Dynamic Measure of Competition and the Dispersion of. Corporate Policy and Performance

Size: px
Start display at page:

Download "Dynamic Measure of Competition and the Dispersion of. Corporate Policy and Performance"

Transcription

1 Dynamic Measure of Competition and the Dispersion of Corporate Policy and Performance Margaret Rui Zhu 1 City University of Hong Kong November 2016 Preliminary, Please do not circulate Abstract This paper develops a dynamic measure of industry competition - Dynamic Market Share (DMS), which is the sum of absolute changes of market share for firms in an industry. A larger value of the DMS measure captures more intensified competition activities within the industry. The DMS measure has low correlation with the commonly used competition measures such as Herfindahl and Lerner Index. The DMS tends to increase when external competition is stronger, proxy by imports growth shock and tariff cuts. Generally, industries with high competition intensity have high cash holding and leverage, low profitability and valuation. More interestingly, based on the DMS measure, we find that the dispersion of corporate policies, such as cash holdings, leverage and cash flow, leads to greater competition activities and higher competition leads to larger performance dispersion in the industry. 1 Contact: Margaret.zhu@cityu.edu.hk. The author thanks the discussant, Jefferson Duarte, and conference participants at 2014 CityU Finance Conference for useful comments. The paper is preliminary and is part of the work in progress project with Tao Shu from University of Georgia and Sheridan Titman from University of Texas at Austin. Please do not cite or circulate this version.

2 1 Introduction There is long standing literature on how product market competition is related to innovation 2, financial constraints 3, managerial incentives 4 and firm performance. 5 Generally, there are two components in industry competition: entry barrier and competitive interactions between existing firms. The existing measures of industry competition emphasize on the existing market power to deter future entry (i.e. Herfindahl Index 6 ) and current pricing strategies, such as Lerner Index (Price- Cost Margin). To complement the existing literature, the paper proposes a dynamic measure to capture the competition activities inside of an industry: Dynamic Market Share (DMS). DMS j,t measure is calculated as the sum of absolute 7 change of market share for all firms in industry j from year t-1 to t. This measure therefore captures how the market shares distribution changes from time t-1 to t, directly evaluating market competition intensity inside of an industry. A larger value of the DMS measure suggests a bigger change in how the market share is distributed within industry, thus indicating a strengthened competition, and a greater pressure faced by managers from existing rivals. The sample consists of all Compustat firms excluding commodity producers, Utilities, Financials, Education and Public Administrative industry, covering 7,632 industry-year observations over the period from 1977 to The DMS measures use segment-level data and adjust for market share 2 Aghion et al (2005) summarizes this line of literature. 3 See Fresard (2007), Phillips (1995) and others. 4 See Aggarwal et al (1999), Raith (2003) and others. 5 See Nickell (1996) and others. 6 Other less used competition measures include number of firms, four firm concentration ratio, and price elasticity, etc. Will be discussed in Section 5. 7 Sum of square term of market share changes is also used as robust. The results do not change significantly. 1

3 changes due to merger and acquisition. The time-series patterns of DMS measure vary across industries. Some industries experience large waves over the time period of 1977 to 2013, such as Machinery, Electronics and Communications in 1990s and 2000s. Other industries are relatively stable over time, such as Chemicals, Rubber, Plastics and Leather. 8 Overall, the DMS measure is higher for the periods of and , is relative low for the periods of , , and The general patterns are consistent with the economic cycles. The DMS measure does not have high correlation with Herfindahl index (-0.204) or Lerner index (-0.096). The DMS measure is positively associated with imports growth and negatively correlated with the growth of the tariff rate in the industry, which suggests that the competition between existing firms will intensify when they face higher outside competition. Generally, we find that industries with lower average ROA, but higher cash holding and leverage are more likely to have higher competition dynamics. More interestingly, these dynamic industries tend to have higher dispersion in financial policies, such as cash holding, leverage and cash flow at previous period. In addition, we also find that high competition activities lead to future dispersion of profitability and performance within the industry. Firms who just experience high level of competition tend to have higher level of leverage, low level of cash holdings and cash flow. In order to detect the dynamics between the DMS measure and the dispersion of corporate policies and performance, we use panel vector autoregression technique. The impulse response functions show a strong reaction of the dispersion of investment policies capital expenditure, R&D and advertising to the impulse of DMS measure. 8 See Section 2 and Figure 2 for detailed industry definition and time pattern of the DMS measure. 2

4 There are many papers using measures of competition in the empirical IO literature. High concentration, measured by Herfindahl index, is used as an indication of weak competition and leads to high prices and high price-cost margins. 9 The Lerner index (Price-Cost-Margin) is another traditional measure of competition. Papers like Aghion et al. (2005) and Nickell (1996) calculate it directly as the profits-sales ratio. Others calculate the optimal PCM for each firm based on estimated demand and cost functions. 10 Corts (1999) criticize the PCM measure and shows that transitory demand shocks could lead to overestimation of competition intensity. In economics literature, people also use factor price elasticity (PE), first introduced by Panzar and Rosse (1987) to measure industry competition. The idea is that if a firm can pass through the input price increase to sales price, it has certain market power. The lower the price elasticity, the more competitive the market is. However, similar to PCM, the PE measure only applies to single manufacturing product line with identifiable input prices, which makes it hard to estimate using standard dataset. The paper complements to the literature by providing a dynamic measure of competition interactions between existing firms within an industry. The DMS measure can be adopted in any industry competing for market shares, not limited to the manufacturing industries with defined input-output data as required by Lerner index and price elasticity. A large finance and economics literature analyzes how competition affects firm performance, innovation and financial policies. Nickell (1996) shows that more intense competition, measured by PCM, leads to more innovation and higher firm performance. Aghion et al. (2005) find an inverted- U relationship between competition and innovation using U.K. data and PCM measurement. Giroud 9 See Scherer and Ross (1990) for an overview. 10 Examples here include Berry et al. (1995), Hausman et al. (1994) and Nevo (2001). 3

