in sectors with rising share of mega firms.8 In contrast to these papers, we use a market based measure of returns on invested capital to characterize star firms, which is also the definition that studies highlighting the emergence of star firms have focused on such as Furman and Orszag [2015], Koller et al. [2017], Council of Economic Advisors [2016]. Our paper is also related to the growing literature on the rise in concentration and markups. Grullon et al. [2017] find that firms in industries that are more concentrated enjoy higher profit margins and positive abnormal stock returns. They proxy price-cost margins by the Lerner Index (operating income after depreciation scaled by total sales) and returns by Return on Assets (ROA) and make no adjustments for intangible capital in their calculations. The focus in their paper is also on increase in industry concentration over time rather than on identifying characteristics of star firms. Barkai [2016] shows that the profit share of the US non-financial corporate sector has increased drastically over the past three decades, enough to offset the decline in both labor and capital shares. Kurz [2017] argues that modern developments in information technology have created higher barriers to entry leading to a rise in market concentration and increasing monopoly errez and Philippon [2017] link the decline in competition to the decrease power of firms. Guti´ in corporate investment. Alexander and Eberly [2018] and Crouzet and Eberly [2018] argue that the rise in intangible investment in retail trade can account for the increase in concentration and decreased investment. By contrast, the focus in our paper is on the characteristics of star firms and providing a link to the rise of intangible capital, increases in market power and industry 8 There is also an international literature where researchers have looked at export markets or labor productivity to document the rise in market power. De Loecker et al. [2016] examine the effect of tariff reductions on competition and markups. Freund and Pierola [2015] show that much of the exports of many countries can be attributed to a small number of firms which they refer to as export superstars. Andrews et al. [2015] highlight the notion of frontier firms, a small number of firms that are much more efficient than the bulk of their competitors. None of these papers look at returns or the role played by intangible capital. 5 concentration. We find little evidence for the hypothesis that these firms are restricting investment. 1 Identifying Star Firms We use data from Compustat that provides detailed financial information on publicly traded firms in the US over an extended period of time. We drop cross listed ADRs and restrict the sample to firms incorporated in the US. We also drop firms in Utilities (SIC 49), Finance, Insurance and Real estate (SIC 60-69) and Public Administration (SIC 90-99), observations with missing SIC codes, negative values for employees, sales, total assets, current assets and current liabilities, fixed assets, cash, and goodwill and missing total assets or sales. The advantage of using Compustat is that we have detailed balance sheet information that allows us to compute intangible capital. The caveat however, is that there are firm selection issues. First, it may be that listed firms, as a class, might not consistently represent star firms. Doidge, Kahle, Karolyi, and Stulz [2018] and Kahle and Stulz [2017] show that there are fewer US listed corporations today than 40 years ago. However, Grullon et al. [2017] argue that the void left by listed firms has not been filled by an increase in the number of private unlisted businesses. Using US Census data that includes both private and public firms, they show that even though more private firms have entered the economy, their marginal contribution to the aggregate product market activity has been relatively small. Public firms also account for one third of total US employment (Davis, Haltiwanger, Jarmin, Miranda, Foote, and Nagypal [2006]) and about 41% sales (Asker, Farre-Mensa, and Ljungqvist [2014]). Also using U.S. Census data, Maksimovic, Phillips, and Yang [2017] show that high initial firm quality at birth predicts subsequent listing decision. These findings suggest that while our sample will not be picking up small and young potential star firms in their private stages, we are targeting the sample of firms among which economically significant stars are highly likely to arise. The second potentially more important issue, as pointed out by Doidge et al. [2018], is that small, young, high-technology firms may benefit from private status where specific financial institutions, such as venture capital partnerships and private equity firms better meet their financing needs than public capital markets. Thus, such firms may be underrepresented in our sample of star firms. To 6 the extent that this listing gap has emerged only since 1999 (see Doidge, Karolyi, and Stulz [2017]), the early part of our sample period is immune to this. We also provide breakdowns of our results by high-technology and other industries. We define star firms as firms that realize high returns for their investors. We begin by using a standard definition of Return on Invested Capital (ROIC) as our measure of returns, where ROIC for firm i in year t is defined as: EBITit + AMit ROICit = (1) Invested Capitalit−1 where EBIT is Earnings before Interest and Taxes (Compustat item EBIT) and AM is Amortization of Intangible Assets (Compustat item AM). ROIC, as used in the Council of Economic Advisors [2016] report and Ben-David et al. [2013], amongst many others, computes the earnings that a corporation realizes over a period, as a fraction of capital that investors have invested into the corporation. The advantage of ROIC is that it measures investment capital as more than physical capital (fixed asset investment) which Doidge et al. [2018] show to be a declining portion of total assets over time in the US. We adopt a relatively conservative definition for Invested Capital as the amount of net assets a company needs to run its business: Invested Capitalit = P P EN Tit + ACTit + IN T ANit − LCTit − GDW Lit − max(CHE − 0.02 × SALE, 0) (2) where PPENT is Net Property, Plant, and Equipment, ACT is Current Assets, INTAN is Total Intangible Assets, LCT is Current Liabilities, GDWL is Goodwill that represents the excess cost over equity of an acquired company, CHE is Cash and Short-term Investments, and SALE is net sales. All these variable labels are the corresponding items in Compustat.9 The intangible assets as registered in Compustat, INTAN, include externally purchased assets like blueprints, copyrights, patents, licenses etc. and goodwill but does not include internal intangible assets like R&D and SG&A. We exclude Goodwill, which are the intangible assets arising out of M&A transactions when 9 We replace missing values of AM and GDWL with 0. 7 one company acquires another for a premium, in the computation of invested capital in equation 2 so that our measure is not distorted by price premiums paid for by acquisitions, allowing for an even comparison of operating performance across companies. Thus, ROIC measures the return that an investment generates for the providers of capital and reflects management’s ability to turn capital into profits. In calculating ROIC we subtract cash stocks in excess of those estimated required for transactions purposes. Following Koller et al. [2017], we treat cash above 2% of sales as excess cash and subtract it from the firm’s invested capital but in section 4.2 we undertake robustness tests allowing for varying percentages. Thus, our principal estimates are not affected by firms’ decisions on whether to stockpile cash in low-tax jurisdictions in order to manage their tax liabilities, as is the case of many large U.S. multinationals. We define ROIC star as a dummy variable that takes the value 1 if the firm’s ROIC is above the 90th percentile of ROIC across all firms in the US economy in a particular year and 0 otherwise. To replicate the figure in previous studies such as Furman and Orszag [2015] and Koller et al. [2017], we restrict our sample to large firms and drop firms with negative invested capital. As noted before, Figure 1 shows that there is a large rise in capital returns over the past three decades where the ratio of the 90th percentile ROIC firm to the median ROIC firm has increased by over 69%. We also see that the divergence of the top decile of firms from the rest of the economy really takes off in the 1990s.10 These results are qualitatively consistent with Furman and Orszag [2015], Koller et al. [2017] and the Council of Economic Advisors [2016], all of which were produced using a proprietary dataset of US firms from McKinsey & Co. 1.1 Role of Human Capital Firms differ in the complexity of tasks that they perform and the product market may reward certain capabilities more than others. To address this issue, we construct industry-level indices of the composition of tasks firms perform and assess how they affect the likelihood of a firm from that industry becoming a star firm. In creating these indices, we draw on a large labor market literature in economics. Autor et al. [2003], Costinot et al. [2011] and Acemoglu and Autor [2011], 10 In an early version of the paper, we find much higher increases in ROIC (over 190%) for the 90th percentile without the sample restrictions to large firms and firms with positive invested capital. 8 have argued that globalization and advances in technology and computerization have increased the comparative advantage of individuals who perform non-routine tasks requiring problem solving, intuition, persuasion, and creativity. To obtain measures of human capital, we use O*NET, a database maintained by the U.S. Department of Labor that provides data on occupation-specific descriptors that define the key features of an occupation such as worker abilities, technical skills, job output, work activities, etc. We focus on the following three measures of human capital: CPS (Complex Problem Solving) which is identifying complex problems and reviewing related information to develop and evaluate options and implement solutions; NRCOG (Non-routine Cognitive Analytical skills) from Keller and Utar [2016] which is the sum of Mathematical Reasoning, Inductive Reasoning, Developing Objectives and Strategies, and Making Decisions and Solving Problems; and RMAN (Routine Manual) from Keller and Utar [2016] which is the sum of Spend time making repetitive motions, Pace Determined by Speed of Equipment, Manual Dexterity, and Finger Dexterity. We merge the occupation level scores to the Occupational Employment Statistics (OES), a US establishment level dataset where workers are classified into occupations on the basis of the work they perform and skills required in each occupation. We compute a weighted average across occupations in each firm weighting by the number of employees in each occupation to obtain a score for each establishment. We then take weighted averages across all establishments in an industry to compute an industry-level skill score. We currently use these scores for manufacturing industries. We separate the manufacturing sample into high and low skill industries based on the CPS, NRCOG, and RMAN scores where high skill is defined as greater than or equal to the median value for each of the skill measures and low skill is defined as less than the median value for each of the skill measures. In Figure 2, we identify star firms in each of these sub-samples as firms in the top 10% of ROIC in that sample in a particular year. We again focus on large firms to be consistent with the sample in Figure 1. Figure 2 shows that the ROIC and the run-up for star firms is higher in industries with high skill as measured by low RMAN, high CPS, and high NRCOG. If this finding is correct, it would imply that firms employing a high skill labor force are also more likely to earn higher returns and that there is a growing divergence between the most profitable of those firms and the other high-skill firms. We also see a large divergence between the ROIC 9 in high skilled versus low skilled industries when we split industries by RMAN, CPS, or NRCOG. However, a concern with these estimates is that in high skilled industries, intangible capital is being mis-measured which reduces total invested capital, thereby inflating ROIC numbers. There is also a large divergence between the ROIC realized by the median cognitively skilled firm (low RMAN) and the median high RMAN firm. 1.2 Mis-measurement of Intangible Capital One of the concerns with the above definition of star firms is that financial statements do not measure intangible assets accurately and the consequent