Advanced Topics in Accounting

Table of Contents

Abstract

List of figures

List of tables

List of abbreviations

1.Introduction

2. Literature review
2.1 Distinction between equity and entity level valuation
2.2 Multiples based valuation
2.2.1 Accuracy of multiples
2.2.2 Selecting comparables
2.2.3 Choosing multiples
2.3 Accounting flows based
2.3.1 Dividend discount model
2.3.2 Free cash flow model
2.3.3 Residual income valuation model
2.3.3.1 Derivation
2.3.3.2 Benefits of the model
2.3.3.3 Performance
2.3.3.4 Implementation issues
2.3.4 Abnormal earnings growth model
2.3.4.1 Derivation
2.3.4.2 AEGM versus the RIVM
2.4 Valuation of defensive and cyclical firms
2.5 Concluding remarks

3. Large sample analysis
3.1 Research question
3.2 Sample selection
3.3 Methodology
3.3.1 Multiples based valuation
3.3.1.1 Selected value driver
3.3.1.2 Identification of comparable companies
3.3.1.3 Estimation of value
3.3.2 Accounting flows based valuation
3.3.2.1 Cost of equity capital
3.3.2.2 Dividend payout rate
3.3.2.3 Forecasted earnings
3.3.2.4 Forecast horizons and continuing value assumptions
3.4 Results and analysis
3.4.1 Sample descriptive statistics
3.4.2 Valuation errors
3.4.2.1 Descriptive statistics of valuation errors
3.4.2.2 Cross sample comparison of valuation errors
3.4.2.3 Comparison of valuation errors within each sample
3.4.3 Interaction between cyclicality and economic state
3.4.4 Price explainability
3.5 Sensitivity tests
3.5.1 Multiples based valuation
3.5.2 Flow based valuation
3.5.2.1 CAPM assumptions
3.5.2.2 Terminal value
3.6 Concluding remarks

4. Small sample analysis
4.1 Research question and hypothesis development
4.1.1 Target prices (Hypothesis 1]
4.1.2 Earnings (Hypothesis 2]
4.1.3 Flow models (Hypothesis 3]
4.1.4 Investment recommendations (Hypothesis 4]
4.2 Sample selection
4.3 Results and analysis
4.3.1 Industry characteristics
4.3.1.1 Education services
4.3.1.2 Tobacco manufacturers
4.3.1.3 Food products manufacturing
4.3.1.4 Health services
4.3.1.5 Utilities
4.3.1.6 Housing contractors
4.3.1.7 Primary metal manufacturing
4.3.1.8 Automobile manufacturing
4.3.1.9 Airlines
4.3.1.10 Paper manufacturing
4.3.2 Tests of Empirical Hypothesis
4.3.2.1 Target prices versus actual price (Hypothesis 1]
4.3.2.2 Earnings as a value driver (Hypothesis 2]
4.3.2.3 Use of flow based models (Hypothesis 3]
4.3.2.4 Analyst investment ratings (Hypothesis 4]
4.4 Concluding remarks

5. Conclusion

6. References

7. Appendices
7.1 Sensitivity of median absolute valuation errors
7.2 Sensitivity of median signed valuation errors
7.3 Stock selection for the small sample analysis
7.4 Analysts use of a cyclically adjusted earnings number
7.5 Models used when earnings are not used
7.6 Brokers reports used

II. Abstract

The empirical research in this thesis aims to better understand two complementary components. First, whether there is a significant difference in the performance of accounting-based valuation models when examined across different industry types (i.e. cyclical or defensive) and economic states (i.e. growth or recession). Second, how analysts use accounting-based valuation to justify investment recommendations and whether this changes across the samples aforementioned. The findings reveal significant insights in both components of the research. Using a large sample setting, the thesis finds that a simple forward earnings to price multiple performs well in all environments, except for cyclical stocks in a recession which are characterised by high levels of volatility. Accounting-flow models are at their best when used in this setting and are therefore great complement to traditional earnings based valuation. The small sample analysis confirms expectations from the large sample, as analysts have a preference for earnings multiples, however recognise that under certain conditions flow models are more meaningful.

