validade dos modelos us ado singles

Upload: monica-cruz

Post on 05-Apr-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    1/15

    ACCOUNTING HORIZONSVol. 18, No. 4

    December 2004pp. 221-240

    What Valuation Models Do Analysts Use?

    Efthimios G Demirakos, Norman C. Strong, and Martin Walker

    SYNOPSIS: This paper adopts a structured positive approach to explaining the valuationpractices of financial analysts by studying the valuation methodologies contained in 104

    analysts' reports from international investment banks for 26 large U.K.-listed companies

    drawn from the beverages, electronics, and Pharmaceuticals sectors. We provide a

    descriptive analysis of the use of altenative valuation models focusing on the value-

    relevant attributes that analysts seek to forecast and the methodologies analysts use to

    convert the forecasts into estimates of firm value. We postulate and test a number of

    hypotheses relating to how the valuation practices of analysts vary systematically across

    industrial sectors. We find that: (1) the use of valuation by comparatives is higher in the

    beverages sector than in electronics or Pharmaceuticals; (2) analysts typically chooseeither a PE model or an explicit multiperiod DCF valuation model as their dominant

    valuation model; (3) none of the analysts use the price to cash flow as their dominant

    valuation model; and (4) contrary to our expectations, some analysts who construct

    explicit multiperiod valuation models still adopt a comparative valuation model as their

    preferred model. We believe the study's findings are important for increasing our

    understanding of the valuation practices of financial analysts. The study also provides a

    basis for further research that tests a richer and more detailed set of hypotheses.

    Data Availability: A list identifying the sampled analysts' reports is available from the

    authors. The reports themselves are publicly available from the Investext database.

    INTRODUCTION

    valuation theorists have studied the theoretical properties of several valuation

    frameworks, and some authors use these theoretical properties to produce normative

    arguments in favor of particular frameworks. Penman (2001) advocates residual income

    valuation (RIV), in preference to discounted cash flow (DCF). Copeland et al. (2000)

    recommend using either the DCF model or the RIV model.' These authors assert that

    DCF is most widely used in practice, but that RIV is gaining in popularity. They also note

    that both methods, properly applied, result in the same

    Efthimios G Demirakos is a Lecturer at Lancaster University; Norman C. Strong and

    Martin Walker are Professors at the University of Manchester.

    The paper has benefited substantially from the comments of two anonymous referees and

    from the comments and advice of the Associate Editor, Stephen Baginski. The authors

    acknowledge the comments of Miles Gietzman, Ken Peasnell, Peter

    Pope, Theodore Sougiannis, and participants at the Spanish Joint Accounting and Finance

    2003 Conference. Professor Demirakos acknowledges a WUN scholarship and a Ph.D.

    grant from the Propondis Foundation. Copeland et al. (2000, 131) refer to these as the

    enterprise DCF model and the economic profit model.

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    2/15

    valuation, and suggest that "the choice is mostly driven by the instincts of the user."

    Palepu et al. (2000) adopt a balanced position; they note that properly constructed RIV

    and DCF models lead to identical valuations, but acknowledge that the preference for one

    approach over the other may depend on the ease of access to acceptable proxies for the

    model constructs.

    Penman (2001), Copeland et al. (2000), and Palepu et al. (2000) all prefer explicit

    multiperiod valuation models based on either discounted cash flows or discounted

    residual income rather than valuations based on single-period comparatives. The

    theoretical superiority of multiperiod valuation models stands in contrast to the evidence

    on valuation models used in practice. Barker (1999), for example, reviews and

    summarizes previous research on the valuation models used by professional investors or

    financial analysts. The most consistent findings are, first "that the [price-earnings ratio]

    is of primary importance," and second "that [DCF] models, technical analysis, and beta

    analysis are of little practical importance to investment decisions" (Barker 1999,197). In

    his own survey of U.K. analysts and fund managers. Barker (1999) fmds that both groups

    rank the PE model and the dividend yield model as the most important, and both groups

    rate the DCF and dividend discount models as unimportant. Barker's findings on the

    importance of PE multiples support thresults of Arnold and Moizer (1984) and Moizer

    and Arnold (1984) for the U.K., Pike et al. (1993) for Germany and the U.K., and Block

    (1999) for the U.S., all of whom investigate the valuation models used by analysts using

    survey-based approaches.

    Barker (1999) and the prior research he reviews are based on interviews and questionnaire

    surveys of investment analysts and fund managers. To overcome some of the subjectivity

    problems associated with interview-based research,^ we adopt an altemative researchdesign based on a content analysis of analysts' equity research reports. This approach

    complements the results of theabove studies with evidence on the equity valuation models

    that analysts use in practice. Govindarajan (1980), Previts et al. (1994), and Rogers and

    Grant (1997) for the U.S., and Breton and Taffler (2001) for the U.K. also employ content

    analysis, but they focus on fmaneial disclosure issues and the general information needs

    of analysts. We differ from these studies principally in our explicit focus on the specific

    valuation models analysts use, along with any other models or frames of reference they

    bring to the specific task of valuation.

