A More General Approach to Estimate the Effects of Competition on Performance

(1)

A More General Approach to Estimate the Effects of

Competition on Performance

Master’s Thesis Finance Derk Mulder

1717685 June 2013

Abstract

In this paper the convergence of excess returns of the 500 US nonfinancial firms listed in the S&P500 index from 1988-2012 to its long-run level is analyzed. Especially, the pattern of convergence of industry excess returns at the level of the entity is of interest. The expected pattern of future excess returns is crucial in estimating company value as a function of returns of the core operating activities. This research takes into account an industry specific and time varying cost of capital and level of long-run excess return. In addition, the effects of net investments on the decay factor of excess returns are taken into account. Industry specific and economy wide estimations of the decay factor of excess returns are biased. Industry excess returns converge quick to their long-run level. By investing companies can slow down the convergence of excess returns as measured by the decay factor of excess returns as expected.

Keywords: Ratio Analysis, Excess Returns, Competition, Mean Reversion, Performance Measurement

(2)

2

A More General Approach to Estimate the Effects of

Competition on Performance

University of Groningen Faculty of Economics and Business

Department of Economics, Econometrics and Finance MSc Finance

Author: Derk Mulder

d.mulder.3@student.rug.nl

(3)

3

1 Introduction

In this paper the convergence of industry excess returns of the 500 US nonfinancial firms listed in the S&P500 index from 1988-2012 to its long-run level is analyzed. The pattern, sign and magnitude of excess returns is important in assessing firm value.

Firms generating returns on investments above their cost of capital will trade at a premium compared to firms that generate zero or negative excess returns. When investors expect that a company has higher excess returns compared to another company, its value, all other things being equal is higher. The increase of value is equal to the earnings before taxes minus the payout ratio. Firms generating returns below their cost of capital will destroy value as they grow. The sign and magnitude of the abnormal earnings and the timeliness of their reversion to an industry or random long-run level are important when measuring value. The value of a company is equal to the sum of the book value of the company’s assets plus the present value of the expected future excess returns. Hence, the expected pattern of future excess returns is crucial for the value of companies (note that the median market to book value of a large US company was 2.4 in 2007, so the second component is about 58% of total value (Koller et al. 2010)).

Industries have well established characteristics that influence both how a firm performs and the returns it earns. The presence of competition erodes current excess returns. Koller et al. (2010) find a gradual convergence of excess returns, whereas Nissim and Penman (2001) find a rather quick convergence of excess returns.

Decreasing positive excess returns could be due to competition or through bad investment decisions (e.g. below present return projects or negative net present value investments). In competitive sectors, the presence of positive excess returns will attract new entrants and imitation will push these excess returns down.

Nevertheless, estimates of previous research imply that no future investments or nonzero investments that generate no excess returns will be made. Therein lays the implicit

(5)

5 future returns are the result of past investments, which is highly unlikely.

First, Koller et al. (2010) and Nissim and Penman (2001) analyze return patterns whereas an analysis of patterns of excess returns is more appropriate. Their analysis accounts for an implicit economy wide constant cost of capital and level of long-run excess returns. Second, Koller et al. (2010) and Nissim and Penman (2001) ignore the effects of net investments on the convergence of excess returns. In this research I include time varying industry specific costs of capital, levels of long-run excess returns and net investments when estimating the effects of competition on excess returns. The performance assessment is at the level of the entity and the equity level. However, the main analysis focuses on the level of the entity. The performance metric at this level is a better analytical tool for understanding the company’s performance than at the equity level because it focuses solely on a company’s operations. Accounting measures such as Return On Assets (ROA) suffer from certain disadvantages, they can be influenced by factors that have no real effect on the economic health of the firm. Tested is whether the erosion of excess returns is industry specific or not. Koller et al. (2010) allow for differences between industries, whereas Nissim and Penman (2001) employ an economy wide estimation. Although Koller et al. (2010) find that companies create value by investing capital to generate future cash flows at rates of return that exceed their cost of capital, their empirical research at enterprise level takes only the different industry returns and their mean reversion into account. Their analysis of firm performance therefore is expected to contain a downward biased decay factor. Additionally, the analysis of Nissim and Penman (2001) is biased because it pools all data into a single estimation. This implies economy wide effects of competition were no industry differences between decay factors of excess returns are taken into account. By assuming that competitive forces, growth

opportunities and costs of capital are equally economy wide the bias of the estimates increases.

In this research the pattern of convergence is analyzed while allowing for industry specific differences between decay factors, returns, costs of capital and levels of long-run excess returns. Additional, the effects of including net investments into the analysis are captured resulting in a general approach to estimate the effects of competition on performance. The inclusion of net investments is expected to reduce the decay factor of excess returns. Improving the assessments of performance contributes in decision making. Helping

(6)

6 0% 3% 5% 8% 10% 13% 15% 18% 20% 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 R ate ROIC WACC

ROIC-WACC Net Investment rate

term and help investors assess the potential value of alternative investments (Koller et al. 2010). Unbiased estimations of firm performance and value are important for price

formation in efficient markets as the focus is on listed companies and especially companies with large market capitalizations. The more accurate the prediction is, the more precise firm value can be assessed. When forecasting the erosion of industry excess returns, the pattern of excess returns is estimated while overcoming the mentioned biases. The academic relevance is on the usefulness of generalized industry forecasting predictors to assess value in a forecasting context and giving an unbiased estimate of firm value. The estimated value of a company is important to analyze the quality of management and to estimate the value of growth opportunities more precisely. Additionally, accounting research is concerned with assessing the information content of financial statements, with a view to develop concrete products that can be used in practice. Capital market based accounting research uses the economic profit approach. In this field the economic profit model is typically labeled as the residual income model. Their expression for value takes account of the potential effects of competition on excess returns, but it ignores the effects of future net investments on residual income (Nissim and Penman, 2001). Figure 1 presents an overview of the median excess returns of the S&P500 index over a period of 25 years from 1988-2012.

Figure 1 - The pattern of excess returns and net investments at the level of the entity.

This figure presents the median values of the return on invested capital (ROIC), excess return on invested capital (ROIC-WACC), the weighted average cost of capital (WACC) and net investments of the S&P 500 index from 1988-2012.

(7)

7 sloping pattern of the risk free rate (Rf). Excess returns and Net Investments follow a similar pattern implying that they are intertwined. Nevertheless, the pattern of the Return On Equity (ROE) is more volatile and seems to be slowly increasing upwards as can be seen in figure A.1 in Appendix A. Additionally, the pattern of the Weighted Average Costs of Capital (WACC) and the cost of equity (Ke) follow the same trend as the risk free (Rf) rate which is presented in figure A.2 in Appendix A.