underestimation of intangible capital is likely to be more important in high skilled industries. This would lead to overestimation of ROIC and biased regression estimates. The concern that conventional measures of invested capital do not properly capitalize the value of intangibles is a long standing one. Earlier attempts to address it include Peles [1971], Grabowksi and Mueller [1978], Hirschey [1982], and Falato et al. [2013]. More recently, Peters and Taylor [2017] have produced firm-level estimates of intangible capital and shown that including intangible capital in the definition of Tobin’s q produces a superior proxy for investment opportunities. They also show that their adjustments are not sensitive to specific assumptions on the depreciation of intellectual capital. Thus, while these measures are, by construction, approximations, they are arguably the best available. Hence, as an alternate definition of invested capital, we replace the IN T ANit in equation (2), with the new definition of intangible capital from Peters and Taylor [2017], ICAPit . T OT Invested Capitalit = P P EN Tit + ACTit + ICAPit − LCTit − GDW Lit − max(CHEit − 0.02 × SALEit , 0) (3) where ICAPit , is defined as the sum of externally purchased intangible capital (Compustat item INTAN ) and internally purchased intangible capital, both measured at replacement cost. Internally purchased intangible capital is in turn measured as the sum of knowledge capital K int know and organization capital K int org. The perpetual-inventory method is applied to a firm’s past research and development expenses (Compustat item XRD ) to measure the replacement cost of its knowledge 10 capital. Similarly, a fraction (0.3) of past selling, general, and administrative (SGA) spending is used as an investment in organization capital, which includes human capital, brand, customer relationships, and distribution systems.11 Correspondingly, we also adjust the profits in the numerator to account for the use of intangible capital in computing invested capital. Thus, the new ROIC is given by: ADJP Rit ROICit = T OT (4) Invested Capitalit −1 where ADJP Rit = EBITit + AMit + XRDit + 0.3 × SGAit − δRD × K int knowit − δSGA × K int orgit (5) where δRD is the depreciation rate associated with knowledge capital and is set to 15% following Peters and Taylor [2017] and δSGA is the depreciation rate associated with organization capital and is set to 20% following Falato et al. [2013]. Note that using an adjustment for intangible capital affects ROIC in two ways. First, it increases the denominator by the amount of the adjustment for intangible capital. Second, R&D and a portion of SGA expenditure, which would previously have been expensed is now treated as additions to capital stock. Thus, it is not subtracted from the firm’s conventionally calculated earnings (EBIT) to obtain the adjusted earnings. However, since the stock of intangible capital is now treated as an asset, an additional depreciation expense is now deducted from EBIT. This second adjustment either increases or decreases the numerator of ROIC, depending on the level of current R&D and SG&A expenditures compared to the stock of intangible capital. After dropping firms with negative invested capital, missing or negative book value of assets or sales, and firms with less than $5 million in physical capital (Compustat variable PPEGT )12 11 Since Compustat item XSGA is the sum of SG&A and R&D, we follow the procedure in Peters and Taylor [2017] to isolate SGA as XSGA-XRD-RDIP where RDIP is In-Process R&D. We replace missing values of XSGA, XRD, and RDIP with 0. 12 We apply the PPEGT filter since Peters and Taylor [2017] recommend that the intangible capital adjustment is not appropriate for firms with less than $5 million in physical capital. 11 and top and bottom 1% outliers in ROIC T OT , we define ROIC T OT Star as a dummy variable that takes the value 1 if the firm’s ROIC T OT is above the 90th percentile of ROIC T OT across all firms in the US economy in a particular year and 0 otherwise. We also focus on the years 1990-2015 for all the figures and tables henceforth since the high run-up in ROIC in Figures 1 and 2 starts around 1990. When we correct invested capital to include intangible capital, we see no run-up in ROIC T OT for the top 10% of firms in Figure 3.13 In Figure 4, we present estimates for high skilled versus low skilled industries. The run-up we saw in Figure 2 in high skilled industries disappears once we adjust for intangible capital. In section 4 of the paper, we discuss various robustness tests and alternate definitions of intangible capital to address concerns with the definition of cash holdings. We are agnostic on how to treat the firm’s cash holdings. For much of the analysis we adopt the practitioner view and subtract cash positions in excess of 2% of revenue from the firm’s invested capital. In section 4.2 we consider alternative treatments of cash and note the outcomes. 2 Estimating Concentration and Market Power As a measure of competition, we define Herfindahl Index (HHI) of market share in each 3-digit NAICS industry in each year. Specifically, in each year t for each 3-digit NAICS industry j, industry concentration is measured as: N HHI = s2 i (6) i=1 SALE where si is market share of firm i given by and N is total number of firms in industry j j SALE in year t. A higher HHI implies weaker competition. While HHI measures industry concentration, it treats all firms in an industry identically. Thus, HHI ignores potential firm-specific indicators of market power such as firm size and market share. We use Log(Assets) as a measure of firm size where assets is the Compustat item AT. Market Share is the ratio of firm i ’s sales to total industry j ’s sales in a particular year, to allow for the possibility 13 As shown in Appendix figure A1, we obtain a similar picture when we restrict the sample to large firms, and extend the time period to 1965 to be consistent with the sample in Figure 1 12 that large market share firms in a concentrated industry realize different returns compared to low market share firms. We also use firms’ markup of price over marginal cost as a firm-level measure of the firm’s pricing power. De Loecker and Eeckhout [2017] show that average markups have increased from 18% above marginal costs (as measured by Cost of Goods Sold) in 1980 to 67% above marginal cost by 2014. To estimate markups, we rely on the framework by De Loecker and Warzynski [2012] and De Loecker and Eeckhout [2017] that is based on cost minimization of a variable input of production, without additional assumptions on firm demand or competition. Heuristically, this measure takes the firm’s capital stock as given, and estimates the markup that the firm can charge customers over its variable costs. A high markup is consistent with market power, as competition would erode the firm’s ability to charge above variable costs. Importantly, since markups do not take into account the cost of tangible and intangible capital, high markups are consistent with both high and low rates of return to invested capital. Intuitively, it is natural to think of high markups resulting from a firm’s exercise of market power by reducing sales and thereby realizing high returns on invested capital. However, it is possible for a firm to have a low markup, high sales per unit of invested capital and to be a star firm. Thus, the extent to which star status is related to high markups is an empirical question. Recent work by Traina [2018] argues that COGS grossly underestimate firms’ variable costs. Other expenses, such as SGA are increasingly a lion’s share of variable costs for US firms. Traina shows that once we include SGA in the calculation of marginal costs, there is no increase in public firm markups. Consistent with Traina [2018], we base our measure of variable inputs on Operating expenses (Compustat item OPEX ) rather than Cost of Goods Sold (Compustat COGS ) as in De Loecker and Eeckhout [2017]. OPEX includes SGA expenses whereas COGS only includes costs of production such as material, labor, and overhead and does not include SGA expenses. COGS has been a declining share of variable costs for US firms as shown in Figure A2 of the Appendix. We differ from Traina [2018] in our adjustment of operating expenses to include the correction for intangible capital. 13 The derivation of markups follows De Loecker and Eeckhout [2017] and Traina [2018] where the markup is simply defined as the ratio of price for the output good over marginal cost: Pit Qit M arkupit = θit × (7) Pit Vit Pit Qit where θit is output elasticity of the variable input, and Pit Vit is the revenue share of the variable input (or simply SALE/OPEX ). To estimate this, we consider 3-digit NAICS industry-specific Cobb-Douglas production functions with variable inputs and capital.14 Thus, for a given industry (3-digit NAICS): SALEit = βv × OP EXit + βk × Kit−1 + ωit + it (8) where SALEit is Log (sales deflated by GDP deflator), OPEX is the variable input and is measured as Log(Operating Expenses deflated by GDP deflator), Kit−1 is Log(Capital Stock), and ωit is Log Productivity which is assumed to follow an AR(1) process ωit = ρωit−1 + ξit . The parameter ξit is the innovation to the firm’s productivity process. We use the perpetual inventory method to construct measures of capital stock. We first initialize the capital stock using the first available entry of gross PPE and then iterate forward on capital using the accumulation equation: Kit = Kit−1 + ∆Iit (9) where ∆Iit is net investment computed using changes to PPENT and deflated by the investment goods deflator.15 The coefficient βv represents the output elasticity of the variable input OPEX. To estimate the above, we follow the literature and adopt a control function approach to address endogeneity concerns due to the potential simultaneity between unobserved productivity shocks and the demand for inputs. If the demand for an input increases with productivity shocks, that input’s demand function can be inverted and the unobserved productivity shocks can be derived as a function of observables. In the first stage, we remove idiosyncratic errors from the production 14 In our current implementation we do not apply the Olley and Pakes [1996] correction. The effect of that correction is somewhat controversial, as discussed by Ackerberg et al. [2015], Ackerberg [2016], Gandhi et al. [2017], and Frank and Yang and footnote 12 in De Loecker and Eeckhout [2017]. As shown by Yasar et al. [2008], the correction is small on this data set. All our results are materially the same if we were to apply the correction 15 GDP deflator is given by line 1 of NIPA Table 1.1.9 and the Non-residential fixed investment good deflator is given by line 9 of Table 1.1.9. 14 process by estimating the following regression: SALEit = βv × OP EXit + βk × Kit−1 + αi + γt + it (10) where αi are firm fixed effects and γt are year fixed effects. From this we obtain sales estimates which are used to derive implied productivity ωit as a function of elasticity parameters β . This function is projected onto its lag which is then used to recover the innovations in the productivity process ξit , again as a function of industry-specific output elasticities, β . Under the assumption that the variable input use responds to productivity shock but its lagged values do not, the elasticity parameters β can be obtained using a standard GMM procedure from the following moment conditions:   ξit (β ) OP EXit−1  E =0 (11) Kit−1 Finally, based on these output elasticity estimates, we re-write the equation for markups as SALEit M arkupit = βv × (12) OP EXit SALEit where the βv are at the industry level and OP EXit are varying at the firm level. Following De Loecker and Warzynski [2012], we correct the markup estimates for measurement error in sales obtained in the first stage. We make two additional modifications to the above. First, since we treat research expenditures as an intangible investment, and the Peters and Taylor [2017] adjustment treats a portion of the SGA as an organizational investment, our calculations of firm markups differs from that in Traina [2018]. Expenditures on R&D and the full amount of the SGA are included in OPEX and treated as variable costs. However, our measure of the firm’s capital stock includes both tangible and intangible capital. This, in turn implies that intangible investments such as R&D and a portion of SGA are subtracted from OPEX in order to obtain our measure of variable costs, OPEX*. Thus we replace OPEX in the above equations with OPEX* where OP EX ∗ = OP EX − XRD − RDIP − 0.3 × SGA (13) 15 Second, we re-define K to include intangible capital as Log (CapitalStock + ICAPit ) where ICAPit , is the sum of externally and internally purchased intangible capital defined in section 1.2. With these adjustments, we re-estimate markups to be: ∗ SALEit M arkupT it OT = βv × ∗ (14) OP EXit ∗ are the output elasticities estimated using OPEX* as the variable input and Log (CapitalStock + where βv ICAPit ) as the capital input in equation 14. Table 1 presents summary statistics of the main variables in our analysis. In addition to the variables discussed above we also use a proxy for firm age which is defined as the number of years since the firm first appears in Compustat following Giroud and Mueller [2010]. The mean ROIC in our sample according to conventional metrics is -32.3% and for just large firms (unreported in the table) it is 24.3. Once we adjust for intangible capital, the mean ROIC T OT is 13%. By definition, 10% of our sample are classified as star firms according to both the ROIC measures, ROIC Star and ROIC T OT Star. Once we take into account intangible capital, the average markup is 1.18. The average Herfindahl industry concentration is 0.09 and the average firm market share is 0.014. 