Is performance driven by industry- or firm-specific
factors? The influence of firm size
By
J.N. Bink
1University of Groningen Faculty of Economics and Business
MSc Finance Date 04-01-2014 Supervisor: Dr. J.O. Mierau
Abstract
In this study I will analyze how firm size affects the relationship between industry- and firm specific factors and long-term firm performance. This study contributes to the debate whether firm performance is driven by industry or firm specific factors. I divided the dataset in three subsamples to estimate the variability in each effect for different firm sizes. I used a dataset which contains of 1500 firms listed at the S&P 1500 covered by a period from 2000 till 2012. I show that size effects counts for about 15% in explaining performance variability. Moreover this study finds that the larger a firm is, the greater firm effects will be. For small firms 37.49% on average and for large firms 49.45% on average. Therefore, it can be concluded that firm size effects do matter in explaining performance variability. Lastly, I find that firm effects dominate industry effects regardless of the size of a firm.
Keywords: Industry Effects, Firm Effects, Firm Performance, Firm Size, Performance Variability, Value Creation.
JEL Classification: G10 and G32
1 Email: nick_bink@hotmail.com, Student number: S1764594, I would like to thank my supervisor J.O
2 Introduction:
There has been a considerable debate about the drivers of firm performance. One fact is obvious clear, firm performance is an important driver of firm value (Koller, 2010). For example, when it is evident from the annual statistics that a firm performs well, immediately the stock price will rise and a firm will be more valuable. The opposite is also true. If a firm performs worsen than expected, the stock price will fall and value of a firm will decrease. But how can it be explained that some firms shows better results and are more valuable than other firms in the same industry? Can industry factors explain this phenomenon? Or are there other factors that can explain this phenomenon?
The last five decades, researchers have shown the importance of the growing interest for firm- and industry specific factors that influence firm performance. The industry organization theory (Cabral, 2000) suggests that performance is driven by the working of markets and industries, therefore the structure of the industry is important. In the opposite of the industry organization theory, the resource-based view (Barney, 1991) suggests that firm specific characteristics drives firm performance. Most of the past research, Rumelt (1991) McGahan and Porter (1997) and Hawawini et al. (2003) find that firm factors dominates industry factors. This means that firm effects are more important in explaining the variability in firm performance. These findings are often generalized to all firms for all industries. However, interest has increased for the fact whether other factors influence this relationship.
3 rate. This could be explained by the fact that both small and large firms could have competitive advantages to outperform the other. For example, Calvo (2006) found that young small firms grow faster than old large companies.
In light of this discussion, this present study will analyze if there is a firm size factor present that influence long-term firm performance. Moreover this study will analyze, by taking into account firm size, whether firm- or industry factors remain the same. In conclusion, this study seeks to explore whether the presence of a firm size factor influence firm performance.
The main findings are as follows. First, In terms of firm size, I find that there is a firm size effect. It counts for about 15.00% in explaining the variability of firm performance. By adding a firm size affect the unexplained firm performance is also reduced. The presence of a firm size effect in the model is a first indication that firm size does matter in explaining performance variability. Moreover, as in line with the first finding, I find that firm effects become more important as firms get larger. This means that firm effects are even more important when the size of a firm is taken into account.
I also find that, similar to the results of other papers, that firm effects dominates industry effects. These findings are regardless of the size of a firm. This means that the resource based view dominates the industry organization theory. This implicates that the firms’ unique characteristics are important in explaining firm performance.
Lastly, I find no big difference between the different performance measures. For all performance measures firm effects dominates. However, for the performance measure market-to-book ratio firm effects are the highest and for the measure return on assets firm effects are the lowest.
4 methodology of this research. In section 5 the findings are presented. Section 6 concludes.
2. Literature review 2.1 Firm size.
Does size matter? One thing is clear, small firms differ from large firms in many ways. For instance, influences of firm size has found in capital markets (Rees, 1995; Cooke 1992) and in bankruptcy predictions (Ohlson, 1980). This paper studies the influence of firm size with respect to the relationship of firm and industry factors on the firm’s long-term performance.
The effect of firm size on firm performance is widely tested with ambiguous results. As already mentioned in the introduction, Gilbrat’s law is one of the main theories in explaining the relationship between firm size and firm performance (Gilbrat, 1931). This theory states that regardless of the size and past growth of a firm, firms have the same average proportional growth rate. Thus, small firms are at no disadvantage for growth, and all are equally likely to double in growth for a given period. Therefore, from Gilbrat’s law it can be concluded that firm size does not matter in explaining firm performance.
5 On the other hand, large firms suffer more from agency problems and they have more hierarchical and less flexible structures, which are needed in the current rapidly changing environments. In addition, there may be problems of effective coordination and control, due to bureaucratic problems (Baumol, 1967).
As discussed, there are different views on the influence of firm size on firm performance. Gilbrat’s Law argues that size does not matter in explaining firm performance. However, the modern view argues that size does matter in explaining firm performance. To be more precisely, firm size could give a firm a competitive advantage. In light of these different backgrounds the research question that plays a central role in this study is defined as follows. Is there a firm size effect which influences the performance variability of a firm? Or does firm size not matter in explaining firm performance.