    Bradshaw (2002) studies the content of 103 U.S. analysts' reports to identify how analysts

    justify their stock price recommendations. He finds that valuations based on PE multiples

    and expected growth are more likely to be used to support favorable recommendations,

    while qualitative analysis of a firm's fundamentals is more likely to be employed to justify

    less favorable recommendations.

    He recommends further research to compare analysts' reports both within and across

    industries.

    Our study complements and extends Bradshaw (2002). First, we provide more detail

    about the particular valuation models analysts use.^ Second, and crucially, we advance

    and test specific hypotheses about the valuation model choices of analysts. In particular

    we test hypotheses about how valuation methodologies vary across industrial sectors.

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    3/15

    The next section explains our methodology and theoretical framework, and it explains the

    hypotheses that we use to guide our evaluation of analysts' reports. We then describe our

    data source, the principles we use in selecting analysts' reports for inclusion in the study,

    and the criteria we apply in scoring the reports in our sample. The next section reports our

    findings on the frequency of use of alternative valuation models, presents formal tests of

    our hypotheses, and offers further sensitivity analyses of our empirical results. Two

    appendices briefly discuss the properties of two new practical valuation approaches

    uncovered in our study of financial analysts' valuation practices.

    METHODOLOGY AND THEORETICAL FRAMEWORK

    A Structured Content Analysis

    We draw on standard discussions of valuation concepts and models to establish a

    structured framework for recording the content of analysts' reports and for generating

    testable hypotheses about how the content of reports varies according to the nature of the

    company analyzed. The conceptual framework of this paper is influenced primarily by

    Penman (2001), with additional insights drawn from Palepu et al. (2000) and Copeland et

    al. (2000). Penman (2001, 11) introduces a five-step process of fundamental analysis.

    1) Knowing the business (strategic analysis).

    2) Analyzing information (accounting and nonaccounting information analysis).

    3) Specifying, measuring, and forecasting value-relevant payoffs.

    4) Converting forecasts to a valuation.

    5) Trading on the valuation.

    The main focus of this paper is on the value-relevant attributes that analysts seek to

    forecast and the methodology analysts use to convert their forecasts into firm value, i.e.,steps 3 and 4 of Penman's process.

    Initially we anticipated finding three main types of valuation analysis: some form of

    single period comparative or benchmark valuation (such as PE multiples), valuation via a

    finite horizon multiperiod DCF model, and valuation via a finite horizon multiperiod RJV

    model. The latter two models figure prominently in Palepu et al. (2000). Penman (2001)

    focuses most attention on implementing the RIV model, while Copeland et al. (2000)

    focus on the DCF model. All three sources mention the widespread practical use of

    single-period comparative valuation techniques, although they all view such techniques as

    low-cost simplifications that are likely to lead to less accurate valuations than a full

    implementation of either the DCF or the RIV models. Penman (2001) is

    particularly scathing of valuation by PE ratios. In our empirical work, we find evidence of

    all three forms of valuation model. Another form of valuation practice we expected to

    encounter was an attempt to use option pricing models to value future growth

    opportunities independently of the valuation of assets in place. However, we encounter

    only limited use of these approaches.

    During the course of our empirical work, several valuation methods and themes emerge to

    a more significant extent than anticipated. As detailed below, various analysts provide

    valuation arguments based on hybrid value creation indicators. These are not complete

    valuation models but are deployed as partial analytical support for a valuation "case."Industry Sectors

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    4/15

    Our study focuses on three industries: beverages, electronics, and Pharmaceuticals

    chosen intentionally to give potential variation in analysts' valuation practices. To confirm

    the differences in industry fundamentals and as background to our empirical hypotheses.

    Table 1 reports growth and volatility characteristics for our three sectors. For each sector,

    we report the median and interquartile range for annualized sales growth, the volatility of

    earnings changes, the ratio of R&D to Sales, and the ratio of market to book value of

    equity over the period 1997-2001. Comparing beverages with

    Pharmaceuticals shows that the former has lower and more stable growth, lower volatility

    of earnings changes, much lower R&D to Sales, and lower and more uniform market to

    book ratios. On most of the indicators, electronics lies between beverages and

    Pharmaceuticals.

    Empirical Hypotheses

    Our paper adopts a structured positive approach to the study of valuation practices. This

    requires us to postulate hypotheses that we can test through this methodology. For this

    paper, we develop four specific hypotheses related to the value-relevant attributes that

    analysts forecast and the methodologies they use to convert their forecasts into firm value.

