Excerpt

## Table of Contents

List of Abbreviations

List of Symbols

List of Figures

List of Tables

List of Appendices

1 Introduction

2 Analysts’ practise

2.1 The character of financial analysts

2.2 Key input factors of recommendations

3 Review of Bradshaw (2004)

3.1 Valuation models

3.1.1 Present value models

3.1.2 Heuristic models

3.2 Fundamentals of the research

3.2.1 Hypothesis

3.2.2 Overview of the data

3.3 Statistical analyses and results

3.3.1 Valuation models and stock recommendations

3.3.2 Valuation models and future exceed returns

4 Critical review

4.1 Criticism referring to Bradshaw (2004)

4.2 Possible extensions based on Bradshaw (2004)

5 Conclusion

Appendices

References

## List of Abbreviations

illustration not visible in this excerpt

## List of Symbols

illustration not visible in this excerpt

## List of Figures

*Figure 1:* Analysts’ decision process

## List of Tables

*Table 1:* Regression of recommendations on three valuation estimates and LTG and Vuong test for relative explanatory power (n = 15,318)

*Table 2:* Regression of changes in recommendations on changes in three valuation estimates and LTG and Vuong test for relative explanatory power (n = 6,258)

*Table 3:* Regression of one-year-ahead size-decile adjusted returns on three valuation estimates and LTG and Vuong test for relative explanatory power (n = 15,024)

## List of Appendices

*Appendix* 1: Trading development from 1980 till 2010

*Appendix* 2: Correlation matrix for recommendations and valuation models

*Appendix* 3: Bivariate regression of stock recommendations on PEG model relative to stock prices and LTG (n = 15,318)

*Appendix* 4: Mean cost of capital estimates across valuation portfolios

*Appendix* 5: Key discussion in recent literature building on Bradshaw’s (2004) outcomes

*Appendix* 6: Timeline of key discussion in recent literature building on Bradshaw’s (2004) outcomes

## 1 Introduction

Since the beginning of the 90s research on issues referring to analysts’ practise grew rapidly to such an extent that even several publications are concerned with giving an overview of this development.^{1} Besides the principal-agent problematic between the firm’s managers and the equity investors, investors are dependent on analysts’ information in times where equity trading soared and the trading turnover in 2008 was 35 times higher than in 1980.^{2} That is why shareholders are not able to analyse the amount of information regarding a company due to lack of time or ability.^{3} Therefore analysts advise investors to make a profitable decision by publishing a report including for instance stock recommendations or earnings forecasts. Another reason why there is so much research about analysts’ practise is the fact that their information influences investors’ trading behaviour.^{4} Thus, it is crucial to know how reliable those statements are and accordingly to be able to assess the quality of the outputs.^{5} However, to answer the question of analysts’ process of transforming various information of stock recommendations have to be examined in detail.

Recent investigations rather focus on the single properties of analysts’ analyses as earnings forecasts and stock recommendations, but did not connect those two values.^{6} Prior studies deal with research questions like the effect of earnings forecasts on the stock prices or the use of stock recommendations to foretell abnormal return.^{7} Bradshaw (2004) is the first research paper which follows the question whether there is a link and if so how analysts incorporate the earnings forecasts into their stock recommendation.^{8} Because of the importance of Bradshaw (2004), this paper reviews the main issues and embeds them into the existing literature concerning the role of analysts.

The rest of this paper is organized as follows. The first chapter focuses on the character of analysts and potential key input factors which might be used by analysts for issuing recommendations. Then a brief review of Bradshaw (2004) is given to present the main results. This enables a discussion about potential and realized extensions in literature which follows in the third chapter. The final chapter concludes.