2.2 Industry organization theory and resource based view.
In this research not only firm size, but also industry- and firm factors play an important role as drivers of firm performance. These two factors can be explained by two underlying theories. Industry factors can explain by the industry organization theory. The industrial organization theory takes the industry structure as driver for firm performance. Firm factors can be explained by the resource-based view. The resource-based theory takes the unique capabilities of a firm as driver for firm performance. In this section both theories are explained more in detail.
6 1959), who developed the structure-conduct-developed (SCP) paradigm. This paradigm is a model in which market structure is seen as more dynamic and evolving (Demsetz, 1973). Maybe the most well know example of this model is the fives forces model of Porter (1980). In this model the five forces (Competitive rivalry, the treat of substitute products and new entrants and the bargaining power of customers and suppliers) influence the performance of a firm in a particular industry. Although there are different schools within the theory of industry organization, they both view the industry as the unit of analysis and are on the assumption that firms within the same industry are homogeneous. Therefore, the industry structure will influence the performance of firms within the industry (Galbreath and Galvin, 2008).
Since the late eighties a new theorem became popular, the resource-based view. This theory became popular due to the fact that the industry organization theory was unable to explain why firm performance differs across firms within same industry (Nelson, 1991 and Wernerfelt, 1984). Whereas in the industry organization theory long-term performance is determined at the industry level of analysis, by developing the resource-based view, the unique resources and capabilities of an individual firm became the focus of explaining long-term firm performance (Hoopes et al. 2003). The resource-based view takes the unique firm characteristics as the basis for competitive advantages. Firms are seen as heterogeneous entities with respect to their resources and capabilities. Therefore, each firm has their own unique and idiosyncratic resources which are hard to imitate by other firms (Mauri and Michaels, 1998; Wernerfelt, 1984).
7 markets, skill acquisition and manipulation of knowledge become important strategic issues.
As a result of these two different theories about the drivers of firm performance, past research tried to give an answer on the question whether firm performance is driven by the dynamics of the industry or if it is driven by the unique resources and capabilities of a firm. In light of the present study, the following research questions can be defined: Does Industry- or firm factors dominate long-term firm performance? Second, given that firm size plays a role in explaining firm performance, how does firm size effects the relationship between industry- and firm specific factors and firm performance?
2.3 Model of Hawawini et al. (2003) and past research
In this section I will explain the empirics of the study of Hawawini et al. (2003) since this study forms the basis for the present study. Moreover, in this section I will also compare the results of different past studies.
8 Hawawini et al. (2003) were not the first one who analyzed the contribution of industry and firm factors to firm profitability. The results of the past studies can be found in Table 1.
One of the first things that can be noticed from Table 1 is the shift from industry effect domination to firm effect domination. Where Schmalensee (1985) found an industry effect of 19.5% and a firm effect of 0.63% in the later studies an industry effect of 8.1% and a firm effect of 35.8% was found. The shift from an industry focus to a firm focus means that there is a shift from the industrial organization theory to the resource-based view. This could be explained by the fact that industry boundaries are becoming more blurred due to the overlap and the convergence of different industries. Therefore it becomes more important for firms to sustain and develop their competitive advantages.
A second obvious observation from Table 1 is that the performance measures change from the use of accounting measures to the use of value based measures. Schmalensee (1985) only used the return on assets, which is an accounting based measure, as indicator for firm performance. In later studies not only the return on assets, but also market-to-book ratios and return on invested capital are used as proxies for firm performance. However, it can be concluded that there are no actually differences in outcomes by using these measures. For example, Rumelt (1991) found an industry effect of 4.0% and a firm effect of 45.8%. Hawawini et al. (2003) found for the return on asset measure an industry effect of 8.1% and a firm effect of 35.8%. Moreover, he found for the Market-to-book ratio an industry effect of 11.4% and a firm effect of 32.5%. Due to the consistency of outcomes between accounting- and value based measures, I do not elaborate on this discussion in this research.
9 unexplained variance lowered to around 10.0%. However, this is really straightforward because future performance has a high correlation with performance that is realized in the year before.
The discrepancy in results between the studies of Schmalensee (1985) and other studies could be explained by the fact that Schmalensee (1985) used market-share as only proxy for firm effects. Moreover he used only a sample that covered one year. Therefore, in the research of Schmalensee (1985) it was difficult to define a composite firm effect. (Rumelt, 1991; Hawawini et al., 2003).
Rumelt (1991) tried to cover this problem by using data from four years. By using a dataset that is covered by more than one year, it was possible for Rumelt (1991) to include a composite firm factor that accounted for all firm effects. This resulted in an increase of firm effects and in a decrease of the unexplained variance. Firm effects counted now for 45 % of the variability in firm performance. The unexplained variance declined from 80 % to 45 %.
In the study of Schmalensee (1985), Rumelt (1991) and McGahan and Porter (1997) there is also a corporate effect included in the model. This corporate effect is defined as the effects that arise when a legal entity operates and owns more than one business-unit (Rumelt 1991). In all studies, these corporate effects are very small. This means that by adding this factor in the model, it hardly helps explaining the variability in firm performance. In addition a corporate effect can have negative role in the model. Because the variance component analysis cannot estimate negative effects, it is very hard to determine these corporate effects. For these two reasons I will exclude the corporate effect of the model in this research.