    TABELA

    Our first hypothesis concerns the choice between valuation based on industry or sector

    single period comparatives and valuation using explicit multiperiod valuation models.

    The prior research discussed above finds that valuation by PE comparatives is the most

    pervasive form of valuation model. This approach could yield a good first approximation

    for industries that have:

    fairly uniform and stable growth; costs of capital, accounting methods and capital structures that are comparable across

    companies;

    and

    transitory earnings items that can be identified and excluded from the analysis.

    In such instances, the simplicity of the PE approach may be attractive to analysts.* A

    similar argument applies to other methods of valuation by single-period comparatives that

    we identify in our study (for example, those based on single-period amounts of sales, cash

    flow, or book values).

    Given our results in Table 1, beverages is a sector characterized by fairly uniform and

    stable growth where valuation by single-period comparatives might yield a reasonable

    first approximation, while electronics and pharmaceutical are sectors for which the ideal

    conditions for valuation by comparatives are much less likely to hold. This gives rise to

    the following hypothesis.

    HI: Use of valuation by single-period comparatives is higher in the beverages sector

    than in electronics or Pharmaceuticals.

    Second, we are interested in why analysts choose accruals-based accounting valuation

    constructs over cash-fiow-based constructs. Note that this choice operates within both the

    single-period comparative valuation and the multiperiod valuation sets of models. We

    first consider the choice between cash-flow-based and residual-income-based multiperiod

    models.

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    5/15

    Multiperiod valuation models can be expressed in terms of projected cash flows, in the

    case of DCF models, or in terms of initial book value and projected abnormal earnings in

    the case of the RIV model. Theoretically the two techniques are equivalent. Since both

    techniques require forecasts of future amountseither residual earnings for RIV or cash

    fiows for DCFease of use does not seem to favor one method over the other.

    In making the choice between RIV and DCF, analysts will consider which technique theyare most familiar with, which technique their clients are most comfortable with, and the

    extent to which the published accounting information of the firm can be used as a credible

    starting point for the analysis. DCF models have a long history of use relative to RIV and

    have been taught extensively in Finance courses. Thus, DCF is in essence a default. For

    analysts to pick RIV over DCF, they have to be confident that the published accounting

    information captures the essence of the business.The question to consider is why and how

    the confidence of analysts in the ability of accounting to faithfully represent the value

    generation processes ofthe business might vary across sectors.

    A number of authors, notably Lev (2001), have pointed out that accounting is relatively

    strong in valuing tangible assets and relatively weak in valuing intangible assets. We

    therefore expect the choice between DCF and RIV to reflect the nature of the firm'sassets. In particular we expect accounting measures of performance to be less relevant for

    intangibles-rich firms or for firms with large portfolios of growth opportunities.

    Consistent with the results reported in Table 1 we characterize pharmaceuticals as falling

    most clearly into this category, with beverages being at the other extreme and electronics

    falling between these two cases.

    These arguments lead to our second hypothesis:

    H2: Use of multiperiod DCF models relative to multiperiod RIV models is higher in

    the pharmaceuticals sector than in the beverages sector, with the electronics sector

    falling between the two extremes.*

    Copeland et al. (2000) note that an important conceptual advantage ofthe RIV model is

    that it focuses on whether the company is generating a return in excess of its cost of

    capital.^ This in turn suggests that the use of RIV and other hybrid accrual models is

    likely to be greatest in sectors where accounting-based measures of profitability are a

    relatively more reliable indicator of economic profitability. As before, we expect

    reliability to be lower for companies with higher amounts of intangibles. This argument

    suggests a more general version of H2:

    H2': Use of multiperiod and hybrid cash fiow models relative to the use of

    multiperiod and hybrid accrual models is higher in sectors with relatively high

    proportions of intangible assets.

    We now consider accrual versus cash fiow in the single-period setting. Penman (2001,

    117) notes that "Free cash fiow does not measure value added in the short mn; valuegained is not matched with value given up." In other words, firm valuation based on a

    multiple of a single year's free cash flow is not sensible because free cash fiow is not a

    defensible proxy for value. This gives our third

    hypothesis:

    H3: Given the limitations of single-period cash fiow as a measure of value

    generation, no analyst will use it as their dominant approach.

    Finally we hypothesize that analysts view valuation by comparatives as a form of

    simplified valuation analysis. If they incur the cost to produce a full-blown multiperiod

    model, they present it as their dominant model. This results in our final hypothesis:

    H4: Analysts who construct a multiperiod valuation analysis of either type do not

    adopt valuation by comparatives as their dominant model.

