Is performance driven by firm- or industry specific factors?

Hele tekst

(1)Master Thesis F.J. Bergsma – August 2012. Is performance driven by firm- or industry specific factors? F.J. Bergsma* ABSTRACT. In this study we contribute to the debate on the magnitude of industry effect versus firm effects on different performance measures. New in this study is that we used a worldwide dataset and therefore we also test for country effects. Secondly, we use different accounting measures and value measures as performance indicators. Finally we also exclude leaders and losers from each industry to study the change of the different effects. We use a dataset of 1200 firms listed at the S&P 1200 over a period from 1990 until 2010, and employ a variance component analysis on this data. We find that firm effects dominate industry effects. The magnitude of this effect is contributed by a couple of firms who dominate or submit in their industry. For firms who are stuck in the middle industry effects are dominant. Therefore we can conclude that resource based view is applicable for the out- and underperformers and for the rest of the industry who are stuck in the middle the industrial organization theory is more dominant.. JEL codes: G10, G11. Keywords: performance measures, firm effects, industry effects, year effects, country effects, return on assets, return on invested capital, weighted average cost of capital, Tobin’s q, Industrial Organisation and Resource Based View.. *Master thesis Folkert Jan Bergsma, student Master of Science Business Administration specialization Finance, Faculty of Economics and Business, University of Groningen. University supervisor: dr. ing. N. Brunia,Student number: S1937251, E-mail: f.j.bergsma@student.rug.nl and folkert_bergsma@hotmail.com. Date: June 2012.

(2) Is performance driven by firm- or industry specific factors?. 1.. Introduction. Performance measurement plays a key role in translating an organization’s strategy into desired behaviours and results, and communicating these expectations, monitoring progress, providing feedback, and motivating employees through performance-based rewards and sanctions (Van der Stede and Chow, 2006). The organisation’s strategy can be measured using financial performance measures. The central question is what drives those performance measures. The industry organisation (IO) approach argues that the industry structure is the most important determinant for performance. On the other hand the resource based view (RBV) argues that firm characteristics drive performance. Past research on this topic Schmalensee (1985), Rumelt (1991) and Hawawini et al. (2003) show mixed results. An important explanation for the mixed results found in previous studies could be the kind of performance measure that is used. Our results show that this is not the case. If firms want to create value they do so by investing capital to generate future cash flows at rates of return that exceed their cost of capital (Koller, 2010).Following this line of reasoning, performance measures should incorporate value creation. Unfortunately value creation is not always incorporated in performance measures. For example the drivers behind return on assets are not substituted in the determents that are used to calculate value creation. However with a few changes this measure can be rewritten in return on invested capital (Koller, 2010) which we also included. And then this performance measure is consistent with value creation.. In this paper we want to address the magnitude of firm or industry factors using several performance measures, and see which of the theories is dominant in explaining firm performance. We test the two theories by examining the impact of firm- and industry effects on several performance measures (return on assets, return on invested capital, and return on invested capital minus weighted average cost of capital, Tobin’s q at enterprise level, tobin’s q at equity level and Tobin’s q as described in the literature).Besides the industry and firm effects we also test for year effects to cover macroeconomic fluctuations. The current paper contributes to the performance driver’s debate in that in contrast to previous research we use a dataset that covers firms around the world. Logically we also add a country effect in the model that we use to examine if and to what extend country effects explain variability in performance measures. This is logical because excluding this effect could result in a higher error in explaining the variability in firm performance. The most relevant contribution of this paper is that we want to examine if a theory or which theory is dominant. This could influence. 2.

(3) Is performance driven by firm- or industry specific factors?. decisions made by a corporate finance manager based on performance measures, because it turns out that firms with above average performance (leaders) and firms with below average performance (losers) are less affected by their industry effects and more influenced by firm effects. Vice versa for the average performing firms stuck in the middle of the industry. Indicating that the performance can change due to changes of the industry- or firm effects they are exposed to. The remainder of the paper is as follows. In Section 2 consists of the literature review. Section 3 presents the data. We elaborate on the methodology in Section 4. In Section 5 we show the results, and in Section 6 we give concluding remarks.. 2.. Literature review. Effects that explain performance variability is a widely discussed topic since the beginning of the previous century. Appendix A contains a table with to our knowledge all previous research; in table 1 the most relevant studies are shown. We first discuss the different effects found in these studies followed by the different measures used in the current and previous studies. Table 1 Variance component comparison with relevant previous studies Measure Industry Firm Year Industryeffects effects effects year effects Schmalensee (1985) ROA 19.46% 0.63% NA NA Rumelt (1991) ROA 4.0% 45.8% NA 5.4% Study. McGahan and Porter (1997) Hawawini et al. (2003) (full sample). ROA ROA EP/CE TMV/CE. 18.7% 8.1% 6.5% 11.4%. 36.0% 35.8% 27.1% 32.5%. 2.4% 1.0% 1.9% 1.3%. NA 3.1% 4.2% 2.9%. Error. 80.54% 44.8% 48.4% 52.0% 60.3% 51.9%. ROA, return on assets; EP/CE, economic profit divided by capital employed; TMV/CE, total market value divided by capital employed. 2.1 The effects The first effect we describe is the industry effect. This effect embodies the industrial organisation theory. Industrial organisation is concerned with the working of markets and industries; in particular the way firms compete with each other, where market structure and competition are important determinants in firm performance (Cabrall, 2000). Furthermore this theory states that structural characteristics such as, economies of scale, barriers to market entry, diversification, product differentiation, and the degree of concentration of firms in the industry are the main determinants for firm performance (Hoskisson et al., 1999; Mauri and Michaels, 1998; Seth and Thomas, 1994). Mason (1939) was one of the first suggesting that industry structure determines firm performance. Masons (1939) view was further developed. 3.

(4) Is performance driven by firm- or industry specific factors?. by Bain (1956, 1959) who produced the structure-conduct-performance (SCP) framework. According to Bain (1956, 1959) firm performance depends on its conduct in such a matter as pricing policy, research and development and investment policy. A firm’s conduct in turn, depends on structural characteristics as concentration, barriers to entry, and industry growth (Caloghirou et al. 2004). This SCP framework is widely used in the strategic management field, and many empirical studies show a relation between industry structure and performance (Scherer and Ross, 1990). In example Weis (1979) found a positive relation between industry concentrations, barriers to entry and profits. This SCP framework is also the building block for Porter’s (1980) well known five forces model (competitive rivalry, the threat of new entrants and substitute products, and the bargaining power of suppliers and customers) on how these forces influence firms in industries. These five forces determine investment requirements, costs and prices and the combination of these five forces drives performances and hence industry attractiveness (Galbreath and Galvin, 2008). All researchers in the field of industrial organisation conclude that because the industry conduct is determined by its structural forces, a firm’s management decisions cannot influence a firm’s performance. Therefore the role of management is ignored as a determinant of performance. In 1980 there was a shift in strategic management due to criticism on the industrial organisation theory. The main reason was that the industrial organisation theory was unable to give an explanation for out- and underperformers within an industry. Over the last decade, much of the strategy literature has emphasized resources internal to the firm as the principal driver of firm performance and strategic advantage. The theory is called the resource based view. The transition occurred for several reasons like; the rate of change in new products, technology, and shifts in customer preferences has increased. Making it difficult formulating a strategy in an increasing dynamic environment (Bettis and Hitt, 1995). Furthermore, the traditional industry boundaries are blurry nowadays, because many industries converge and overlap (Bettis and Hitt, 1995; Hamel and Prahalad, 1994). While industrial organisation theory is based on a stable industry, like many strategic analysis tools, including competitor analysis, strategic groups, and diversification typologies. The increasing rate of change has put increasing pressure on firms to react more quickly, as time is often seen as source of competitive advantage (Stalk and Hout, 1990). This suggests that firms may look inwardly for strategic opportunities, while, at the same time, must reconceptualise how they think of industries and define competitors (Kostopoulos el al. 2002).. 4.