10 industry who are not winners nor leaders, the industry effect mare more. Reason for this could be the ‘stuck in the middle’ argument. This implies that these firms have average managerial capabilities and performance. This implicates that the average firm depend more on the performance of the industry in general. However, still for the average performers the firm effects dominates the industry effects.
11 4. Model and Methodology
4.1 Model
The model that will be used in this research largely follows the variance component analysis model (VCA) of Hawawini et al. (2003) and Rumelt (1991). However, I will add firm size as new effect to the model. Equation 1 presents the descriptive model that forms the basis for the research model in this paper
𝑅𝑖𝑗𝑡𝑠 = 𝜇 + 𝛼𝑖 + 𝜌𝑗 + 𝛾𝑡+ 𝛽𝑠 + 𝛼𝛾𝑖𝑡+ 𝜀𝑖𝑗𝑡𝑠 (1) 𝑅𝑖𝑗𝑡𝑐 = Performance measure 𝜇 = Constant 𝛼𝑖 = Industry Effect 𝜌𝑗 = Firm effect 𝛾𝑡 = Year effect
𝛽𝑠 = Firm size effect
𝛼𝛾𝑖𝑡 = industry year interaction effect
𝜀𝑖𝑗𝑡 = Error term (Unexplained variance term)
𝑅𝑖𝑗𝑡𝑠 is the performance measure, in this research the return on assets, return on invested capital and a market-to-book ratio, for industry i, firm j, year t and size s. The first right hand side variable 𝜇 is a constant that incorporates the average performance over all firms, 𝛼𝑖 is the random industry effect, 𝜌𝑗 is the random firm effect, 𝛾𝑡 is the random year effect, 𝛽𝑠 is the random firm size effect, 𝛼𝛾𝑖𝑡 is the interaction between the main year and industry effect also called the transient effect. The last variable 𝜀𝑖𝑗𝑡𝑠 denotes the random error term.
12 4.2 Methodology
The method that will be used to examine the different effects is the variance component analysis. As already mentioned this method is also used by Hawawini et al. (2003). The variance component analysis estimates how much of variance in dependent variable is explained by the variance of the different independent variables in the model. However, there is one problem concerning the use of a variance component analysis. By using the variance component analysis I do not test whether the different independent effects have a statistical significant influence on the performance variable. Therefore, before performing a variance component analysis, I will first test if the differences between groups of effects are statistically significant. When the effect is statistically significant this effect can be included in the variance component analysis.
The first test that is used for testing significance of the independents effects is the One-way analysis of variance (One-One-way ANOVA). The One-One-way ANOVA tests the hypothesis that the samples of the different groups have the same mean values. Two important assumptions of the One-way ANOVA test is that the data of each variable is normally distributed and variances of populations are equal. If these two assumptions do not hold the Kruskall-Wallis and Mood’s median test can be used for testing significance of the independents effects.
The Kolmogorov-Smirnov test is used for assessing the normality of the data. In case when the data is non-normal distributed I will use the Kruskal-Wallis test to test whether the independent effects have a statistical significant influence on the performance measure. The Kruskal-Wallis test is a non-parametric version of the one-way ANOVA test (Kruskal and Wallis, 1952). According to Conover et al. (1981) in most cases the Kruskal-Wallis test use the same procedures as the one-way ANOVA. The major difference between the two tests is that the Kruskal-Wallis test use ranks of data instead of the complete data sample itself.
13 variability in rank scores for that particular effect. However, I want to test the effects size by taking into consideration the other effect sizes. This is eventually tested with the variance component analysis.
The second assumption of the one-way ANOVA test is that the variances are equal among groups. The median and the non-parametric Levene’s test can used for assessing whether the variances are equal between all groups. In this research the non-parametric Levene’s test is used for testing equal variances among groups, because this test has more statistical power if the sample is skewed and size is unequal (Norstokke and Zumbo, 2010). If the variances are not equal between all groups the Mood’s median test can be used for testing if the independent effects are statistical significant. (Brown and Mood, 1951). The Mood’s median test tests whether the median rank of the different dependent variables differs between the groups. Because the Mood’s median does nog assume equal variances among groups, the Mood’s median test is less powerful than the Kruskal-Wallis test (Gibbons and Chakraborti, 2003).
Like Hawawini et al. (2003) I will use the variance component analysis to measure the contribution of each effect to the variability of the dependent variable. The variance component analysis transforms these values to relative percentages of the total variability of the dependent variable. Equation 2 shows the decomposition of the variability of the performance measures in the effects. This equation is based on the descriptive model of Equation 1.
𝜎𝑅2 = 𝜎
𝛼2+ 𝜎𝜌2+ 𝜎𝛾2+ 𝜎𝛽2+ 𝜎𝛼𝛾2 +𝜎𝜀2 (2)
14 Besides only testing for the complete sample, I also divide the full sample in three subsamples. The modified sample is classified according to firm size and therefore each sample consists of small, medium or large firms. In this way it is possible to test the influence of the different effects on firm performance with respect to firm size. Total assets are taken as proxy for firm size. Like the whole sample I will use the models of Equations 1 and 2 for assessing the variability in the different effects.