List of figures

Figure 1 - The business cycle

Figure 2 - FCF volatility

Figure 3 - Absolute valuation errors

Figure 4 - Signed valuation errors

List of tables

Table 1 - Large sample descriptive statistics

Table 2 -Descriptive statistics of errors

Table 3 - Cross sample analysis of errors

Table 4 - Within sample analysis of errors

Table 5 - Interaction between cyclicality and economic state

Table 6 - Price explainability using OLS regression

Table 7 - Difference between target and actual price

Table 8 - Using earnings as a value driver

Table 9 - Use of a flow based model

Table 10 -Analysts recommendations

List of Abbreviations

illustration not visible in this excerpt

1. Introduction

Between late-2007 and mid-2009, the United States (herein referred to as the US) suffered the worst economic recession since the Great Depression of the 1930's (Gascon, 2009). Its global impact is still felt to this day, with many economies still in (or having returned to) a recession at the time of writing^[1]. In a recession, stock markets are characterised by high levels of volatility and large decreases in market value (Schwert, 2011). The US S&P 500 index lost 56 percent of its value and the VIX index more than tripled between 2007 and 2009^[2] (Manda, 2010). The term business cycle is used to refer to fluctuations in economic activity; essentially a business cycle is an alternating pattern of recession and recovery (ECRI, 2012), it is captured by 4 stages shown below in Figure 1 (NBER, 2012). Note that in many textbooks a recession is defined as two consecutive quarters of decline in real gross domestic product (GDP), this thesis however, uses the National Bureau of Economic Research's (NBER) definition of a recession, which takes into account real GDP, real income, employment, industrial production, and wholesale-retail sales (NBER, 2012).

Figure 1 -The Business Cycle

A simple diagram showing the business cycles, the Y axis shows level of economic activity and the X axis shows time. This is used for illustration purposes only as actual economic activity behaves stochastically.

illustration not visible in this excerpt

Some industries are more affected by changes in economic activity than others. Analysts refer to industries whose cash flows and earnings follow the ebb and flow of economic cycles, such as car manufacturers, as cyclical industries and industries which do not, such as utility companies, as defensive^[3] industries (Damodaran, 2002). They are called 'defensive' industries because in an economic recession investors can buy these stocks to defend themselves against volatile markets. As a rule of thumb, defensive industries tend to include companies which provide goods or services the population will buy regardless of disposable income.

This thesis aims to better understand how accounting based valuation models are used to value these two very different industry types. First, it reviews the literature surrounding accounting based valuation models and valuation of cyclical and defensive industries. Using a large sample, it then analyses the performance of different valuation models depending on industry type and economic state (specifically in growth and recession periods). Finally, a small sample analysis is used to connect theory developed in the literature and findings from the large sample to what analysts use in practice.

2. Literature Review

The role of accounting numbers in equity valuation is discussed in depth in the literature. Much of the empirical research has revolved around analysing historical and forecasted accounting numbers to reach a value estimate (Richardson and Tinaikar, 2004). This review will critically discuss the issues surrounding accounting based valuation, with a focus on multiples (Section 2.2) and accounting-flows based valuation (Section 2.3). Where possible, it will identify the links between valuation models and the advantages and disadvantages of each. This chapter is structured as follows, first a brief distinction between how a company is valued at entity and equity level is made. Then a review of the literature surrounding multiples and accounting flows based valuation will be carried out, before examining the specific issues facing the valuation of cyclical and defensive industries.

2.1 Distinction between Entity Level and Equity Level Valuation

The distinction between entity level and equity level valuation can be drawn from Modigliani and Miller (1958) who state that the value of a company's economic assets must equal the value of the claims against those assets. Thus, to calculate the value of equity of a company, an analyst can either value the company's operations and subtract all non-financial claims (i.e. net debt), or value the equity holders' claims directly. The former is what is referred to as entity level valuation and the latter is what is referred to as equity level valuation. An analyst may wish to value a firm at entity level to take into account financing risk (Penman, 2010). This thesis examines the performance of valuation models from an equity level to allow comparison to key literature. To show the difference in treatment between entity and equity level flow models, the free cash flow model in this chapter is looked at from an entity perspective and all other models from an equity level.