(5) Is performance driven by firm- or industry specific factors?. The firm effect embodies the resource based view. In contrast to industrial organisation theory where performance is determined by its industry and external forces for example described by Porter (1980), the resource based view states that performance is driven by internal forces, also called unique resources. The resource based view assumes that each organization is a collection of unique resources and capabilities that provides the basis for its strategy and serves as the primary source of firm performance. In that way sustainable value creation opportunities are derived from firm-specific, specialized factors that are difficult to imitate or substitute. Profitability differences between firms are therefore caused by the ability to sustain fundamental differences between firm’s resource endowments (Porter, 1980).Wernerfelt (1984) defined the term resource based view as the performance variation among firms according to idiosyncratic and unique resources rather than the industry’s structural characteristics. Suggesting that the firm’s resources are the base for value creation, these resources are characterized by the properties of scare distribution among the firms within the industry, protection from competition and imperfect mobility of these resources (Barney, 1991; Peteraf, 1993). The resource based view is described as a combination of tangible and intangible resources including management, organisational process and routines, and the knowledge and information it reserves within the firm (Galbreath and Galvin, 2008; Michalisin et al., 1997; Amit and Schoemaker, 1993). Barney (1991) found that a firm can only be efficient and effective if it exploits resources that are rare and valuable. In that way the firm can create a competitive advantage to its competitors and can create sustainable value. Therefore, the firm’s objective is to develop and deploy their resources in such a way that they cannot be imitated by their competitors. If this goal is achieved, performance advantages are created and sustained (Galbreath and Galvin, 2008). To summarize, the resource based view states that firms should focus on their unique resources and capabilities, and manage them dynamically in pursuit of sustainable competitive advantage so that it increases firm performance (Lee et al., 2001; Markides, 1999). Besides the industry and firm effect elaborated above there is a third effect that has a potential relation with performance. This effect is named the year effect. The year effect covers macroeconomic year to year fluctuations that influence all firms equally. Like for instant the economic booms and slumps. The fourth effect on performance is the industry-year effect which is also called a transient industry effect. This effect reflects the year to year sensitivity of the performance measure to. 5.

(6) Is performance driven by firm- or industry specific factors?. the impact of business cycles on the industry. Like the changes in car sales for the automotive industry. The main contribution of this paper is that it makes use of a global sample while other studies only use data from a single country (mostly US data). Because we use a dataset that covers almost all the developed countries throughout the world we also examine if country effects explain performance variation. There are two approaches that can explain differences in countries and their impact on performance. First, the comparative advantage theory (Ricardo, 1817) which is in line with industry organization theory it states that countries differ in the availability, prices, and the intensity of use of resources. The second approach, the comparative advantage of nations is in line with the resource based view. The comparative advantage of nations (Porter, 1990) theory highlights to the importance of individual countries for the technological development and innovation from which firms employ their advantages to compete against foreign rivals in domestic and foreign markets. Firms can achieve a superior performance in distinct industries because these countries have greater capacity to help their firms improve and innovate faster than foreign rivals in a particular industry (Makino et al. 2004). So a country effect can be a consequence of the resource based view or the industrial organizational view. However, we cannot allocate the country effects to the resource based view or the industrial organisation theory. Makino et al. (2004) is one of the first who did research on country effects. However, he investigated country effects of foreign subsidiaries in one particular country. Nevertheless, he concluded that country effects matter and that they do influence firm performance. Chen (2007) found that the effects of individual country’s characteristics are more important than the effects of industries on performance of firms in knowledge-intensive service industries worldwide, supporting evidence that these firms have gained in global competitiveness by using fairly sound microeconomic fundamental strategies with their countries intervention to ensure economic activity across industry boundaries. An important note is that Chen (2007) did not include firm effects in his research. Furthermore, Goldszmidt et al. (2011) found significant country effects in performance evaluation. Therefore we cannot ignore this effect in our research. So besides the firm- industry and year effects also country effects are admitted in our model. Previous research on firm- and industry effects did not test for firm effects because all this research is done in a specific country.. 6.

(7) Is performance driven by firm- or industry specific factors?. The following effects are not tested but we still discuss them briefly to strengthen the conclusions of this paper. The dataset we use contains only large firms throughout the world because large firms create economies of scale and scope and therefore can create more value compared to smaller firms within the industry. Therefore we question if size could be an explanatory factor in performance variance. Firm size gives advantages of scale due to the reduction of the average cost per unit and scope because the average cost of units decrease, and therefore the firm can create more value. The causal relationship between profitability and firm size is widely tested and the results are mixed and these findings are small and insignificant, (Baumol, 1967; Hall and Weiss, 1967; Cinca et al. 2005). Another effect that could be of influence is a firm’s market share. The results on the relation with performance are mixed. Szymanski et al. (1993) found a small market share-profitability relationship that, on average, market share has a positive effect on business profitability. Gale (1972) concludes that high market share is associated with high rates of return and that the effect of share on profitability depends on other firm and industry characteristics. Indicating that the relationship is context specific (Prescott et al., 1986). Also Schmalensee (1985) argues that market share has a negligible effect on performance. 2.2 Performance measures Almost all previous research on firm and industry effects used return on assets (ROA) as performance measure. Return on assets is an accounting measure that suffers from conceptual disadvantages that arise from accounting conventions (Hawawini et al., 2003). First, because accounting conventions like General Accepted Accounting Principles (GAAP) exclude intangible assets from the balance sheet. Second, net income divided by the total amount of assets excludes the effect of the sources of financing and ignores the benefits of operating liabilities that could reduce the amount of capital required for investors (Koller, 2010). Third, accounting values such as return on assets are calculated based on historic prices rather than their true replacement values. This results in that these accounting measures cannot provide insight in the firm’s historical economic or future profitability. Harcourt (1965) and Fisher and McGowan (1983) both argues that accounting measures cannot infer anything about the economic performance of a firm. In this paper we also test for value based performance measures besides the accounting measure return on assets. These measures are based on residual income (adjusted for capital costs and time value of money).These value measures are also called economic measures. These measures are a better analytical tool in. 7.

(8) Is performance driven by firm- or industry specific factors?. understanding firm performance because they are focused on the firm’s operations. Accounting measures such as return on assets do not incorporate the cash flow, the cost of capital or the true value of the assets. The drivers behind return on assets are not substituted in determinants that are used to compute value measures. So this performance measure is not consistent with value creation theory. However we do include return on assets in our study for comparison with previous studies. Return on assets (Needles, 2008) is calculated by net income divided by the average of total assets Return on assets is a profitability measure for a firm and is mainly used to compare firms within the same industry.. Stewart (1991) introduced value performance measures like economic value added (EVA) and market value added (MVA). Economic value added is the residual income after operating profit and a full and fair return on capital (Stewart, 1991), this is obtained by multiplying the spread between the return on invested capital and the weighted average cost of capital by invested capital. In this paper we test for the return per dollar invested and the economic value added per dollar invested as performances measures. Advantages are that economic value added is that it is relative easy to calculate, it can be used as a management compensation tool and it can help to improve performance. Disadvantages of economic value added is that it is subject to accounting anomalies and analyst adjustments, it requires accurate estimates of its components and it does not necessarily measure shareholder value.. Hawawini et al. (2003) was the first one who used value measures as performance indicator to test for firm and industry effects. In this study we also use value measures, however we cannot say if we calculated these measures similar to the one Hawawini et al. (2003) uses because he does not specify the adjustments he has made in his calculations.. Return on invested capital is a measure similar to the return on assets only this measure is calculated on the enterprise level. And shows how efficient the firm is in investing capital in profitable investments. Besides growth, also return on invested capital determinants value creation. Return on invested capital (Damodaran, 2007) is earnings before interest, tax and amortization (EBITA) times 1-taxrate divided by average invested capital. Return on invested capital is the return a firm earns on each dollar invested (Koller 2010). The advantage of this value measure is that it can compare companies with different capital structures because it looks at all the sorts of financing the firm uses. It also does not only take into account the. 8.