15 3. Data
The Dataset contains US non-financials firms from the S&P Small Cap 600, the S&P Mid-Cap 400 and the S&P 500 covering the years 2000 till 2012. All data is extracted from the Thomson Reuters Datastream database. The S&P 600 is a capitalization-weighted index that covers the small cap range of $350 million to $1.6 billion US stocks, that’s about three percent of the total US stock market. The S&P 500 index is a free-floated capitalization index that covers the 500 leading stocks in the US. The latest one, the S&P Mid-Cap 400 is a value-weighted index that covers mid cap range of $750 million to $3 billion US stocks.
In this research I will use three different variables as proxies for firm performance. The first performance measure I will use is the return on assets. Return on assets is calculated as the net income divided by the total assets. In most past studies, the return on assets is used as measure for firm performance. The second measure I will use as performance indicator is the return on invested capital. Return on invested capital is calculated as the net income divided by the total invested capital for a given period. The third and last performance measure I will use in this paper is a market-to-book ratio, called the Tobin’s Q ratio. This ratio was developed by Tobin (1969) and is calculated by the firm’s total market value divided by the book value of its total assets. Table 2 presents information about the variable definitions and the sources of the different variables. Table 2 also shows the different calculations for the different variables
A bias that could arise in this research is the survivor bias. This is the tendency for failed firms to be excluded from the study because they do not longer exist. This data set contains only firms that survived the time period. However, the assumption of random variables means that the results could be generalized if there is indeed a survivorship bias.
16 first level consists of 10 industries, partitioned into 19 super sectors, which are further divided into 41 sectors and eventually these 41 sectors includes 114 subsectors. In this paper I will classify the firms in the dataset according to the level 2 super sector classification (See Appendix A2.1). Due to the relative small data sample size I cannot use a more narrow industry classification. If I will use a more narrow industry classification this will leads to sectors or subsectors which contains not enough data.
Moreover, I make some adjustments regarding the dataset. First, firms who are listed under the utility and financial sector are excluded from the dataset. This is due to the fact that utility firms are different regulated and financial firms have different balance sheets. Therefore these two sectors are not comparable to the other sectors in the sample. Second, all observations that contains missing values are excluded from the sample. Thirdly, in some cases the market-to-book ratio has negative observations. These negative observations are excluded from the sample. Fourthly, I will remove statistically outliers from the sample. In contrast to Hawawini et al. (2003) I will not test for economic outliers, this includes out- and underperformers. I will exclude these outliers according to the method of Cohen et al. (2003). Cohen et al. (2003) argue that all data that lies more than three standard deviations away from the median need to be excluded from the sample. Lastly, I categorized the firm size data. The component variance analysis can only estimate the sizes of an effect if the data of the independent effects are categorical data. Therefore I categorized the data of firm size into three groups (Small, medium and large firms). Total assets are taken as indicator for firm size.
After adjusting the data the new dataset contains 13493 observations for the ROA, 13787 observations for the ROIC and 13340 for the market-to-book ratio. These observations are spread amongst 13 years, 14 different industries and 1151 different firms. Table 3 presents the descriptive statistics for all dependent variables.
17 Moreover, Figure A3.1 shows these fluctuations and trends throughout the years. The skewness for the return on assets and return on invested capital is negative meaning that the mean is left of the median. For the market-to-book ratio the skewness is positive which indicates that the mean is right of the median. However, in all cases the skewness is a small number which indicates that there is very little skewness present. In contrast to the skewness, the kurtosis is larger than three indicating that the distribution has a sharper peak and fatter tails compared to a normal distribution. In addition the Kolmogorov-Smirnov test statistics are significant. This indicates that the data from the sample is not normal distributed, which implies that there are significant difference between the groups.
Table 4 presents the correlations between the different performance measures. Overall the correlations between the different firm performance measures are not high and are under the 0.5. The correlation between return on invested capital and the market-to-book ratio is -.039 and is significant at a significance level 0.01 (two-tailed).
In addition, I also calculated the descriptive statistics for respectively each year, industry and firm size. Although I do not use these results for further research, it may be interesting for the reader. These descriptive statistics can be found in appendix A4.
5. Results
Prior to performing the variance component analysis I will first test if the independent factors are statistical significant and if these effects can be included in the model. It is not possible to add the transient effect therefore this effect is could not be tested. In Table 5, Table 6 and Table 7 the results of the Non-parametric Levene’s test, Kruskal-Wallis test and Mood’s median tests can be found. From the results it can be concluded that the effects are for all performance measures statistically significant. This means that all effects can be included in the variance component analysis.
18 second model measured the variability in which the factor firm size is excluded from the model. The results are presented as the effects in percentage of the total variance.
From the results in the first model it is obvious that there is a firm size effect present. For the performance measures ROIC and MtB the firm size effects are respectively 14.09% and 14.06%. Only for the performance measure ROA the firm size effect is not significant high, namely 0.02%. Moreover, due to the presence of a firm size effect the unexplained variance is lower compared to the second model without a firm size effect. This reduction in the unexplained variance is 15.36% and 13.80% this is mainly due to the firm size effect. The presence of a firm size effect in the model is a first indication that firm size does matter in explaining performance variability.