5 and Mueller (2011) find that weak governance firms have poor performance only in noncompetitive industries. Fresard (2010) uses shocks from import tariff to measure competition and finds that large cash reserves lead to future market share gains. The results is consistent with my findings that firms tend of have higher cash holdings and large variation in cash holdings in industry with active competition measures. Haushalter, Klasa and Maxwell (2007) argue that, when deciding upon their optimal amount of cash, firms take into account the product market competition, measured indirectly using proxy for predation risk. With the proposed dynamic measures of industry competition, we can directly examine the relationship between firm performance, policies, stock returns and the internal industry competition. While the literature mainly focus on how competition affects the mean value of the firms investment and performance, the dispersion of corporate policies and performance are less embraced. The paper also contributes to the literature by providing new evidence on how industry competition is related to the dispersion of corporate policies and performance. The rest of the paper is organized as follows: section 2 describes the data; Section 3 develops the measures; Section 4 examines the association between internal competition activities and external competition shocks; Section 5 analyzes the relationship between the dynamic measures of competition and the dispersion of corporate policies and performance; Section 6 concludes with a discussion. 4

6 2. Data The paper uses standard data sources. The firm and industry characteristics data comes from Compustat. The data used to develop the dynamic measures of competition are from Compustat and Compustat-Segments data. I exclude firms with negative values in Total Assets, Total Liabilities, Sales, COGS, or Cash to avoid data error. To identify industries with meaningful competition, I excludes commodity producers (SIC ), financial industries (SIC ), Utilities (SIC ), Public Administration industry (SIC ) and industries with only one firm. To obtain enough dynamics, I exclude firms with less than three years of data and firms make more than two times of industry switches within five years. 11 The final sample consists of 122,748 firmyear observations and 7,556 industry-year observations from 1977 to The definitions of all variables are reported in Appendix A. The paper defines industries mainly at four-digit SIC level with adjustment based on business nature and links in market competition. We excludes SIC industry code which is too broad and consist of more than one competing industries, for example, Beverages (SIC 2080) 12, Chemicals & Allied Products (SIC 2800) and other Miscellaneous industries. We combine four-digit SIC industries which have different SIC code but compete in the same market, for example, Eating and Drinking Place (SIC 5810) and Eating Place (SIC 5812), Food Stores (SIC 5400), Grocery Store (SIC 5411) and Convenience Store (SIC 5412). The industry-by-industry report in Appendix B is 11 Industry switch is identified when firms report different three-digit SIC code at time t, compared to t-1. Multiple industry switches within five years indicate data error and add noises to dynamic competition measures. 12 Beverages (SIC 2080) contains both non-alcohol and alcohol beverages, which do not share the same consumer base. 5

7 presented by two-digit SIC industries groups. 13 There are two popular alternative industry classifications in the literature. Fama-French 48 industry classification defines industry based on the product and correlated effects on asset prices. It classifies vertical industries together, such as Forestry, Lumber products and Hard surface floor. It might be preferable in asset pricing analysis because the vertical industries tend to experience same shocks and have correlated stock returns. However, it is not desirable to examine industry competition effects because the vertical industries do not compete with each other. For example, when Lumber industry experiences high level of competition, its market share distribution changes and the margin variation decreases. But the downstream industry - Hard Floor products may not have the same intensified competition. Instead, the Hard Floor industry could possibly have decreased level of competition because the decreased cost resulting from the increased competition of upstream lumber industry. Another alternative industry classification is Text based industry classification (TNIC) developed by Hoberg and Phillips (2013). The TNIC is a pair wise identification of how closely two firms products relate to each other. It is useful to analyze, at the firm level, how one firm s activities are related to its competitors. However, because TNIC is a pair wise classification, it does not have a clear definition of industry, which makes it hard to calculate market share at industry level or analyze activities within or across industry. For example, according to TNIC classification, firm A is a rival to firm B and firm B is a rival to firm C, but firm A is not necessarily a rival to firm C. 13 See Appendix for details at industry groups. 6

8 The imports tariff data and import penetration data is from Peter Schott website. 14 The sample period for Tariff growth data is from 1974 to 2005 and 1993 to 2007 for Imports growth data. The paper also used merger and acquisition data covering to adjust for market share changes due to M&A. The data comes from Thomason ONE (SDC) database of Thomason Reuters. 3 Dynamic Measures of Competition 3.1 Adjust for Business Segments and M&As When a sample firm has multiple segments, the firm s SIC captures only its main segment. We therefore use the segment-level sales data to calculate competition measures for an industry. That is, to calculate DMS for an industry, we use all the single segment firms in this industry together with the segment-level data of conglomerate firms segments in this industry. Additionally, merger and acquisition activities can also lead to a mechanical change in market share without actual competition activities. We adjust the merger and acquisition activities in SDC as follows: when firm i in our sample acquired a private firm, or acquired a public firm in a different industry in year y, then firm i' s sales is contaminated, and therefore firm i was dropped from the calculation of the industry competition measures in year y. For firm i that acquired a public firm j in the same industry in year y, then we add firm j s sales to firm i in year y-1. That is, we treat firms i and j as one firm in the year prior to merger