(9) Is performance driven by firm- or industry specific factors?. interest of the shareholder but it embodies the long term interests of all the stakeholders. A downside is that the measure cannot specify the sources of the value creation.. In this paper we also test for return on invested capital minus the weighted cost of capital is a similar measure to the economic value added Hawawini et al. (2003) uses. The weighted cost of capital is the rate of return investors except to earn from investing in a firm (Koller, 2010). By subtracting this from the return on invested capital you get the value added per dollar invested.. In this study we use the name Tobin’s Q at enterprise level (TOB1) as name for market value added. Market value added is a measure of the value a firm has created in excess of its resources that are already committed to the firm. Market value added (Stewart, 1991) is the enterprise divided by invested capital. Because shareholder value maximization is the primary goal, shareholder wealth is maximized by maximizing the difference between enterprise value and invested capital (Brigham and Ehrhard, 2010). Tobin’s q at enterprise level is the capital markets representation of the net present value of all the firms past and projected investment projects (Moyer et al, 2010).. Beside the performance measures Hawawini et al. (2003) used, that we tried to replicate we also include other performance measures into our research. The first additional performance measure is the Tobin’s Q at equity value, also known as the market to book ratio. This is the value of common shares minus the book value of equity divided by the book value of equity. The last performance measure we use is the Tobin’s q (Tobin, 1969) as described in the literature and is defined as the value of the firm relative to its replacement value. The Tobin’s Q ratio is calculated as the enterprise value divided by total assets.. Variable name BVTA BVD MVD NI BVE EV IT PI Beta. Table 2 Variable definitions and sources* Description Sources Book value total assets Datastream code WC02999 Book value debt Datastream code WC03255 Market value debt Book value debt Net income before preferred Datastream code WC01651 dividends Book value common equity Datastream code WC03501 Net Enterprise value Datastream code WC18100 Income taxes Datastream code WC01451 Pretax income Datastream code WC01401 Beta Datastream code Beta897E. 9.

(10) Is performance driven by firm- or industry specific factors?. MC IE EBITA SALES CCE RF MP SP. Market capitalization Interest expense Earnings before interest, tax and amortization1 Sales Cash and cash equivalents US 10 year treasury bond yield Market premium2 Spread based on ICR over RF3. Datastream code WC08001 Datastream code WC01251 Datastream code WC01250 Datastream code WC01001 Datastream code WC02001 www.damodaran.com www.damodaran.com www.damodaran.com. *all data is in US dollars. EC ICR COE COD TR WACC IC ROA ROIC ROIC-WACC Tobins Q1. Table 3 Variable definitions and calculations* Working cash 2% Excess cash.

(11) , 0 Interest coverage ratio / Cost of equity . Cost of debt Tax rate Weighted average cost of 1 capital ! ! Invested capital. ! ! # Return on assets $% /2 1 Return on invested capital $% /2 Value per dollar invested ! Tobins Q at enterprise level . Tobins Q2. Tobins Q at equity level. Tobins Q3. Tobins Q as described in the literature. ! ! !. *all data is in US dollars. 2.3 Previous studies Studies on the impact of industry- and firm effects that explain performance variations are relatively new (see table 1 for the most relevant ones). The first one to examine the fundamental disagreement between the industry and firm effects was Schmalensee (1985) who used a cross sectional dataset to analyse the contribution of firm or industry effects. Schmalensee (1985) used return on assets as performance measurement. His study showed that industry effects played an important role in explaining performance and that firm effects were insignificant. The paper by Wernerfelt and Montgomery (1988) supported this result. In 1. Represents the difference between sales and total operating expenses Yearly average market risk premium (return of the market – risk free return) 3 Spread based on the interest coverage ratio indicating the spread over the risk free rate. This table used is the most recent from 2010 tables from previous years could not be found. 2. 10.

(12) Is performance driven by firm- or industry specific factors?. contrast, Rumelt (1991), found that firm effects (measured as business unit effects) were dominant in explaining performance. Therefore his result was the first indication that firms within an industry can be fundamentally different and that those differences are economically and statistically relevant and therefore can explain variation in performance (Eriksen and Knudsenb, 2003). The studies of Schmalensee (1985) and Rumelt (1991) are both based upon Federal Trade Commission (FTC) data, which is limited to very large manufacturing firms. Later research of Roquebert et al. (1996), McGahan and Porter (1997), Mauri and Michaels (1998),Chang and Singh (2000) and Hawawini et al. (2003)with similar methodology used Compustat data which included service industries, smaller firms, and contained data at the level of business segments. They all found that firm effects were dominant (i.e. has more explanatory power in explaining performance variation) in explaining firm performance. An explanation can be the data that is used, and that in large manufacturing firms industry plays an important role in explaining performance variation, while in other industries with smaller firms the resources of a firm are more dominant. Although the sizes of the effects differ, the latest results all rule in favour of the resource based view that the firm’s unique resources have a larger impact in explaining performance compared to industrial organization theory. Due to the research that is in favour of the resource based view the question is if the industrial organization theory is still important in explaining firm performance. Knowing that both theories can explain performance we want to know the size of these effects. Excluding an effect will increase the error and therefore these effects are not mutually exclusive. Roquebert et al. (1996), McGahan and Porter (1997), Mauri and Michaels (1998),Chang and Singh (2000) and Hawawini et al. (2003) found some industry effects, however these effects are significantly smaller than the firm effects. Hawawini et al. (2003) confirms previous findings that industry effects do not matter much for the firm performance. 2.4 Industry winners and losers Hawawini et al. (2003) observed that a few firms per industry outperform the rest of the firms (hereafter called leaders). Besides leaders, every industry also has firms that underperform (hereafter called losers). He identified these leaders and losers by selecting the two firms with the highest or lowest average performance measure over the entire sample period. If a sustainable competitive advantage results in superior performance, a competitive disadvantage results in below average performance. Therefore these leaders and losers should influence the results and the conclusion. Because firm effects drive the performance of leaders and losers, interesting is what drives the performance of the firms who aren’t leaders and 11.

(13) Is performance driven by firm- or industry specific factors?. losers? Therefore we also examine if the leaders and losers influence the relative magnitude of the firm- and industry effects, and the changes in the effects. Hawawini et al. (2003) found that industry effects increased due to removal of the two leaders and losers, and for some performance measures industry effects even dominated firm effects. Hawawini et al. (2003) suggested that a few firms in an industry create or destroy a large part of the industry value and therefore firm effects dominate industry effects. Resulting, that industry only matters when a firm is not a leader or a loser within his industry. McNamara et al. (2005) criticised the sample selection methodology of Hawawini et al. (2003) because they argue that with increasing artificial restrictions in example excluding leaders and losers the wrong way you can obtain biased results. They also argue that the method of detecting outliers employed by Hawawini et al. (2003) is not appropriate. Instead of removing a fixed amount leaders and losers in each industry McNamara et al. (2005) argue that outliers must be identified by taking at least three standard deviations from the industry average performance, following Cohen et al.’s (2003) guidelines of re-estimating variance component effects after excluding firms with average performance at least three standard deviations from their industry’s average performance. Hawawini et al. (2005) responded to this criticism, that removing outliers based on at least three standard deviation leads to the same result. And therefore the result of McNamara et al. (2005) only strengthens the findings of Hawawini et al. (2003). On the topic that the range of restrictions on the firm-level variance that were employed are less than ideal Hawawini et al. (2005) concludes that in every industry there are a limited number of winners and losers, which in turn have an important influence on the relationships of firm and industry effects on firm performance, and that applying range restriction on other effects like McNamara et al. (2005) did does not alter but supports the findings of Hawawini et al. (2003). In this paper outlier selection and the mean idea of removing those outliers is identical to the motivations of Hawawini et al. (2003). 3.. Data and methodology. 3.1 Data The data used in this paper to calculate the dependent variables (return on assets, return on invested capital, and return on invested capital minus weighted average cost of capital, Tobin’s q at enterprise level, tobin’s q at equity level and Tobin’s q as described in the literature)and the independent variables (industry-, firm-, year-, industry year- and country effects) are extracted from the Thomson Reuters DataStream database. The dataset. 12.