Measuring economy wide performance forces the estimated impact of current firm performance on future firm performance to be uniform across all firms, irrespective of industry. An economy wide estimation model is appropriate if, on average, all firms’ excess returns converge at the same rate to economy wide benchmarks (Stigler, 1963). It seems highly unlikely that different industries have on average similar competitive structures, performances and costs of capital. When benchmarking the historical decay of the

company’s returns, it is important to segment results by industry (Koller et al., 2010). This industry segmentation is especially important because industries are used as a proxy for sustainability of competitive advantage. Figure B.1 and B.2 in Appendix B show that there are huge industry differences between excess returns at the level of the entity and at the equity level and expected is that an industry specific analysis is appropriate.

In this research the excess returns over the cost of capital are considered. The difference between industries is measured and the rate of convergence is analyzed. In addition to previous research by Koller et al. (2010), Nissim and Penman (2001) and Fairfield et al. (2009) the effect of net investments on the decay factor of excess returns is estimated. The focus of the analysis is mainly on the convergence of excess returns at the level of the entity. Koller et al. (2010) argue that ROIC is a better analytical tool for understanding the company’s performance than ROE because it focuses solely on a company’s operations. An important disadvantage is the inability of accounting measures; in particular ROA, to conceptually reflect economic performance (Hawini et al., 2001). However, due to the frequent usage of ROE in empirical research and to benchmark the results of this paper with previous research this metric is also employed. Based on empirical research about the effects of competition and the process of mean reversion1 of returns I develop the best

1

To illustrate the concept of mean reversion, consider a simple univariate time-series model of the form: X i,t = α + βX i,t−1. If β = 0, X is

(8)

8 linear unbiased estimator of firm performance. The effects of including time varying

nonconstant industry specific costs of capital and net investments on the convergence of excess returns are analyzed. The following research question is formulated to test the impact of explicitly including costs of capital and net investments in the analysis:

What are the effects of excess returns and net investments on the convergence of returns when analyzing performance?

Extending the research question gives the following testable implications:

(1): Industry specific non constant excess returns at enterprise and equity level decay faster to the mean than returns do.

(2): Including the effects of net investments at enterprise and equity level reduces the estimated decay factor.

The objective of this study is the convergence of excess returns of the S&P 500 over the period of 1988-2012 in the US. The influence of taking into account an industry specific time varying cost of capital and level of long-run excess return on the convergence of abnormal profits is analyzed. In particular, the effect of including net investments on the decay factor of excess returns is of interest. I find a quick convergence of industry excess returns and this confirms the findings of Nissim and Penman (2001). The estimation of the decay factor of industry excess returns is biased downwards when the cost of capital is assumed to be constant and homogenous. Including a nonconstant heterogeneous cost of capital increases the decay factor of industry specific excess returns. By omitting net investments, the

industry decay factor of excess returns is biased upwards. As expected, industry net

investments on average slow down the convergence of excess returns and increase the level of long-run excess returns. Including explicitly the cost of capital and net investments

(9)

9

2 Literature Review

The principle that growth creates value only when the return on investments exceeds the opportunity cost of capital is well established. It is already included in Marshall (1890) and Preinreich (1937). However, the concept is popularized in the economic profits models in the nineties by Stewart (1991) and Copeland et al. (1990). This shift in valuation towards giving excess returns a more prominent role in determining the value of a business is based on the underlying principle that growth unaccompanied by excess returns creates no value or is even value destructive (Damodaran, 2007). The shift towards excess returns has increased the focus on measuring and forecasting returns earned by businesses on both investments made in the past and expected future investments. Previous empirical research finds different patterns of the convergence of excess returns. Koller et al. (2010) find a gradual convergence of excess returns when employing an industry specific approach using a sample of over 5,000 US nonfinancial companies from 1963-2008. Nissim and Penman (2001) find a rather quick convergence of excess returns whilst employing an economy wide estimation using NYSE and AMEX firms listed on the combined COMPUSTAT (Industry and Research) files for the 37 years from 1963 to 1999. The sign and magnitude of excess returns and the timeliness of their reversion to a long-run level are important in measuring value. The value of a company is equal to the sum of the book value of the company’s assets plus the present value of the expected future excess returns. Hence, the expected pattern of future excess returns is crucial for the value of companies. Previous empirical research by Koller et al., (2010), Nissim and Penman (2001) and Fairfield (2009) does not take into account the effects of net investments on the decay factor of excess returns and therefore their pattern of convergence is biased. Expected is that net investments reduce the decay factor of excess returns and thereby increase the length of convergence of excess returns to their long-run level. Reducing the bias in estimations of firm value is useful due to the usage of generalized industry forecasting predictors in assessing value in a forecasting context. The estimated value of a company is important to analyze the quality of management and to estimate the value of growth opportunities more precisely.

2.1 Excess returns, net investments and firm value

(10)

10 income. The identification of current and future residual income drivers is overlaid with a distinction between transitory and permanent features of ratios. Transitory ratios bear only on the present while permanent features forecast drivers of the future (Nissim and Penman, 2001). Estimates of previous research by Koller et al. (2010), Nissim and Penman (2001) and Fairfield (2009) imply that no future investments or nonzero investments that generate no excess returns will be made. Therein lays the implicit assumption that the convergence of excess returns is fully due to competitive forces instead of for example the gradual eroding return on new investments. The convergence of excess returns is partially determined by the magnitude of competition. Companies need to invest to, at least partially sustain future excess returns. The omission of net investments implies that all future returns are the result of past investments, which is highly unlikely.

Firm value is a function of its expected cash flows resulting from current and future investments. To generate these cash flows firms have to raise and invest capital in assets. However, most investments are funded with retained earnings, though this capital is not costless. The extent to which the cash flows exceed the costs of capital creates value for a business. In effect, the value of a business can be simply stated as a function of the excess returns that it generates from both existing and new investments (Damodaran, 2007). By earning excess returns firms will trade at a premium when compared with similar firms that earn zero or even negative excess returns.

(11)

11 The sign of excess returns is important, while it is possible that managers of some firms can embark on a destructive path for extended periods, assuming that they will do so forever is an extreme assumption. After all, there is an alternative to reinvesting in the business

(Damodaran, 2007). An asymmetric decay factor is the implication when current profitability is insufficient.

Second, firms invest in projects in a hierarchical order choosing the highest net present value projects over lower net present value projects; future investments are likely to yield lower returns than past investments.

Third, firms do not operate in a vacuum. Industries have well established characteristics that influence both how the firm performs and the returns it earn (Damodaran, 2007). Sustaining high returns in a mature sector with lots of competition is more challenging for an otherwise similar firm in a growing sector with significant barriers to entry. Profitability will draw in new competitors over time, pushing out industry excess returns. Based on previous research of Nissim and Penman (2001) the expectation is that excess returns converge faster when the cost of capital are explicitly accounted for. Nevertheless, companies can sustain excess returns over a certain period by making capital investments to support growth. Therefore, including the effects of net investments is expected to reduce the decay factor of excess returns and lengthen the process of convergence to the long-run level.