9 3.2 Dynamic Market Share (DMS) The DMS measure is calculated as follows n DMS j,t = MS ij,t -MS ij,t-1 i=1 where DMS j,t is the dynamic measure of competition for industry j at time t. MSij,t is the market share for firm i in industry j at time t. The market share is defined as sales of firm i divided by total sales of all the Compustat firms in the same defined industry. 15 Table 1 defines the main variables used in this paper. The DMS measure captures the changes of Market Share distribution from t-1 to t. A larger value of the DMS measure suggests a bigger change in how the market share is distributed within industry, thus indicating an intensified competition. The higher the changes in the market share distribution, the more intensive competition the industry has been through. Theoretically, the measure could range from 0 to 2. In the sample, the DMS measure range from 0 to The summary statistics of the measure are reported in Table 1. Table 1 reports the summary statistics of different competition measures at industry level. Panel A shows the statistics for the full sample and panel B for manufacturing industries only. There are 7,632 industry year observations in the main sample and 4,536 observations for manufacturing industries only. The median number of firms in each industry is 11 for the main sample and 10 for manufacturing industries. The mean value of DMS is The magnitude is equivalent to the case where one firm losing 11.2% in market share and the other firms gaining 11.2%. The measure shows 15 I use total industry sales from Bureau of Economic Analysis as a robustness check. The results do not differ significantly. 8

10 a significant skewness of 3.40 with a median value of 0.087, which indicates that there are a small number of industry-years experiencing especially high level of competition. The observation is intuitively consistent with the notion that most of the industries at normal times are stable, while some industries are more dynamic in certain periods. Panel C and D in table 1 presents the summary statistics for subsample period of and , respectively. DMS measures are not different in mean, but small increase in standard deviations from early to later period. The time-series patterns of DMS measure vary across industries. See Figure 2 for plots for selected industries. Some industries experience large waves over the time period of 1977 to 2013, such as Machinery, Electronics and Communications in 1990s and 2000s. Other industries are relatively stable over time, such as Chemicals, Rubber, Plastics and Leather. Overall, the DMS measure is higher for the periods of and , is relative low for the periods of , , and The general patterns are consistent with the economic cycles. We examine the transition probabilities of the DMS measure in one-year, three-year and fiveyear period. The results are reported in Table 3. It shows that the stable industries with the least competition activities are sticky for several years, around 45% likely to stay as stable industry within three years. Industries in the middle three quintiles are more likely to move around, but slightly tend to stay in the middle quintiles. The dynamic industries with the most competition activities are less sticky than stable industries, but more than 50% will stay in quintile 4 and 5 for one- and three-year period. 9

11 3.3 Comparison with Other Competition Measures The dynamic competition measure, DMS, has three advantages compared to traditional competition measures, namely, Herfindahl, Lerner index (PCM), and Price Elasticity (PE). First, DMS captures the dynamic nature of industry competition and directly measures the effects from firms competing actions. Herfindahl is a measure of organizational concentration, which does not recognize the industry market share distribution change from {0.2, 0.3, 0.4, 0.1} to {0.3, 0.2, 0.1, 0.4}. PCM and PE are measures of product price margin, which varies across different industry. It is difficult to compare competition across industries. Second, DMS can be calculated using standard dataset where sales and cogs data are available. PCM and PE need estimation and detailed factor input price data, which is not available in large scale. Third, PCM and PE are for product level data, where input and output need to be separately identified. But firms could have a series of products with different PCM and PE for each product line. For example, a machinery firm could produce low-quality and high-quality machines with distinct PCM and PE for each lines. DMS, on the contrary, can simply use total sales and cogs of the combined product line as long as industry can be clearly defined. For similar reasons, DMS can be used for non-manufacturing industries, such as financial and services, while PCM and PE can only be used for manufacturing industries. Table 1 presents the summary statistics for different industry competition measures. Herfindahl has a mean of for the full sample and for the manufacturing industries. Lerner index has a mean of and for full sample and the manufacturing industries, respectively. Panel C and D in table 1 presents the summary statistics for subsample period of and , respectively. DMS measures are not different in mean, but small increase in standard deviations 10

12 from early to later period. The Herfindahl index increases from to and Lerner index decreases from to Herfindahl indicates an increasing in concentration over time, but Lerner index suggest an increasing in competition from early period to later period. Figure 1 plots the time series of average value of the three competition measures over the period It is consistent with subsample statistics that Herfindahl is increasing over time. DMS and Lerner index, on the other hand, have more dynamics than simple trend. Table 2 reports the Pairwise Pearson correlation coefficient between different measures of competition. The DMS measure is negatively correlated with Herfindahl index (-0.204) and with Lerner index (-0.096). To further analyze the correlation between different competition measures, we divide all the industries in the sample into quintiles based on DMS measures each year. The mean values of different competition measures for each quintile are reported in the top part of Table 4, panel A. Herfindahl index shows a general decreasing pattern with DMS quintiles, while number of firms and Lerner index is increasing. The observed relationships show that the dynamic competition interactions within an industry (DMS) have a positive relationship with the external competition pressure (Herfindahl) and a negative relationship with the ability to extract rent from customers in the industry (Lerner index). 11