(14) Is performance driven by firm- or industry specific factors?. covers1200 firms who are included in the Standard and Poor’s 1200 global index, also written as S&P 1200 at 15-03-2012. This index is a free float weighted stock market index that captures approximately 70% of the capital markets in the world. The index is a composite of 31 local markets of 7 headline indices: S&P 500 (United States), S&P Europe 350, S&P TOPIX 150 (Japan), S&P/TSX 60 (Canada), S&P/ASX All Australian 50, S&P Asia 50, and S&P Latin America 40.From this dataset 21 years of data is gathered from 1990 until 2010. We removed outliers from the data following Cohen et al (2003) guidelines for estimating variance component effects meaning that we removed data points that deviated more than three standard deviations from the firms mean.. A bias in this kind of performance studies is the survivor bias. This is the tendency for failed firms to be excluded from performance studies. Also our dataset uses only the firms that survived or were admitted in the S&P 1200 during the time period. This may result in a performance measure that is more skewed. This leads to obvious biases in first and second moments and cross moments including Beta (Brown et al., 1992). Therefore we can partly explain the post-earning-drift-phenomenon which is the tendency of a share price to drift upor downwards weeks after a firm has announced an earnings up- or downgrade noted by Ball and Brown (1968). The dataset used in this paper does include data at a business level. Due to the fact that only firm level is available. To divide the firms into industries we used the Industry Classification Benchmark (ICB) this system is developed FTSE and Dow Jones and its main purpose is to segregate firms into groups within the macro economy. The ICB is a classification system with four levels. At the first level 10 industries, the second level gives 20 so called super sectors, at the third level contains 41 sectors and the final fourth level exists of 114 subsectors. The super sectors (level 2) are used in this paper to identify the different industries. We also tested the same sample using sectors (level 3) and found similar results. In this paper we only used level 2 classifications (see appendix B) because we also want to remove leaders and losers and if we divide the firms according to level 3 some sectors would be excluded. The ICB is a global used system used by the NASDAQ, NYSE and other exchanges around the world. We also included a country effect to test for the country effects because we use a worldwide dataset. We used the FSTE country code classification which is based upon where the headquartersof each firm is situated. The firms are divided into 27 groups(see appendix C), each group representing a country except Belgium and Luxembourg and Chile and Brazil that we merged into one group due to the fact of small number of. 13.

(15) Is performance driven by firm- or industry specific factors?. observations. We also divided the firm into continental groups. However there where zero to very small effects found. Therefore we did not further examine these effects.. 3.2. Model. Study. Table 4 Data and methodology of relevant previous studies Dataset Method. Schmalensee (1985). Rumelt (1991). McGahan and Porter (1997). Hawawini et al. (2003). 1975 US Federal Trade Commission line of business dataset with 1,775 observations 1974-1977 US Federal Trade Commission line of business dataset with 6,932 observations 1981-1994 US Compustat Business Segment Reports data with 72,742 observations 1987-1996 Stern Stewart dataset covering US firms with 5,620 observations. Ordinarily least squares and variance component analysis ANOVA and variance component analysis ANOVA and variance component analysis variance component analysis. The model we use in this study to examine the industry, firm, country, industry year and year effects is the variance component analysis (VCA) model that is also used by Schmalensee (1985), Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003). We however added country effects as an independent variable to explain the performance variation that is contributed by the country the firm is based. The model we use for our estimations is presented as follows. &'()* + ,…. /' 0( 1) 2* /1') 3'()*. 1. &'()* + 4&56&789 7 :& ;48;8< =&>? ,…. + 968 <8< /' + >8;: <&@ 559< 0( + 5>&7 559< 1) + @& 559< 2* + 96:8<&@ 559< /1') + >8;: <&@ @& >8<&9<>68 559< 3'()* + 86&7 &8;67 ;> <&>?:<>68 The dependent variable &'()* is the performance measure (return on assets, return on invested capital, return on invested capital minus weighted average cost of capital, Tobin’s q at 14.

(16) Is performance driven by firm- or industry specific factors?. enterprise level, Tobin’s q at equity level and Tobin’s q as described in the literature) for industry >, firm j, year t and country c. The first right hand side variable is µ.... which is a constant that incorporates the average performance over all firms during the whole period (the four dots stand for the averages of the i, j t and c indices). /' is the random industry effect and i = 1 . . . r represents the industries. 0( is the random firm effect and where j = 1 . . . ni displays the firm and ni the number of firms within industryi. 1) is the random year effect where t is one year.2* is the random country effect. And/1') is the random industry year interaction effect also called the transient industry effect, and the last variable 3'()* denotes the random error term.Other interaction effects were also tested but did this not present any results therefore we excluded them from our analyses. The model distinguishes five sources of variations in the performance measures: stable and transient industry effects, stable firm effects, effects of macroeconomic fluctuations captured by the year effect, the country effect and the random error. In this paper the firm effects consists of both corporate and business unit effects. Rumelt (1991) states that corporate effects arise from differences in the quality of monitoring and control, differences in resource sharing and other types of synergy, and differences in accounting policy and therefore are similar to our firm effects. And business unit effects represent persistent differences among business-unit returns other than those due to industry and corporate membership.Therefore we are not able to test for intra-industry heterogeneity. The firm effects we use capture firm specific factors like heterogeneity among firms in fixed and variable assets due to the differences in managerial skills, organizational process, operational effectiveness and reputation. The transient industry effect reflects the sensitivity of the performance measures to the impact of industry specific cycles on the industry. The stable industry effects measure the influence of structural characteristics of industries on the performance of firms. The macroeconomic effects are covered by the year effect in the model. And the effects of the location of the headquarters are covered by the country effect.. 3.3 Methodology First we compute the descriptive statistics of all the performance measures. And we execute the test for normality with the Kolmogorov-Smirnova test of normality (Chakravarti, Laha, and Roy, 1967). The Kolmogorov-Smirnova test calculates the discrepancy between the distribution of the data used and an ideal Gaussian distribution also called a normal distribution. Larger test statistics indicate higher discrepancies between the variables. The. 15.

(17) Is performance driven by firm- or industry specific factors?. statistics are not very informative by themselves, but are used in calculating the p-value also known as the significance value. This value is used to test the hypothesis if the data is normally distributed. The results are presented in table 5.. Table 5 Descriptive statistics ROA. Statistic. ROIC. ROICWACC. TOB1. TOB2. TOB3. N. Statistic. 20621. 20783. 18173. 20242. 20623. 20708. Minimum. Statistic. -1.51. -16.06. -16.20. .02. .00. .00. Maximum. Statistic. .88. 228.23. 228.11. 139202.11. 359.08. 73.47. Mean. Statistic. .05. .22. .13. 10.67. 3.35. 1.40. Std. Deviation. Statistic. .08. 1.72. 1.80. 978.62. 5.80. 1.65. Skewness. Statistic. -2.49. 113.24. 112.28. 142.17. 23.43. 10.97. Kurtosis. Statistic. 48.31. 14881.71. 14114.98. 20221.82. 1005.90. 300.67. Statistic KolmogorovSmirnova. df Sig.. .14. .43. .41. .50. .30. .18. 17100. 17100. 17100. 17100. 17100. 17100. .00. .00. .00. .00. .00. .00. ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. The means of the performance measures return on assets, return on invested capital, return on invested capital minus weighted average cost of capital, Tobin’s q at enterprise level,Tobin’s q at equity level and Tobin’s q as described in the literature are 5%, 22% , 13%, 10.67, 3.35 and 1.40. The standard deviation of the variable Tobin’s q 1 at enterprise level is extremely high, indicating that the data is spread over a large value range. The skewness for all the performance measures except return on assets is large and positive is indicating that the right tail of the distribution is longer and the mass is concentrated on the left. The high kurtosis indicates that the distributions have a sharper peak and longer and fat tails compared to a normal probability distribution. These extreme values are not surprising because we tested if the measures were normally distributed with the Kolmogorov-Smirnova test of normality (Chakravarti, Laha, and Roy, 1967). From the results you can see that all of the performance data is not normally distributed i.e. (Gaussian distribution). Therefore we can reject the hypothesis that these performance measures are normally distributed because for all measures the p-value is smaller than 0.05. This is an indication of extreme values (outliers) in the data. Therefore we further examine the data with non-parametric tests for non-normal data like the non-parametric Levine’s test (Nordstokke and Zumbo, 2010), Kruskal-Wallis test (Kruskal and Wallis, 1952), and the variance component analysis (Searle, Casella and McCulloch, 1992) and (Cox and Solomon, 2002). These tests will be discussed hereafter.. 16.