2.2 Competition and industry structure

Koller et al. (2010) analyze the convergence of returns; however, their analysis implicitly captures a constant cost of capital and long-run level of excess returns. They find that excess returns tend to fade faster at firms that reinvest more (higher reinvestment rates) than at firms that reinvest less. It is unrealistic to assume that companies are able to sustain high levels of excess returns while reinvesting large amounts, meaning that their value measured as a function of excess returns of the core operating activities decreases.

At the individual firm level, luck, management quality and competitive advantages, amongst other things, explain differences in performance. In their industry analysis Koller et al. (2010) find that differences in returns across sectors have widened.

(12)

12 determinant of firm performance. Hawawini et al. (2001) discuss that the structure-conduct-performance (SCP) model proposes the existence of a deterministic relationship between market structure and profitability. Structural characteristics of an industry are said to constrain the behavior of its component firms, which in turn leads to industry-specific performance differentials between firms. In this framework, the industry structure in which a firm operates is the main reason offered to explain variations in firm profitability. Firms early in their product life cycle often report low returns. They make large investments and excess returns are at this stage often zero or have a negative sign. In the maturity stage, margins are declining and returns are eroded. The profitability of companies with short product lives that face constant innovation tends to be more volatile. In an industry with a continuous changing environment and short product lives, returns are less stable and firms need to reinvest more. Therefore, the decay factor of excess returns is higher in these industries and the convergence of excess returns happens faster. Sectors with high returns tend to have a preponderance of young firms generating high returns relative to their cost of capital. Sectors with returns below the cost of capital tend to have more mature firms in decline or younger firms with heavy infrastructure investments (Damodaran, 2007).

Second, accounting inconsistencies can affect returns significantly by the way of

classification of operating, capital and financial expenditures. A permanent level of nonzero excess returns reflects permanent abnormal real profitability in the sense that the firm can invest always in nonzero net present value projects. However, it can also be induced by conservative accounting. Always keeping book values low by expensing R&D for example, will yield permanent positive excess returns even with zero or negative NPV projects (Nissim and Penman, 2001).

(13)

13 barriers are lower all other things held constant. The convergence of excess returns in these industries is therefore expected to be gradual.

Firms that earn high excess returns and report high growth rates are expected to see their growth rates decrease quickly but excess returns remain high. The reason for the decline in growth rates is twofold. First, the expected growth rate cannot exceed the growth rate of the economy in which the firm operates in perpetuity. Second, to keep growing companies have to reinvest and their returns on invested capital need to exceed their cost of capital (Koller et al. 2010).

Economic differences across industries, such as product demand, barriers to entry, or business risk, induce differences in the level or persistence of firm performance and the support for industry-specific models. Additionally, firms within an industry could have

heterogeneous cost structures in equilibrium, which could point at sustainable differences in profitability among firms in the same industry (Mills and Schumann, 1985).

Over longer forecast horizons, firms’ investment decisions are equally influenced by changes in demand at the industry level. It is possible that significant economic differences exist between firms within industries and that even within an industry certain firms could be more alike than others (Porter, 1979). Hawawini et al. (2001) find that a significant proportion of the absolute estimates of the variance of firm-specific factors are due to the presence of a few exceptional firms in an industry. They conclude that for most firms, the industry effect turns out to be more important for performance than firm-specific factors. This implies that industry effects are relevant when using generalized forecasting predictions in estimating firm value in a forecasting context. Industries have significant differences in performance, returns and net investments; therefore the industry analysis receives more attention than the economy wide analysis.

According to Koller et al. (2010) returns are driven by competitive advantages that show up as price premium and cost advantages. They state that industry structure is an important but not exclusive determinant of returns; individual companies’ capabilities also determine sources of advantage. Industries are heterogeneous meaning that there is large variation in competitive factors, excess returns, costs of capital and net investments. Nevertheless, within each industry there is significant variation in the rates of return for individual

(14)

14 economic, industry, and company conditions. This especially holds in the case of industries that enjoy relatively long product life cycles. To compare the results from this research with previous research by Koller et al. (2010) and Nissim and Penman (2001), the analysis is divided into different models to answer the research question. Koller et al. (2010) analyze the convergence of industry specific returns by implicitly taking into account a constant, homogenous cost of capital and level of long-run excess returns. However, their estimation of the decay factor is expected to be biased downward by leaving out an explicit cost of capital and excluding net investments. Nissim and Penman (2001) estimate the convergence of excess returns with and without an explicit cost of capital. Nevertheless, they estimate an economy wide convergence of excess returns and do not take into account the effects of net investments. This is expected to bias the decay factor upwards. Excess returns evolve as:

(1) (at – kt ) x At-1 = (1 – δ) x (at-1 – kt-1 – γ ) x At-2 + (it – kt ) x It-1

Where at is the return on capital in year t, kt is the cost of capital in year t, At-1 is the

invested capital at the end of the year t-1, it is the return on net investment in year t, and It-1

the net investment in year t-1. All variables are measured at the level of the equity and at the level of the entity. Note that invested capital and net investment are related as At-1 = At-2

+ It-1. The decay rate of excess return is denoted as δ, and the long-run excess return as γ.

Equation (1) can be rewritten as2:

(2) (at – kt ) = β0 + β1 x (at-1 – kt-1 ) / (At-2 / At-1 ) + β2 x (It-1 / At-1 )

Where β0 = – (1 – δ) x γ, β1 = (1– δ), and β2 = (i – k) + (1 – δ) x γ. Assuming that the excess

return on net investments is constant, equation (2) can easily be estimated. The estimated coefficients can be used to determine the fundamental parameters δ, γ, and (i – k). These fundamental parameters are likely to be industry specific. In the above γ is interpreted as a long-run excess returns, however, since invested capital is measured at historical prices γ also reflects (conservative) accounting biases. Based on existing literature, the following influences on the decay factor of excess returns are expected, as denoted in Table 1 on the following page. Table 2 summarizes the relevant literature on the convergence of excess returns.

2

(15)

Table 1 – Expected influences on the decay factor of excess returns

Influence Expected sign Intuition

Returns +/- Higher returns will reduce the decay factor of excess returns, thereby slowing down the process of convergence. However, industries in which excess returns are generated attract competition.

Competition + Higher competition increases the decay factor of excess returns. Depending on the entry and mobility barriers competition might drive away excess returns.

Cost of capital + To earn excess returns, returns need to be higher than the cost of capital. The higher the cost of capital, the higher the returns need to be and the less sustainable they are expected to be. Firms and industries with higher costs of capital are expected to have higher decay factors.

Net investments - Expected is that net investments reduce the decay factor of excess returns. However, Koller et al. (2010) find that excess returns tend to fade faster at firms that reinvest more (higher reinvestment rates) than at firms that reinvest less.

Table 2 – Relevant literature related to the convergence of excess returns

Author (year) Level Performance metric Sample years Regression analysis Industry

separation Classification Remarks Findings

Process of Convergence

Koller et al. (2010) Entity ROIC 1963 - 2008 No Yes GICS

Assumption of an implicit cost of capital,

net investments are not taken into account

There are large variations in rates of ROIC between and within industries. Individual company ROICs gradually tend toward their industry medians over time but are fairly persistent. Entities' ROICs revert to the mean but

at a much slower rate than seen in the full sample.