2.2 Multiples Based Valuation

Multiples based valuation (sometimes referred to as comparables or comps) involves the use of comparable companies' multiples to value the target company. It is implemented in three steps as follows. Firstly, a comparable company or set of companies should be chosen which have similar characteristics to the target company. There are ongoing discrepancies in the literature over the best way to choose comparable companies which will be discussed in more detail, although generally comparables should have a similar growth rate, product line, and size to the target company. Secondly, a set of measures should be identified as value drivers and used to calculate the multiples that are used in the valuation process. Again, there are ongoing discrepancies in the literature over the best multiples to use. Finally, the mean, median, harmonic mean, or another measure of central tendency should be used on the comparable multiple. The averaged multiple should then be multiplied by the value driver to estimate the target firm value, as shown in Equation 1. If the multiples used are entity level multiples, then the net debt should be subtracted from the value estimate to give the value estimate of equity.

illustration not visible in this excerpt

Where:

VEi = value estimate VDi = value driver

M- multiple derived from comparable firms

A survey of Morgan Stanley analysts revealed that the most widely used valuation method by analysts was the P/E ratio, with over 50 percent of analysts using this method (Morgan Stanley Research, 1999: cited in Fernández, 2002). Barker (1999) found that out of all other valuation models the price-earnings (P/E) and dividend yield valuation was considered to be the most useful to analysts in helping reach an investment decision, more influential than the dividend discount model and the free cash flow (FCF) model. The importance and key advantage of multiples based valuation can perhaps be attributed to the simplicity of implementing such a method. It avoids forecasting analysis when looked at using historical numbers and minimises the amount of information that needs to be collected. Both Lee and Tweedie (1981) and Breton and Taffler (1995) cast doubt on whether market participants are sophisticated users of accounting information, making simplicity a key factor in their choice of valuation model. Kim and Ritter (1999) highlight the benefit of multiples in valuing IPOs, where it is often difficult to forecast future cash flows for a cash flow valuation due to the target company's lack of track record.

2.2.1 Accuracy of Multiples Based Valuation

Multiples using comparables relies on the economics 'law of one price' principle. That is, that two assets offering the same measure of value should sell at the same price. However, as Lee et al. (1999) point out, the process by which price adjusts to intrinsic value requires time, and price does not always reflect the intrinsic value of a firm. Furthermore, while multiples valuation can reduce the probability of misvaluing a firm relative to its comparables, it provides no safeguard against the entire sector being over or undervalued (Kim and Ritter, 1999) as sometimes occurs with economic bubbles. Lee et al. (1999) found that an intrinsic value measure that includes a time varying interest rate component produces a value to price ratio with a much better tracking ability and predictive power.

2.2.2 Selecting Comparables

Selecting comparable companies is an issue that is discussed extensively in the literature. It is the most difficult part of multiples based valuation to implement correctly. This is because it is unlikely that a company will be a 'pure-play'^[4] company, and finding fitting comparables can prove challenging. Often a sum of the parts approach needs to be used to value parts of the business separately with comparable parts. Bhojraj and Lee (2002) have been pioneers in developing a scientific approach to choosing comparables to make the process more objective, they have found that selecting comparables based on their "warranted multiples" produces a more accurate valuation than a more traditional approach.

In other studies, such as Alford (1992), the standard industrial classification (SIC) code is used to identify comparables. Alford (1992) found that narrowing down to the three digit SIC code improved valuation accuracy, but the accuracy of using four digits is no better than using three. It should, however, be noted that companies can sometimes be listed under more than one SIC code.