(18) Is performance driven by firm- or industry specific factors?. Second, we calculated a correlation matrix on the performance measures which is presented in table 6. There is a high correlation of 0.96 between return on invested capital and return on invested capital minus weighted average cost of capital and a correlation of 0.88 between Tobin’s q at enterprise level and Tobin’s q at equity level. What is somewhat surprising because these performance measures are calculated on different levels (enterprise and equity level). The average correlation over all the performance measures is 0.73. Hawawini et al. (2003) found an average correlation of 0.81. All the correlations are all larger than0.5 and significant at a 0.01 level (two-tailed). Table 6 Correlations ROA. ROIC. ROICWACC. TOB1. TOB2. TOB3. ROA. 1.00. .76. .73. .63. .50. .68. ROIC. .76. 1.00. .97. .69. .56. .56. ROIC-WACC. .73. .97. 1.00. .65. .54. .52. TOB1. .63. .69. .65. 1.00. .88. .82. TOB2. .50. .56. .54. .88. 1.00. .74. TOB3. .68. .56. .52. .82. .74. 1.00. All correlations are significant at a 0.01 level (two-tailed) ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. Further, we calculated the means and medians of each industry, country and year. We do not use these results any further nor draw any conclusion on them, but for the interesting reader these results are presented in appendix D.. The first non-parametric test we perform is the non-parametric Levine’s test (Conover, Johnson, and Johnson, 1984).It involves performing a traditional Levine F test on pooled ranked data. This is the most robust and powerful test when analyzing non distributed data (data that does not follow a normal distribution) that is skewed and has unequal sample sizes (Nordstokke and Zumbo, 2010). This test is used to assess variance homogeneity. If the pvalue is smaller than 0.05 then variances of the performance measures between groups of each effect (industry, firm, country and year) are significantly different from each other. Therefore you can reject the hypothesis of equal variances between the groups within the effect. The following test we apply is the Kruskal-Wallis analysis (Kruskal and Wallis, 1952) which is a main generalization of the Mann Whitney U test and the Wilcoxon Rank-Sum test; however with the Kruskal-Wallis test it is possible to compare more than two groups. Thistest. 17.

(19) Is performance driven by firm- or industry specific factors?. is used when the data is non-normally distributed which is the case. If the p-value is smaller than 0.05 you can conclude that there are differences between the mean values of the tested groups of each effect. Then we can reject the hypothesis of no differences between the mean values of the different groups. Another useful output from the Kruskal-Wallis test is that you can estimate the independent effect size estimate by taking the Chi-Square divided by # 1,where N is the sample size. The effect size estimate is a percentage that explains the variability in rank scores is accounted for that effect tested (industry, firm, country and year). We want to emphasize that all the effect are independently tested for their effect size estimates so there is no interaction between the different effects. This is however tested by the final test the variance component analysis (VCA). The variance component analysis (Searle, Casella and McCulloch, 1992; Cox and Solomon, 2002) estimates the contribution of each effect to the variance of the dependent variable (performance measure) and transforms these values to calculate relative percentage of the total variability of each of the effects. The method we use to test these effects is the variance component analysis. The variance component analysis is also used by Schmalensee (1985) Rumelt (1991) and Hawawini et al. (2003). The descriptive model displayed in equation 1 is used to calculate the variance component estimates by decomposing the variability of the performance measure in its effects as shown in equation 2. C ABC + ADC AEC AFC AGC ADF AHC. 2. The 0, 1 and 2 are the main effects and /1 is the interaction effect of the model. All theeffects. follow. a. normal. random. distribution. with. a. zero. mean. and. C varianceADC ,AEC ,AFC ,AGC ,ADF , and AHC also known as 30, A C . The effects of the model described. and used in this paper, are all random independent effects. These effects are calculated by random processes that are independent of each other, indicating that that every effect is an independent random solution forms an underlying population which is normally distributed. In this study we use VARCOMP procedure in SPSS (PASW) statistics software. The dataset used in this paper is a sample therefore it is impossible to cover all firms in all industries and because we have unequal groups with missing values the variance component analysis is very suitable test because it allows us to generalize the results over the whole population and not just the sample, which is not allowed under the assumption that one or more effects are fixed. Because firm effects drive the performance of leaders and losers, what drives the performance. 18.

(20) Is performance driven by firm- or industry specific factors?. of the firms who aren’t leaders and losers? Therefore we also examine if the leaders and losers influence the relative size of the firm- and industry effects. We identify the leaders and losers by taking their average performance during the sample period (21 years). And exclude the top five and ten leaders and losers in each industry based on average performance. We first tested the whole sample for effects (1200 firms), second we exclude the five leaders and losers from each industry (1020 firms) and finally we exclude the ten leader and losers (840 firms). The advantage of assuming random effects are that the results can be generalized over the whole population and not only on the sample. A disadvantage is that variance component analysis does not provide reliable tests for the significance of the independent effects. However Schmalensee (1985), Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003) argue a test for significance is not the primary goal of their study, that is the magnitude of the effects can be used as an indicator that the parameter is non-zero (Roquebert et al.1996). This significant testing problem can be overcome by using fixed effects; however it restricts the general assumption of randomness of the independent effects. And therefore the results and conclusions cannot be generalized. Therefore Schmalensee (1985), Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003) all use the random effects model that assumes that all the independent effects are generated by a random process, consistent with the variance component assumptions.. 19.

(21) Is performance driven by firm- or industry specific factors?. 4.. Results. The first result we shortly discuss is the non-parametric Levine’s test (Nordstokke and Zumbo, 2010) (Appendix E). This test is used to assess variance homogeneity. The results from this test indicate that among all four effects and all performance measures the hypothesis of equal variances between groups can be rejected because all p values are smaller than 0.05. The second test we applied is the Kruskal-Wallis analysis (Kruskal and Wallis, 1952) displayed in table 7. We could not add the interaction effect so this effect is not tested. If the p- value is smaller than 0.05 you can conclude that there are differences between the medians of the tested groups. From the results we can conclude that we can reject the hypothesis of no differences between the mean values of the different groups. Table 7 Kruskall-Wallis test statistics Industry Chi-square df Asymp. Sig.. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 3950.8. 4299.6. 2953.4. 5391.8. 2819.0. 7133.7. 17. 17. 17. 17. 17. 17. .000. .000. .000. .000. .000. .000. Country ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 1819.3. 1999.1. 1349.8. 1693.4. 1806.3. 2234.7. 26. 26. 26. 26. 26. 26. .000. .000. .000. .000. .000. .000. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 11474.8. 12535.0. 10187.8. 13654.0. 12018.8. 15723.8. df. 1186. 1173. 1150. 1167. 1184. 1178. Asymp. Sig.. .000. .000. .000. .000. .000. .000. Chi-square df Asymp. Sig.. Firm Chi-square. Year Chi-square df Asymp. Sig.. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 656.4. 382.9. 434.4. 769.6. 1034.7. 572.0. 19. 20. 20. 20. 20. 20. .000. .000. .000. .000. .000. .000. N (sample sizes) N (sample sizes). ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 20621. 20783. 18173. 20242. 20623. 20708. ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. Another useful output from the Kruskal-Wallis test is that you can estimate an effect size estimate by taking the Chi-Square dividing the sample size minus 1, the effect size estimates are presented in table 8.. 20.