Moreover from Table 8 it can be concluded that for both models firm effects dominates the long term performance. For all performance measures, ROA, ROIC and MtB, the firm effects are greater than the total industry effects. The total industry effects are the sum of the stable industry effects and the transient effects. The total industry effects are 2.93%, 3.02% and 6.85% for the first model and 2.94%, 2.10% and 6.61% for the second model. Related to this, the firm effects are 27.54%, 35.28% and 53.40% for the first model and 27.59%, 32.32% and 51.38% for the second model. Therefore, I can conclude that the resource based view, as discussed in the literature section, dominates the industrial organization theory. This means that the performance variation among firms is mostly due to the unique and intangible firm characteristics. Furthermore, the year effects that are captured in the model are very small. This year effect varies from 1.27% to 1.80%. Therefore it can be questioned whether it is meaningful to add the year factor in the model.
19 modified sample. Like the variance estimates in Table 8, the results are presented as the effects in percentage of the total variance.
Table 9 provides evidence that firm effects become more important if firms are larger. In term of variance estimates, the variance estimates goes from an average firm effect of 37.49% for small firms to 49.45% for large firms. In contrast to this, the variance estimates for the total industry effects remains almost the same. For all three categories the industry effect average is between 2.50% for small firms and 3.50% for large firms. Moreover, for all measures and all firm sizes the firm effect dominates the total industry effect.
As in line with Table 8, firm size has no effect on the variation in firm performance for the measure ROA. All effects remain the same regardless of the size of a firm. Moreover the effects of the modified sample are almost the same as for the full sample. However, the findings for the ROA are in opposite of the findings for the measures ROIC and MtB. For the measures ROIC and MtB, firm effects in the modified sample are larger than the firm effects in the full sample. The firm effects are 22.89% and 22.00% larger for medium-sized firms and 15.00% and 14.47% larger for large firms.
Comparing the results of Table 8 and Table 9 it can be concluded that firm size does matter in explaining firm performance. Firm size influence firm performance in the way that if firms become larger firm effects are more important in explaining firm performance. This is in line with the view of Baumol (1967) and Hall and Weiss (1967). They suggest that firm size could lead to a competitive advantage for a firm.
20 effects have the largest variability followed by the year effect, industry effect and size effect. Like the findings in Table 8, firm effects dominate the other effects.
In table 11 the results of the present study is compared with the results of past research. With the exception of the study of Schmalansee (1985), all past research and the present study found that firm effects dominate the total industry effects. When using the performance measure ROA the firm effects are somewhat lower compared to other research. In addition, industry effects found in this research are lower compared to the industry effects in the research of Hawawini e al. (2003). However, for all studies the impact of the industry effects are not high, ranging from 0.6% to 19.5%. Moreover, due to the firm size effect, the unexplained variance in this study is lower than the unexplained variance in past research. A possible explanation for these deviating results could be that I used a different way of classifying the industries. Hawawini et al. (2003) used 55 industries in his research and I used 14 industries in my research. Moreover I used a different sample, the S&P 1500 instead of the S&P 500. Lastly the sample is covered by a different time span which could influence these deviating results.
6. Conclusion
21 According to the results, I find that there is a firm size effect present. Only for the ROA this effect is small. Moreover, like other studies the results suggest that firm effects dominate industry effects. This result is robust for all three firm performance measures. For the year effect I find a relative small variance effect. Lastly, I find that firm effects become more important if firms become larger. For all the three firm size categories firm effect dominates. However, this domination became even larger for medium and large firm. This is between 15 and 20 percent point larger than for the full sample. This means that for medium and large firms, firm effects have a larger impact on firm performance compared to the smaller firms where the firm effects are lower.
This study contributes to the literature on the topic about which factors drives the long-term performance of a firm. As far as I know, this is the first study that investigates the influence of a firm size factor as a driver of firm performance. As other studies like Schlamansee (1985) Hawawini et al. (2003) only used a sample which covered data from large firms, this sample also covered small and medium sized firms. Therefore with the results of this research it is possible to easier generalize the results to small and medium firms. Although you should always be careful when making this kind of assumptions.
For practitioners like CFO’s or other financial managers this study enables them to make better capital allocating decisions and moreover it gives them a better understanding of which effects have an impact on long-term firm performance. But also for other corporate stakeholders it is important to know which factors drives firm performance in order to get a more accurate value of the firm. Moreover, also for external parties like consultants this study could be interesting. With the results of this study it will give external parties like consultant more knowledge about making strategic decisions. Eventually this leads to better understanding why firms within the same industry differ in performance.
22 generalized on all US firms because I used a random selection of the underlying population.
Second, the definition of firm size leaves room for discussion. In this paper I used total assets as indication for firm size. However, there are other measures that could be used as indicator for firm size. A measure that is often used as indicator for firm size is the market capitalization. Another measure that could be used as indicator for firm size is the number of employees. Testing if the results would be the same in case when market capitalization or the number of employees would be used as indicator for firm size would be a good topic for further research.
23
This table presents the results of past studies. ROA: Return of assets, ROIC: Return on invested capital, WACC: Weighted cost of capital, ROIC-WACC: Added economic value, M/B: Unadjusted market-to-book ratio at enterprise level, M/B2: Unadjusted market-to-book ratio at equity level, TOBIN Q: Enterprise value divided by replacement value of assets.