## 2 Analysts’ practise

### 2.1 The character of financial analysts

Financial analysts act as information provider for investors who want to evaluate the attractiveness of several securities. One can distinguish between two types of analysts: sell-side and buy-side analysts.^{9} Generally buy-side analysts work for institutional investors who value stocks in their own interest, in particular as basis for their investment decisions. In contrast to this, sell-side analysts work for banks, brokerage or dealer firms. Their analysis is more detailed as the one from buy-side analysts and the final results serve external interested parties. Thus, analysts’ reports from sell-side analysts address a higher number of people which indeed makes their recommendations even more important. Accordingly this paper focuses on sell-side analysts and their technique to produce stock advices.

### 2.2 Key input factors of recommendations

The day-to-day business of sell-side analysts is to obtain as much information as possible concerning a certain security. Therefore they interview different external and internal stakeholders (e.g. suppliers, customers, management, financials, and staff) of a company, they study market and industry circumstances and analyse external historical accounting statements.^{10} These findings are probably key input factors which lead to stock recommendations and forecasts. Although it is possible to observe the single inputs and outputs of analysts’ work, the process of transforming information into the results remains unobservable as a ‘black box’ which is illustrated in figure 1.^{11}

*Figure 1*: Analysts’ decision process^{12}

illustration not visible in this excerpt

A possibility to fill the problematic of the ‘black box’ is linked to indirect evidence by analysing the relationship between inputs and outputs.^{13} While figure 1 shows that forecasts are one of the main results of analysts’ decision process, in the process of generating stock advices they also act as necessary input factors.^{14} This can be explained by a connection of company’s intrinsic value. On the one hand forecasts are utilized to determine the firm’s value because they give a tendency about the future success. On the other hand analysts have to consider the firm’s value and other relevant information as a target price in their decision whether the stock price is under- or overvalued.^{15} Thus, it seems obvious that analysts’ forecasts are associated with stock recommendations through valuation. This is one probable explanation for the relation between analysts’ information and their results.

Before starting to analyse the transformation process it is essential to differentiate between two types of forecasts. They can be either earning- or cash flows-based.^{16} Earnings are based on information of the accounting system as income and expenses. In contrast, cash flows symbolise the firm’s liquidity by deducting cash outflows from cash inflows. In general those values are not equal because on the one hand expenses and income can be non-cash items and on the other hand cash flows do not have to be income relevant.^{17} Depending on the type of forecasts there are suitable valuation measures. Two representative approaches in the category of present value (PV) models are discounted cash flow (DCF) models which consider forecasted cash flows and the residual income (RI) models that incorporate earnings forecasts.^{18}

Despite extensive research on earnings forecasts and a raising demand on cash flow forecasts, the literature shows that analysts’ earnings forecasts are more precise than cash flow forecasts.^{19} The following chapters neglect cash flow forecasts.

## 3 Review of Bradshaw (2004)

In 2004 Bradshaw investigates with his article “How do analysts use their earnings forecasts in generating stock recommendation” a direct relation between analysts’ earnings forecasts and their stock recommendations. The purpose of the study is to fill the unobservable ‘black-box’ while using valuation models as transforming instrument.^{20} Further, Bradshaw observes the correlation between future exceed returns and valuation estimates to compare these results with the association between valuation models and stock recommendations.^{21}

The author refers to different valuation models which can be differentiated into two categories: PV and heuristic models. PV models contain two different approaches of RI models with respect to the terminal value. Heuristic measures include the price-to-earnings-to-growth (PEG) model and the long-term growth (LTG) projection.^{22}

The overall results are first that stock recommendations are associate with heuristic valuation models especially with LTG models and second that future exceed returns are explained by RI models. This would lead to the conclusion that relying on analyst’s stock recommendations which incorporate growth models does not result in high returns, because those are related to RI models.

### 3.1 Valuation models

Before introducing the theoretical background for each valuation examined by Bradshaw (2004), it is interesting to know whether these approaches are accepted by financial analysts in practise. Although PV models are very important in theory, in practise they are just used by 50 percent of the respondents.^{23} A plausible reason could be the uncertainty of variables, like discount rates and assumptions for dividend-paying policy which are needed to calculate PV.^{24} To avoid those complications analysts prefer to look at the business’ LTG potential.^{25} Therefore the research considers the theoretical favoured approach as well as the practical one explained in the next section.