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    6/15

    DATA, SAMPLE SELECTION, AND SCORING CONVENTION

    This section reports our data source, our sample selection criteria, and the scoring

    convention we adopt when analyzing the contents of analysts' reports.

    Data

    We download the analysts' reports from Investext Plus. Investext Plus is a database of

    reports and forecasts by top Wall Street and intemational investment firms and analysts.

    The service covers over 11,000 U.S. and intemational companies from 53 industries. The

    current study examines reports by intemational investment firms regarding U.K.

    companies.

    Sample Selection

    The reports selected for the study are from the period January 1997 to October 2001 and

    consist of reports exceeding 15 pages in length. They cover listed companies in the

    London Stock Exchange's beverages, electronics and electrical equipment (which we

    abbreviate to electronics), and pharmaceuticals sectors. All companies are constituents of

    the FTSE All-Share Index. Where an analyst publishes more than one report for a

    particular company in the same year, we select the largest report.

    From all the reports satisfying these conditions, we select a manageable final sample of

    104 sellside analysts' equity research reports covering 26 companies.* We have between

    32 and 38 reports per sector. Table 2 reports summary statistics for the sample of

    companies and reports. Of the 26 companies in the sample, 9 are among the 100 largest

    U.K.-quoted companies. The 4 (12, 10) beverage (electronics, pharmaceutical) companies

    we study represent 57.1 (57.1,43.5) percent of the beverage (electronics, pharmaceutical)

    companies appearing in the FTSE All-Share index.

    The selection of reports for analysis reflects our focus on the most detailed researchreports. The Investext Plus database contains a wide range of report lengths, from just a

    few pages to over 100 pages. The very short reports contain little by way of analysis and

    often focus on the implications of a particular event or update a previous eamings

    forecast. We focus on comprehensive equity research reports of over 15 pages. Table 2

    shows that the length of our sampled reports ranges from 15 to 176 pages with the median

    length being around 28 pages for all three sectors. This characteristic of our data selection

    permits us to extend and complement the analysis of previous studies that

    mainly analyze the content of reports containing only a few pages.' We choose reports

    from multiple investment houses so that a particular house does not dominate the results.

    But we also require each investment house to be included in more than one sector because

    differences in valuation methodologies across sectors should reflect genuine difl Ferences

    between the sectors, not differences in which investment houses cover each sector.

    Scoring Convention

    Table 3 describes the valuation perspectives and modelsand their definitionsfound in

    our analysis of the valuation content of sell-side analysts' equity research reports. It shows

    that analysts employ single-period comparative valuation models, hybrid valuation

    models, and multiperiod valuation models. We classify the various price or value

    multiples as single-period comparative valuation models, various value creation

    indicators along with real option valuation techniques as hybrid

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    7/15

    TABELA

    TABELA

    valuation models, and the explicit discounted cash flow and residual income valuation

    models as multiperiod valuation models. Table 3 provides a short definition of each

    valuation model. We give the report a score of 1 for each valuation model in Table 3 only

    if the analyst uses and discusses that particular valuation model in the main text of the

    report. Any tables contained in analysts' reports are only analyzed when analysts refer to

    the content in their narrative. This scoring convention assumes that only the arguments

    presented in the narrative are value relevant and useful.'"

    We classify a valuation model as dominant if it is most closely associated with the

    analyst's own stock price recommendation. Where a report uses only one model, we score

    this valuation model as dominant. Where a report uses more than one model, we first

    check the valuation section of the report to see if it reveals the analyst's preference. We

    also examine the first page of the report or the executive summary to see which valuation

    model is highlighted. If these initial assessments yield no clear view, we calculate the

    differences between the analyst's alternative value estimates and the analyst's final target

    price. We select the dominant model as the one closest to the target price.

    However, some reports do not have a target price and some valuation techniques do not

    produce a specific value estimate. In the rare cases where we are unable to determine the

    dominant valuation model using the above criteria, we use the amount of space in the

    report devoted to the analysis of each model to select the dominant model. In our formal

    tests, we assign a score of 1 when a model is used as the sole dominant valuation model,

    and we assign a score of 0.5 to a valuation model when it is used jointly with anothermodel as the dominant valuation model.

    MAIN FINDINGS

    This section reports the main findings of our empirical analysis. We begin by describing

    the frequency of use of the various types of valuation methodologies we encounter. We

    then test our hypotheses. Finally, we report a number of sensitivity analyses.

    Descriptive Analysis

    Table 4 presents descriptive evidence on the range of valuation models analysts employ.