13 4. Internal Competition Dynamics and External Competition shocks The industry competition has two components: entry from outside competitors and competitive interactions between existing firms. It is interesting to examine the relationship between the internal and external competitions. We use two measures for outside competition: Imports penetration growth and tariff rate growth. 16 Large imports growth suggests an increase in outside competition and tariff cuts indicates a potential entry from international competition. The imports growth shocks are identified when imports penetration growth rate is two times higher than the median growth rates. The tariff cut shocks are identified when tariff growth rate is two times lower than the median growth rates. The regression results of DMS, Herfindahl and Lerner index on lagged external competition shocks are presented in Table 5. The tariff cut shocks lead to increase in DMS measure for the next period, which is about 10% increase at the mean level. The imports penetration shocks lead to growths in DMS (30% at the mean level). The coefficients are significant statistically and economically. On the contrary, the external competition shocks do not affect the Herfindahl or Lerner index significantly, shown in table 5, column (4)-(6). The effects are negative as expected, but not statistically significant at 10% level. The results suggest that firms compete more aggressively with each other when they face higher outside competition. 16 Refer to Table 1 for detailed definition of the measures. 12

14 5. The Dispersion of Corporate Policies, Performances and Competition Dynamics In this section, we investigate how the DMS measure is related to corporate policies and performance. Specifically, we test what kind of industries tend to have high competition dynamics in section 5.1, how the dispersion of corporate policies and performances affects (section 5.1) and is affected by (section 5.3) the dynamic competition measure and what is the effects of DMS on future corporate policy and performance (section 5.4). 5.1 Industry Characteristics and Competition Table 4, panel B presents the summary of average industry characteristics for the quintiles of DMS intensity. Industries with high sales growth, lower ROA, less investment, high R&D and large cash holdings are more likely to have higher competition dynamics. The regression results are shown in Table 5 and Table 6. Consistent with the univariate test in table 4 and the literature, we find that industries with higher competition dynamics at time t tend to have higher cash holdings (10% higher), leverage (3-7% higher) and sales growth (4.5% higher) at the previous year. Those industries also tend to have lower ROA ( % lower), but higher R&D (8% higher) at the previous year. The observations are consistent with the literature arguing that firms increase cash holding and leverage on expectation of future competition Corporate Policy Dispersion and Competition Table 6 reports how dynamic competition is affected by corporate policy dispersion. We find that larger dispersion in cash holding, leverage and cash flow leads to higher dynamic competition. 17 See Freshad (2010), Haushaltera, Klasa and Maxwell (2007), Hoberg and Phillips (2014) 13

15 The results indicate that industries tend to have more dynamics in competition when the financial situations are more diverse for different firms within the industry. The results are consistent with the literature showing that deep-pocket firms tend to compete more aggressively when they face shallow-pocket rivals. In order to detect the dynamics between the DMS measure and the dispersion of corporate policies and performance, we use panel vector autoregression technique. The results are displayed in table 8 and figure 3. The impulse response functions show significant reactions of DMS measure to the impulse of the dispersion of cash holding and leverage ratio. 5.3 Dynamic Competition and Corporate Performance Dispersion Table 7 panel A reports how dynamic competition affects future corporate policy and performance dispersion. We find that after experiencing strong competition interactions in the industry, firms tend to have larger dispersion profitability (ROA) and valuation (Tobin s Q). Firms also have marginally higher dispersion of cash holding, leverage and R&D expenditures after intensified competition. The results may have two indications. One is that deep-pocket competitors choose to keep higher financial slack to prepare for the next round of competition or deter other rivals to have further predatory or revenging competition. The other indication is that firms with financial constraints became more tight-in-hand after a period of intensified competition. They just cannot keep up with the cash flow even if they wanted to. The two indications could and need to be tested in the next draft of the paper. The impulse response functions also show strong reactions of 14

16 the dispersions of financial policies and investment policies capital expenditure, R&D and advertising to the impulse of DMS measure. (See table 8 and figure 3 for details.) 5.4 Dynamic Competition and Future Corporate Policies and Performance Table 7, panel B reports how dynamic competition affects future corporate policy and performance average in the industry. We find that after experiencing strong competition interactions in the industry, firms have lower profitability (ROA) and valuation (Tobin s Q) 18. On average, firms also have lower cash holding, leverage and cash flow after intensified competition. The results suggest that more intensified competition leads to decreased profitability, financial situation and valuation. The results are consistent with the literature arguing that competition negatively affects firm performance. 6. Conclusion The paper complements the existing literature by proposing a dynamic measure to capture the competition interactions inside of an industry: Dynamic Market Share (DMS). DMS j,t measure is calculated as the sum of absolute 19 change of market share for all firms in industry j from year t-1 to t. This measure therefore captures how the market shares distribution changes from time t-1 to t, directly evaluating market competition intensity inside of an industry. A larger value of the DMS measure suggests a bigger change in how the market share is distributed within industry, thus indicating a strengthened competition, and a greater pressure faced by managers from existing rivals. 18 contemporaneous results are reported in Table 6, column (1) and (2) 19 Sum of square term of market share changes is also used as robust. The results do not change significantly. 15