(22) Is performance driven by firm- or industry specific factors?. Table 8 Kruskall-Wallis independent effect size estimates Effect. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. Industry. 19.2%. 20.7%. 16.3%. 26.6%. 13.7%. 34.5%. Firm. 55.6%. 60.3%. 56.1%. 67.5%. 58.3%. 75.9%. Country. 8.8%. 9.6%. 7.4%. 8.4%. 8.8%. 10.8%. Year. 3.2%. 1.8%. 2.4%. 3.8%. 5.0%. 2.8%. ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. The different effects size estimates are independently tested, indicating that each effect size calculated only explains the variation found for a single effect, and therefore does not take other effects into account. This is why the sum of the effects for each measure is not hundred percent.. For all the performance measures the firm effects have the largest effect size estimate, followed by industry effect, country effect and finally the year effect. For all the measures we can rank the calculated effect size estimates from large to small: >&7 I 8;: <&@ I 6:8<&@ I J&. This is the first indication that firm effects dominate industry effects. However these estimates are calculated separately from each other. In the variance component analysis we calculate the variance components of all the effects in one test.. The final test discussed is the variance component analysis (Searle, Casella and McCulloch, 1992)( Cox and Solomon, 2002) which is used to determine the variability for different effects for each of the performance measures and transform these values to calculate relative percentage of the total variability of each of the effects. In appendix F the variance estimates are displayed in absolute values. When dividing the variance estimates of each effect by the total variance you transform these variances into percentages displayed in table 9.. Table 9 Variance Estimates (whole sample as a percentage) Effect Industry. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 5.4%. 0.6%. 0.5%. 0.0%. 3.1%. 14.2%. 29.3%. 21.8%. 21.4%. 12.1%. 63.4%. 40.5%. Country. 5.8%. 0.1%. 0.1%. 0.0%. 0.9%. 1.1%. Year. 1.8%. 0.0%. 0.0%. 0.0%. 0.7%. 1.6%. Firm. Industry*Year. 3.1%. 0.0%. 0.0%. 0.0%. 0.8%. 3.6%. Error. 54.5%. 77.5%. 78.0%. 87.9%. 31.1%. 39.1%. Total. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. 21.

(23) Is performance driven by firm- or industry specific factors?. ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. Below we briefly analyze most interesting findings in table 9. For all performance measures the firm effect is the largest in magnitude (29.3%, 21.8%, 21.4%, 12.1%, 63.4% and 40.5%) and that the industry effect (5.4%, 0.6%, 0.5%, 0.0%, 3.1% and 14.2%) is relatively small compared to the firm effect for all the performance measures. Firm effects dominate industry effects by at least the factor 2.9. Therefore we can conclude that resource based view dominates the industrial organization theory. The performance variation among firms is due to idiosyncratic and unique resources rather than the industries structural characteristics (Wernerfelt, 1984) i.e. >&7 559< I 8;: <&@ 559< → ! I . In this study we also examined year effects that should capture the macroeconomic fluctuations. These effects are relative small and are most pronounced in return on assets (1.8%) of all the performance measures. This indicates that the performance variability for firms in our sample is very small. The country effects measured are also relatively small compared to firm effects. This indicates that the country where the firm is situated does not explain performance variability to a large extend. Therefore we can conclude it does not matter much in explaining performance variability. This is the result due to globalisation and open economies. However, when we use return on assets as the performance measure we see that the country effect (5.8%) is relatively large in magnitude and it also dominates the industry effect (5.4%).The industry year effect is relatively large with return on assets (3.1%) and Tobin’s q (3.6%) as the performance measures. This is remarkable because they are both accounting measures. The other performance measures all show an industry year effect smaller than 0.8%. Remarkable is that if we use the performance measures return on invested capital, return on invested capital minus weighted cost of capital, and Tobin’s Q at enterprise level all show effect estimates smaller than 0.6% for the industry, country, year and industry-year effects.. Now we compare our results with that of Schmalensee (1985), Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003). A comparison of the results can be found in Table 10. The firm effects found when using return on assets as a performance measure have a somewhat smaller impact compared to the studies of Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003) and are larger compared to the study of Schmalensee (1985). This is also the case with the performance measures return on invested capital and Tobin’s q at enterprise level who Hawawini et al. (2003) calls EP/CE and TMV/CE. Also the. 22.

(24) Is performance driven by firm- or industry specific factors?. examined industry effects are smaller compared to the effects of the other studies. One explanation can be that Schmalensee (1985), Rumelt (1991), McGahan and Porter (1997) and Hawawini et al. (2003) all used a dataset that covered the US, while this study uses a worldwide dataset, indicating that firms within industries within a country are more connected with each other compared to firms within an industry global, therefore our industry effects are less pronounced.. Table 10 Variance component comparison with previous studies Measure Firm Industry Year Country Industryeffects effects effects effects year effects Schmalensee (1985) ROA 0.63% 19.46% NT NT NT Rumelt (1991) ROA 45.8% 4.0% NT NT 5.4% McGahan and Porter (1997) ROA 36.0% 18.7% 2.4% NT NT ROA 35.8% 8.1% 1.0% NT 3.1% Halawini (2003) EP/CE 27.1% 6.5% 1.9% NT 4.2% (full sample) TMV/CE 32.5% 11.4% 1.3% NT 2.9% ROA 29.3% 5.4% 1.8% 5.8% 3.1% ROIC 21.3% 0.6% 0.0% 0.1% 0.0% ROIC21.4% 0.5% 0.0% 0.1% 0.0% This study WACC TOB1 12.1% 0.0% 0.0% 0.0% 0.0% TOB2 63.4% 3.1% 0.7% 0.9% 0.8% TOB3 40.5% 14.2% 1.6% 1.1% 3.6% Study. Error. 80.54% 44.8% 48.4% 52.0% 60.3% 51.9% 54.5% 77.5% 78.0% 87.9% 31.1% 39.1%. ROA, return on assets; EP/CE, economic profit divided by capital employed; TMV/CE, total market value divided by capital employed; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature; NT, not tested.. In table 11 we discuss the impact of performance leaders and losers (the absolute values are presented in appendix F) and the change in firm- or industry effects. If we exclude the top five leaders and losers of each industry, the industry effects increase for each performance measure except return on invested capital. Also the firm effects decrease with all performance measures except for Tobin’s q at enterprise level. However, the firm effects still dominate the industry effects. If we exclude the top 10 leaders and losers the industry effects show even larger percentages. The firm effect of Tobin’s q at enterprise level increases from 0.0% to 16.0%. And the for the Tobin’s q as described in the literature (TOB3) the industry effect dominates firm effects. Table 11 Variance Estimates (sample excluding top 5 leaders and losers as a percentage) Effect Industry Firm Country. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. 9.6%. 0.5%. 0.6%. 0.0%. 10.4%. 22.6%. 19.5%. 18.6%. 20.3%. 12.3%. 30.9%. 29.6%. 3.0%. 0.1%. 0.1%. 0.1%. 1.8%. 1.5%. 23.