2.2.3 Choosing Multiples

Deciding what multiples or selection of multiples to base a company valuation on is again debated extensively in the literature. Bhojraj and Lee (2002) use multiples with sales or book value in the denominator; this allows companies with negative earnings to be valued as these numbers are unlikely to be negative. Beaver (1968) found that there is high information content in earnings numbers, which can change an investor's outlook on price. This supports research from Liu et al. (2002), which finds that multiples based on earnings, in particular forward earnings forecasts, performs better than cash flow measures such as a RIV based multiple. In Lie and Lie's (2002) research, they found that asset based multiples produced more accurate estimates of value than sales or earnings multiples, especially for financial companies that have a high amount of liquid assets and can be easily valued.

2.3 Accounting Flows Based Valuation

The idea behind flow based valuation is that the value of equity is equal to the present value of the expected flows to equity, discounted at the required rate of return; the methodology can be dated back to Williams (1938).

Broadly speaking this is shown with the formula below:

illustration not visible in this excerpt

The required rate of return on the stock is a function of the security's level of risk. If p is calculated using the capital asset pricing model then this risk is captured through the beta component. Therefore, if the required rate of return is relatively high on a stock we can deduce that the stock has a higher level of risk. As discussed in section 2.1, a DCF approach can also be used to value the entity as a whole. We can value the firm as the expected flows to the entity discounted at a firm's cost of capital. The cost of capital is the discount rate used, which takes into account financing risk from a firms debt obligations.

2.3.1 Dividend Discount Model (DIV)

The dividend discount model involves forecasting future dividends and discounting them by the cost of equity to give their present value at the valuation date.

This is shown by the below formula:

illustration not visible in this excerpt

Where:

[illustration not visible in this excerpt] - value of equity

d - dividends at a certain point in time

pE - one plus cost of equity at a certain point in dime

However, in a finite world it is impossible to forecast dividends to infinity. In practice, there are three different ways to value a firm using dividends. These are shown below.

With a forecast of price at finite horizon T:'

illustration not visible in this excerpt

With a flat perpetuity continuing-value term:

illustration not visible in this excerpt

With a growing perpetuity continuing-value term:

Where:

g - perpetual growth rate

The last equation of the three is what is commonly referred to as the Gordon growth model, which assumes that dividends grow at a constant rate after a certain period of time, forever (Gordon, 1959). A key benefit of the DIV model is that it is simple to implement and can work effectively for companies that have steady growth and a predictable dividend policy. The models benefits, however, are overshadowed by some crucial problems. There are many stocks which do not actually pay any dividends, of all S&P 500 stocks, 122 of them do not pay any dividends (Reese, 2011). Moreover, Modigliani and Miller (1961) highlight dividend policy irrelevancy, their research shows that a dollar of dividend actually displaces a dollar of marketvalue.

2.3.2 Free Cash Flow Valuation (Entity Level)

Free cash flow is the amount of cash that the operating entity of the company pays to, or receives from, the pool of net financial assets of the company (Penman, 2010). Free cash flow can be calculated in a number of ways, the equation below shows its relationship with operating income.

illustration not visible in this excerpt

Where:

illustration not visible in this excerpt

The free cash flow model is one of the most widely used valuation models by analysts and there are clear advantages and disadvantages of using such a model (Demirakos et al., 2004). First, equation 8 shows how the free cash flow model is calculated in a growing perpetuity context, similar to equations 4 and 5 it can also be calculated as a flat perpetuity or over a finite horizon.

illustration not visible in this excerpt

Where:

pF = weighted average cost of capital

Note that the weighted average cost of capital (WACC) is used to discount free cash flow instead of the cost of equity, this is because Formula S gives the firm level FCF value. The WACC is the "company's opportunity cost of funds and represents a blended required return by the company's debt and equity holders" (Koller et al., 2005: 106). A further adjustment is made, based on Modigliani and Miller's (1958) research, to give the value of equity.

illustration not visible in this excerpt

Where:

illustration not visible in this excerpt

From the free cash flow equation one can argue that free cash flow is a "suspect concept" of gauging value added (Penman, 2010) as it treats investment spending as a loss of value, when in fact this can add value in the long run. As such, long forecast horizons are required to recognise the future inflows from investments. This places a high weight on the continuing value term^[5], which is the most speculative component of the valuation process (Penman, 2010).