Gradual

Nissim and Penman (2001) Equity,

Entity ROE, RNOA 1963 - 1999 No No Three digit SIC

Constant cost of capital, equity risk premium of 6%. Net investments are not taken into account

ROE converges quick to an economy-wide average Quick

Hawawini et al. (2001) Entity

Economic Profit, Market

Value, ROA

1987 - 1996 No Yes Three digit SIC

Performance is industry specific rather than company specific for the average company. Industry specific factors can have a different meaning for different types

of firms within an industry. Industry factors can have a large impact on the performance of the 'also-ran' firms, while for the industry leaders and losers that firm

factors dominate.

NA

Fairfield et al. (2009) Equity ROE 1980 -

1992 Yes Yes GICS

No cost of capital taken into account. No

net investments.

ROE quickly converges to an economy-wide average. Industry specific models are generally more accurate in predicting firm growth but not profitability. The effect

of industry membership on firm performance is dependent on the specific performance measure.

(16)

16 t j t j t j t j j j t j A A r r _, 2 , 1 , 1 , 1 0 , /          

3 Methodology

This paper focuses on an empirical assessment of future firm performance when valuing companies and making forecasts. The effects of competition and the persistence of excess returns are analyzed at entity and equity level3.Koller et al. (2010) argue that ROIC is a better analytical tool for understanding the company’s performance than ROE because it focuses solely on a company’s operations. Therefore, the analysis is mainly focused at the level of the entity. The estimations in previous empirical research by Koller et al. (2010), Nissim and Penman (2001) and Fairfield et al. (2009) are biased. The effects of reducing these biases on the economic meaningful coefficients as for example the decay factor of excess returns are analyzed. The regression coefficients are a mix of the decay rate of excess returns, the level of long-run excess returns, and excess returns on net new investments. First, Koller et al. (2010) and Nissim and Penman (2001) measure the convergence of returns assuming implicitly that the cost of capital and the steady state of long-run excess returns are constant. Koller et al. (2010) allow for industry specific coefficients whereas Nissim and Penman (2001) do not. Equation (3) shows the model which captures these assumptions at the level of the entity.

(3)

Where rj,t is the book return on net operating assets in year t, rj,t-1 is the book return on net

operating assets in year t-1, Aj,t-1 is the industry book value of total assets in year t-1 and Aj,t-2

is the industry book value of total assets in year t-2. The book return on net operating assets in year t-1 is divided by the lagged growth in assets to get a direct estimate of the decay factor of excess returns. The estimated coefficients are used to determine the fundamental parameters. The coefficient β0 encapsulates the level of long-run excess return; the steady

state of excess returns and β1 the decay factor of return. Note that the decay factor of

excess returns can be estimated as β1j = (1– δj) / (At-2 / At-1), where δj is the industry decay

factor. However, by dividing the book return on net operating assets by the lagged growth in assets the decay factor can be estimated as β1j = (1– δj). The subscript j stands for the

corresponding industry, whereas t stands for time measured in years. Koller et al. (2010) find

3_{In the models Total Assets (A) is shown. However, at equity level Total Assets is replaced by Total}

(17)

17









jt t j t j j t j t j t j j j t j A I A A k r k r _, 1 , 1 , 2 2 , 1 , 1 , 1 0 , /                 









jt t j t j t j j j t j A A k r k r _, 2 , 1 , 1 , 1 0 , /            

that returns at the level of the entity gradually converge toward their industry medians over time and remain fairly persistent. However, their analysis assumes implicitly a sample wide constant cost of capital and a constant long-run excess return and does not take into

account time and entity variety. Nissim and Penman (2001) analyze returns as well as excess returns at the equity level. They find that an entity’s return at the equity level quickly

converges to an economy-wide average. Equation (4) shows the model that takes into account explicitly the cost of capital on the pattern of convergence of excess returns.

(4)

Where at the level of the entity (r – k)j,t is ROIC over the WACC (excess return) in year t and

(r – k)j,t-1 is the excess return in year t-1. The coefficient β0 encapsulates the level of long-run

excess return and β1 the decay factor of excess return. Equation (3) can be seen as a

restricted form of equation (4). The restriction is that in equation (3) the cost of capital and the long-run level of excess return are both assumed to be constant. This however, implies that no future investments or nonzero investments that generate no excess returns will be made. Therein implicitly assumed is that the convergence of excess returns is fully due to competitive forces instead of for example the gradual eroding return on net investments. Expected is that firms can slow down the convergence of excess returns by making investments. However, it seems unlikely that firms can sustain levels of return on new investments and that their returns gradually fade. Hence, it is expected that the decay factor of excess returns is reduced by accounting for net investments. The omission of net

investments implies that all future returns are the result of past investments, which is highly unlikely. Equation (5) does take into account excess returns on net new investments.

(5)

The coefficient β2j captures the effects of excess returns of net new investments.

Additionally, Ij,t-1 is net industry investments in year t-1 and Aj,t-1 is the industry book value of

total assets in year t-1. Including the effects of net investments in the estimation is expected to reduce the decay factor of excess returns, thereby slowing down the process of

(18)

18



j



j

j  

₀ 1 

new investments. The long-run excess return γ is captured by the coefficient β0j as shown in

equation (6). (6)

Where δj is the decay factor of excess returns in industry j and γ is the long-run excess return

in industry j. The decay factor of excess returns is captured by the coefficient β1j.

(7)

The excess return on net new investments is captured by the coefficient β2j.

(8)

Where (ij – kj) is the excess return on new investments in industry j, ij is the return on new

investments in industry j and kj is the cost of capital in industry j. In this paper the

relationship between industries and their performance is of interest. A parsimonious first-order autoregressive forecasting model for returns and excess returns is estimated. The current excess returns are modeled as a linear function of lagged excess returns. Panel data analysis is used which increases the power of the test by increasing the number of

observations. Additional variation is introduced by combining time series and cross sectional series the number of observations is increased which helps to overcome problems of

multicollinearity that can arise if time series are modeled individually. By using Panel data sets individual entity’s heterogeneity is controlled for. Not controlling for these unobserved individual specific effects leads to a bias in the resulting estimates. Additionally, panel data sets are also better able to identify and estimate effects that are simply not detectable in pure cross-sections or pure time series data. First of all, a pooled regression is estimated, aggregating all the data together into one equation:

(9)

Where ƴi,t is a vector with entity’s (excess) returns in year t, α is the intercept term which

embeds the level of long-run excess returns, β is the coefficient of the entity’s lagged (excess) returns ƴi,t-1, ƴi,t-1 is a vector with entity’s (excess) returns in year t-1 and ui,t is a

vector with error terms. A Limitation of this approach is that by aggregating all the data into a pooled estimation the explicit assumption is that the average values of the variables and the relationships between them are constant over time and across all of the cross-sectional

(19)

19 units in the sample. This implies economy wide effects of competition and the assumption that there are no industry differences between decay factors of excess returns, returns and costs of capital. Tests for equality of means and medians are performed to statistically test if there are significant differences in excess returns, decay factors and net investments

between industries. To be more specific, ANOVA F-tests are executed to test if there is a significant difference among the mean values of the variables between different industries. Furthermore, Kruskal-Wallis tests are performed to test if there is a significant difference among the median values of the variables between different industries.