3 Empirical Findings 3.1 Is there a rise in Markups? There has been a great deal of controversy on the rise in market power in the US. Using marginal costs measured by COGS, De Loecker and Eeckhout [2017] document a stunning rise in markups in the US over the past three decades. Traina [2018] however argues that COGS are a declining share of firm costs and once we use operating expenses that includes COGS and SGA, there has been no rise in firm markups. This is an important policy question as it also speaks to the discussion on the rise in industrial concentration and decline in labor share (see Grullon et al. [2017] and Autor et al. [2017]). 16 We begin by comparing markups generated using different measures of marginal costs - COGs as in De Loecker and Eeckhout [2017], OPEX as in Traina [2018], and our own measure of operating expenses adjusted for investments in intangible capital following Peters and Taylor [2017], OPEX*. In Figure 5, we plot the sales-weighted average markups each year estimated using COGS, OPEX, and OPEX* as variable inputs. We plot the markups over the period 1950-2015 for comparison to the evidence in De Loecker and Eeckhout [2017] and Traina [2018]. The figure confirms that the rise in markups exists only when we define variable costs in terms of COGS rather than OPEX. The rise in COGS markups in the figure mimics the rise in markups shown in De Loecker and Eeckhout [2017]. Our magnitudes are smaller because unlike in their paper, we do not include financials, real estate, and utilities in our sample and also drop foreign incorporated firms. The rise in COGS markups are even higher with the inclusion of these other sectors and foreign incorporated firms. However, importantly the figure also shows that once we correct the definition of OPEX for intangible capital and use OPEX* as an input, we get a similar rise in markups. In Figure 6, we estimate the evolution of markups using capital adjustments, M arkupsT OT , in the US economy over our sample period. We see an upward trend only for the 90th percentile firms. To see if there is dispersion in markups by industry skill, we look at industries that rely heavily on routine manual tasks versus those that do not rely heavily on routine manual tasks in Figure 7 and find that markups are higher in high skilled industries (low RMAN) than low skilled industries for the top 10% of firms. Overall, this section shows that there has indeed been a rise in markups once we adjust operating expenses for investment in intangible capital. While there is just a modest divergence between the top 10% of firms with highest markups and the rest of the economy, we see these differences amplified in high skilled industries where the measurement of intangible capital is even more important. 17 3.2 Explaining the Rise of Star Firms To explore the incidence of star firms, for firm i in industry j in year t, we estimate the following regression: Starijt = a + β1 × Log (Assets)it + β2 × Log (Age)it + β3 × M arket Shareit + β4 × M arkupsit + β5 × HHIjt + φj + γt + ijt (15) where Star is a dummy variable that takes the value 1 if the firm is a star firm and 0 otherwise. We first estimate the equation using ROIC Star as the definition of a star firm and using Markups based on operating expenses, OPEX. We then re-estimate the equation using corrections for intangible capital using ROIC T OT Star and M arkupsT OT . Log(Assets) and Log(Age) are measures of firm size and size, HHI is Herfindahl Index measure of industry concentration, and Market Share is firm level market share. The regressions include industry (3-digit NAICS level) fixed effects, and year fixed effects and all standard errors are clustered at the firm level. All the regressions are estimated using ordinary least squares (linear probability models) but we get similar results using Logit estimation. We don’t use firm fixed effects in our main specification because we are interested in understanding time invariant human capital or skill characteristics that explain variation in ROIC. Instead, we employ year and industry dummies and cluster the standard errors at the firm level to capture the lack of independence among the residuals for a given firm across years (Petersen [2009]). The main coefficient of interest in the above specification is β4 which shows the sensitivity of star status to firm markups. In Table 2, we estimate specification 15 to examine firm and industry characteristics that are associated with star status. In columns 1-5 we focus on the full sample of firms, in column 6 we look at only manufacturing firms for which we have data on skills, in column 7 we look at large firms (defined as firms with more than $200 million in assets in real terms obtained by deflating Compustat item AT by GDP deflator) and in column 8 we focus on young firms (defined as firms that are less than five years of age). We drop firms with negative invested capital in all regressions. The results in Table 2 show that ROIC Star firms are on average larger, younger, have higher 18 markups and higher market share than rest of the firms in the economy. Industry concentration does not seem to affect star status. Given the concerns with mis-measurement of intangible capital, we repeat the specifications in Table 2 with the new measures of ROIC and Markups, ROIC T OT and M arkupsT OT , that include adjustments for intangible capital from Peters and Taylor [2017]. Next, we repeat the estimation in Table 2 using ROIC T OT and M arkupsT OT . The results in columns 1-5 of Table 3 show that after correcting for intangible capital, we find strong evidence that high markups predict ROIC T OT . The effects are also economically significant. There is a 5 percentage point increase in the probability of being a star firm when markups go up by one standard deviation. We also see that high ROIC T OT firms are on average large and young. These results on size, age, and markups hold in the different sub-samples in columns 6-8 for manufacturing, large firms, and young firms respectively. We find limited evidence that firm level market share or industry concentration at the 3-digit NAICS level predicts star status. HHI is never significant in any of the specifications and market share comes in significant only for large firms. In unreported tests, we find similar results if we use logit estimation for all the regressions instead of linear probability models. It is likely that the effect of market power indicators may vary across levels of ROIC T OT . In panel B of Table 3, we re-estimate the full model (column 5 of panel A) using quantile regressions. This approach also has the advantage of being directly suggested by the original star firm hypothesis, which is formulated focusing on the differences between the top 10% of firms and the rest. We use the generalized quantile regression estimator developed in Powell [2016] that allows us to estimate unconditional quantile effects in the presence of additional covariates. A possible advantage of this approach may be that the estimates are not sensitive to the extreme outliers, that is firms with very little invested capital due to write-offs and/or atypical revenue windfalls. The results show that the profitability of firms at the top of the distribution of ROIC T OT appears more sensitive to markups than that at the bottom. Henceforth, in all regression specifications we use the new definitions of ROIC and Markups (ROIC T OT and M arkupsT OT ) that incorporate intangible capital. To explore the impact of human capital, we interact industry indices of human capital skill with 19 the key variables of interest as shown in the equation below: Starijt = α0 + β1 × Log (Assets)it + β2 × Log (Age)it + β3 × M arket Shareit + β4 × M arkupsit + β5 × HHIjt + β6 × Log (Assets)it × Industry Skillj + β7 × M arket Shareit × IndustrySkillj + β8 × M arkupsit × Industry Skillj + β9 × HHIjt × Industry Skillj + φj + γt + ijt (16) In Table 4, we examine the role of industry skill in predicting star status. We have the industry skills data only for manufacturing following Ayyagari and Maksimovic [2017] and hence these regressions are based on just the manufacturing industries. Across all regressions, the main effects on size and markups are positive and significantly associated with predicting superstar status. In columns 1-4 we look at the role of RMAN, in columns 5-8 we look at the role of CPS, and in columns 9-12 we look at the role of NRCOG. The interaction of RMAN with size, markups and HHI are all significant at the 1% level in columns 1-3 suggesting that in industries that rely heavily on routine manual skills (low skilled industries), large firms, firms with high markups and those in concentrated industries are less likely to be in the top 10% of ROIC firms in a year. We find similar results with industries that rely on high complex problem solving skills in columns 5-8 where large firms and firms with high markups in these industries are more likely to star firms. In columns 9-12 we see that in industries that rely on high non-routine cognitive analytical skills, large firms are more likely to be ROIC stars. In unreported tests, we estimate unconditional quantile effects of the interaction of skill and markups (models 2, 6, and 10 of panel A) and find that the impact of skill and markups is stronger at higher levels of quantiles, especially for large firms. 3.3 Future Performance of ROIC Stars Thus far, we have defined a firm as a star firm in a given year if its return on invested capital is in the top 10% of firms in that year. It could be the case that there is a lot of churning in this top 10% of firms each year with different sets of firms randomly realizing high returns each year. In the 20 first part of this sub-section we explore if these high returns are persistent and if being a superstar is associated with superior performance. To this end, we construct five non-overlapping panels: 1990-1995, 1995-2000, 2000-2005, 2005-2010, and 2010-2015 and examine if being a superstar is associated with higher average performance in the subsequent five year period. Specifically, the regression we estimate is as follows: P erf ormanceijt = α0 + β1 × Log (Assets)it−5 + β2 × Log (Age)it−5 + β3 × HHIjt−5 + β4 × M arket Shareit−5 + β5 × M arkupsit−5 + β6 × ROIC T OT Starit−5 T OT + β7 × ROICit − 5 + φj + γt + ijt (17) We look at the following five performance measures: 5-year average ROIC T OT , Sales growth com- puted as the five year log difference in sales divided by 5, Employment growth computed as the five year log difference in employment divided by 5, 5-year average Labor Productivity, and 5-year average Tobin’s QT OT over these periods. Using stacked panel regressions, we examine the asso- ciation between each of these measures and star firms identified at the beginning of each panel. We also control for size, age, market share, HHI, and markups at the beginning of each panel. All regressions also include industry and year fixed effects. Column 1 of Table 5 shows that while higher ROIC T OT is on average positively associated with higher average ROIC T OT in the subsequent five year period, star firms have lower ROIC T OT suggesting a regression to the mean. The predicted value of average 5-year ROIC T OT for firms that were superstars five years ago is 55.3 compared to 20.5 for firms that were not superstars five years ago. Column 2 shows that while past ROIC T OT is not associated with sales growth, past ROIC T OT stars have higher sales growth in the future but not necessarily higher employment growth in column 3. In column 4, ROIC T OT is associated with higher average labor productivity whereas star firms are not necessarily more productive than other firms in the economy. Column 5 shows that firms that are classified as ROIC T OT stars five years ago have high average Tobin’s Q in the 5 years hence. Overall these results suggest that star status is associated with higher future performance and there is a fair degree of persistence in star status as firms that were ROIC stars five years ago have 21 higher average returns over the subsequent five-year period than firms that were not ROIC stars. 