(25) Is performance driven by firm- or industry specific factors?. Year. 2.4%. 0.0%. 0.0%. 0.0%. 2.9%. 2.0%. Industry*Year. 4.5%. 0.0%. 0.0%. 0.0%. 4.0%. 4.6%. Error. 60.9%. 80.7%. 79.0%. 87.6%. 50.0%. 39.7%. Total. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. Variance Estimates (sample excluding top 10 leaders and losers as a percentage) Effect. ROA. ROIC. ROIC-WACC. TOB1. TOB2. TOB3. Industry. 11.4%. 9.5%. 5.1%. 16.0%. 15.1%. 26.8%. Firm. 14.1%. 27.0%. 73.1%. 28.9%. 23.0%. 22.2%. Country. 1.6%. 1.0%. 0.5%. 0.5%. 1.5%. 1.0%. Year. 2.7%. 0.6%. 0.6%. 0.9%. 3.7%. 2.3%. Industry*Year. 4.5%. 3.3%. 1.1%. 2.8%. 4.7%. 5.2%. Error. 65.7%. 58.6%. 19.6%. 50.9%. 52.1%. 42.6%. Total. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. 100.0%. ROA, return on assets; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature.. Therefore the results imply that in every industry a few firms are able to create or destroy industry value and dominate intra-industry variance and firm effects. Hawawini et al. (2003) also found similar results (see table 12). They experienced a large decrease in firm effects and the industry effect doubled for each of his performance measures. We experienced somewhat less decreases in firm effects and increases in industry effect, however we think that due to the use of a larger dataset the exclusion has a smaller impact compared to the results of Hawawini et al. (2003). Table 12Modified variance component comparison with previous studies Measure Firm Industry Year Country Industryeffects effects effects effects year effects Halawini (2003) ROA 17.6% 12.4% 1.9% NT 5.8% (modified sample excluding EP/CE 17.0% 30.2% 2.5% NT 5.0% the top and bottom 2 from TMV/CE 16.7% 16.0% 1.1% NT 4.1% each industry) ROA 19.5% 9.6% 2.4% 3.0% 4.5% ROIC 18.6% 0.5% 0.0% 0.1% 0.0% This study modified sample ROIC20.3% 0.6% 0.0% 0.1% 0.0% excluding top 5 leaders and WACC losers from each industry TOB1 12.3% 0.0% 0.0% 0.1% 0.0% TOB2 30.9% 10.4% 2.9% 1.8% 4.0% TOB3 29.6% 22.6% 2.0% 1.5% 4.6% Study. This study modified sample excluding top 10 leaders and losers from each industry. ROA ROIC ROICWACC TOB1 TOB2 TOB3. Error. 61.0% 45.3% 62.1% 60.9% 80.7% 79.0% 87.6% 50.0% 39.7%. 14.1% 27.0% 73.1%. 11.4% 9.5% 5.1%. 2.7% 0.6% 0.6%. 1.6% 1.0% 0.5%. 4.5% 3.3% 1.1%. 65.7% 58.6% 19.6%. 28.9% 23.0% 22.2%. 16.0% 15.1% 26.8%. 0.9% 3.7% 2.3%. 0.5% 1.5% 1.0%. 2.8% 4.7% 5.2%. 50.9% 52.1% 42.6%. ROA, return on assets; EP/CE, economic profit divided by capital employed; TMV/CE, total market value divided by capital employed; ROIC, return on invested capital; ROIC-WACC, return on invested capital minus weighted average cost of capital; TOB1, Tobin’s q at enterprise level; TOB2, Tobin’s q at equity level TOB3, Tobin’s q as described in the literature; NT, not tested.. 24.

(26) Is performance driven by firm- or industry specific factors?. Hawawini et al. (2003) also concluded that the high firm effects are the result of a few firms that substantially and consequently deviate from the rest of the industry. So including only a few leaders and losers firm specific factors matter more than industry factors. For the firms who aren’t leaders or losers within their industry the industry effect becomes more important and increases for firms who are “stuck in the middle”. 5.. Conclusions. This study examined the importance and magnitude of firm and industry effects on the performance measures return on assets, return on invested capital, return on invested capital minus weighted average cost of capital, Tobin’s q at enterprise level, Tobin’s q at equity level and Tobin’s q as described in the literature. Because we used a worldwide dataset we also included country effects in our analysis, and to capture macroeconomic influences we added the year effect. First we tested the whole sample for these effects and in de second and third run we excluded the top five and later the top ten leaders and losers from each industry. The results are in line with the results from Hawawini et al. (2003). First, we see for all the performance measures that firm effects dominate industry effects. Second, there is not a large difference in magnitude between the effects on accounting- or value measures. Overall the firm effects dominate industry effects by at least the factor 2.9. Therefore we cannot conclude that the use of return on assets in previous research (Schmalensee, 1985; Rumelt, 1991; McGahan and Porter, 1997 and others) was a limiting factor by calculating the firm and industry effects.. For the year and country effects we found relatively small variances for the performance measures, the year effect was smaller than 5.8% and the country effects were less than 1.8% indicating that the home country effect is not very important in explaining performance variation. An explanation could be the increasing internationalization. Second, we examined the change in the industry- and firm effects of the firms who did not out- or underperform in their industry (the so called leaders and losers). We excluded the top five and top ten leaders and losers of each industry from our sample. The results show that industry effects increase and for some performance measures even dominate firm effects. Therefore these results indicate that industry effects have different impact on firms within an industry. Industry factors have a larger impact on the performance of the firms who are stuck in the middle of their industry compared to the outperformers and down performers where the industry effect is low and the firm effects have a large effect. This is the same for all performance measures. 25.

(27) Is performance driven by firm- or industry specific factors?. A driver for these effects could be that there are outperforming or down performing industries. The key empirical basis for the high firm effects observed in previous studies is that the intra industry variance was notably larger compared to the inter industry performance variance Hawawini et al. (2003).. Like Hawawini et al. (2003) we built on the empirical basis that most of the intra industry variance is caused by a few under- and outperformers so that removing a whole industry does not alter the results. Our results show a large amount of variance error, we found an average error of 61%, however previous studies found the following variance error: Hawawini et al. (2003) 55% Rumelt (1991) 45% and Schmalensee (1985) 80%. We want to emphasize that the firm effects only dominate the explained variations in the performance measures. There is a large amount of performance variation unexplained, indicating that there is room for other effects that could explain the performance variation. Effects not mentioned in this study are the interaction variables country-year industry-country, country-firm, firm-year and the industry-firm effects because we found small to zero effects.. Hawawini et al. (2003) argued that ICB industry coding is supply based and ignores other dimensions like consumer segmentation on the demand side. Therefore you can argue if industry classification is the right way to test for effects used in this study. The second limitation is that we used the S&P 1200 for our sample which consists of only large firms. Therefore we cannot generalize this finding on medium and small firms. Indicating more research is needed on the effects that influence performance variability for medium and small firms, because it is interesting to know if our results also hold for medium and small firms. Second, there is a lot of performance variance unexplained and it is interesting to research this unexplained performance variability.. 26.

(28) Is performance driven by firm- or industry specific factors?. References Amit, R. and P. Schoemaker, 1993, Strategic assets and organizational rent, Strategic Management Journal 14, 33–46. Arend, R., 2009, Industry effects and firm effects: No effect is an island, Journal of Business Research, 62, 651–659. Ball, R and P. Brown, 1968, An empirical examination of accounting income numbers, Journal of Accounting Research 6, 159–178. Bain, J.S., 1956, Barriers to New Competition (Harvard University Press, Cambridge, MA). Bain, J.S., 1959, Industrial Organization (John Wiley & Sons, New York, NY). Barney, J., 1991, Firm resources and sustained competitive advantage,Journal of Management 17, 99– 120. Baumol,W.J., 1967, Business Behaviour, Value and Growth (Harcourt Brace and World Inc., New York, NY). Bettis, R.A. and M.A. Hitt,1995, The new competitive landscape, Strategic Management Journal16, 7–19. Brigham, E. F and M. C. Ehrhard, 2010,Financial Management Theory and Practice (South-Western Cengage Learning, Mason, OH). Brown, S.J., W. Goetzmann, R.G. Ibbotsen and S. A Ross, 1992, Survivor bias in performance studies,The Review of Financial Studies 5, 553–580. Brush, T.H. and P. Bromiley, 1997, What does a small corporate effect mean? A variance components simulation of corporate and business effects, Strategic Management Journal 18, 825–835. Brush, T.H., Bromiley, P. and M. Hendrickx, 1999, The relative influence of industry and corporation on business segment performance: an alternative estimate, Strategic Management Journal20, 519– 547. Cabrall, L.M.B., 2000, Introduction to Industrial Organization, (MIT Press: Cambridge, MA). Caloghirou, Y., Protogerou, A., Spanos, Y. and L. Papagiannakis, 2004, Industry-versus firm-specific effects on performance: contrasting SMEs and large-sized firms, European Management Journal 22, 231–243. Chakravarti, I.M., Laha, R.G. and J. Roy, 1967, Handbook of Methods of Applied Statistics, Volume I (John Wiley and Sons, New York, NY). Chang, S.J. and H. Singh, 2000, Corporate industry effects on business unit competitive position, Strategic Management Journal21, 739–52.. 27.