Previous research finds that performance is industry specific rather than company specific for the average company (Hawawini et al. 2001) and that entity’ returns gradually tend toward their industry steady state (Koller et al. 2010). Therefore, based on theoretical and statistical evidence, the dataset is divided into separate industries. Industry dummy variables are included in the estimation to capture the industry specific effects. Technically, different industries might have different intercepts and different slopes. In economic terms industries might have different decay rates, levels of long-run excess returns, and excess returns on net investments. An economy wide estimation as in Nissim and Penman (2001) pools firms across industries thereby forcing the estimated impact of current firm performance on future firm performance to be uniform across all firms, irrespective of industry. An economy wide estimation model is appropriate if, on average, all firms’ returns converge at the same rate to economy wide benchmarks, for example, due to general competitive forces (Stigler, 1963). However, research finds that firm profitability converges to industry specific

benchmarks (Fairfield et al., 2009).

There are broadly two classes of panel estimator approaches that can be employed, namely, fixed effects models and random effects models. The effects of competitions on excess returns, as measured by the decay rate, might differ between industries and might differ between profit-making and loss-giving companies. Additionally, the effects of net

(20)

20 given point in time. Due to the industry dummy variables, cross section fixed effects cannot be employed in the industry specific estimation.

The entities in the sample cannot be thought of as having been randomly selected from the population as they effectively constitute the entire population. Additionally, unobserved omitted variables that are correlated with the included explanatory variables do not have to be taken into account in the fixed estimation procedure whereas the random effects model would be inconsistently estimated. The Hausman test helps analyzing whether there is significant correlation between the unobserved firm specific random effects and the regressors. In addition the Redundant Fixed Effects Likelihood Ratio Test is employed. Based on theoretical and empirical grounds, the fixed effects estimation approach is

preferred. By excluding the technology dummy variable the dummy variable trap is avoided which is described as perfect multicollinearity between the dummy variables and the

intercept term. By excluding the technology industry dummy variable the industrial features in the data are captured, although the interpretation of the coefficients changes. The

estimated coefficients on the other industry dummy variables represent the average deviations of the dependent variables for the included industries from their average values for the technology industry dummy variable.

The assumption related to the ordinary least squares method is that the explanatory

variables are not correlated with one another. Therefore, the correlation between different variables is measured using Spearman’s rank correlation coefficient and Pearson’s product-moment correlation coefficient. Spearman’s rho is a nonparametric measure of statistical dependence; it assumes no specific distribution. While Pearson’s rank correlations test underlying assumption is a linear distribution.

The following equation is employed to test the hypotheses, including the industry dummy variables and period fixed effects:

(10)

Where ƴj,t is (excess) return in industry j in year t, α is the intercept term in industry j, βj is

the coefficient of excess return in industry j, ƴj,t-1 is (excess) return in industry j in year t-1. μj

is the coefficient of the dummy variable of industry j, Dj is the dummy variable of industry j

and νj,t is the error term of industry j in year t. Note that this equation is general as it only

shows the estimation procedure. In line with previous research (Fairfield et al. 2009), the adjusted R2 is used in the estimation output because the models presented have varying

(21)

21 numbers of independent variables. To compensate for the addition of variables to the

model, the adjusted R2 corrects for the number of independent variables.

Unlike the mean, median values are not influenced by outliers at the extreme of the data set, especially when the underlying distribution is not normally distributed. This is the case, and therefore mostly median values are used.

In addition to the above basic fixed effects specification, two additional estimation

techniques will be used. First, panel-corrected standard error (PCSE) estimates are applied in which a heteroscedastic error structure is combined. Period specific variance with an AR(1) process where the correlation parameter is unique for each period is allowed for. The resulting coefficients are exactly the same as with the fixed effects approach, however, the standard errors are smaller. Potential problems with heterosckedastic errors and

(22)

22

4 Data

This research is based upon the companies listed in the S&P 500 index in the period 1988-2012. The composition of the S&P500 changes at more or less regular intervals. These changes are taken into account by using all firms that currently are included and looking back till 1988. The S&P 500 is a free-float capitalization-weighted stock market index. The database Thomson Reuters DataStream is used to retrieve the financial data for each of the 500 companies constituting the index. The dataset consists of an unbalanced panel due to missing observations. However, this has no effect on the estimations because they are automatically accounted for by the software package.

To subdivide the dataset into different industry classes the Industry Classification Benchmark (ICB) is used in contrast to the usage of the Global Industry Classification Standard (GICS) in research by Fairfield et al. (2009) and Koller et al. (2010) and the usage of the Standard Industrial Classification (SIC) by Nissim and Penman (2001) and Hawawini et al. (2001). These classifications are not used in this paper due to their unavailability in DataStream.

(23)

23

Table 2 - Data selection procedure

S&P 500 Number of firms

Total number of firms 500

Financial firms (ICB 8000) 84

Incomplete data 14

Total firm used from sample 402

This table shows the number of firms that constitute the analysis. The firms in the financial industry are excluded as well as incomplete data records.

4.1 Variables

In this section the variables are defined, the definition and the Mnemonic as provided by DataStream are given. Koller et al. (2010) define economic profit as whether a company is earning its cost of capital and how its financial performance is expected to change over time. Economic profit will be used in this research at enterprise and equity level. Economic profit at enterprise level can be stated as:

(11)

Where ROICstands for the Return On Invested Capital, WACC is the cost of capital and IC is the invested capital. In this paper excess returns at the level of the entity are calculated as: (12)

The ROIC is defined as the return that the company earns on each dollar invested in the business. The ROIC is calculated as follows:

(13)

Where Earnings before interest and taxes (EBIT)[WC18191] is calculated by taking the pre-tax income and adding back interest expense on debt and depreciation and amortization and subtracting interest capitalized. Net Debt (D)[WC18199] represents the book value of Total Debt minus cash and Total Shareholders’ Equity (E) *WC03995+ represents the sum of

Preferred Stock and Common Shareholders Equity at book value. Finally, Txo is the operating tax rate. The value of the operating taxes is calculated by using the following equation: (14)

Where Txr [WC01451] stands for the reported tax rate, Txm stands for the US marginal tax rate; this rate is extracted from KPMG4 as of 5 January 2013 and IEO[WC01251] is the companies’ interest and other expenses.