3.4 Markups, Output, and Investment in Star Firms A central policy question is whether high profits, and in particular, star firm status, are the result of monopoly power. This possibility is suggested by De Loecker and Eeckhout [2017] who calculate that markups in the US increased from 18% above marginal cost in 1980 to over 67% in 2014. Markups could rise due to outright monopoly or due to monopolistic competition where with high markups but also high fixed costs, firms actually earn low profits. Grullon et al. [2017] also argue that the increase in industry concentration in the US is related to high return on assets and this is mainly driven by firm’s higher profit margins rather than asset utilization. 16 Our results above are more nuanced. We find that a measure of pricing power, a high markup, statistically predicts star status for a firm, but there has been little increase in markups over the past several decades. To gain further intuition about the role of markups, we explore the relation between markups, ROIC, and a measure of output - the ratio of sales to invested capital - graphically for star firms and for all other U.S. public firms in general. In Figure 8 we present a histogram of markups for firms that were classified as ROIC T OT stars and for all other firms and for each of those sub-samples, a non-parametric smoothed scatter plot of ROIC T OT against markups using kernel weighted local polynomial smoothing. The figure shows that while firms are distributed across the range of markups even when we look at just the star firms, the tails are thin so there are few firms with very low markups and very high markups for both star firms and all other firms. In general, we see a monotonically increasing relationship between ROIC T OT and markups suggesting that high profits are associated with pricing power. However, for star firms the plot is quite flat, indicating that there is no visible association between markups and ROIC T OT within the sample of firms that have passed the threshold to be classified as star firms. In addition, in Figure 9, when we define superstars as firms that have ROIC T OT in the top decile in 3 or more years over this period or 5 or more years over the period, we find the 16 Blonigen and Pierce [2016] study mergers and show that mergers are not necessarily associated with increase in efficiency but rather with an increase in market power as seen in the 15-50% increase in markups associated with mergers. 22 scatter plot for superstars to be similar to that of the star firms.17 In Figure 10, we present non-parametric smoothed scatter plot of Sales/Invested Capital against markups using kernel weighted local polynomial smoothing over the period 1990 to 2015 for all firms in the economy and for ROIC stars. We find that the relation between sales/invested capital in non-monotonic for both sets of firms. Beyond a certain level of markups, there is a decline in Sales/Invested Capital. Such an association of high markups and low output might arise if firms with market power restrict output to increase markups. However, for star firms, Sales/Invested Capital is higher at each level of markup than for all firms in general, suggesting that these firms are not restricting output more than other firms with the same markups. The difference is particularly high at lower markups, suggesting that low-margin star firms, in particular, are adopting a high volume marketing strategy. In Table 6, we explore the relation between star status, markups, output and investment more formally in a multivariate regression framework controlling for Log Assets, Log Age, profitability (Tobin’s Q ), industry and year fixed effects. For output, we use Sales/Invested Capital and for investment, we use both physical investment Capex/Invested Capital and intangible investment (XRD/Invested Capital ). All the independent variables are lagged by one period. Columns 1, 3, and 5 shows that ROIC T OT star firms have higher Sales/Invested Capital and greater investment, both CAPEX, and Intangible (R&D) Investment compared to all other firms. Higher markups are associated with lower Sales/Invested Capital, higher CAPEX, and lower R&D Investment in these regressions. We find these results to be robust to a number of checks including industry x year fixed effects, scaling CAPEX by PPENT and XRD by intangible capital (ICAP ). Given the skewness in the R&D variable, we also find similar results if we were to use both the dependent and independent variables in terms of logs. When we focus on the the interaction of ROIC T OT and markups, we find the interaction to be negative and significant for Sales/Invested Capital. The economic significance of these interaction terms can be seen in the predictive margin plots. Figure 11 shows that ROIC T OT star firms have steeper declines in Sales/Invested Capital with markups but their investments do not seem to be 17 We obtain similar outcomes when we define superstar status more narrowly by requiring five years of star status in the period 2005 to 2015. 23 very significantly different from that of the non-star firms. These results are also robust to using firm and year fixed effects in all regressions. In all cases, we find no evidence that the star firms are cutting output or investment more than the other firms in the economy. The analyses thus far jointly show that the evidence attributing high profits of star firms to errez and Philippon [2017] that the market power is modest. Moreover, the concern raised by Guti´ decline CAPEX is attributable to increasing market power is not supported. Overall, our results indicate that while markups strongly predict high profits, not all star firms have high mark-ups and that star firms are restricting output or investment less than other firms with the same markups. The conclusion that the exercise of market power by star firms is relatively modest contrasts with the popular public policy debate in the US that has been dominated by anecdotal evidence of a few star firms - Facebook (FB), Amazon.com (AMZN), Apple (AAPL), Microsoft(MSFT) and Alphabet (GOOGL). These firms are often accused of using monopoly power as a result of proprietary technology and increasing returns to scale. To take a close look at this, we examine the returns to capital and markups of these in relation to the rest of the economy. Figure 12 shows that these firms (especially Apple) have abnormally high returns to capital which exceed even the top 10% of ROIC T OT firms. Their markups in Figure 13 are however not abnormally high (except for Facebook) and are below the 90th percentile of markups in our sample for most of the sample period. Therefore, surely a small number of superstar firms are truly diverging from the rest and disrupting conventional business models in the process. For these firms, their markups may be understating their market power. Indeed, in some cases these firms might be limiting their short- run profits in the hopes of realizing future market dominance. An example of this might be Amazon. In his letter to Amazon shareholders in 1997, Jeff Bezos stated that Amazon makes decisions and weighs tradeoffs differently than most other firms: We believe that a fundamental measure of our success will be the shareholder value we create over the long term. This value will be a direct result of our ability to extend and solidify our current market leadership position. The stronger our market leadership, the more powerful our economic model. Market leadership can translate directly to higher revenue, higher profitability, greater 24 capital velocity, and correspondingly stronger returns on invested capital. Our decisions have consistently reflected this focus. We first measure ourselves in terms of the metrics most indicative of our market leadership: customer and revenue growth, the degree to which our customers continue to purchase from us on a repeat basis, and the strength of our brand. We have invested and will continue to invest aggressively to expand and leverage our customer base, brand, and infrastructure as we move to establish an enduring franchise. (Emphasis added)18 Thus, Amazon prioritized growth over profits to achieve enough scale that was central to their business model. This suggests that even for some of the most capable star firms like Amazon, metrics such as ROIC and markups may understate their potential market power. By the same token, these firms are not exercising that potential market power in ways that harm consumers in the short run. Of course, firms that follow this strategy are likely hoping that their dominant position will enable them to profit from their market dominance in the future. As seen in Figures 12 and 13, ROIC and markups of most of these elite firms seem to be reasonable initially when they are in the ”franchise” building stage and then explode for a couple of firms that have built up a large enough market, which compounds the measurement issues. Khan [2016] also argues that the current anti-trust laws their focus on short-run consumer welfare are just not equipped to recognize the anti-competitive nature of Amazon’s predatory pricing and ability to use its dominance in one sector to gain market share in another. Building a franchise in the expectation of future profits is not new, and these star firms of today may be likened to the superstars in the early part of the 20th century like US Steel, Standard Oil and Sears, and Roebuck and Company who have passed into history. This suggests that the critical concern for policy is not only to control the exercise of market power by these few firms, but to ensure that markets remain contestable and that entrants with new technologies are able to challenge the current market leaders. Policy measures could include limitations of acquisitions of new technologies through mergers. For instance, see Cunningham et al. [2018] for a discussion of mergers and the subsequent liquidation of new technologies by incumbent firms in order to maintain 18 See Damodaran (2018, April 26). Amazon: Glimpses of Shoeless Joe? [Blog post]. Retrieved from http://aswathdamodaran.blogspot.com/2018/04/amazon-glimpses-of-shoeless-joe.html 25 market dominance. 4 Robustness In this section, we subject our findings to a series of robustness tests. At the outset, our results are crucially dependent on the adjustment for intangible capital in the measurement of ROIC and markups. In unreported robustness tests we investigate whether our results are affected by if the adjustment is partial. We vary the intangible capital adjustment from 25% to 75% of the amount recommended by Peters and Taylor [2017] and repeat the specifications in Table 4. All our results are materially similar suggesting that our results are robust to smaller adjustments to intangible capital. 