Koller et al. (2005) examine the discounted cash flow model in the context of cyclical firms. They argue that at first glance the share price of cyclical companies appears too volatile to be consistent with a FCF valuation. They find that although the underlying cash flows are volatile, the final discounted value is not, because no single year's performance has an impact on the value of a company. This is demonstrated more clearly in Figure 2; periods with high cash flows cancel out low ones (Koller et al., 2005: 654). The volatility of cyclical stocks can be explained by the uncertainty surrounding the industry cycle (Koller et al., 2005).

Figure 2 -Figure showing how FCF volatility does not have an effect on DCF value

An illustration from Koller et al. (2000) that shows how volatility in FCF does not have an effect on the DCF valuation.

illustration not visible in this excerpt

2.3.3 Residual Income Valuation Model

The concept of residual income^[6] is not a recent discovery; it can be attributed to papers from Preinreich (1938), Edwards and Bell (1961), and Peasnell (1982). Contemporary research on the residual income valuation model, however, can be attributed to Ohlson (1995). The RIV model estimates value of equity as the book value of equity plus the present value of future residual income. Residual income is defined by Ohlson (1995: 667) as return on the capital invested at the beginning of the period minus a charge for the use of that capital. Equation 9 shows the residual income equation from an equity perspective:

illustration not visible in this excerpt

Where:

The residual income valuation relationship is a forward-looking relationship that links economic value, book value, and expected future residual incomes to firm value. It does not encompass backward-looking accounting numbers and value creation. However, consulting firm Stern Stewart & Co. propose an economic value added (EVA™) approach which does take into account historical performance (O’Hanlon and Peasnell, 2002). This dissertation will not examine EVA™ in any detail, but it is important to note that there are variations to the RIV model.

2.3.3.1 Derivation of ResiduallncomeModel

What links the RIV model and the DIV model is the clean surplus relationship (CSR). If CSR holds, then closing book value equals opening book value plus earnings minus dividends (net of equity issues). The RIV model and the abnormal earnings growth model (AEG) model can both be derived from the DIV model, which is as follows:

illustration not visible in this excerpt

The RIV model and AEG model can be derived by adding the following zero-sum expression:

illustration not visible in this excerpt

Where [illustration not visible in this excerpt]. Alternatively,y could be a series that is expected to end at time T, where yt= 0. Addition of the zero-sum expression to the DIV model yields:

illustration not visible in this excerpt

To derive the RIV model we need to define / in Equation 12 in terms of book value of equity B. The term B0is the current book value and is the models 'anchor'. This gives:

illustration not visible in this excerpt

As mentioned previously, under CSR, earnings can be written as below. Earnings are defined as income in excess of a normal return on the book value of equity capital:

illustration not visible in this excerpt

Residual earnings can thus be written as:

illustration not visible in this excerpt

Substituting equation 14 into equation 12 gives the RIV model at an equity level:

illustration not visible in this excerpt

2.3.3.2 Benefits of the RIV Model

The RIV model has gained popularity with both practitioners and academics in the last decade (Higgins, 2010). Its advantage over the DIV model and FCF model is its links to fundamental accounting information (Higgins, 2010). Previous research has shown how important accounting measurements are to market participants, especially earnings (Burgstahler and Dichev, 1997). Penman (2010) highlights some key benefits of using accounting information. Firstly, the model can benefit from accruals accounting, whereby value is recognised ahead of cash flows (Lui et al., 2002). This avoids investments being treated as value losses, as seen with the FCF model. Shorter forecast horizons can generally be used, as more value is recognised in the immediate future with accruals. This reduces the amount of speculation that is needed and places less weight on the speculative continuing value component. Secondly, the RIV model is geared towards key value drivers, such as profitability of investment and growth in investment. Essentially, residual earnings are the return on common equity expressed as a dollar excess return rather than a ratio.