Study Measure Industry effect Firm effect Year effect Transient effect Country effect R&D effect Past Performance effect Patent effect Unexplained variance Schmalensee (1985) ROA 19.5% 0.63% NA NA NA NA NA NA 80.5% Rumelt (1991) ROA 4.0% 45.8% NA 5.4% NA NA NA NA 44.8% McGahan and Porter (1997) ROA 18.7% 36.0% 2.4% NA NA NA NA NA 48.4% Hawawini et al. (2003) Full sample ROA 8.1% 35.8% 1.0% 3.1% NA NA NA NA 52.0% ROIC-WACC 6.5% 27.1% 1.9% 4.2% NA NA NA NA 60.3% M/B 11.4% 32.5% 1.3% 2.9% NA NA NA NA 51.9% Bergsma (2012) Full sample ROA 5.4% 29.3% 1.8% 3.1% 5.8% NA NA NA 54.5% ROIC 0.6% 21.3% 0.0% 0.0% 1.0% NA NA NA 77.5% ROIC-WACC 0.5% 21.4% 0.0% 0.0% 1.0% NA NA NA 78.0% M/B 0.0% 12.1% 0.0% 0.0% 0.0% NA NA NA 87.9% M/B2 3.1% 63.4% 0.7% 0.8% 0.9% NA NA NA 31.1% TOBIN Q 14.2% 40.5% 1.6% 3.6% 1.1% NA NA NA 39.1%
Van den Berg (2013) M/B 0.3% 2.0% 1.0% 1.3% NA 2.0% 81.7% 0.0% 11.57%
24 Table 2: Variable Definitions and sources
Description** Symbol Source
Return on assets ROA Datastream code WC08326
Return on invested capital ROIC Datastream code WC08376
Enterprise Value EV Datastream code WC18100
Book Value Total Assets BTA Datastream code WC02999
This table present the different definitions and sources of the variables used in this reserach.* Calculations made by Datastream. ** All data is in US dollars and are end of the year market data.
Table 3: Descriptive statistics
Statistic Return on assets Return on invested capital Market -to-book ratio N 13493 13787 13340 Minimum -68.06 -128.17 .00 Maximum 77.32 144.96 6.97 Mean 6.63 9.75 1.58 Median 6.93 9.85 1.25 Std. Deviation 10.02 15.44 1.09 Skewness -1.29 -0.75 1.90 Kurtosis 10.26 14.60 4.22 Kolmogorov- Smirnov 0.15 0.16 0.15 Sig. 0.00 0.00 0.00
The table shows the mean, minimum, maximum, standard deviation (Std. Deviation.), Skewness, Kurtosis and Kolmogorov-Smirnov for the variables Return on assets, Return on invested capital and the Market-to-book ratio.
Description** Symbol Calculation
Market-to-book ratio MtB EV/BVTA
Firm Size Size BTA
Return on Assets* ROA Net Income/ BTA
25 Table 4: Correlations Return on assets Return on invested capital Market-to book ratio Return on assets 1.00 0.00 0.00 Return on invested capital 0.00 1.00 -0.04* Market-to-book ratio 0.00 -0.04* 1.00
This table presents the correlations between the different dependent variables. *Correlation is significant at the 0.01 level (2-tailed)
Table 5: Non parametric Levene’s test
This table presents the Non parametric Levene’s test. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio. All effects are statistically significant.
26 Table 6: Kruskal Wallis test
This table presents the Kruskal Wallis test. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio. All effects are statistically significant.
Table 7: Mood’s median test
This table presents the Mood’s median test. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio. All effects are statistically significant.
27 Table 8: Variance Estimates
ROA ROIC MtB ROA ROIC MtB
Firm effect 27.54 35.28 53.40 27.59 32.31 51.38
Industry effect 1.01 1.05 5.85 1.00 1.40 5.48
Year effect 1.80 1.27 1.42 1.79 1.84 2.97
Transient effect 1.93 1.05 0.76 1.93 1.62 1.37
Firm Size Effect 0.02 14.09 14.06 - - -
Error 67.70 47.27 24.50 67.69 62.83 38.80
This table presents the variance estimates calculated by the variance component analysis. Effects are in percentage of the total dependent variance. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio.
Table 9: Variance Estimates if data is divided in three subsamples.
This table presents the variance estimates calculated by the variance component analysis. Effects are in percentage of the total dependent variance. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio
Table 10: Independent variance sizes estimated by Krukal-Wallis test
This table presents the Independent effect sizes, which are calculated by the kruskal-Wallis test. ROA is the return on assets; ROIC is the return on invested capital and MtB is the market-to-book ratio.
Small Firms Medium Firms Large Firms
ROA ROIC MtB ROA ROIC MtB ROA ROIC MtB
28 Table 11: Variance Estimates compared with past research.
This table presents the variance estimates of the different studies including the present study. ROA: Return of assets, ROIC: Return on invested capital, WACC: Weighted cost of capital, ROIC-WACC: Added economic value, M/B: Unadjusted market-to-book ratio at enterprise level, M/B2: Unadjusted market-to-book ratio at equity level, TOBIN Q: Enterprise value divided by replacement value of assets. MtB is the same as TOBIN Q.