4.1 Alternate Classification of Superstars We next narrow our definition of star firms to see whether the role of market power and size differs for the most profitable firms. Appendix Figure A3 plots the mean ROIC for the Top 100 and Top 150 firms each year and confirms that there is no run-up in ROIC over time for even the top 100 or 150 firms. In unreported robustness tests, we vary our definition of star firms to the Top 100 firms and Top 150 firms in terms of ROIC and repeat our estimations from Table 4. Although we do not have as much power in these tests, the results confirm that in high skilled industries, large firms are more likely to be in the top 100 or 150 firms earning super normal returns. While the above definitions are based on returns to equity, as an alternate definition, we define stars in terms of Tobin’s Q. Again following Peters and Taylor [2017], we define QT OT as the ratio of Firm value to TOTCAP which is the sum of physical (PPENT ) and intangible capital (ICAP ): Vit QT it OT = (18) T OT CAPit where V is the market value of the firm defined as the market value of equity (=total number of common shares outstanding (Compustat item CSHO ) times closing stock price at the end of the fiscal year (Compustat item PRCC ) plus the book value of debt (sum of Compustat items DLTT 26 and DLC ) minus the firm s current assets (Compustat item ACT ) which includes cash, inventory, and marketable securities. After dropping top and bottom 1% outliers in QT OT , we define Q star as a dummy variable that takes the value 1 if the firm’s QT OT is above the 90th percentile of QT OT across all firms in the US economy in a particular year and 0 otherwise. While QT OT has the advantage of using a market valuation of the firm’s prospects, the measure is prospective in that it captures the value of the firm’s investment opportunities given the market’s view of its investment plans (e.g. Novy-Marx [2007]). Thus, it may not measure current product market performance, and we do not use it as the primary metric of star status. In Table 7, we use the alternate definition of star status based on Tobin’s Q defined in equation (14). The estimates in Table 7 confirm our previous results on the role of industry skill as a determinant of star status. In high skilled industries (high complex problem solving skill, high non- routine cognitive skills, and low routine manual skills), large firms and those with high markups are more likely to be top performers in terms of Tobin’s Q. 4.2 Measurement of Excess Cash There is a great deal of controversy in how to treat a firm’s cash holdings in the computation of a firm’s invested capital. It is standard financial reporting practice to include a firm’s cash holdings in the definition of its invested capital. However, financial analysts routinely subtract a large fraction of cash holdings, say any cash in excess of 2% of annual revenues, from the firm’s calculated investment capital (e.g. Koller et al. [2017]). The rationale for that is that the excess cash is unnecessary to support operations and confounds valuations of product market opportunities. This view is also supported by a large body of academic work (e.g. Jensen [1986]; Harford et al. [2008]; Dittmar and Mahrt-Smith [2007]) which argues that large cash holdings are a reflection of agency conflicts between managers and firms shareholders, and are not relevant to the valuation of a firm’s operations. A second reason to subtract excess cash from invested capital is to circumvent the policy of many large U.S. multinationals to stockpile cash in low-tax jurisdictions in order to manage their tax liabilities (e.g. Faulkender and Petersen [2012]; Faulkender et al. [2017]. Against that, there are 27 numerous findings that high cash positions occur typically in R&D intensive firms, and that these cash holdings may economically rational (see Martin and Santomero [1997]; Boyle and Guthrie [2003]; Bates et al. [2009]; and Harford et al. [2014]). In particular, to the extent that R&D intensive firms face higher operational risks, and that intellectual capital cannot be easily used as collateral for bank loans, high cash positions are economically motivated. Moreover, from the perspective of the firms’ owners, the relevant returns should be calculated as function of all the capital committed, not just the portion which would have been committed under an alternative corporate governance regime. Moreover, as Damodaran [2005] notes, the 2% ratio has been used as a rule of thumb amongst analysts and does not have a deep theoretical basis. This ratio can be higher or lower depending on the working capital needs of a business. In this section, we examine whether our findings are sensitive to the treatment cash holdings. Hence as an alternate variation, we define invested capital to only include working capital and physical and intangible capital. Thus CASH Invested Capitalit = P P EN Tit + ACTit + ICAPit − LCT it − GDW Lit (19) Analogously we define ROIC with this new adjustment as: CASH ADJP Rit ROICit = CASH (20) Invested Capitalit In Figure 16, we present four ROIC T OT graphs where ROIC T OT is re-computed using cash above 1% of sales, 5% of sales, 10% of sales, and 20% of sales respectively as excess cash. Across all the figures, we see that there is no run-up in ROIC T OT for the top 10% of firms as in Figure 3. In Table 8, we repeat the specifications in Table 4 but re-estimating ROIC T OT using different treatment of cash. We consider excess cash to be any cash over 1% of sales in panel A and over 10% of sales in panel B. In Panel C, we use the firm’s total cash holdings in computing ROIC, which we term ROIC CASH . Across the three panels, we obtain similar results wherein large firms and those with high markups in high skilled industries are more likely to be superstar firms. 28 5 Conclusion In this paper, we assess publicly-listed star firms in the U.S. We we use financial statement data as conventionally presented, a small percentage of star firms seem to be pulling away from other firms in the economy over time in terms of their return on capital. In particular, star firms in highly skilled industries seem to be pulling away from the others. However, conventional financial statements do not capitalize R&D expenditures or organiza- tional capital. Once we adjust firms’ returns to capital to address these shortcomings, we see that the differences in firm returns in highly skilled and other industries shrink dramatically and the gap between star firms and other firms does not widen over time. Furthermore, once we adjust markups based on operating expenses for investment in intangible capital, we do find an increase in market power especially in high skilled industries. Star firms tend to be larger, younger, and have higher markups. While they may have more pricing power than other firms, at each level of markup star firms tend to produce more than other firms. In a companion paper Ayyagari et al. [2018], we examine the influences of market power and human capital in an international sample of firms. Overall, our results indicate that there islittle evide nce that the most profitable 10% of firms are pulling away from the rest of the economy. However, there is reason for concern regarding a smaller subset of elite publicly-listed firms. The usual suspects for membership in such an elite group are Apple, Facebook, Google, Amazon, and Microsoft. When we examine these firms individually, the ROIC and markups of most of these elite firms do not seem extraordinary initially and then explode but again only for a couple of firms that have built up a large enough market. However, for these firms, the critical policy concern may not only be the regulation of their use of market power today, but also the need to maintain contestable markets that allow the creation of independent technologies in the future. 29 References Daron Acemoglu and David Autor. Skills, tasks and technologies: Implications for employment and earnings. In Handbook of labor economics, volume 4, pages 1043–1171. Elsevier, 2011. Daniel A Ackerberg. Timing assumptions and efficiency: Empirical evidence in a production func- tion context. University of Michigan mimeo, 2016. Daniel A Ackerberg, Kevin Caves, and Garth Frazer. Identification properties of recent production function estimators. Econometrica, 83(6):2411–2451, 2015. Lewis Alexander and Janice Eberly. Investment hollowing out. IMF Economic Review, 66(1):5–30, 2018. Dan Andrews, Chiara Criscuolo, Peter Gal, et al. Frontier firms, technology diffusion and public policy: Micro evidence from OECD countries, volume 2. OECD Publishing, 2015. John Asker, Joan Farre-Mensa, and Alexander Ljungqvist. Corporate investment and stock market listing: a puzzle? The Review of Financial Studies, 28(2):342–390, 2014. Andrew Atkeson and Patrick J Kehoe. Modeling and measuring organization capital. Journal of Political Economy, 113(5):1026–1053, 2005. David Autor, David Dorn, Lawrence F Katz, Christina Patterson, and John Van Reenen. The fall of the labor share and the rise of superstar firms. NBER Working Papers 23396, National Bureau of Economic Research, May 2017. URL http://www.nber.org/papers/w23396. David H Autor, Frank Levy, and Richard J Murnane. The skill content of recent technological change: An empirical exploration. The Quarterly Journal of Economics, 118(4):1279–1333, 2003. Meghana Ayyagari and Vojislav Maksimovic. Fewer and less skilled? human capital, competition, and entrepreneurial success in manufacturing. SSRN Working Paper Series, 2017. Available at SSRN: https://ssrn.com/abstract=3093443. Meghana Ayyagari, Asli Demirguc-Kunt, and Vojislav Maksimovic. Corporate inequality and hu- man capital accumulation. University of Maryland mimeo, 2018. 30 Jonathan B Baker and Steven C Salop. Antitrust, competition policy, and inequality. Geo. LJ Online, 104:1, 2015. Simcha Barkai. Declining labor and capital shares. Stigler Center for the Study of the Economy and the State New Working Paper Series, 2, 2016. e M Stulz. Why do us firms hold so much more cash Thomas W Bates, Kathleen M Kahle, and Ren´ than they used to? The Journal of Finance, 64(5):1985–2021, 2009. Itzhak Ben-David, John R Graham, and Campbell R Harvey. Managerial miscalibration. The Quarterly Journal of Economics, 128(4):1547–1584, 2013. Bruce A Blonigen and Justin R Pierce. Evidence for the effects of mergers on market power and efficiency. NBER Working Papers 22750, National Bureau of Economic Research, Inc, 2016. Nicholas Bloom and John Van Reenen. Measuring and explaining management practices across firms and countries. The Quarterly Journal of Economics, 122(4):1351–1408, 2007. Glenn W Boyle and Graeme A Guthrie. Investment, uncertainty, and liquidity. The Journal of Finance, 58(5):2143–2166, 2003. Carol A Corrado and Charles R Hulten. How do you measure a technological revolution? American Economic Review, 100(2):99–104, 2010. Arnaud Costinot, Lindsay Oldenski, and James Rauch. Adaptation and the boundary of multina- tional firms. The Review of Economics and Statistics, 93(1):298–308, 2011. Council of Economic Advisors. Benefits of competition and indicators of market power. Issue Brief, 2016. Nicolas Crouzet and Janice Eberly. Intangibles, investment, and efficiency. In AEA Papers and Proceedings, volume 108, pages 426–31, 2018. Colleen Cunningham, Florian Ederer, and Song Ma. Killer acquisitions. Mimeo, 2018. Aswath Damodaran. Dealing with cash, cross holdings and other non-operating assets: ap- proaches and implications. SSRN Working Paper Series, 2005. Available at SSRN: https://ssrn.com/abstract=841485. 31 Steven J Davis, John Haltiwanger, Ron Jarmin, Javier Miranda, Christopher Foote, and Eva Nagypal. Volatility and dispersion in business growth rates: Publicly traded versus privately held firms [with comments and discussion]. NBER macroeconomics annual, 21:107–179, 2006. Jan De Loecker and Jan Eeckhout. The rise of market power and the macroeconomic implications. NBER Working Paper 23687, National Bureau of Economic Research, August 2017. URL http: //www.nber.org/papers/w23687. Jan De Loecker and Frederic Warzynski. Markups and firm-level export status. American Economic Review, 102(6):2437–71, 2012. Jan De Loecker, Pinelopi K Goldberg, Amit K Khandelwal, and Nina Pavcnik. Prices, markups, and trade reform. Econometrica, 84(2):445–510, 2016. Amy Dittmar and Jan Mahrt-Smith. Corporate governance and the value of cash holdings. Journal of Financial Economics, 83(3):599–634, 2007. e M Stulz. The US listing gap. Journal of Financial Craig Doidge, G Andrew Karolyi, and Ren´ Economics, 123(3):464–487, 2017. e M Stulz. Eclipse of the public Craig Doidge, Kathleen M Kahle, G Andrew Karolyi, and Ren´ corporation or eclipse of the public markets? Journal of Applied Corporate Finance, 30(1):8–16, 2018. Andrea L Eisfeldt and Dimitris Papanikolaou. The value and ownership of intangible capital. American Economic Review, 104(5):189–94, 2014. Antonio Falato, Dalida Kadyrzhanova, and Jae Sim. Rising intangible capital, shrinking debt capacity, and the US corporate savings glut. FEDS Working Paper No.2013-67, 2013. Available at SSRN: https://ssrn.com/abstract=2350863. Michael Faulkender and Mitchell Petersen. Investment and capital constraints: repatriations under the american jobs creation act. The Review of Financial Studies, 25(11):3351–3388, 2012. Michael Faulkender, Kristine Hankins, and Mitchell Petersen. Understanding the rise in corporate 32 cash: Precautionary savings or foreign taxes. NBER Working Papers 23799, National Bureau of Economic Research, Inc, 2017. Murray Frank and Keer Yang. Does finance flow to high productivity firms? University of Min- nesota mimeo. Caroline Freund and Martha Denisse Pierola. Export superstars. The Review of Economics and Statistics, 97(5):1023–1032, 2015. Jason Furman and Peter Orszag. A firm-level perspective on the role of rents in the rise in inequality. Presentation at A Just Society Centennial Event in Honor of Joseph Stiglitz Columbia University, 2015. Amit Gandhi, Salvador Navarro, and David Rivers. How heterogeneous is productivity? a com- parison of gross output and value added. Journal of Political Economy, 2017. Xavier Giroud and Holger M Mueller. Does corporate governance matter in competitive industries? Journal of Financial Economics, 95(3):312–331, 2010. Henry G. Grabowksi and Dennis