Koller et al. (2010) define the WACC as the rate of return that investors expect to earn from

4_{The US marginal tax rate in the is subtracted from the website of KPMG:}

(24)

24 investing in the company adjusted for corporate taxation. The WACC represents rates of return required by the company’s debt and equity holders blended together, and as such is the company’s opportunity cost of funds. The WACC is calculated as:

(15)

Where Kd stands for the cost of debt and Ke stands for the cost of equity. Kd is approximated by using the following formula:

(17)

Where rf stands for the risk free rate in year t and τ stands for the implied industry’s credit spread. The risk free rate is approximated by using the historical US Treasury Bond rate with a maturity of 10 years5. The implied credit spread is calculated as the difference between the estimated cost of debt per industry6 and the risk free rate. The cost of equity capital Ke is defined as the opportunity cost of equity funding and is calculated as:

(18)

Where ß* is the Blume adjusted beta of the individual company in year t representing its incremental risk to a diversified investor and λ is the market risk premium in year t. Koller et al. (2010) find that different forms of measurement converge on an appropriate range of market risk premium of 4.5 to 5.5 percent. Therefore, first a constant 5 percent is used in the analysis. As a robustness check the results are also crosschecked with a market risk premium of 4 and 6 percent.

Koller et al. (2010) state that according to the CAPM, a stock’s expected return is driven by beta, which measures how much the stock and entire market move together. Beta cannot be observed directly; therefore its value must be estimated. First the raw beta must be

estimated using regression, and smoothing techniques need to be used to improve the estimate. I estimate the individual raw beta by calculating the monthly continuous

compounded returns of the Total Return Index [RI] of the individual firms and the S&P500. Annual betas are estimated by using 12 months returns. The following formula is used to calculate the beta of the individual securities:

5

The risk free rate is subtracted from the website of the Federal Reserve: http://www.federalreserve.gov/releases/h15/data.htm

6_{The estimated cost of debt per industry is extracted from the website:}

(25)

25 (19)

Where βi,t is a measure of the risk of security i in relation to the market, ri,t is the monthly

return of security i, rmt is the monthly return of the S&P500 and σ² (rmt) is the variance of

the return of the S&P500. Estimated betas will be partly determined by its fundamental value and partly by sampling error. To smooth the beta the Blume adjustment is performed. This adjustment is based on empirical research by Blume (1975) who found that estimated beta coefficients tend to regress towards the mean over time. The following equation clarifies the beta adjustment:

(20)

A complete list of the raw variables can be found in table C.1 in Appendix C, a table of the constructed variables and their abbreviations can be found in Appendix D, table D.1.

4.2 Outliers

Outliers are unusual data points that occur randomly in a dataset and are caused by measurement error such as coding errors or wrong data-collection procedures. In the OLS framework, ignoring the underlying causes of these unusual observations is a form of model misspecification and a potential correlated omitted variables problem. The variables of companies in the dataset which have a negative cost of capital are excluded because these values would lead to an increase in returns which does not make sense economically. In line with Fairfield et al. (2009), all values of the dependent and independent variables are winsorized and truncated. Winsorizing and truncating data alters legitimately occurring extreme observations that can contain information that improves estimation efficiency. In a truncated estimator the extreme values are discarded, in a Winsorized estimator the

extreme values are replaced by certain percentiles (the trimmed minimum and maximum). First of all, all dependent and independent variables are winsorized at 1% and 99% to remove outliers. Furthermore, to ensure that winsorized values do not drive the reported results, the observations in the 1% and 99% levels are truncated. This process is repeated using 5% and 95%. These results are qualitatively similar and are reported in this report because these values do not contain economic unrealistic values.

(26)

26

4.3 Descriptive statistics

Table 3 reports descriptive statistics for the full sample at the level of the entity, Panel A and Panel B show the descriptive statistics and the Pearson and Spearman correlation matrix respectively. The descriptive statistics and correlations at the equity level can be found in the Appendix E table E.1 and the industry descriptive statistics can be found in table F.1 in Appendix F. The descriptive statistics show that the variables are not normally distributed. However, by employing panel estimation the normality assumption is irrelevant due to the large number of observations. Additionally, the results of tests of equality of industry

medians and means of the returns, costs of capital, excess returns, excess returns divided by the growth in total assets and net investments are presented in table F.2 in Appendix F. The results show significant industry differences in excess returns, cost structures and net investments.

Table 3 - Descriptive statistics and Correlations at Enterprise Level

Panel A: Descriptive statistics

ROICt-1 WACCt-1 (ROIC-WACC)t-1 (ROIC-WACC)t-1

/(At-1/At-2) It-1/At-1 Mean 0.165 0.086 0.076 0.068 0.092 Median 0.144 0.084 0.057 0.051 0.075 Maximum 0.517 0.147 0.423 0.363 0.413 Minimum -0.029 0.047 -0.144 -0.126 -0.123 Std. Dev. 0.097 0.021 0.100 0.088 0.105 Skewness 1.122 0.533 0.930 0.886 0.675 Kurtosis 4.375 2.843 4.008 3.854 3.249 Jarque-Bera 2114 274 1059 907 577 Probability 0.000 0.000 0.000 0.000 0.000 Observations 7328 5677 5677 5628 7338

Panel B: Pearson and Spearman correlations coefficients

ROICt-1 WACCt-1 (ROIC-WACC)t-1 (ROIC-WACC)t-1

/(At-1/At-2) It-1/At-1 ROICt-1 1.000 -0.067 0.975 0.960 0.192 WACCt-1 -0.058 1.000 -0.288 -0.301 0.085 (ROIC-WACC)t-1 _0.957 _-0.309 _1.000 _0.989 _0.165 (ROIC-WACC)t-1 /(At-1/At-2) 0.950 -0.318 0.996 1.000 0.074 It-1/At-1 0.206 0.070 0.179 0.110 1.000

(27)

27

5 Results

Table 4 presents the results of the economy wide and industry specific panel regression analysis. As in the empirical research by Koller et al. (2010) and Nissim and Penman (2001) the convergence of returns to its long-run level with an implicit cost of capital and a constant long-run level of excess returns is analyzed in model A. Model B relaxes the assumption of a constant and equal cost of capital of all companies in that each company has an individual WACC and an individual level of long-run excess returns as in the empirical research by Nissim and Penman (2001). Finally, Model C presents the results of the model employed in this research. Net investments are included in the analysis and the effects of investments on the convergence of excess returns to its long run level are analyzed. Panel A shows the regression output of the economy wide regression similar to Nissim and Penman (2001) and Panel B the regression output of the industry analysis similar to Koller et al. (2010). Note that the technology dummy variable is excluded from the industry estimation to overcome

problems with multicollinearity. Additionally, the regression coefficients in panel B are relative due to the omission of the technology dummy variable. The intercept term captures the level of long-run excess return, the coefficient ROICt-1 / (At-2/At-1) captures the decay

factor of returns with an implicit cost of capital as in Koller et al. (2010) and Nissim and Penman (2001), the coefficient (ROICt-1 – WACCt-1)/ (At-2/At-1) captures the decay factor of

excess returns with an explicit heterogeneous and nonconstant cost of capital as in Nissim and Penman (2001) and the coefficient It-1/At-1 captures the excess returns on new

(28)

28 factor of excess returns increases from 0.575 to 0.777 when the costs of capital are explicitly modeled. After accounting for the net investments the coefficient reduces from 0.777 to 0.502. Again, the relationships are unexpected; accounting for an explicit cost of capital reduces the economy wide decay factor of excess returns. Whereas including net

investments into the analysis increases the decay factor of excess returns. The coefficient that embeds the excess returns on net investments is positive with a value of 0.054. By including net investments into the economy wide analysis, the variance explained by the model increases from 0.586 to 0.598.