#### 3.1.1 Present value models

One way to estimate the firm’s value is the use of RI models which can be derived from the dividend discount model (DDM).^{26} Instead of using expected dividend payments, as the DDM does, the RI model values the company in terms of accounting information.^{27} More precisely the value equals the current book value and the discounted sum of all future RI.^{28} RI is defined as earnings forecasts minus costs of capital invested.^{29} In conclusion, analysts’ earnings forecasts have an important influence on the firm’s value.^{30} Moreover the time horizon has to be split into a detailed forecasts period and a terminal value due to the practical application of finite planning horizon.^{31} Within the forecast period analysts’ earnings predictions can be used and the book value is foreseen in consideration of the clean surplus relation.^{32} Due to the fact that the terminal value is significantly dependent on the forecasted RI, Bradshaw examines two approaches.^{33} First of all, he assumes that the residual value reverts to zero (RI-1) and second that it equals perpetuity (RI-2).^{34} The mathematical application is demonstrated below:^{35}

Abbildung in dieser Leseprobe nicht enthalten^{36}

In summary equation (1) refers to the theory which has to be adjusted for the practise due to a finite forecast horizon presented in equation (2). Equation (3) shows with a terminal value fading towards zero represented as a regression by using the autoregressive parameter .^{37} A more optimistic suggestion is demonstrated by equation (4) with growing earnings in the long-run which is computed as perpetuity.

#### 3.1.2 Heuristic models

The second method to value a company is the heuristic procedure. Bradshaw distinguishes between PEG models and LTG projections. While PEG models incorporate earnings forecasts, LTG is just a variable concerning the firm’s growth potential that analysts predict. The PEG model consists of the well-known price-to-earnings ratio and LTG projection uses a three- to five-year time horizon.^{38} If a stock is fairly traded, than the PEG ratio will be one. By assuming that equilibrium exists, the value can be calculated as following:^{39}

illustration not visible in this excerpt

### 3.2 Fundamentals of the research

Preliminary information referring to Bradshaw (2004) are given in section 3.2.1 and 3.2.2 to understand the research proceedings in section 3.3.

#### 3.2.1 Hypothesis

The intention of Bradshaw’s research is to examine the direct link between analysts’ earnings forecasts and their stock recommendations by applying valuation models. In particular his general hypothesis is: “Analyst’s recommendations are more (less) favourable for relative higher (lower) earnings-based estimates of value”.^{40} This suggestion implies the more recommendations are based on earnings forecasts, for instance via using valuation models, the more future return can be generated when relying on analysts advices.

#### 3.2.2 Overview of the data

The sample is built up of the First Call data base and a consolidation of Compustat and CRSP data.^{41} In sum there are information concerning consensus earnings forecasts, projections of LTG, stock recommendations, book values, dividend payouts, shares and market prices. The more information the analysts get over the year, the more often they revise their forecasts. Therefore it is necessary to observe monthly-based subsamples controlling for serial correlation.^{42} All in all 46,209 companies are observed from January 1994 till June 1998 that demonstrates 4,421 different firms.^{43}

**[...]**

^{1} Cf. Bradshaw (2011), p. 5 and for a broader overview see Ramnath / Rock / Shane (2008) and Kothari (2001).

^{2} According to appendix 1.

^{3} Cf. Bradshaw (2011), p. 2.

^{4} Cf. Bradshaw (2004), p. 27.

^{5} Cf. Hall / Tacon (2010), p. 19.

^{6} Cf. Bradshaw (2004), p. 29 and Lustgarten / Tang (2008), p. 377.

^{7} Cf. for example Womack (1996), p. 137 and Frankel / Lee (1998), p. 283-284.

^{8} Cf. Hall / Tacon, (2010), p. 19 and Yu / Ke (2009), p. 1.

^{9} Cf. hereinafter Schipper (1991), p. 106.