Study Measure Industry effect Firm effect Year effect Transient effect Country effect Firm Size Effect Unexplained variance Schmalensee (1985) ROA 19.5% 0.63% NA NA NA NA 80.5% Rumelt (1991) ROA 4.0% 45.8% NA 5.4% NA NA 44.8% McGahan and Porter (1997) ROA 18.7% 36.0% 2.4% NA NA NA 48.4% Hawawini et al. (2003) Full sample ROA 8.1% 35.8% 1.0% 3.1% NA NA 52.0% ROIC-WACC 6.5% 27.1% 1.9% 4.2% NA NA 60.3% M/B 11.4% 32.5% 1.3% 2.9% NA NA 51.9% Bergsma (2012) Full sample ROA 5.4% 29.3% 1.8% 3.1% 5.8% NA 54.5% ROIC 0.6% 21.3% 0.0% 0.0% 1.0% NA 77.5% Tobin's Q 14.2% 40.5% 1.6% 3.6% 1.1% NA 39.1%
Current study ROA 1.01% 27.54% 1.80% 1.93% NA 0.02% 67.69%
ROIC 1.05% 35.28% 1.27% 1.05% NA 14.09% 47.27%
29 7. References
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
Journal 16, 7–19.
Boyer, K.K., Ward, P.T and G.K. Leong, 1996, Approaches to the factory of the future. An empirical taxonomy. Journal of Operations Management 14, 297–314
Brown, G.W. and A.M. Mood, 1951, Proceedings of the Second Berkeley Symposium on
Mathematical Statistics and Probability (University of California Press, Berkeley,
CA).
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,
30 Chang, S.J. and H. Singh, 2000, Corporate industry effects on business unit competitive
position, Strategic Management Journal21, 739–52.
Cohen, J., Cohen, P., West, S.G. and L.S., Aiken, 2003, Applied Multiple 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.
Cooke, T.E., 1992, The impact of size, stock market listing and industry type on disclosure in the annual reports of Japanese listed corporations. Accounting and
Business Research 22, 229-237.
Cox, D.R. and P.J. Solomon, 2003, Components of Variance (Chapman &Hall, London, UK).
Fisher, F, and J. McGowan, 1983, On the misuse of accounting rates of return to infer monopoly profits, American Economic Review 73, 82–97.
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.
Gibbons, J.D. and S. Chakraborti, 2003, Non Parametric Statistical Inference (Marcel Dekker, New York, NY).
Hall, M. and L. Weiss, 1967, Firm size and profitability, The Review of Economics and
31 Hawawini, G., Subramanian, V. and P. Verdin, 2003, Is performance driven by industry
or firm specific factors? A new look at the evidence, Strategic Management
Journal 24, 1–16.
Harcourt, G.C., 1965, The accountant in a golden age, Oxford Economic Papers 17, 65– 80.
Hoopes, D.G., T.L. Madsen, and G. Walker, 2003, Why is there a resource-based view? Toward a theory of competitive heterogeneity, Strategic Management Journal 24(10), 889-902.
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., Goedhart, M. and D. Wessels, 2010, Valuation Measuring and Managing the
Value of Companies, (John Wiley & Sons, Hoboken, NJ).
Kruskal, W.H. and W.A. Wallis, 1952, Use of ranks in one-criterion variance analysis,
Journal of the American Statistical Association 47, 583–621.
Mason, E.S., 1939, Price and production policies of large-scale enterprises, American
Economic Review 29, 61-74.
McGahan, A.M. and M.E. Porter, 1997, How much does industry matter, really?,
Strategic Management Journal 18, 5–14.
Nelson, R. R., 1991, Why do firms differ, and how does it matter? Strategic Management
32 Nordstokke, D.W. and B.D. Zumbo, 2010, A new nonparametric Levene’s test for equal
variances, Psicológica 31, 401–430.
Ohlson, J., 1980, Financial Ratios and the Probabilistic Prediction of Bankruptcy, Journal
of Accounting Research 18, 109-131.
Porter, M.E., 1980, Competitive Strategy (The Free Press, New York, NY).
Rees, B., 1995, Financial Analysis, (Prentice Hall, Hertfordshire)
Roquebert, J.A., Phillips, R.L. and P.A. Westfall, 1996, Market vs. management: what ‘drives’ profitability?, Strategic Management Journal 17, 653–664.
Rumelt, R.P., 1991, How much does industry matter?, Strategic Management Journal12, 167–185.
Schmalensee, R., 1985, Do markets differ much?, American Economic Review 75, 341– 351.
Searle, S.R., Casella, G. and C.E. McCulloch, 1992, Variance Components (John Wiley and Sons, New York, NY).
Seth, A. and H. Thomas, 1994, Theories of the firm: implications for strategy research,
Journal of Management Studies 31, 165–191.
Stewart, G.B., 1991, The Quest for Value (Harper Business, New York, NY).
Tobin, J., 1969, A general equilibrium approach to monetary theory, Journal of Money,
Credit, and Banking 1, 15–29.
Teece, D.J., Pisano, G. and A. Shuen, Dynamic capabilities and strategic management,
33 Van der Stede, W.A. and C.W. Chow, 2006, The use and usefulness of nonfinancial
performance measures, Management Accounting Quarterty 7, 1-8.