17 Using a full sample consisting of all Compustat firms, covering 7,632 industry-year observations over the period from 1977 to 2013, we find that the DMS measure has large variation across industries and over time. Overall, the DMS measure is higher for the periods of and , is relative low for the periods of , , and The DMS measure does not have high correlation with Herfindahl index (-0.204) or Lerner index (-0.096). The DMS measure is positively associated with imports growth and negatively correlated with the growth of the tariff rate in the industry, which suggests that the competition between existing firms will intensify when they face higher outside competition. Generally, we find that industries with lower average ROA, but higher cash holding and leverage are more likely to have higher competition dynamics. More interestingly, these dynamic industries tend to have higher dispersion in financial policies, such as cash holding, leverage and cash flow at previous period. In addition, we also find that high competition activities lead to future dispersion of profitability and performance within the industry. Firms who just experience high level of competition tend to have higher level of leverage, low level of cash holdings and cash flow. In order to detect the dynamics between the DMS measure and the dispersion of corporate policies and performance, we use panel vector autoregression technique. The impulse response functions show a strong reaction of the dispersion of investment policies capital expenditure, R&D and advertising to the impulse of DMS measure. The paper complements to the literature by providing a dynamic measure of competition interactions between existing firms within an industry. The DMS measure can be adopted in any industry competing for market shares, not limited to the manufacturing industries with defined 16

18 input-output data as required by Lerner index and price elasticity. The paper finds positive relationship between internal competition intensity with external competition shocks. The paper also contributes to the literature by providing new evidence on how industry competition is related to the dispersion of corporate policies and performance. One puzzling fact, surprisingly, is that there is no clear pattern for the relationship between dynamic competition and advertising expenditures or the dispersion of it. The summary statistics across DMS quintiles (Table 4) shows a decreasing pattern of advertising as competition activities intensified. But the relationship is not significant for the regression results for both mean value and dispersion of advertising expenditures. 17

19 Appendix A. Variable Definition Variable Definition Firm Characteristics Size Natural log of total asset of the firm i Cash holding Cash and cash equivalent divided by total assets Leverage Sum of short term debt and long term debt divided by total assets Investment Capital Expenditure divided by total assets Sales growth Total revenue at time t divided by total revenue at time t - 1 minus 1 ROA EBIT divided by total assets R&D R&D expenses divided by total revenue Advertising Advertising expenses divided by total revenue Market to Book Market capitalization divided by book value of asset minus book value of debt Industry Competition Measures DMS The sum of absolute change of market share in industry j at time t Herfindahl Herfindahl index calculated by sum square of market share in defined industry Lerner Index Mean Price-Cost Margin in industry j at time t, defined as sales minus cost of goods sold divided by total sales Number of Firms Number of firms in industry j at time t. Imports growth Tariff growth Growth rate of total Imports in four-digit SIC industry Growth rate of average tariff rate on imported goods in four-digit SIC industry 18

20 References Aggarwal, R., and Samwick, A., Executive compensation, strategic competition, and relative performance evaluation: Theory and evidence. Journal of Finance 54, Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith, and Peter Howitt, 2005, Competition and Innovation: an Inverted-U Relationship, Quarterly Journal of Economics 120, Beiner, S., M. Schmid, and G. Wanzenried, 2011, Product market competition, managerial incentives and firm valuation, European Financial Management 17, Bloom, N., and J. Van Reenen, 2007, Measuring and explaining management practices across firms and nations, Quarterly Journal of Economics 122 Boone, Jan, 2008, A New Way to Measure Competition, The Economic Journal 118, Blundell, Richard, Rachel Griffith and John Van Reenen, 1999, Market Share, Market Value and Innovation in a Panel of British Manufacturing Firms, Review of Economic Studies 66, Campello, Murillo, 2003, Capital Structure and Product Markets Interactions: Evidence from Business Cycles, Journal of Financial Economics 68, Chevalier, Judith, and David S. Scharfstein, 1995, Liquidity Constraints and the Cyclical Behavior of Markups, American Economic Review Papers and Proceedings 85, DeFond, M., Park, C., The effect of competition on CEO turnover, Journal of Accounting and Economics 27, Domowitz, I., G. Hubbard, B. Petersen, 1986, Business cycles and the relationship between concentration and price-cost margins. Rand Journal of Economics 17, Fisher, F.M. 1987, On the misuse of the profits-sales ratio to infer monopoly power, RAND Journal of Economics 18. Fresard, Laurent, 2010, Financial Strength and Product Market Behaviors: The Real Effects of Corporate Cash Holdings, Journal of Finance 65,