Mueller. Industrial research and development, intangible capital stocks, and firm profit rates. Bell Journal of Economics, 9(2):328–343, 1978. URL https: //EconPapers.repec.org/RePEc:rje:bellje:v:9:y:1978:i:autumn:p:328-343. Gustavo Grullon, Yelena Larkin, and Roni Michaely. Are US industries becoming more concentrated? SSRN Working Paper Series, 2017. Available at SSRN: https://ssrn.com/abstract=2612047. an Guti´ Germ´ errez and Thomas Philippon. Declining competition and investment in the US. NBER Working Papers 23583, National Bureau of Economic Research, Inc, 2017. Robert E Hall. New evidence on the markup of prices over marginal costs and the role of mega-firms in the us economy. NBER Working Papers 24574, National Bureau of Economic Research, Inc, 2018. Jarrad Harford, Sattar A. Mansi, and William F. Maxwell. Corporate governance and firm cash 33 holdings in the US. Journal of Financial Economics, 87(3):535–555, 2008. URL https:// EconPapers.repec.org/RePEc:eee:jfinec:v:87:y:2008:i:3:p:535-555. Jarrad Harford, Sandy Klasa, and William F. Maxwell. Refinancing risk and cash holdings. Journal of Finance, 69(3):975–1012, 2014. URL https://EconPapers.repec.org/RePEc:bla:jfinan:v: 69:y:2014:i:3:p:975-1012. Mark Hirschey. Intangible capital aspects of advertising and r&d expenditures. Journal of Industrial Economics, 30(4):375–90, 1982. URL https://EconPapers.repec.org/RePEc:bla:jindec:v: 30:y:1982:i:4:p:375-90. Michael Jensen. Agency costs of free cash flow, corporate finance, and takeovers. American Eco- nomic Review, 76(2):323–29, 1986. URL https://EconPapers.repec.org/RePEc:aea:aecrev: v:76:y:1986:i:2:p:323-29. Kathleen M. Kahle and Ren M. Stulz. Is the US public corporation in trouble? Journal of Economic Perspectives, 31(3):67–88, 2017. URL https://EconPapers.repec.org/RePEc:aea:jecper:v: 31:y:2017:i:3:p:67-88. ale Utar. International trade and job polarization: Evidence at the worker- Wolfgang Keller and Hˆ level. (22315), 2016. Lina M Khan. Amazon’s antitrust paradox. Yale LJ, 126:710, 2016. Tim Koller. What is value-based management? The McKinsey Quarterly, (3), 1994. Tim Koller, Marc Goedhart, and David Wessels. Valuation: Measuring and Managing the Value of Companies. JohnWiley & Sons, 6 edition, 2017. Mordecai Kurz. On the formation of capital and wealth. Stanford Institute for Economic Policy Research (SIEPR) Working Paper, 2017. Vojislav Maksimovic, Gordon Phillips, and Liu Yang. Do public firms respond to investment opportunities more than private firms? the impact of initial firm quality. NBER Working Papers 24104, National Bureau of Economic Research, Inc, 2017. URL https://EconPapers.repec.org/ RePEc:nbr:nberwo:24104. 34 J. Spencer Martin and Anthony M. Santomero. Investment opportunities and corporate demand for lines of credit. Journal of Banking & Finance, 21(10):1331–1350, 1997. URL https:// EconPapers.repec.org/RePEc:eee:jbfina:v:21:y:1997:i:10:p:1331-1350. Ellen R McGrattan and Edward C Prescott. Technology capital and the us current account. American Economic Review, 100(4):1493–1522, 2010. Robert Novy-Marx. An equilibrium model of investment under uncertainty. The Review of Finan- cial Studies, 20(5):1461–1502, 2007. G Steven Olley and Ariel Pakes. The dynamics of productivity in the telecommunications equipment industry. Econometrica, 64(6):1263–97, 1996. URL https://EconPapers.repec.org/RePEc: ecm:emetrp:v:64:y:1996:i:6:p:1263-97. Yoram Peles. Rates of amortization of advertising expenditures. Journal of Political Economy, 79 (5):1032–58, 1971. URL https://EconPapers.repec.org/RePEc:ucp:jpolec:v:79:y:1971:i: 5:p:1032-58. Ander Perez-Orive, Andrea Caggese, et al. Capital misallocation and secular stagnation. In 2017 Meeting Papers. Society for Economic Dynamics, 2017. Ryan H Peters and Lucian A Taylor. Intangible capital and the investment-q relation. Journal of Financial Economics, 123(2):251–272, 2017. Mitchell A Petersen. Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies, 22(1):435–480, 2009. David Powell. Quantile treatment effects in the presence of covariates. RAND Labor and Population Working Paper, 2016. Carl Shapiro. Antitrust in a time of populism. International Journal of Industrial Organization, page 135, 2018. James Traina. Is aggregate market power increasing? production trends using financial statements. Stigler Center Working Paper Series, 2018. 35 John Van Reenen and C Patterson. The rise of star firms has been better for investors than for employees. Harvard Business Review, 2017. Mahmut Yasar, Rafal Raciborski, and Brian Poi. Production function estimation in stata using the olley and pakes method. Stata Journal, 8(2):221–231, 2008. URL https:// EconPapers.repec.org/RePEc:tsj:stataj:v:8:y:2008:i:2:p:221-231. 36 Figure 1: Rise in Star Firms This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC) in each year across all public firms in the US economy.Detailed variable definitions are in the Appendix. 37 Figure 2: Differences in Human Capital This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC) in each year in low and high routine manual (RMAN) manufacturing industries. Detailed variable definitions are in the Appendix. 38 Figure 2: Differences in Human Capital (Continued....) This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC) in each year in low and high complex problem solving skill (CPS), and low and high cognitive skilled (NRCOG) manufacturing industries. Detailed variable definitions are in the Appendix. 39 Figure 3: Rise in Star Firms - correcting for intangible capital This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC T OT ) in each year across all public firms in the US economy. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. Detailed variable definitions are in the Appendix. 40 Figure 4: Differences in Human Capital - correcting for intangible capital This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC T OT ) in each year in low and high routine manual (RMAN) manufacturing industries. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. Detailed variable definitions are in the Appendix. 41 Figure 4: Differences in Human Capital - correcting for intangible capital (Continued....) This figure plots the 25th , 50th , 75th , and 90th percentile of Return on Invested Capital (ROIC T OT ) in each year in low and high complex problem solving skill (CPS), and low and high cognitive skilled (NRCOG) manufacturing industries. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. Detailed variable definitions are in the Appendix. 42 Figure 5: Markups based on different variable inputs OPEX* Markups are M arkupsT OT which use operating expenses with intangible capital adjust- ments as a measure of variable cost. OPEX Markups are M arkups which use operating expenses without intangible capital adjustments as a measure of variable cost. COGS Markups use cost of goods sold (COGS) as a measure of variable cost. 43 Figure 6: Markups in the US Economy This figure plots the 25th , 50th , 75th , and 90th percentile of M arkupsT OT in each year across all public firms in the US economy. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost. Detailed variable definitions are in the Appendix. 44 Figure 7: Markups in the US Economy - Differences in Human Capital This figure plots the 25th , 50th , 75th , and 90th percentile of M arkupsT OT in each year in low and high routine manual (RMAN) manufacturing industries.M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost in estimation of markups. Detailed variable definitions are in the Appendix. 45 Figure 8: Distribution of Markups across Star firms and all other firms This figure plots the histogram of M arkupsT OT for ROIC T OT Stars and all other firms. The figure also shows the smoothed values of a kernel-weighted local polynomial regression of ROIC T OT on M arkupsT OT . ROIC T OT stars are firms that are in the top 10% of ROIC T OT in a particular year. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost in estima- tion of markups. Detailed variable definitions are in the Appendix. 46 Figure 9: ROIC and Markups This figure plots the smoothed values of a kernel-weighted local polynomial regression of ROIC T OT on M arkupsT OT for ROIC T OT stars, all other firms, for firms that are ROIC T OT stars for at least three years over the period 2000-2015 and for firms that are ROIC T OT stars for at least five years over the period 2000-2015. ROIC T OT stars are firms that are in the top 10% of ROIC T OT in a particular year. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost in estimation of markups. Detailed variable definitions are in the Appendix. 47 Figure 10: Sales over Invested Capital - Stars vs all other firms This figure plots the smoothed values of a kernel-weighted local polynomial regression of Sales/Invested Capital on M arkupsT OT for ROIC T OT stars and all other firms. ROIC T OT stars are firms that are in the top 10% of ROIC T OT in a particular year. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost in estimation of markups. Detailed variable definitions are in the Appendix. 48 Figure 11: Output and Investment - ROIC Stars x Markups This figure plots the predicted margin graphs of the interaction effects in Table 6. ROIC T OT stars are firms that are in the top 10% of ROIC T OT in a particular year. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of variable cost in estimation of markups. Detailed variable definitions are in the Appendix. 49 Figure 12: ROIC of Elite Firms (Apple, Facebook, Amazon, Microsoft, Google) This figure plots the 90th percentile of Return on Invested Capital (ROIC T OT ) in each year across all public firms in the US economy as well as the ROIC T OT for five firms referred to as superstars anecdotally. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. Detailed variable definitions are in the Appendix. 50 Figure 13: Markups of Elite Firms (Apple, Facebook, Amazon, Microsoft, Google) This figure plots the 90th percentile of M arkupsT OT in each year across all public firms in the US economy as well as the M arkupsT OT for five firms referred to as superstars anecdotally. M arkupsT OT use operating expenses with intangible capital adjustments as a measure of vari- able cost in estimation of markups. Detailed variable definitions are in the Appendix. 51 Figure 14: Measurement of Excess Cash - Robustness This figure plots the 25th , 50th , 75th , and 90th percentile of alternate definitions of ROIC T OT in each year across all public firms in the US economy. The alternate definitions correspond to using cash above 1% of sales, 5% of sales, 10% of sales, and 20% of sales respectively as excess cash rather than the 2% of sales used in the rest of the paper. ROIC T OT includes the Peters and Taylor [2017] adjustment for intangible capital. Detailed variable definitions are in the Appendix. 52 Table 1: Summary Statistics This table reports the summary statistics of the key variables used in our analysis. All variable definitions are in the Appendix. Variable Obs Mean Std. Dev. Min Max ROIC Star 97,241 0.100 0.300 0.000 1.000 ROIC 97,241 -32.354 250.464 -2955.128 443.913 Markups 78,668 0.971 0.227 0.150 1.763 ROICTOT Star 93,438 0.100 0.300 0.000 1.000 ROICTOT 93,438 12.943 27.207 -129.511 150.167 Log(Assets) 93,438 5.448 1.988 -6.908 12.906 Log(Age) 93,438 2.666 0.744 1.099 4.205 MarkupsTOT 83,842 1.182 0.338 0.226 2.635 Market Share 91,683 0.014 0.036 0.000 0.323 HHI 91,966 0.093 0.082 0.028 0.596 RMAN 46,997 0.382 0.323 -0.050 1.380 CPS 46,997 0.318 0.352 -0.706 0.962 NRCOG 46,997 -0.181 0.300 -1.076 0.281 53 Table 2: Who are America’s Stars? This table reports estimates from the following regression model: ROIC Starijt = α0 + β1 × Log (Assets)ijt + β2 × Log (Age)it + β3 × HHIjt + β4 × M arket shareijt + β5 × M arkupsijt + φj + γt + εijt ROIC Star is a dummy variable that takes the value 1 if the firm i ’s ROIC is above the 90th percentile of ROIC across all firms in a particular year and 0 otherwise. Log(Assets) is the logarithm of total assets and Log(Age) is the number of years the firm is in the database. HHI is Herfindahl Index of market share in each industry in each year. Markups are estimated using operating expenses as a variable input of production. Market Share is the ratio of the firm i ’s sales to total industry j ’s sales in a particular year. Columns (1)-(5) include the full sample; column (6) is manufacturing sub-sample, column (7) is large firm sample (Real value of assets is ≥ USD 200million) and column (8) is sample of young firms (Age≤5 years). All regressions are estimated using ordinary least squares with standard errors clustered at the firm level. Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. (1) (2) (3) (4) (5) (6) (7) (8) ROIC Star ROIC Star ROIC Star ROIC Star ROIC Star ROIC Star ROIC Star ROIC Star Full Sample Full Sample Full Sample Full Sample Full Sample Manufac Large Firms Young Firms Log(Assets) 0.013*** 0.013*** 0.004*** 0.002 0.014*** 0.000 0.001 0.004 (0.001) (0.001) (0.001)(0.001) (0.002) (0.001) (0.004) (0.003) 54 Log(Age) -0.037*** -0.037*** -0.032*** -0.030*** -0.038*** -0.034*** -0.028*** -0.175*** (0.003) (0.003) (0.003)(0.003) (0.005) (0.003) (0.005) (0.036) HHI -0.022 -0.011 -0.026 0.056 -0.097 (0.031) (0.038) (0.087) (0.056) (0.089) Markups 0.280*** 0.287*** 0.348*** 0.325*** 0.301*** (0.015) (0.016) (0.024) (0.027) (0.023) Market Share -0.025 0.248*** 0.413** 0.374*** 0.297 (0.084) (0.096) (0.186) (0.119) (0.251) N 97241 97241 78668 96237 77875 41100 39539 6703 Adj. R-sq 0.059 0.059 0.081 0.058 0.082 0.061 0.120 0.089 Fixed Effects ——————————————————— Industry, Year ——————————————————— Table 3: Who are America’s Stars? Correcting for intangible capital This table reports estimates from the following regression model in panel A: T OT ROICijt = α0 + β1 × LogAssetsijt + β2 × LogAgeit + β3 × HHIjt + β4 × M arket shareijt + β5 × M arkupsijt + φj + γt + εijt ROIC T OT Star is a dummy variable that takes the value 1 if the firm i ’s ROIC T OT is above the 90th percentile of ROIC T OT across all firms in a particular year and 0 otherwise. Log(Assets) is the logarithm of total assets and Log(Age) is the number of years the firm is in the database. HHI is Herfindahl Index of market share in each industry in each year. M arkupsT OT are estimated using operating expenses as a variable input of production and include correction for intangible capital. Market Share is the ratio of firm i ’s sales to total industry j s sales in a particular year. In panel A, columns (1)-(5) include the full sample; column (6) is manufacturing sub-sample, column (7) is large firm sample (Real value of assets is ≥ USD 200Mil and column (8) is young firm sample (Age ≤ 5years). All regressions in Panel A are estimated using ordinary least squares with standard errors clustered at the firm level. Panel B presents generalized quantile regressions for the 25th , 50th , 75th , and 90th of ROIC T OT for specification (5) of panel A. Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. Panel A: OLS Estimation (1) (2) (3) (4) (5) (6) (7) (8) ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT Star Star Star Star Star Star Star Star Full Full Full Full Full Large Young 55 Manufac Sample Sample Sample Sample Sample Firms Firms Log(Assets) 0.022*** 0.022*** 0.017*** 0.018*** 0.022*** 0.017*** 0.006** 0.020*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) (0.003) (0.003) Log(Age) -0.086*** -0.086*** -0.078*** -0.077*** -0.088*** -0.069*** -0.075*** -0.206*** (0.003) (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) (0.023) HHI -0.021 -0.050 -0.000 -0.030 0.078 (0.038) (0.040) (0.109) (0.053) (0.127) Markups 0.149*** 0.149*** 0.124*** 0.175*** 0.198*** (0.010) (0.010) (0.012) (0.015) (0.017) Market Share 0.056 0.131 -0.107 0.374*** -0.146 (0.081) (0.085) (0.104) (0.104) (0.320) N 93438 91966 83842 91683 81950 41887 44611 9652 Adj. R-sq 0.089 0.089 0.099 0.089 0.099 0.068 0.129 0.105 Fixed Effects ——————————————————— Industry, Year ——————————————————— Table 3: Who are America’s Stars? Correcting for intangible capital (Continued...) Panel B: Generalized Quantile Estimation of Model (5) in Panel A (1) (2) (3) (4) ROICTOT ROICTOT ROICTOT ROICTOT Quantile 25 50 75 90 Markups 18.079*** 17.665*** 25.242*** 38.081*** (0.894) (1.103) (1.570) (1.920) N 83842 83842 83842 83842 Fixed Effects Industry, Year 56 Table 4: Human Capital and Star Status This table reports estimates from the following panel regression model: T OT ROICijt = α0 + β1 × LogAssetsijt × Skillj + β2 × LogAgeit + β3 × HHIjt × Skillj + β4 × M arket shareijt × Skillj + β5 × M arkupsijt × Skillj + φj + γt + εijt ROIC T OT Star is a dummy variable that takes the value 1 if the firm i s ROIC T OT is above the 90th percentile of ROIC T OT across all firms in a particular year t and 0 otherwise. Log(Assets) is the logarithm of total assets. HHI is Herfindahl Index of market share in each 3-digit NAICS industry in each year. M arkupsT OT are based on the estimation in De Loecker and Eeckhout (2017) using operating expenses as a variable input of production and include correction for intangible capital. Market Share is the ratio of the firm i s sales to total industry j s sales in a particular year. Skill is industry-level measure of routine manual skills (RMAN) in columns 1-4, complex problem solving skills (COMPLEXPS) in columns 5-8 and non-routine cognitive analytical skills (NRCOG) in columns 9-12. All regressions include industry and year fixed effects and are estimated using ordinary least squares with standard errors clustered at the firm level. Panel B presents generalized quantile regressions of models 2,6, and 10 of panel A for the 25th , 50th , 75th , and 90th quantile of ROIC T OT . Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT Star Star Star Star Star Star Star Star Star Star Star Star Skill RMAN CPS NRCOG Log(Assets) 0.020*** 0.017*** 0.017*** 0.017*** 0.014*** 0.017*** 0.017*** 0.017*** 0.018*** 0.017*** 0.017*** 0.017*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) 57 Log(Age) -0.069*** -0.069*** -0.069*** -0.070*** -0.069*** -0.070*** -0.070*** -0.070*** -0.069*** -0.070*** -0.070*** -0.070*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) HHI 0.034 0.036 0.363* 0.034 0.036 0.038 0.021 0.037 0.040 0.035 -0.096 0.037 (0.080) (0.080) (0.207) (0.080) (0.081) (0.081) (0.082) (0.083) (0.081) (0.080) (0.143) (0.082) Market Share -0.058 -0.122 -0.126 0.017 -0.037 -0.109 -0.110 -0.111 -0.078 -0.107 -0.106 -0.138 (0.108) (0.104) (0.103) (0.197) (0.105) (0.104) (0.104) (0.106) (0.104) (0.105) (0.105) (0.162) Markups 0.122*** 0.129*** 0.124*** 0.125*** 0.123*** 0.115*** 0.125*** 0.125*** 0.122*** 0.126*** 0.126*** 0.125*** (0.012) (0.012) (0.012) (0.012) (0.012) (0.014) (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) Log(Assets) x Skill -0.009*** 0.008*** 0.006* (0.002) (0.003) (0.003) Markups x Skill -0.024* 0.019* -0.009 (0.013) (0.012) (0.016) HHI x Skill -0.466* -0.095 -0.216 (0.238) (0.181) (0.210) Market Share x Skill -0.212 -0.005 -0.062 (0.234) (0.303) (0.335) N 42355 42355 42355 42355 42355 42355 42355 42355 42355 42355 42355 42355 Adj. R-sq 0.070 0.069 0.070 0.069 0.070 0.069 0.069 0.069 0.069 0.069 0.069 0.069 Fixed Effects —————Industry, Year————–— —————Industry, Year————–— —————Industry, Year————–— Table 5: Are Star Firms Persistent Performers? This table reports estimates from the following panel regression model: P erf ormanceijt = α0 + β1 × LogAssetsijt−5 + β2 × LogAgeijt−5 + β3 × HHIjt−5 + β4 × M arket shareijt−5 +β5 × M arkupsijt−5 + β6 × ROIC T OT Star[ ijt − 5]φj + γt + εijt Performance is Sales growth/Employment growth (each defined as the 5-year log difference in sales or employment respectively divided by 5), Labor Productivity, Tobins Q (QT OT ) or ROIC T OT averaged over 5 years. Log(Assets) is the 5-year lagged value of the logarithm of total assets. HHI is 5-year lagged value of Herfindhal Index of market share in each 3-digit NAICS industry. M arkupsT OT is the 5-year lagged value of Markups based on the estimation in De Loecker and Eeckhout (2017) using operating expenses as a variable input of production and include correction for intangible capital. Market Share is the 5-year lagged value of the ratio of the firm i s sales to total industry j s sales in a particular year. ROIC T OT Star is a dummy variable that takes the value 1 if firm i s 5-year lagged ROIC T OT was above the 90th percentile of ROIC T OT across all firms 5 years back and 0 otherwise. The regressions are 5-year stacked panel regressions: 1990-1995, 1995-2000, 2000-2005, 2005-2010, and 2010-2015 and include industry and year fixed effects with standard errors clustered at the firm level. Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. 1 2 3 4 5 5-year average 5-year change 5-year change 5-year average 5-year average ROIC T OT in Log(Sales) in Log(Empl) Labor Produc- Tobin’s QT OT tivity L5.ROICTOT Star -2.780*** 0.038*** 0.003 0.897 0.985*** 58 (0.611) (0.007) (0.007) (10.998) (0.060) L5.ROICTOT 0.633*** -0.000*** 0.001*** 1.459*** 0.003*** (0.009) (0.000) (0.000) (0.144) (0.001) L5.Market share -1.675 0.030 0.079** -4.537 0.734** (1.900) (0.040) (0.037) (83.102) (0.299) L5.MarkupsTOT 9.592*** 0.130*** 0.087*** 41.803** 0.793*** (0.546) (0.009) (0.007) (17.092) (0.058) L5.Log(Assets) 0.733*** -0.006*** -0.009*** 28.268*** 0.033*** (0.067) (0.001) (0.001) (2.157) (0.007) L5.Log(Age) 0.276* -0.025*** -0.018*** -39.109*** -0.191*** (0.144) (0.002) (0.002) (4.735) (0.015) L5.HHI -4.985** -0.011 0.025 202.495*** 0.153 (2.196) (0.035) (0.037) (60.034) (0.185) N 16392 10762 10415 16106 16001 Adj. R-sq 0.732 0.092 0.088 0.402 0.227 Fixed Effects —————————–Industry, Year————————–— Table 6: Output and Investment in Star Firms This table reports estimates from the following panel regression model: Output or P erf ormanceijt = α0 + β1 × LogAssetsijt−1 + β2 × LogAgeijt−1 + β3 × M arkupsT OT ijt−1 +β4 × ROIC T OT Starijt−1 + β5 × M arkupsT OT ijt−1 × ROIC T OT Starijt−1 φj + γt + εijt The dependent variable is Sales/Invested Capital or CAPEX/Invested Capital or XRD/Invested Capital. ROIC T OT Star is a dummy variable that takes the value 1 if the firm i s ROIC T OT is above the 90th percentile of ROIC T OT across all firms in a particular year t and 0 otherwise. Log(Assets) is the logarithm of total assets. M arkupsT OT are based on the estimation in De Loecker and Eeckhout (2017) using operating expenses as a variable input of production and include correction for intangible capital. QT OT is Tobins Q corrected for intangible capital. All regressions include industry and year fixed effects and are estimated using ordinary least squares with standard errors clustered at the firm level. Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. (1) (2) (3) (4) (5) (6) Sales/Invested Sales/Invested Capex/InvestedCapex/InvestedXRD/Invested XRD/Invested Capital Capital Capital Capital Capital Capital L. ROICTOT Stars 0.804*** 1.378*** 0.014*** 0.038*** -0.008*** -0.028*** (0.028) (0.093) (0.002) (0.006) (0.002) (0.006) L. MarkupsTOT -1.052*** -0.992*** -0.001 0.001 0.048*** 0.046*** (0.039) (0.039) (0.003) (0.003) (0.003) (0.003) 59 L. ROICTOT Stars x L.MarkupsTOT -0.452*** -0.019*** 0.016*** (0.062) (0.004) (0.005) L.Log(Assets) 0.058*** 0.057*** 0.004*** 0.004*** -0.001*** -0.001*** (0.005) (0.005) (0.000) (0.000) (0.000) (0.000) L.Log(Age) -0.014 -0.013 -0.008*** -0.008*** -0.012*** -0.012*** (0.013) (0.013) (0.001) (0.001) (0.001) (0.001) L.QTOT -0.021*** -0.019*** 0.008*** 0.008*** 0.006*** 0.006*** (0.005) (0.005) (0.000) (0.000) (0.000) (0.000) N 70094 70094 69582 69582 70667 70667 Adj. R-sq 0.355 0.357 0.359 0.360 0.465 0.466 Fixed Effects — ———————————- Industry, Year ———————————- — Table 7: Human Capital and Star Status Robustness: Tobin’s Q This table reports estimates from the following panel regression model: QT OT ijt = α0 + β1 × LogAssetsijt × Skillj + β2 × LogAgeit + β3 × HHIjt × Skillj + β4 × M arket shareijt × Skillj + β5 × M arkupsijt × Skillj + φj + γt + εijt QT OT Star is a dummy variable that takes the value 1 if the firm i s Tobin’s QT OT is above the 90th percentile of QT OT across all firms in a particular year t and 0 otherwise. Log(Assets) is the logarithm of total assets. HHI is Herfindahl Index of market share in each 3-digit NAICS industry in each year. M arkupsT OT are based on the estimation in De Loecker and Eeckhout (2017) using operating expenses as a variable input of production and include correction for intangible capital. Market Share is the ratio of the firm i s sales to total industry j s sales in a particular year. Skill is industry-level measure of routine manual skills (RMAN) in columns 1-4, complex problem solving skills (COMPLEXPS) in columns 5-8 and non-routine cognitive analytical skills (NRCOG) in columns 9-12. All regressions include industry and year fixed effects and are estimated using ordinary least squares with standard errors clustered at the firm level. Detailed variable definitions are in the Appendix. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. 