An industry specific analysis is more appropriate in estimating the convergence of excess returns as will be elaborated upon in the analysis of the fundamental parameters.

In Panel B, almost all coefficients are significant at a level of significance of 1% meaning that the dependent variables have a statistically significant impact in explaining the independent variable. Again, the three models fit the data well given their significance at 1%. The

intercept term in which the industry level of long-run excess return is captured is significant at a significance level of 1% in the three models. In model A in panel B, all coefficients are significant at a significance level of 1% except for the health care industry which is significant at a significance level of 10%. Therefore, the coefficient of the healthcare industry needs to be interpreted with caution. In model B, all coefficients are significant at a level of

(29)

29 increases the level of long-run excess returns which is in line with expectations.

The excess return coefficient that embeds the decay factor of excess returns is positive and reduces from 0.804 to 0.767. This means that the decay factor of excess returns increases when the cost of capital is explicitly modeled. The excess return coefficient of model C is positive with a value of 0.777, meaning that including net investments into the analysis decreases the decay factor of excess returns. The changes in the decay factor of excess returns are confirmed at the equity level as can be seen in table G.2 in Appendix G. However, most of the coefficients are insignificant in the analysis at the equity level meaning that the independent variables have no statistical impact in explaining the dependent variable. This result shows that the performance metric at the entity level is a better analytical tool for understanding the company’s performance than at the equity level because it focuses solely on a company’s operations. The analysis at the equity level mixes up operating and financial performance and therein distorts the analysis of economic profit. Finally, the net investment coefficient is positive and significant at 1% with a value of 0.069. From the net investment coefficient the excess return on net investments can be inferred. By including net

(30)

30

Table 4 – Coefficients of the convergence of excess returns

Panel A: Economy wide Panel B: Industry specific

Variable A B C A B C Intercept 0.079*** 0.026*** 0.040*** 0.065*** 0.045*** 0.032*** (-38.245) (-38.932) (-26.112) (-17.087) (-11.738) (-8.687) ROICt-1/(At-2/At-1) 0.575*** 0.804*** (-42.722) (-73.119) (ROIC-WACC)t-1/(At-2/At-1) 0.777*** 0.502*** 0.767*** 0.777*** (-32.533) (-31.792) (-61.859) (-64.162) It-1/At-1 0.054*** 0.069*** (-5.129) (-7.118)

Oil & Gas -0.030*** -0.029*** -0.024***

(-6.654) (-5.618) (-5.327) Basic Materials -0.030*** -0.024*** -0.016*** (-5.897) (-4.408) (-3.401) Industrials -0.022*** -0.020*** -0.013*** (-5.615) (-4.456) (-3.208) Consumer Goods -0.022*** -0.017*** -0.010** (-5.242) (-3.741) (-2.531) Health Care -0.009* -0.009 -0.005 (-2.126) (-1.740) (-1.164) Consumer Services -0.012*** -0.011** -0.004 (-2.898) (-2.392) (-0.998) Telecommunications -0.042*** -0.043*** -0.036*** (-5.398) (-5.142) (-4.787) Utilities -0.042*** -0.039*** -0.030*** (-9.145) (-7.624) (-6.583) Adjusted R-squared 0.618 0.586 0.598 0.564 0.517 0.535 Prob(F-statistic) 0.000 0.000 0.000 0.000 0.000 0.000 Observations 6351 5242 4861 6351 5242 4861 Firms 403 393 391 403 393 391

Table 4 presents the result of the economy wide analysis of the convergence of excess returns in Panel A and the results of the industry specific analysis in Panel B. The cost of capital is explicitly modeled and the effect of including net investments on the convergence of excess returns is analyzed. Model A is the model that is used in the empirical research by Koller et al. (2010), Model B is the model that is used by Nissim and Penman (2001) and Model C is the model employed in this research. The t-statistic is presented under each coefficient and in parentheses, additionally, *** denotes significance at 1%, ** at 5 %, * at 10%.

Table 5 presents the fundamental parameters in this research. Panel A present the economy wide parameters similar to Nissim and Penman (2001) and Panel B the industry specific parameters similar to Koller et al. (2010). Koller et al. (2010) find a gradual

convergence of excess returns whereas Nissim and Penman (2001) and Fairfield et al. (2009) find a quick convergence of excess returns. Expected is that explicitly including a

(31)

31 level of long-run excess returns.

The values of the industry specific parameters in this table are not relative in that they represent average deviations from the excluded technology industry. In this table, γ is the long-run level of excess returns, δ is the decay factor of excess returns and (i-k) is the excess return on new investments.

The empirical findings of a quick convergence of economy wide excess returns of Nissim and Penman (2001) are confirmed which can be seen when looking at the decay factors of the three models in Panel A. The gradual convergence that is found by Koller et al. (2010) is not confirmed due to the slight deviation in the estimation approach employed.

The results of the economy wide estimated parameters in model A and model B show that when the cost of capital explicitly is taken into account the decay factor (δ) of excess return reduces against expectations from 0.425 till 0.223. By explicitly modeling the costs of capital and relaxing the restriction of a homogeneous constant cost of capital and level of long-run excess returns, the convergence of excess returns as measured by the decay factor of excess returns changes. Additionally, by including net investments into the economy wide analysis the decay factor of excess returns increases from 0.223 to 0.498, an even higher decay factor than found in the analysis in model A. Expectations is a reduction in the decay factor of excess returns when net investments are accounted for.

When accounting for the explicit costs of capital, the economy wide level of long run excess returns (γ) increases from -0.137 to -0.034. After including net investments into the analysis the level of long-run excess return remains negative and reduces from -0.034 to -0.079 meaning that the economy wide steady state of firms’ excess returns on average has a negative sign over the period of 1988-2012 which seems unrealistic. Important to notice is that the economy wide excess return on new investments (i-k) is also negative with a value of -0.015 meaning that on average the net investments are below their cost of capital and therefore destroying value.

Nevertheless, the quick convergence of excess returns that Nissim and Penman (2001) find is confirmed at the economy wide estimation. However, the change of the decay factor of excess returns is opposite to what is expected. Expected is that the decay factor of explicit heterogeneous stationary excess returns is higher than the decay factor of implicit constant homogeneous excess returns and that the level of long-run excess returns reduces.