    This table serves as a comparison with prior work. The table shows that almost all the

    sampled reports contain some form of valuation by reference to a multiple of earnings. In

    the electronics sector, only four reports (out of thirty-four) contain no valuation by

    reference to earnings. Two of these reports focus on DCF models, one on a hybrid model,

    and one on a multiple of sales. In the pharmaceuticals sector, eight reports (out of thirty-

    eight) contain no valuation by reference to earnings. Of these eight cases, five rely on

    some form of cash flow valuation analysis, one combines price to sales with DCF,another

    combines price to sales with some form of option-pricing analysis, and one focuses on

    option-pricing analysis. These results suggest that using earnings as a basis of valuation is

    the prevalent form of analysis for the beverages sector, but that it is less prevalent for the

    pharmaceuticals sector, with electronics falling between.

    As in previous studies such as Barker (1999) and Bradshaw (2002), we find widespreaduse of PE models, but we also fmd that the attention given to PE models varies

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    8/15

    systematically across sectors in understandable ways. In contrast to prior studies, we find

    considerable use of explicit multiperiod DCF models.

    Tests of Empirical Hypotheses

    Hypothesis 1

    Hypothesis 1 states that valuation by single-period comparatives is greater in the

    beverages sector than in electronics and pharmaceuticals due to differences in the growth

    characteristics of the three sectors. Table 5 presents the results of a formal test of this

    hypothesis. Panel A shows that

    TABELA

    TABELA

    with HI, it is not surprising given the heavy use of comparative valuation techniques.

    Financial analysts probably feel a need to use comparative valuation techniques as a

    starting point even if they are not their preferred valuation choice.

    A more powerful test of the difference between the stable and high-growth sectors in the

    prevalence of valuation by single-period comparatives compared with multiperiod

    valuation models is based on dominant valuation models. Panel B of Table 5 reports the

    results of this test. In beverages, valuation by comparatives is dominant in 25.5 reports,

    while multiperiod models are dominant in only 3 reports. In electronics and

    pharmaceuticals, the corresponding numbers are 41 and 19 reports. A Chi-square test of

    the relative proportions of reports in which single-period and multiperiod valuation

    models are dominant reveals a significant difference between stable (beverages) and high-growth sectors (electronics and pharmaceuticals) at the 5 percent level (x^ = 4.62;

    p-value = 0.032). This result is consistent with Hl.'^ An altemative approach to testing

    this hypothesis is to see how many of the reports that use a multiperiod valuation model

    choose this as their dominant valuation model. Table 5, Panel C shows that, in beverages,

    13 reports use multiperiod valuation models, but they are dominant in only 3 (23.1

    percent) of these reports. Corresponding figures for pharmaceuticals are multiperiod

    valuation models dominant in 9 out of 13 reports (69.2 percent), and for electronics

    multiperiod valuation models dominant in 10 out of 14 reports (71.4 percent). Consistent

    with HI, we find a significantly greater use of multiperiod models as the dominant model

    in pharmaceuticals and electronics than in beverages

    (,x^ = 7 9; p-value = 0.005).'"

    Hypothesis 2

    Hypothesis 2 suggests that use of the DCF model relative to the RIV model should be

    higher in the pharmaceuticals sector than the beverages sector due to the difference in the

    extent to which financial statements properly capture the value of a firm's tangible and

    intangible assets. Table 4 shows that only one report in beverages, one report in

    electronics, and no reports in pharmaceuticals use RIV. Interestingly, both instances of

    RIV are implemented jointly with DCF and produce the same valuation estimates as the

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    9/15

    DCF models. Perhaps analysts who use RIV perceive a need to back up this "new"

    approach with the more widely accepted DCF. RIV is never employed as the sole

    dominant valuation model, and it is used jointly with DCF as the dominant valuation

    methodology in only one report (in electronics). This empirical evidence clearly shows

    that financial analysts prefer DCF to RIV. While finding one instance of RIV for both

    beverages and electronics and no instances of RIV for pharmaceuticals is directionally

    consistent with H2, a formal Chi-square test lacks power to reject the null hypothesis.

    Hypothesis H2' broadens the issue of choice between RIV and DCF, to the more general

    issue of the use of multiperiod and hybrid cash flow versus multiperiod and hybrid

    accrual models. Table 5, Panel D reports our test of this more general hypothesis. We find

    that the use of RIV and other hybrid accrual models for valuation purposes varies

    markedly across sectors. We define hybrid accrual models as comprising accounting rate

    of return (ARR) and Economic Value Added (EVA^"^) while hybrid cash flow models

    include cash recovery rate (CRR). Consistent with H2', the beverage sector

    analysts use hybrid and multiperiod accrual models more often than analysts covering the

    pharmaceuticals sector (statistically significant at the 1 percent level), while the use of

    cash flow models is constant across sectors.

    Insight into the combined choice of forecasting system and valuation model can be

    achieved by considering the extent to which the use of accounting profitability analysis

    varies across sectors.