1 2 3 4 5 6 7 8 9 10 11 12 QTOT QTOT QTOT QTOT QTOT QTOT QTOT QTOT QTOT QTOT QTOT QTOT Star Star Star Star Star Star Star Star Star Star Star Star Skill RMAN CPS NRCOG Log(Assets) 0.005*** 0.002 0.002 0.002 -0.000 0.002 0.002 0.002 0.004** 0.002 0.002 0.002 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Log(Age) -0.065*** -0.065*** -0.065*** -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.065*** -0.066*** -0.066*** -0.066*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) HHI 0.083 0.086 0.467** 0.079 0.085 0.087 0.104 0.091 0.092 0.089 0.068 0.088 60 (0.074) (0.075) (0.190) (0.074) (0.076) (0.075) (0.077) (0.077) (0.075) (0.075) (0.142) (0.076) Market Share 0.268* 0.203 0.201 0.441 0.279* 0.222 0.219 0.229 0.277* 0.211 0.221 0.132 (0.153) (0.149) (0.148) (0.309) (0.149) (0.149) (0.149) (0.153) (0.150) (0.151) (0.149) (0.188) Markups 0.162*** 0.171*** 0.164*** 0.165*** 0.164*** 0.154*** 0.165*** 0.165*** 0.159*** 0.164*** 0.165*** 0.165*** (0.016) (0.016) (0.016) (0.016) (0.016) (0.018) (0.016) (0.016) (0.017) (0.017) (0.016) (0.016) Log(Assets) x Skill -0.009*** 0.007*** 0.010*** (0.002) (0.003) (0.003) Markups x Skill -0.032** 0.024* 0.021 (0.015) (0.013) (0.018) HHI x Skill -0.544** 0.118 -0.027 (0.221) (0.197) (0.226) Market Share x Skill -0.373 -0.298 -0.202 (0.306) (0.409) (0.387) N 41085 41085 41085 41085 41085 41085 41085 41085 41085 41085 41085 41085 adj. R-sq 0.072 0.071 0.072 0.071 0.071 0.071 0.071 0.071 0.072 0.071 0.071 0.071 Fixed Effects —————Industry, Year————–— —————Industry, Year————–— —————Industry, Year————–— Table 8: Human Capital and Star Status Robustness: Measurement of Excess Cash This table reports estimates from the following panel regression model: T OT ROICijt = α0 + β1 × LogAssetsijt × Skillj + β2 × LogAgeit + β3 × HHIjt × Skillj +β4 × M arket shareijt × Skillj + β5 × M arkupsijt × Skillj + φj + γt + εijt ROIC T OT Star is a dummy variable that takes the value 1 if the firm i s ROIC T OT is above the 90th percentile of ROIC T OT across all firms in a particular year t and 0 otherwise. Log(Assets) is the logarithm of total assets. HHI is Herfindahl Index of market share in each 3-digit NAICS industry in each year. M arkupsT OT are based on the estimation in De Loecker and Eeckhout (2017) using operating expenses as a variable input of production and include correction for intangible capital. Market Share is the ratio of the firm i s sales to total industry j s sales in a particular year. Skill is industry-level measure of routine manual skills (RMAN) in columns 1-4, complex problem solving skills (COMPLEXPS) in columns 5-8 and non-routine cognitive analytical skills (NRCOG) in columns 9-12. All regressions include industry and year fixed effects and are estimated using ordinary least squares with standard errors clustered at the firm level.In Panel A, ROIC T OT Star is computed using 1% of sales as excess cash, in panel B ROIC T OT Star is computed using 10% of sales as excess cash and in panel C we use non-cash working capital in computing ROIC T OT Star firms. All regressions include industry and year fixed effects and are estimated using ordinary least squares with standard errors clustered at the firm level. (∗∗∗ ), (∗∗ ), (∗ ) denote statistical significance at 1%, 5%, and 10% levels respectively. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT Star Star Star Star Star Star Star Star Star Star Star Star Skill RMAN CPS NRCOG Log(Assets) 0.021*** 0.018*** 0.018*** 0.017*** 0.015*** 0.018*** 0.017*** 0.017*** 0.019*** 0.017*** 0.017*** 0.017*** 61 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Log(Age) -0.068*** -0.068*** -0.068*** -0.069*** -0.068*** -0.069*** -0.069*** -0.069*** -0.068*** -0.069*** -0.069*** -0.069*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) HHI 0.054 0.056 0.394* 0.052 0.055 0.057 0.029 0.056 0.059 0.055 -0.106 0.056 (0.085) (0.085) (0.216) (0.085) (0.086) (0.086) (0.088) (0.088) (0.085) (0.085) (0.153) (0.087) Market Share -0.117 -0.183* -0.189* 0.045 -0.100 -0.170 -0.171 -0.173 -0.139 -0.166 -0.166 -0.141 (0.113) (0.110) (0.108) (0.213) (0.110) (0.110) (0.109) (0.112) (0.110) (0.110) (0.110) (0.174) Markups 0.123*** 0.130*** 0.124*** 0.126*** 0.124*** 0.118*** 0.126*** 0.126*** 0.122*** 0.127*** 0.127*** 0.126*** (0.013) (0.013) (0.012) (0.013) (0.013) (0.014) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log(Assets) x Skill -0.010*** 0.008*** 0.006* (0.003) (0.003) (0.003) Markups x Skill -0.023* 0.016 -0.012 (0.014) (0.012) (0.017) HHI x Skill -0.481* -0.161 -0.263 (0.249) (0.190) (0.222) Market Share x Skill -0.363 0.042 0.070 (0.250) (0.331) (0.362) N 39651 39651 39651 39651 39651 39651 39651 39651 39651 39651 39651 39651 Adj. R-sq 0.070 0.069 0.069 0.069 0.070 0.069 0.069 0.069 0.069 0.069 0.069 0.069 Fixed Effects —————Industry, Year————–— —————Industry, Year————–— —————Industry, Year————–— Table 8: Human Capital and Star Status Robustness: Measurement of Excess Cash (Continued...) Panel B: ROIC T OT computed using 10% of sales as excess cash (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT ROIC T OT Star Star Star Star Star Star Star Star Star Star Star Star Skill RMAN CPS NRCOG Log(Assets) 0.021*** 0.018*** 0.018*** 0.017*** 0.015*** 0.018*** 0.017*** 0.017*** 0.018*** 0.017*** 0.017*** 0.017*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Log(Age) -0.069*** -0.069*** -0.069*** -0.070*** -0.070*** -0.070*** -0.070*** -0.070*** -0.070*** -0.070*** -0.070*** -0.070*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) HHI 0.067 0.069 0.481** 0.065 0.068 0.071 0.054 0.069 0.072 0.068 -0.025 0.069 (0.082) (0.083) (0.211) (0.082) (0.083) (0.083) (0.085) (0.085) (0.083) (0.082) (0.152) (0.084) Market Share -0.091 -0.162 -0.169 0.076 -0.077 -0.147 -0.148 -0.149 -0.119 -0.142 -0.144 -0.133 (0.113) (0.108) (0.107) (0.229) (0.110) (0.108) (0.108) (0.109) (0.109) (0.109) (0.109) (0.210) Markups 0.131*** 0.138*** 0.132*** 0.134*** 0.132*** 0.125*** 0.134*** 0.134*** 0.131*** 0.135*** 0.135*** 0.134*** (0.013) (0.013) (0.013) (0.013) (0.013) (0.015) (0.013) (0.013) (0.014) (0.013) (0.013) (0.013) Log(Assets) x Skill -0.010*** 0.008*** 0.005 (0.003) (0.003) (0.003) Markups x Skill -0.026* 0.018 -0.013 (0.014) (0.012) (0.017) HHI x Skill -0.588** -0.091 -0.154 62 (0.239) (0.193) (0.225) Market Share x Skill -0.375 0.035 0.034 (0.275) (0.384) (0.415) N 39659 39659 39659 39659 39659 39659 39659 39659 39659 39659 39659 39659 Adj. R-sq 0.072 0.070 0.071 0.070 0.071 0.070 0.070 0.070 0.070 0.070 0.070 0.070 Fixed Effects —————Industry, Year————–— —————Industry, Year————–— —————Industry, Year————–— Table 8: Human Capital and Star Status Robustness: Measurement of Excess Cash (Continued...) Panel C: ROICTOT computed using no excess cash 1 2 3 4 5 6 7 8 9 10 11 12 Cash Cash Cash Cash Cash Cash Cash Cash Cash Cash Cash Cash ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT ROICTOT Star Star Star Star Star Star Star Star Star Star Star Star Skill RMAN CPS NRCOG Log(Assets) 0.021*** 0.018*** 0.018*** 0.017*** 0.015*** 0.018*** 0.018*** 0.017*** 0.018*** 0.017*** 0.018*** 0.017*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Log(Age) -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.066*** -0.067*** -0.066*** -0.066*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) HHI 0.079 0.082 0.497** 0.075 0.081 0.082 0.065 0.078 0.084 0.079 0.005 0.079 (0.087) (0.087) (0.208) (0.086) (0.088) (0.087) (0.090) (0.090) (0.087) (0.087) (0.152) (0.089) Market Share 0.036 -0.027 -0.037 0.282 0.049 -0.017 -0.017 -0.022 0.007 -0.008 -0.014 0.077 (0.127) (0.122) (0.121) (0.260) (0.123) (0.122) (0.122) (0.125) (0.122) (0.123) (0.123) (0.214) Markups 0.070*** 0.076*** 0.071*** 0.073*** 0.071*** 0.067*** 0.073*** 0.073*** 0.071*** 0.075*** 0.074*** 0.073*** (0.010) (0.010) (0.010) (0.010) (0.010) (0.012) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Log(Assets) x Skill -0.009*** 0.007*** 0.004 (0.002) (0.003) (0.003) Markups x Skill -0.019 0.012 -0.020 (0.012) (0.010) (0.015) HHI x Skill -0.593** -0.101 -0.127 63 (0.234) (0.189) (0.229) Market Share x Skill -0.499* 0.205 0.216 (0.294) (0.381) (0.411) N 42384 42384 42384 42384 42384 42384 42384 42384 42384 42384 42384 42384 Adj. R-sq 0.051 0.050 0.051 0.050 0.051 0.050 0.049 0.049 0.050 0.050 0.049 0.049 Fixed Effects —————Industry, Year————–— —————Industry, Year————–— —————Industry, Year————–— Figure A1: Rise in Star Firms - Large Firm Sample Figure A2: COGS are a declining share of firm variable costs 64 Figure A3: ROIC T OT of Top 100 and Top 150 firms 65 Table A1: Variable Definitions Variables Definition Invested Capital Invested Capital = PPENT + ACT + INTAN - LCT - GDWL - max(CHE-0.02 x SALE, 0) where PPENT is Net Property, Plant, and Equipment, ACT is Current Assets, IN- TAN is Total Intangible Assets (excluding Goodwill), LCT is Current Liabilities, GDWL is Goodwill that represents the excess cost over equity of an acquired com- pany, CHE is Cash and Short-term Investments, and SALE is net sales/turnover. This definition does not include the Peters and Taylor (2017) correction for intangi- ble capital ROIC (EBITt+AMt)/Invested Capitalt-1 where EBIT is Earnings before Interest and Taxes and AM is Amortization of Intangibles ROIC Star Dummy variable that takes the value 1 if the firms ROIC is above the 90th percentile of ROIC across all firms in the US economy in a particular year and 0 otherwise. SGA SGA= XSGA-XRD-RDIP where XRD is Research and Development Expense, RDIP is in-process R&D expense, XSGA is Selling, General, and Administrative Expense Invested CapitalTOT Invested CapitalTOT = PPENT + ACT + ICAP - LCT - GDWL - max(CHE-0.02 x SALE, 0) where PPENT is Net Property, Plant, and Equipment, ACT is Current Assets. ICAP is defined as the sum of externally purchased intangible capital (INTAN) and inter- nally purchased intangible capital, both measured at replacement cost. Internally purchased intangible capital is in turn measured as the sum of knowledge capi- tal (K int know) and organization capital (K int org). LCT is Current Liabilities, GDWL is Goodwill that represents the excess cost over equity of an acquired com- pany, CHE is Cash and Short-term Investments, and SALE is net sales/turnover. ROICTOT ROICTOT = (EBIT + AM + XRD + 0.3 x SGA - δRD x K int know - δSGA x K int org)/Lagged value of Invested CapitalTOT where EBIT is Earnings before Interest and Taxes, AM is Amortization of Intangibles, XRD is Research and Development Expense, δRD is the depreciation rate associated with knowledge capital and is set to 15% following Peters and Taylor (2017) and δSGA is the depreciation rate associated with organization capital and is set to 20% following Falato, Kadyrzhanova, and Sim (2013) and Peters and Taylor (2017). K int know and K int org are the firms intangible capital replacement cost and organization capital replacement cost respectively from Peters and Taylor (2017) ROICTOT Star Dummy variable that takes the value 1 if the firms ROICTOT is above the 90th percentile of ROICTOT across all firms in the US economy in a particular year and 0 otherwise Markups Markups following the estimation in De Loecker and Eeckhout (2017) using Operating Expenses (OPEX) as a variable input MarkupsTOT Markups following the estimation in De Loecker and Eeckhout (2017) using Operating Expenses* (OPEX*) as a variable input where OP EX ∗ = OP EX − XRD − RDIP − 0.3 × XSGA where OPEX is Total Operating Expenses, XRD is Research and Devel- opment Expense, RDIP is in-process R&D expense, XSGA is Selling, General, and Administrative Expense COGS Markups Markups following the estimation in De Loecker and Eeckhout (2017) using Cost of Goods Sold (COGS) as a variable input Log(Assets) Logarithm of total assets Log(Age) Log(1+Firm Age) where Firm Age is the number of years the firm has appeared in Compustat 66 Table A1: Variable Definitions Variables Definition Market share Ratio of firm i’s sales to total industry j’s sales in a particular year HHI Herfindahl-Hirschman Index defined as the sum of squares of the market shares of the firms within each 3-digit NAICS industry Sales/IC Sales/Invested CapitalTOT Capex/IC Capital Expenditures/Invested CapitalTOT Tobin’s QTOT QTOT = V/TOTCAP where V is the market value of the firm defined as the market value of equity (=total number of common shares outstanding (Compustat item CSHO) times closing stock price at the end of the fiscal year (Compustat item PRCC F) plus the book value of debt (Compustat items DLTT + DLC) minus the firms current assets (Compus- tat item ACT) which includes cash, inventory, and marketable securities. TOTCAP is sum of Property, Plant and Equipment (Compustat item PPENT) and Intangi- ble Capital (ICAP). ICAP is defined as the sum of externally purchased intangible capital (INTAN) and internally purchased intangible capital, both measured at re- placement cost. Internally purchased intangible capital is in turn measured as the sum of knowledge capital (K int know) and organization capital (K int org). CPS Identifying complex problems and reviewing related information to develop and eval- uate options and implement solutions.Source: O*NET NRCOG Mathematical Reasoning + Inductive Reasoning + Developing Objectives and Strate- gies + Making Decisions and Solving Problems. Source: O*NET RMAN Spend time making repetitive motions + Pace Determined by Speed of Equipment + Manual Dexterity + Finger Dexterity. Source: O*NET 67