21 Fresard, Laurent, and Philip Valta, 2012, Competitive Pressure and Corporate Policies, working paper, SSRN Goldberg, P., 1995, Product differentiation and oligopoly in international markets: the case of the US automobile industry, Econometrica 63, Harris, M., 1998, The association between competition and managers business segment reporting decisions, Journal of Accounting Research 36, Haushaltera, David, Sandy Klasa, William F. Maxwell, 2007, The influence of product market dynamics on a firm s cash holdings and hedging behavior, Journal of Financial Economics 84 (2007) Hermalin, B., The effects of competition on executive behavior, Rand Journal of Economics 23, Hoberg, Gerard and Gordon Phillips, 2014, Product Market Threats, Payouts, and Financial Flexibility, Journal of Finance, Volume 69, Issue 1, pages , February 2014 Hoberg, Gerard and Gordon Phillips, 2012, Product Market Synergies and Competition in Mergers and Acquisitions: A Text-Based Analysis, Review of Financial Studies Karuna, Christo, 2007, Industry product market competition and managerial incentives, Journal of Accounting and Economics 43, Kovenock, Dan, and Gordon M. Phillips, 1997, Capital Structure and Product Market Behavior, Review of Financial Studies 10, Nalebu, B., and Stiglitz, J., 1983, Information, competition, and markets, American Economic Review Papers and Proceedings 73, Nevo, A. 2001, Measuring market power in the ready-to-eat cereal industry, Econometrica 69. Nickell, Stephen J., 1996, Competition and Corporate Performance, Journal of Political Economy 104,

22 Nickell, S. 1999, Product markets and labor markets, Labor Economics 6 Opler, Tim, and Sheridan Titman, 1994, Financial Distress and Corporate Performance, Journal of Finance 49, Ornstein, Stanley I., 1975, Empirical Uses of the Cost-Price Margin, Journal of Industrial Economics 24, Phillips, Gordon M., 1995, Increased Debt and Industry Product Markets: An Empirical Analysis, Journal of Financial Economics 37, Raith, M., Competition, risk and managerial incentives. American Economic Review 93, Scharfstein, D., Product market competition and managerial slack. Rand Journal of Economics 19, Schmidt, K., Managerial incentives and product market competition. Review of Economic Studies 64, Schott, Peter K., The Relative Sophistication of Chinese Exports. Economic Policy January Zingales, Luigi, Survival of the Fittest or the Fattest? Exit and Financing in the Trucking Industry, Journal of Finance 21

23 Table 1, Summary Statistics of Industry Competition Measures This table displayed the summary statistics of different industry competition measures. The statistics reported are mean, median, standard deviation, 25 percentile, 75 percentile. DMS is the sum of absolute change of market share in industry j from time t-1 to time t. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to The manufacturing industries in panel B. are industries with SIC between Refer to section 2 for detailed definition of industries. Variable N Mean St. Dev. 25th Median 75th Panel A. the whole sample DMS 7, Herfindahl 7, Lerner Index 7, Number of firms 7, Panel B. manufacturing industries only DMS 4, Herfindahl 4, Lerner Index 4, Number of firms 4, Imports growth 1, Tariff growth 1, Panel C. sample DMS 3, Herfindahl 3, Lerner Index 3, Number of firms 3, Imports growth Tariff growth 1, Panel D. sample DMS 3, Herfindahl 3, Lerner Index 3, Number of firms 3, Imports growth Tariff growth

24 Table 2, Correlation between different competition measures This table reports the pairwise correlation coefficients for different measures of competition: DMS, Herfindahl, Lerner Index, Number Firms, Imports growth and Tariff growth. DMS is the sum of absolute change of market share in industry j from time t-1 to time t. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to The manufacturing industries in panel B. are industries with SIC between Refer to section 2 for detailed definition of industries. DMS Herfindahl Lerner Number of Imports Tariff Firms growth growth DMS Herfindahl Lerner Number of Firms Imports growth Tariff growth

25 Table 3, Transition Probability of DMS measure categorization This table reports the transition probability for DMS quintiles. The rows are DMS quintiles at t and the columns are DMS measures in t+1 for Panel A, t+3 for Panel B, and t+5 for Panel C. DMS is the sum of absolute change of market share in industry j at time t. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to Least Competition Quintile 2 Quintile 3 Quintile 4 Most competition Panel A. One-Year Transition probabilities Least Competition Quintile Quintile Quintile Most competition Panel B. Three-Year Transition probabilities Least Competition Quintile Quintile Quintile Most competition Panel C. Five-Year Transition probabilities Least Competition Quintile Quintile Quintile Most competition

26 Table 4, Industry Characteristics of DMS Intensity This table reports the firm and industry characteristics for DMS quintiles, measured each year. Panel A. summarizes Competition Measures, Panel B summarizes the mean value of industry characteristics and Panel C summarizes the dispersion measure of industry characteristics. DMS is the sum of absolute change of market share in industry j at time t. Refer to section 3 for detail construction of the measures. Disp(x) is the standard deviation of variable x divided by the mean value of x in defined industry at time t. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to *, ** and ***indicates 10%, 5% and 1% significance, respectably. Least Competition Quintile 2 Quintile 3 Quintile 4 Most Competition T-stat of Most-Least Competition Panel A. Competition Measure DMS *** Herfindahl *** Lerner Index *** Number of firms *** Imports growth Tariff growth Panel B. Mean value of industry characteristics Sales growth *** ROA *** Tobin's Q Investment ** R&D *** Advertising expenses *** Cash holding *** Leverage Panel C. Dispersion measure of industry characteristics Disp(Sales Growth) ** Disp(ROA) ** Disp(Q) *** Disp(Investment) ** Disp(R&D) *** Disp(Advertising) Disp(Cash) ** Disp(Leverage) * Disp(Cash Flow) ** 25