^{10} Cf. Balog (1991), p. 47 and Womack (1996), p. 138

^{11} Cf. Bradshaw (2011), p. 6.

^{12} Source: According to Bradshaw (2011), p. 51.

^{13} Cf. Bradshaw (2011), p. 6.

^{14} Cf. Schipper (1991), p. 113 and hereinafter Bradshaw (2004), p. 25.

^{15} Cf. Bradshaw (2004), p. 25 and Bradshaw (2002), pp. 27-28: Analysts determine target prices by using valuation models and compare them with the actual stock prices.

^{16} Cf. Block (1999), pp. 88-89 who demonstrates that earnings and cash flows are the most important inputs in analysing stocks.

^{17} Examples are depreciations, provisions or prepaid expenses.

^{18} Present value models are recommended, Cf. Demirakos / Strong / Walker (2004), p. 222.

^{19} Cf. Penman / Sougiannis (1998), p. 343 and Givoly / Hayn / Lehavy (2009), p. 1879 for detailed explanations.

^{20} According to section 2.2.

^{21} The comparison and the single results are discussed in detail in section 3.3.

^{22} Cf. hereinafter Bradshaw (2004), p. 26.

^{23} Cf. Block (1999), pp. 87, 92.

^{24} Cf. Bradshaw (2004), p. 27 or Block (1999), p. 92.

^{25} Cf. Block (1999), p. 89: Financial analysts rank ‘growth potential’ as No. 1 for determining the price-to-earnings ratio.

^{26} Cf. Bradshaw (2004), p. 30: The DDM is based on the idea that the firm’s stock price equals the discounted sum of all expected dividend payments.

^{27} Cf. Frankel / Lee (1998), p. 286.

^{28} According to formula (1).

^{29} Cf. Bradshaw (2004), p. 31.

^{30} Cf. Ohlson (1995), p. 662 with the explanatory statement of Frankel / Lee (1998), p. 286: The company generates value if the future earnings exceed the costs of capital invested. Otherwise its value equals its current book value.

^{31} Cf. Bradshaw (2004), p. 31: He applies a five-year horizon because this is consistent with the analysts’ long-term earnings predictions. Compare equation (1) and (2).

^{32} Cf. Frankel / Lee (1998), p. 286 and Bradshaw (2004), p. 31; Clean surplus relation can be represented by the following equation: where = dividend per share.

^{33} Cf. Lundholm /O’Keefe (2001), p. 321 and Frankel / Lee (1998), p. 289.

^{34} It depends on the belief of what will happen after the forecasted period. Either the company is able to grow on a constant rate or the income will tend to zero because of competition at the market; Cf. Bradshaw (2004), pp. 31-32 and Penman (1991), p. 237.

^{35} All numbers are regarded on a per share basis; Cf. Bradshaw (2004), pp. 30-33 for detailed derivation of the final formulas in equation (1) till (4).

^{36} Based on a three-factor industry-specific discount rate as represented by Fama / French (1996), p. 55.

^{37} Cf. Bradshaw (2004), p. 32 for further explanations.

^{38} Cf. Bradshaw (2002), p. 29 and Bradshaw (2004), p. 33.

^{39} Solving the equation for and using a two-year ahead earnings forecast because the results of Bradshaw’s investigation are more significant; Cf. Bradshaw (2004), p. 34.

^{40} Bradshaw (2004), p. 30.

^{41} First Call data base recognises the change in the consensus at real-time; Cf. Bradshaw (2004), p. 34.

^{42} Cf. Wooldridge (2009), p. 845: Serial correlation denotes the correlation between error terms in different time periods. Without a monthly based sample the consensus forecasts would consist of many analysts’ forecasts that are related to each other because the predictions are based on the same information.

^{43} Cf. Bradshaw (2004), p. 34.

- Quote paper
- Master of Science Malwina Woznik (Author), 2011, Discussion and review of Bradshaw (2004): "How do analysts use their earnings forecasts in generating stock recommendations", Munich, GRIN Verlag, https://www.grin.com/document/231304

Comments