2.3.3.3 Performance of RIV Model

A number of studies have highlighted the model's ability to explain stock prices. Notably, Frankel and Lee (1998) found that RIV estimates explain more than 70% of cross-sectional variation in stock prices. This is supported by research from Francis et al. (2000), whose comparison of the DIV model, FCF model, and RIV model, show that the RIV model explains stock prices best. This holds true for each sample by Francis et al. (2000)^[7]. Moreover, their research shows that valuation is more accurate for the RIV model, as shown by absolute valuation errors (see Formula 20 for calculation of absolute errors).

Penman and Sougiannis (1998: 365) understandably suggest that accruals in themselves may introduce error. A manager's choice of accounting methods and accrual estimates can affect earnings and book value numbers, thus these numbers can be different for two otherwise identical firms (Palepu et al,. 2008).

However, as long as analysts are aware of these differences then adjustments can be made in their valuation (Palepu et al., 2008). Francis et al. (2000) find no differences in the performance of the RIV model between their low accruals and high accruals sample. Penman and Sogiannis (1998) find that accrual accounting supplies some of the missing value in a cash flow analysis. The managerial discretion of accruals accounting is actually one of the model's key advantages, as economic events are accounted for as they occur.

Conceptually, the RIV, DIV, and FCF models are equivalent (Penman, 1998). They, theoretically, yield identical value estimates under the same payoff assumptions. However, in practice value estimates differ between models because of differences in forecasted discount rates, forecasted growth assumptions, and violations to the CSR relationship (Lundholm and O'Keefe, 2001 a). The theoretical similarities between the models sparked debate over whether it is justified to treat one model as being superior to another under finite forecast horizons. Penman and Sougiannis (1998), Francis et al. (2000), and Corteau et al. (2001) argue that the RIV model is superior to the DIV and FCF model. Lundholm and O'Keefe (2001 a: 315) refute this and state that "any claim of the [RIV] model's superiority over the [FCF] model is mistaken". Penman (2001) responds by arguing that the differences in the models lie with the choice of accounting. "The discounted cash flow model is of the same form as the residual income model; only the substance of the accounting differs" (Penman, 2001: 683). Lundholm and O'Keefe (2001 b) is a response to Penman (2001), where they reaffirm their previous paper (2001 a). Richardson and Tinaikar (2004) bring some conclusion to this debate by confirming that both are correct.

Penman and Sougiannis (1998) are correct in that the FCF model is not complete on its own, and requires accrual information in order to recover the missing piece: expensed operating assets. Lundholm and O'Keefe (2001 a b) are correct in that, if one assumes full pro-forma financial statements are available the missing piece is recovered and the choice of models does not matter.

2.3.3.4 Implementation Issues with the RIVModel

One particular 'shortcoming' of the RIV model is that it needs an accounting system that satisfies the CSR (Lo and Lys, 2000). However, as Lo and Lys (2000) point out, even in cases where the accounting system does not satisfy CSR (e.g. U.S. GAAP), it is possible to restate earnings in terms of comprehensive income, that is, change in the book value of equity minus net capital contributions. Lo and Lys (2000: 342) found that violations of CSR may be "substantial" under US GAAP. The issue of CSR violation was highlighted by Ohlson (2000 and 2005) himself as one of the issues of the RIV model. This has led to the development of the aforementioned AEGM (Ohlson and Juettner-Nauroth, 2005).

[...]

---

^[1] For example, Greece, Spain, Italy, Portugal, and Cyprus (BBC, 2012)

^[2] The Standard and Poor 500 Index is a market value weighted index which tracks the 500 largest companies in the US. The VIX index shows the market's expectation of 30-day volatility; it is a widely used measure of market risk.

^[3] Often referred to as acyclical.

^[4] A pure play company is a company that has, or is very close to having, a single business focus.

^[5] The continuing value term captures all cash flows that happen after time T. It is represented by the last part of the equation in equation S.

^[6] Also known as abnormal earnings and residual earnings. The model itself is also known as the discounted abnormal earnings model and the Edwards-Bell-Ohlson model.

^[7] They compare high and low R&D spend as a percentage of assets, high and low accruals, precision and predictability of attribute, each at a0 percent and 4 percent growth level.