    Financial statement analysis textbooks teach students to focus on firm profitability ratios,

    typically represented by return on net operating assets, and then introduce the

    disaggregation of this ratio into a multiple of the operating profit margin and the ratio of

    sales to net operating assets (see, e.g.. Penman 2001,354). Fairfield and Yohn (2001)show that disaggregating the change in return on net operating assets can improve

    forecasts of future profitability. On the other hand, we know that accounting profitability

    ratios are problematic in high-growth, intangibles-rich industries due to the

    accounting treatment of R&D expenditures and intangible assets. Hence we expect

    profitability analysis to be more prominent in an industry such as beverages where value

    comes mainly from assets in place than in an industry such as pharmaceuticals where a

    large component of value comes from growth opportunities.

    Table 6 reports our examination of the profitability analyses contained in analysts' reports.

    Table 6 shows, as expected, that there are differences in the importance of profitability

    analysis across the three sectors. An analysis of accounting profitability is more

    prominent in beverages and electronics than in pharmaceuticals. Analysts employ a ratio-

    based profitability decomposition more frequently in beverages and electronics. But, even

    in these sectors, the decomposition is rudimentary, with the

    depth of analysis restricted to a consideration of ARR and of profit margins for some

    individual products. In pharmaceuticals, analysts devote little space to accounting and

    financial analysis. Instead, an analysis of strategic issues and of R&D projects is the

    critical part of the valuation process.

    Except for pharmaceuticals, analyzing accounting profitability is clearly more popular

    than analyzing cash fiows. For example, in beverages, the ARR is used in 23 out of 32reports compared with 1 report that uses the CRR. Corresponding figures in electronics

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    10/15

    are ARR in 20 out of 34 reports and CRR in 3 reports. In most cases, the analysts

    compare the ARR with the cost of capital. In contrast to beverages and electronics, both

    the ARR and the CRR are each referred to in only one pharmaceutical report.

    From Table 6, the proportion of analyst reports containing a profitability analysis based

    on profit margins is significantly lower (at the 1 percent level) in pharmaceuticals (57.9

    percent) than in either beverages (100.0 percent) or electronics (91.2 percent). Similarly,

    the proportion of analyst reports containing a profitability analysis based on ARRs is

    significantly lower (at the 1 percent level) in pharmaceuticals (2.6 percent) than in either

    beverages (71.9 percent) or electronics (58.8 percent).

    TABELA

    Hypothesis 3

    Table 7 lists the types of models that we classify as being dominant in each report. Table

    7 shows that PE multiples are the dominant valuation model in 68.8 percent ofN reports

    in the beverages sector, in 39.7 percent of reports in the electronics sector, and in 52.6

    percent of reports in pharmaceuticals.

    Across all sectors, a PE model is the dominant model in 55.5 (53.4 percent) ofthe reports,

    and a multiperiod DCF models is the dominant approach in 21.5 (20.7 percent) of cases.

    Consistent with H3, we fmd no report that uses a single-period cash flow multiple as its

    dominant valuation methodology.

    This result is important in relation to public debates and pronouncements on valuation

    issues.

    One often comes across the slogan "cash is king," but we fmd no evidence to support this

    view in our data. Nevertheless, we do fmd several instances of price to cash flow being

    used as a sensitivity analysis. Moreover we fmd greater use of price to cash flow as a

    sensitivity check in beverages than in the other two sectors. The evidence reported inTable 4 shows that a cash flow multiple is used in 11 out of 32 reports in the beverages

    sector and only twice out of 72 reports in the other two sectors.

    This difference is statistically significant at the 1 percent level. These fmdings appear to

    conflict with H3. However, in sectors of relatively low growth and reasonably stable and

    predictable levels of capital investment, operating cash flows provide a viable altemative

    estimate of sustainable earnings that is not prone to manipulation via discretionary

    accounting accruals. Our data are thus consistent with analysts valuing the firm based

    primarily on an estimate of core earnings, but they support this valuation with a

    sensitivity check based on a multiple of operating cash flows.

    Hypothesis 4

    We expect analysts who produce a complete multiperiod valuation model to identify it astheir dominant model. A total of 40 reports produce a valuation using a multiperiod

    model. Of the 40 reports that implement a multiperiod valuation model. Table 7 shows

    that 20 reports identify a multiperiod model as their sole dominant model, while 4 other

    reports use DCF along with other models. In 4 cases out of 40, the dominant model is not

    clear from the text. This leaves 12 cases out of 40 (30 percent) where the analysts prefer

    some other form of valuation. Of these 12 cases, 10 reports favor valuation based on PE

    multiples, and 2 cases are based on the technology value'^ of the company (both

    pharmaceuticals firms). These 12 cases are inconsistent with H4. Perhaps these analysts

    believe users of their reports prefer to focus on valuation by comparatives.