(32)

32 decay factor and increase the level of long-run excess returns. The results show that

accounting for the explicit costs of capital increases the level of long-run excess returns and reduces the decay factor, both unexpected. By accounting for net investments the level of long-run excess returns reduces and the decay factor increases to a higher level than in the first model, which is also unexpected.

The empirical findings of a quick convergence of excess returns of Nissim and Penman (2001) are also confirmed in the industry specific estimation which can be seen when looking at the industry specific decay factors presented in the three models in Panel B. The gradual

convergence of industry excess returns that Koller et al. (2010) find is not confirmed. Model A, the estimation model employed in the empirical research by Koller et al. (2010), also shows a quick convergence of excess returns. The decay factor differs per industry and confirms the empirical findings of different industry performances of Koller et al. (2010). When an industry specific approach is employed instead of an economy wide approach the decay factors of excess returns are reduced (note that the economy wide decay factor of excess returns was 0.498 in model C in table 4). The fundamental parameters shown in panel B are not relative in that they represent average deviations from the excluded industry. Comparing model A and Model B in Panel B confirms the expectation that explicitly

accounting for the costs of capital increases the industry decay factor of excess returns. By explicitly modeling the costs of capital and relaxing the restriction of a homogeneous

constant cost of capital and level of long-run excess returns as in Nissim and Penman (2001), the convergence of excess returns as measured by the decay factor of excess returns

changes. The industry levels of long-run excess returns also change after employing explicit costs of capital. Against expectations the industry levels of convergence increase after taking the costs of capital into account. Taking into account an industry-specific time-varying cost of capital increases the decay factor of industry’ excess returns. However, the level of long-run excess returns increases also, which is against expectations.

(33)

33 industry decay factors that are found employing the model of Koller et al. (2010) increase by accounting for net investments. The expected relationship of a reduction in the decay factor of excess returns when net investments are included in the analysis is found after accounting for explicit industry costs of capital. The decay factor of excess returns decreases when net investments are taken into account. The Oil & Gas industry confirms this image, the decay factor of excess returns decreases from 0.261 in model B to 0.248 in model C.

Except for the Telecommunications industry, the industries on average converge to a negative level of long-run excess return. This level is slightly less negative in the estimation employed in this research compared to the research models employed by Koller et al. (2010) and Nissim and Penman (2001). However, this level remains against expectations and

economic theories negative.

Due to their positive level of long-run excess returns in model C, the Telecommunications industry presents a clear image. By taking into account an explicit stationary cost of capital and level of long-run excess returns as in Koller et al. (2010), the decay factor of excess returns increases as expected from 0.239 to 0.275. However, the long-run level of excess returns increases also which is against the expectations from -0.067 to -0.045. Introducing the net investments into the analysis has the expected effect on the decay factor of excess returns, which is reduced from 0.275 to 0.259. The level of long-run excess returns is positively influenced by net investments as it increases from -0.003 to 0.005. Additionally, the excess return on new investments has a negative sign in the Telecommunications industry with a value of -0.073. The negative excess returns in this industry are even higher than the other industries, which is remarkable considering their positive level of long-run excess returns. Although the decay factors of industry excess returns show the expected relationships, the negative level of industry long-run excess returns and the negative excess returns on industry new investments are unexpected and not in line with economic theories. Perhaps within industry estimations could give a more accurate and realistic estimation of the industry level of long-run excess returns at the level of the entity. Further research might include omitted variables or introduce a more advanced estimation procedure to resolve this problem.

Assuming a constant and equal cost of capital for all companies’ economy wide

(34)

34 slowed down. Additionally, the steady state of returns as measured by the level of long-run excess returns is underestimated thereby predicting a length of convergence of excess returns that is too short. This is contrasting to the expected relationship. Including the net investments into the analysis increases the process of convergence and reduces the long-run level of excess returns. Net investments seem to be value destructive in the short term by increasing the decay factor of excess returns. However, they increase the level of long-run excess returns and thereby lengthen the convergence of excess returns. On average the net investments are negative NPV projects and thereby value destructive. Statistical and

theoretical evidence shows that an economy wide estimation as in Nissim and Penman (2001) is not correct. Measuring economy wide performance forces the estimated impact of current firm performance on future firm performance to be uniform across all firms,

irrespective of industry. First, the results of the ANOVA F-tests and Kruskall-Wallis tests in table F.2 in Appendix F show that there are significant differences in mean and median values of returns, excess returns, costs of capital and net investments. Second, the untabulated outcomes of the Redundant Fixed Effects Likelihood Ratio Test show that pooled sample estimation is inappropriate to employ. Third, the industry specific panel estimation shows that there are industry differences in levels of long-run excess returns, decay factors and excess returns on new investments. Additionally, Fairfield et al. (2009) finds that industry membership is a fundamental determinant of firm performance. When benchmarking the historical decay of the company’s returns, it is important to segment results by industry (Koller et al., 2010). In this research industry segmentation is especially important because industries are used as a proxy for sustainability of competitive advantage. Therefore, the industry specific analysis of the convergence of excess returns is more

appropriate than the economy wide analysis.

When the industry specific approach is employed, the long term excess return reduces after taking into account the explicit cost of capital and reduces even further when net

(35)

35 capital is assumed to be constant and the same for all firms. After correcting for this bias the estimated decay factor is still biased due to exclusion of net investments from the analysis. By omitting net investments the industry decay factor of excess returns is biased upwards. Industry net investments on average slow down the convergence of excess returns and increase the level of long-run excess returns, these results are confirmed at the equity level which can be seen in table G.2 in Appendix G. The value of future excess returns is

(36)

36

Table 5 - Industry analysis, the fundamental parameters

Table 5 presents the fundamental parameters of the industry analysis of the convergence of excess returns. The cost of capital is explicitly modeled and the effect of including net investments on the convergence of excess returns is analyzed. Panel A presents the results of the economy wide analysis similar to Nissim and Penman (2001), Panel B presents the results of the industry analysis similar to Koller et al. (2010). In this table k is the WACC, γ is the long-run level of excess returns which is captured by the intercept, δ is the decay factor of excess returns which is captured by the excess return coefficient and (i-k) is the excess return on net investments which is embedded in the net investments coefficient. The fundamental parameters are not relative in that they represent average deviations from the excluded industry. Model A is the model that is used in the empirical research by Koller et al. (2010), Model B is the model that is used by Nissim and Penman (2001) and Model C is the model employed in this research.

A B C

Panel A: Economy wide

γ -0.137 -0.034 -0.079

δ 0.425 0.223 0.498

(i-k) -0.015

Panel B: Industry Specific

Oil & Gas

A More General Approach to Estimate the Effects of Competition on Performance

A More General Approach to Estimate the Effects of

Competition on Performance

A More General Approach to Estimate the Effects of

Competition on Performance

Table of Contents

1 Introduction

2 Literature Review

3 Methodology





















4 Data

5 Results