(29) Is performance driven by firm- or industry specific factors?. Chen, Y., 2009, Determinants of industry performance: region vs. country effects in knowledgeintensive service industries,The Service Industries Journal 29, 297–316. Chen, Y. and F. Lin, 2006, Sources of superior performance: Industry versus firm effects among firms in Taiwan, European Planning Studies 14, 733–751. Cinca, C. S., Molinero C. M. and J.L. GallizoLarraz, 2005, Country and size effects in financial ratios: A European perspective, Global Finance Journal 16, 26–47. Claver, E., Molina, J. and J. Tari, 2002, Firm and industry effects on profitability: a Spanish empirical analysis,EuropeanManagement Journal 20, 321–328. Cohen, J. Cohen, P. West S.G. and L.S. Aiken, 2003, AppliedMultiple Regression Correlation Analysis for the Behavioral Sciences,(Lawrence Erlbaum Associate,Mahwah, NJ). Conover, W.J., Johnson, M.E. and M.M. Johnson, 1981, A comparative study of tests for homogeneity of variances, with applications to the outer continental shelf bidding data, Technometrics 23, 351–361. Cox, D.R. and P.J. Solomon, 2002, Components of Variance, (Chapman and Hall/CRC, New York, NY) Damodaran, A., 2007, Return on Capital (ROC), Return on Invested Capital (ROIC) and Return on Equity (ROE): Measurement and Implications, Working paper, New York University. Eriksen, B. and T. Knudsenb, 2003, Industry and firm level interaction Implications for profitability, Journal of Business Research 56, 191–199. Fisher, F, and J. McGowan,1983,On the misuse of accounting rates of return to infer monopoly profits,AmericanEconomic Review 73, 82–97. Galbreath, J., 2004, Determinants of firm success: a resource based analysis, Unpublished doctoral thesis, Curtin University of Technology. Galbreath, J. and P. Galvin, 2008, Firm factors, industry structure and performance variation: new empirical evidence to a classic debate, Journal of Business Research 61, 109–117. Gale, B. T., 1972, Market share and rate of return, The Review of Economics and Statistics54, 412– 423. Goldszmidt, R.G.B.,Brito L.A.L. and F.C. de Vasconcelos, 2011, Country effect on firm performance: A multilevel approach,Journal of Business Research 64, 273–279. Gonzalez-Fidalgo, E. and J. Ventura-Victoria, 2002, How much do strategic groups matter?,Review of Industrial Organization 21, 55–71. Hall, M. and L. Weiss, 1967, Firm size and profitability,The Review of Economics and Statistics 49, 319–331. Hansen, G. and B. Wernefelt, 1989, Determinants of firm performance: the relative importance of economic and organizational factors,Strategic Management Journal 10, 399–411. 28.

(30) Is performance driven by firm- or industry specific factors?. Hamel G. and C.K Prahalad, 1994, Competing For TheFuture, (Harvard Business School Press, Boston, MA). Hawawini, G., Subramanian, V. and P. Verdin,2003, Is performance driven by industry or firmspecific factors? A new look at the evidence, Strategic Management Journal 24, 1–16. Hawawini, G., Subramanian, V. and P. Verdin, 2005,Is performance driven by industry or firmspecific factors? A reply to McNamara, Aime and Vaaler, Strategic Management Journal 26, 1083– 1086. Harcourt, G.C., 1965, The accountant in a golden age,Oxford Economic Papers 17, 65–80. Hoskisson, R.E., Hitt, M.A., Wan, W.P. and D. Yiu, 1999, Theory and research in strategic management: swings of a pendulum, Journal of Management25, 417–456. Koller, T., M. Goedhart and D. Wessels, 2010, Valuation Measuring and Managing the Value of Companies, (John Wiley & Sons, Hoboken, NJ). Kostopoulos, Κ.,Spanos, Y. E. and G. P. Prastacos, 2002, The resource based view of the firm and innovation: identification of critical linkages, Working paper Athens university of Economics and Business. Kruskal, W.H. and W.A. Wallis, 1952, Use of ranks in one-criterion variance analysis, Journal of the American Statistical Association47, 583–621. Lee, C., Lee, K. and J.M. Pennings, 2001, Internal capabilities, external networks, and performance: a study on technology-based ventures, Strategic Management Journal 22, 615–640. Makino,S., Isobe T. and C. M. Chan, 2004, Does country matter?,Strategic Management Journal 25, 1027-1043. Markides, C.C., 1999, A dynamic view of strategy, Sloan Management Review 40, 55–72. Mason, E.S., 1939, Price and production policies of large-scale enterprises, American Economic Review 29, 61-74. Mauri, A.J. and M.P. Michaels, 1998, Firm and industry effects within strategic management: an empirical examination, Strategic Management Journal 19, 211–219. McGahan, A.M. and M.E. Porter, 1997, How much does industry matter, really?,Strategic Management Journal 18, 5–14. McGahan, A.M. and M.E. Porter, 2005, Comment on ‘industry, corporate and business segment effects and business performance: a non-parametric approach’ by Ruefli and Wiggins, Strategic Management Journal 26. 873–880.. 29.

(31) Is performance driven by firm- or industry specific factors?. McGahan, A.M., 1999, The performance of US corporations: 1981-1994, The Journal of Industrial Economics 47, 373–398. McNamara, G., Aime, F. and P.M. Vaaler, 2005, Is performance driven by industry or firm-specific factors? A response to Hawawini, Subramanian, and Verdin, Strategic Management Journal 26, 1075– 1081. Michalisin, M.D., Smith, R.D. and D.M. Kline, 1997, In search of strategic assets, The International Journal of Organizational Analysis 5, 360–387. Moyer, R.C., McGuigan, J.R., Rao, R., and W.J. Kretlow, 2010,Contemporary Financial Management(South-Western Cengage Learning, Mason, OH). Needles,jr, B.E. and M. Powers, 2008, Financial Accounting (South-Western Cengage Learning, Mason, OH). Needles, B.E., Powers, M. and Cengage Learning, Mason, OH).. S. V. Crosson, 2008,. Principles of Accounting(South-Western. Neter, J., Kutner, M. Wasserman, W. and C. Nachtsheim, 1996,Applied Linear Statistical Models (The McGraw-Hill/Irwin, New York, NY). Nordstokke, D.W. and B.D. Zumbo, 2010, A new nonparametric Levene’s test for equal variances, Psicológica31, 401–430. Peteraf, M., 1993,The cornerstones of competitive advantage: a resource-based view,Strategic Management Journal 14, 179–191. Porter, M.E., 1979, The structure within industries and companies performance, The Review of Economics and Statistics 61, 214–227. Porter, M.E., 1980, Competitive Strategy (The Free Press, New York, NY). Porter, M.E., 1981, The contributions of industrial organization to strategic management, The Academy of Management Review 6, 609–620. Porter, M.E., 1990, The Competitive Advantage of Nations (Free Press, New York, NY). Prescott, J.E., Kohli A. K. and N. Venkatraman, 1986, The market share-profitability relationship: An empirical assessment of major assertions and contradictions,Strategic Management Journal7, 377– 394. Ricardo, D., 1817, Principles of Political Economy and Taxation (John Murray, London, UK). Roquebert, J.A., Phillips, R.L. and P.A. Westfall, 1996, Market vs. management: what ‘drives’ profitability?,Strategic Management Journal 17, 653–664. Ruefli, T.W. and R.R. Wiggins, 2003, Industry, corporate and segment effects and business performance: a non-parametric approach, Strategic Management Journal 24, 861–879.. 30.

No results found