    TABELA

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    11/15

    SENSITIVITY ANALYSES

    Length of Sampled Equity Research Reports

    Our interest in comprehensive equity research reports that offer a detailed analysis of the

    firm differs from previous studies that mainly base their results on the content analysis of

    short reports of only a few pages in length.'* While prior work and our results report a

    preference for valuation by single-period comparatives, the longer reports in our sample

    contain more examples of sophisticated valuations than in the prior evidence. To examine

    the effect of excluding reports of less than 15 pages, we analyze a smaller sample of 22

    short reports (with average length 8.1 pages per report) for 15 of the firms in our initial

    sample. The reports are written by analysts from the same investment houses as in our

    initial sample. Seven (nearly one third) of these reports do not highlight any valuation

    model in their main text. Two reports base their recommendation on the DCF model,

    while the remaining thirteen use some form of single-period comparative valuation to

    generate a stock recommendation. This limited analysis of shorter reports suggests a

    greater use of valuation by comparatives than in longer reports, although multiperiod

    DCF models are still used. Prior results based on short reports may understate the

    sophistication of the analysts.

    Firms Reporting Losses

    Differences in the use of DCF multiperiod models across sectors could be due to variation

    in the incidence of reported losses. The presence of losses may force analysts to base their

    valuations on something other than PE multiples. We find that half of the sampled

    pharmaceutical reports refer to firms reporting losses, raising the possibility that our

    results could be due to the incidence of losses rather than the high-tech nature of the

    industry. In order to discriminate between these two possibilities, we examine anotherhigh-tech sector that does not have a high incidence of losses during our

    sample period. Specifically, we examine 28 reports for 10 firms in the information

    technology hardware sector.'^ Of the 28 reports, 24 refer to profitable firms. We find:

    1) Sixteen (57.1 percent) of the IT hardware reports use DCF, and a multiperiod DCF

    model is the dominant model in 11.5 (41.1 percent) reports.

    2) There are 40 instances of single-period comparative valuation techniques and 18

    instances of multiperiod models. The greater use of multiperiod valuation models relative

    to singleperiod models in information technology hardware compared to beverages is

    significant

    (X^ = 4.52; p-value = 0.034).

    3) Single-period comparative valuation models are the dominant choice in 14 IT hardware

    reports, with multiperiod valuation models dominant in 11.5 reports. The greater use of

    single-period comparatives as the dominant model in beverages compared to IT hardware

    is statistically significant (x^ = 819; p-value = 0.004).

    In the light of these additional findings, we conclude that the use of multiperiod valuation

    models and DCF models in particular is greater in high-growth sectors irrespective of the

    incidence of reported losses.

    Brokerage Firms' House Styles

    Bradshaw (2002, 40) suggests that "It would be interesting to examine the extent to whichanalysts' reports systematically differ across brokerage houses ..." As mentioned earlier in

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    12/15

    the descriptive analysis section, ahnost all the equity research reports include some form

    of single- period comparative valuation analysis. However, investment houses might

    differ in their preferences for DCF and accounting-based economic profitability models.

    Panel A of Table 8 reports the frequency of employing DCF analysis at each house;

    Dresdner Kleinwort Wasserstein (69.2 percent), Credit Suisse First Boston (68.4 percent),

    and HSBC (45.5 percent) use DCF the most.'* Table 8,

    Panel B offers a sell-side analysts' ranking based on the use of accounting-based

    economic profitability models (rating to economic profit, accounting rates of return,

    economic value added, and residual income valuation model). HSBC uses some form of

    economic profitability analysis for valuation purposes in 72.7 percent of its reports,

    followed by Merrill Lynch (42.9 percent). Credit Suisse First Boston (36.8 percent), and

    UBS Warburg (36.4 percent).

    TABELA

    lype ofRecommendation

    In choosing the sample, we ignored the investment recommendation given in the report.

    However, when carrying out the analysis, we record the recommendations made by the

    analysts. Twelve types of recommendation appear in the reports. Six types of

    recommendation (Strong Buy, Buy, Accumulate, Market Outperform, Add, Undervalued)

    are positive. The 104 reports contain 64 positive recommendations. The categories

    Market Perform, Neutral, and Hold are neutral recommendations.

    The sample contains 25 of these. Only 15 reports contain one of the remaining three

    negative categories (Market Underperform, Reduce, Sell). Counting the neutralrecommendations as weak negatives, the sample contains 64 positives and 40 negatives.

    A standard binomial test rejects the hypothesis of an equal number of positive and

    negative recommendations at the 1 percent level.