Weiss, L.W., 1979, The Structure-Conduct-Performance paradigm and antitrust, The
University of Pennsylvania Law Review 127, 1104–1140.
34 Appendices:
Appendix A1: Research Sample and methods past studies Table A 1.1: Research Sample and methods past studies:
Research Sample Method
Schmalansee (1985) FTC US manufacturing firms in 1975 OLS and VCA
Firm Effects: 0.62, Industry Effects: 19.46 Unexplained Variance 81% Rumelt (1991) FTC US manufacturing firms between 1974 -1977 ANOVA and VCA
Firm Effects: 46%, Industry Effects: 4%, Unexplained Variance 45% Roquebert et al. (1996) Compustat US manufacturing firms between 1985 - 1991
VCA Firm Effects: 55%, Industry Effects: 10%, Unexplained Variance 35% McGahan and Porter (1997) Compustat US manufacturing firms between 1984 - 1994
VCA Firm Effects: 36%, Industry Effects: 19%, Unexplained Variance 49%
Hawawini et al. (2003) (ROA)
Stern Stewart data US Firm between 1987 - 1996
VCA Firm Effects: 36%, Industry Effects: 8%, Unexplained Variance 58%
Bergsma (2012) (ROA)
Thomson Reuters Datastream S&P global 1200 firms between 1990-2012
VCA Firm Effects: 29%, Industry Effects: 5.4%, Unexplained Variance 65.6%
35 Appendix A2: Classification firms
Table A2.1: Classification of firms into industries
This table shows the different industries. In addition it shows the number of firm within each industry. In total there are 1151 firm divided across 14 industries
Industry Name Number of firms
Automobiles & Parts 14
Basic Resources 38
Chemicals 41
Construction & Materials 33
Food & Beverage 48
Health Care 134
Industrial Goods & Services 259
Media 36
Oil & Gas 87
Personal & Household Goods 86
Retail 121
Technology 185
Telecommunications 15
Travel & Leisure 54
36 Appendix A3: Relations performance measures
Figure: A3.1: Relations between the different performance measures from 2000 till 2012
37 Appendix A4: Descriptive statistics
Table A4.1: Descriptive statistics per year.
Year Variable N Mean Std. Dev
38
The table displays the sample size, mean and standard deviation (Std. Dev) for each year. In total there are 13 years. ROA is the return on assets, ROIC is the return on invested capital and MtB is the market-to-book ratio.
ROIC 1120 10.39 13.12
MtB 1124 1.41 0.95
Total ROA 13493 6.63 10.02
ROIC 13787 9.75 15.44
39 Table A4.2: Descriptive statistics per industry.
Industry Variable N Mean Std. Dev
Automobiles & Parts ROA 159 6.76 5.68
ROIC 158 10.08 7.06
MtB 150 1.30 0.93
Basic Resources ROA 430 4.66 10.61
ROIC 433 7.45 15.64
MtB 417 1.12 0.65
Chemicals ROA 488 6.70 7.41
ROIC 490 10.16 11.87
MtB 482 1.29 0.63
Construction & Materials ROA 396 5.10 7.17
ROIC 394 7.75 9.82
MtB 395 1.19 0.59
Food & Beverage ROA 515 9.08 8.62
ROIC 552 13.73 14.29
MtB 526 1.70 1.02
Health Care ROA 1554 6.02 13.15
ROIC 1605 8.23 17.66
MtB 1540 2.02 1.37
Industrial Goods & Services ROA 3107 6.84 7.32
ROIC 3165 10.30 11.72
MtB 3115 1.41 0.91
Media ROA 374 5.45 9.52
ROIC 402 8.79 18.10
MtB 373 1.55 1.06
Oil & Gas ROA 989 6.67 9.12
ROIC 1014 9.69 13.33
MtB 990 1.30 0.67
Personal & Household Goods ROA 1074 7.28 9.37
40
The table displays the sample size, mean and standard deviation (Std. Dev) for each industry. In total there are 14 industries. ROA is the return on assets, ROIC is the return on invested capital and MtB is the market-to-book ratio.
ROIC 169 5.88 19.35
MtB 155 1.26 0.74
Travel & Leisure ROA 643 7.62 9.44
ROIC 635 10.15 14.79
MtB 606 1.70 1.16
Total ROA 13493 6.63 10.02
ROIC 13787 9.75 15.44
41 Table A4.3: Descriptive statistics per firm size category.
The table displays the sample size, mean and standard deviation (Std. Dev) for each firm size category. In total there are three different firm size categories, Small, Medium and Large firms. ROA is the return on assets, ROIC is the return on invested capital and MtB is the market-to-book ratio.
Industry Variable N Mean Std. Dev
Small Firms ROA 5575 6.83 11.30
ROIC 5690 7.09 18.23
MtB 5598 1.59 1.14
Medium Firms ROA 5343 6.30 9.01
ROIC 5494 11.14 13.40
MtB 5379 1.60 1.08
Large Firms ROA 2575 6.86 9.00
ROIC 2603 12.59 11.42
MtB 2363 1.56 1.03
Total ROA 13493 6.63 10.02
ROIC 13787 9.75 15.44