27 Table 5, Competition Measures and External Competition Shocks This table reports the regression results for competition measures DMS, Herfindahl and Lerner Index on external competition measures: Imports shocks and Tariff Cuts. The dependent variables are DMS at t+1 for column (1)-(3), Herfindahl at t+1 for columns (4)-(5) and Lerner Index for column (6). DMS is the sum of absolute change of market share in industry j at time t. Imports shocks are dummy variable which equals to 1 when imports growth rates are at least twice as the median growth rate. Tariff cut shocks are dummy variable which equals to 1 when tariff negative growth rates are twice as the median growth rate. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to Heterogamous robust T-statistics are reported in parenthesis *, ** and ***indicates 10%, 5% and 1% significance, respectably. (1) (2) (3) (4) (5) (6) DMS t+1 DMS t+1 DMS t+1 Herfindahl t+1 Herfindahl t+1 Lerner t+1 Tariff cut shocks 0.010** 0.012** (2.22) (2.47) (-1.61) (-0.70) Imports shocks 0.032** (2.07) (-0.43) Average sales growth *** (1.61) (0.11) (0.27) (-0.15) (-1.07) (-3.17) ROA * ** *** (-1.82) (-2.28) (-1.30) (-0.17) (0.40) (10.55) Tobin's Q *** ** (-1.07) (-0.14) (-0.98) (-2.63) (-2.09) (-1.09) Investment ** * *** ** (-2.43) (-1.99) (-1.14) (-4.09) (-2.55) (0.34) Cash 0.066* 0.051* 0.111** *** (1.96) (1.85) (2.17) (-0.43) (-0.76) (-5.93) Leverage 0.048* 0.048* ** (1.87) (1.89) (1.17) (-0.96) (-2.48) (1.57) Advertising *** (-0.03) (3.27) Constant 0.105*** 0.097*** 0.110*** 0.374*** 0.462*** 0.092*** (7.89) (5.98) (6.23) (12.24) (12.06) (4.76) Observation Adjusted R-square

28 Table 6, OLS Regression of DMS Measures on lagged Corporate Policy Dispersion This table reports the regressions of DMS measures on lagged corporate policy dispersion. The dependent variables are DMS at t+1. DMS is the sum of absolute change of market share in industry j from time t-1 to time t. Column (1) and (2) reports the contemporaneous results. Column (3) and (4) reports the results for DMS regression on one lagged dispersion measures of corporate policies. Column (5) and (6) reports the results for DMS regression on two lagged dispersion measures of corporate policies. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to Heterogamous robust T-statistics are reported in parenthesis *, ** and ***indicates 10%, 5% and 1% significance, respectably. (1) (2) (3) (4) (5) (6) DMS t DMS t DMS t+1 DMS t+1 DMS t+2 DMS t+2 Disp(Cash) * 0.004** 0.008** (-1.29) (-1.65) (2.32) (2.22) (-0.14) (-0.61) Disp(Leverage) ** 0.005* 0.012* (0.65) (0.11) (2.44) (1.97) (1.65) (1.09) Disp(Cash Flow) 0.020*** 0.040*** 0.013** 0.019** (3.97) (4.37) (2.32) (2.25) (-0.23) (-0.74) Disp(ROA) *** * (-3.26) (-1.69) (-1.06) (0.20) (-0.16) (0.38) Disp(Q) 0.005*** 0.007*** 0.002* 0.003* 0.003** 0.004* (2.96) (3.12) (1.67) (1.66) (2.04) (1.86) Disp(Investment) * * (-0.65) (-0.22) (-1.77) (-1.90) (0.27) (1.25) Disp(Advertising) * (-1.78) (-1.10) (-1.21) Disp(R&D) ** (-0.19) (0.20) (2.26) Sales Growth 0.072*** 0.061*** 0.045*** 0.045*** 0.027*** 0.045*** (7.51) (5.83) (5.22) (3.15) (3.18) (3.34) ROA *** ** ** * (-5.28) (-2.47) (-2.18) (1.92) (-0.87) (0.17) Tobin's Q *** *** * * * (-5.75) (-4.87) (-1.82) (-1.68) (-0.44) (-1.78) Investment * (-1.93) (0.56) (-0.56) (1.25) (-0.11) (0.67) Cash holding 0.107*** 0.189*** 0.115*** 0.145*** 0.124*** 0.169*** (4.84) (5.76) (4.97) (4.20) (5.28) (5.54) Leverage 0.039*** 0.070*** 0.043*** 0.069*** 0.049*** 0.069*** (3.57) (4.29) (3.81) (4.12) (4.10) (4.17) Advertising (0.23) (-0.69) (-0.10) R&D 0.077** 0.082** (1.99) (2.18) (1.53) Constant 0.104*** 0.078*** 0.085*** 0.067*** 0.079*** 0.066*** (17.49) (9.29) (14.48) (8.16) (12.34) (7.66) Observation 6,836 3,832 6,624 3,738 6,416 3,637 Adjusted R-square

29 Table 7, OLS regression of corporate policy dispersion on lagged DMS competition This table reports the regressions of corporate policy dispersion and future firm performance on lagged DMS measures. All independent variables are lagged at t-1. Panel A reports the results of the dispersion of corporate policy and performance on DMS. Panel B reports the results of the mean value of corporate policy and performance on DMS. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to Heterogamous robust T- statistics are reported in parenthesis *, ** and ***indicates 10%, 5% and 1% significance, respectably. Panel A. The Dispersion of Corporate Policy and performance on DMS (1) (2) (3) (4) (5) (6) (7) (8) Disp(ROA) Disp(Q) Disp(Investment) Disp(R&D) Disp(Advertising) Disp(Cash) Disp(Leverage) Disp(Cash Flow) DMS 0.291* 1.684*** * ** 0.265* (1.85) (3.64) (1.51) (1.94) (0.15) (2.17) (1.93) (1.52) ROA * (-0.89) (-1.56) (0.18) (-0.94) (0.61) (0.14) (-1.75) (-0.71) Tobin's Q 0.063** 0.499*** (2.00) (6.31) (1.55) (1.46) (1.05) (0.56) (-0.71) (1.19) Investment 1.395*** 4.322*** 0.641*** 0.336*** 0.249** 0.642*** 0.647*** 1.341*** (3.19) (3.98) (5.44) (2.78) (2.00) (3.79) (3.53) R&D ** *** * * (-0.57) (-2.27) (-1.41) (3.39) (-1.69) (-1.72) (0.04) Cash holding ** *** 0.889** (1.30) (2.06) (1.32) (0.12) (0.61) (5.20) (1.35) Leverage 0.261** ** *** 0.403*** 0.208** (2.25) (0.30) (2.20) (-0.11) (1.33) (3.12) (2.00) Cash Flow (-1.11) (-0.76) (-0.94) (-1.62) (-0.89) (-0.85) (-0.94) (-1.18) Constant *** *** *** *** ** ** (-2.73) (-4.54) (-3.33) (-1.05) (-1.52) (-3.49) (-2.06) (-2.06) Observation Adjusted R-square

30 Panel B. Mean value of Corporate Policy and performance on DMS (1) (2) (3) (4) (5) (6) (7) (8) ROA Tobin's Q Investment R&D Advertising Cash Leverage Cash Flow DMS ** *** * 0.041*** ** (-2.17) (-2.66) (-0.50) (1.60) (-1.50) (-1.93) (-1.96) ROA 0.446*** *** (6.46) (1.49) (-0.22) (-1.12) (0.41) (-0.01) Tobin's Q *** 0.003*** 0.004*** 0.009*** 0.003* * (-1.58) (21.86) (4.02) (3.02) (6.40) (1.73) (-1.68) Investment *** *** *** 0.215*** (-1.09) (-0.45) (46.52) (0.43) (2.71) (-0.70) R&D *** 0.772*** *** 0.654*** *** 0.134*** *** (-4.79) (4.33) (-4.10) (15.21) (-5.77) (7.17) (-0.71) (-3.99) Cash holding *** 0.460*** 0.021*** 0.086*** *** *** ** (-2.97) (2.92) (3.80) (4.98) (-0.44) (49.84) (-3.18) (-2.34) Leverage ** *** ** *** *** 0.778*** ** (-2.13) (-0.12) (3.09) (-2.46) (-5.51) (-2.93) (-2.33) Cash Flow *** 0.021*** ** *** (1.56) (-3.24) (3.29) (-0.48) (-2.18) (-0.94) Constant 0.073*** 0.548*** 0.004** *** 0.026*** 0.060*** 0.045*** (5.34) (11.51) (2.56) (0.04) (11.84) (5.97) Observation Adjusted R-square

31 Table 8, Panel Vector Autoregression of DMS and Corporate Policy and Performance Dispersion This table reports the panel vector auto regression of DMS and corporate policy and performance dispersion. Column (1) reports the coefficient of lagged variables response to DMS. Column (2) reports the variables response from lagged DMS. DMS is the sum of absolute change of market share in industry j at time t. Active is a dummy variable, which equals to 1 when DMS measure is above its median. Refer to section 3 for detail construction of the measures. The other variables are defined in Appendix A. The sample is all Compustat firms excluding commodity producers (SIC>1499), Utilities (SIC ), Financials (SIC ), Education and Public Administrative industry (SIC>8800) over the period from 1977 to Heterogamous robust T-statistics are reported in parenthesis *, ** and ***indicates 10%, 5% and 1% significance, respectably. (1) (2) Variables Response to Response from DMS t+1 DMS t-1 DMS 0.313*** (10.58) Dispersion of Cash 0.013* 0.527*** (1.94) (3.74) Dispersion of Leverage 0.005* 0.935*** (1.88) (4.10) Dispersion of Cash Flow *** (0.99) (2.46) Dispersion of ROA ** (0.85) (2.18) Dispersion of Q 0.003*** 0.520*** (2.77) (4.77) Dispersion of Investment *** (1.44) (4.30) Dispersion of R&D (0.39) (1.55) Dispersion of Advertising *** (1.18) (5.79) 30

32 Figure 1. Average Competition Measures over time

33 Figure 2. Plot of DMS measure for selected industries 2-digit SIC 26, 27 paper and printing 2-digit SIC 32. Stone and concrete 2-digit SIC 35,36 Machinery and Electronics 2-digit SIC 48 communications 2-digit SIC 51, wholesale non-durable 2-digit SIC 70, Hotels 32

34 Figure 3, Impulse Response of DMS Measure and Corporate Policy and Performance Dispersion Dispersion of Cash Holding and DMS Dispersion of Leverage and DMS Dispersion of Cash Flow and DMS Dispersion of ROA and DMS Dispersion of Tobin s Q and DMS Dispersion of Investment and DMS Dispersion of R&D and DMS Dispersion of Advertising and DMS 33