    These findings are consistent with previous work indicating a tendency for analysts'

    recommendations to be biased toward a buy.

    Looking at the individual sectors, we find that the proportion of positive

    recommendations is 46.9 percent in beverages, 67.6 percent in electronics, and 68.4

    percent in pharmaceuticals. These differences could drive the choice of valuation model.

    We therefore examine whether the choice of DCF as a valuation model varies

    significantly across types of recommendation. Of the 64 (40) reports that have a positive

    (neutral/negative) recommendation, 27 (13) use DCF analysis i.e., 42.2

    percent (32.5 percent) of the reports. A Chi-square test reveals that this difference is not

    significant i;^^ = 0.98; p-value = 0.323). For reports containing a DCF valuation, the

    DCF valuation is dominant in 15 out of 27 (55.6 percent) for the positive recommendation

    cases, and 6.5 out of 13 (50 percent) for the negative recommendation cases. This

    difference in the proportion of reports for which DCF is the dominant valuation model

    across different types of recommendations is not significant (^^ = 0.11; p-value = 0.74).

    These results suggest that the type of recommendation does not drive the choice of

    valuation model.SUMMARY

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    13/15

    The main message to emerge from this content analysis of financial analysts' reports is

    that analysts appear to tailor their valuation methodologies to the circumstances of the

    industry. PE models remain the mainstay of valuation practice, but other forms of analysis

    complement these as circumstances demand. In some cases DCF models are used, and in

    others, more detailed analyses of price-to-sales multiples, growth options, or profitability

    analysis are used. Another finding is that use of the RIV model is extremely limited, but

    analysts frequently use accounting data in single-period comparative and hybrid models.

    Analysts appear to vary the choice of valuation methodology in understandable ways with

    the context in which the valuation is made, but analyst familiarity with a valuation model

    and its acceptability to clients is a strong driving force.

    In terms of our positive approach to explaining valuation practices, we examine four

    hypotheses.

    The pervasiveness of comparative valuation techniques results in no significant difference

    in their use across sectors (HI). However, a more discriminating test shows that a

    multiperiod valuation model rather than a single-period method of comparatives is more

    likely to be the analysts' dominant model in the electronics and pharmaceuticals sectors

    compared with beverages. This result is consistent with the hypothesis that comparative

    valuation models are more popular in more stable sectors

    where conventional accounting does a better job of capturing the value of the firm.

    Insufficient instances of RIV mean that we cannot reject the null hypothesis on the

    relative use of DCF and RIV across sectors (H2). However, if we broaden this hypothesis

    to compare the use of multiperiod and hybrid valuation models based on cash and

    accruals, respectively, we find a significantly greater use of accruals models in beverages

    than in pharmaceuticals. Analysis of the combined choice of forecasting system andvaluation model shows that profitability analysis is more prominent in reports for the

    beverages sector than for electronics or pharmaceuticals.

    We find no report that uses a single-period cash flow multiple as its dominant valuation

    model (H3), consistent with analysts understanding the limitations of using a single-

    period cash flow number. Some analysts use price to single-period cash flow as a

    sensitivity check in sectors where the rate of growth is relatively stable. Finally, we find

    that over half of the analysts who construct a multiperiod valuation analysis choose it is as

    their dominant model. However, we also find that over one-quarter of these analysts

    subsequently adopt valuation by single-period comparatives as their dominant model,

    inconsistent with H4. We conjecture that this latter finding is due to valuation by

    comparatives, with the implicit support of a more sophisticated model, being preferred by

    the analysts' clients.

    This research suggests that careful study of comprehensive analysts' reports can improve

    our understanding of the variations in valuation practice. Specifically, the types of

    valuations used to justify analysts' recommendations depend on characteristics of the

    company being analyzed. In this paper, we focus on differences across industrial sectors,

    but we anticipate that valuation behavior may vary in other contexts. Further insights may

    emerge from studying analysts' reports for firms involved in IPOs, mergers, and major

    capital issues. Also, analysts may employ different models for

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    14/15

    dividend payers versus nonpayers. The special problems of valuing firms that report

    losses is also worthy of further work. Finally, sensitivity analyses suggest that prior

    results based on short, less comprehensive reports may understate the sophistication of the

    valuation models used by analysts.

    APPENDIX A

    DISCOUNTED FUTURE EARNINGS (DFE)

    When analysts value a firm based on a PE multiple, they control for the effects on

    earnings of nonrecurring events, transitory components, and accounting conservatism.

    Where a firm has negative, very low, or very high earnings that are unlikely to continue,

    financial analysts try to normalize earnings. The DFE approach to valuation, given by the

    following equation, is one such technique:

    V, = ^{EBITDA,^^)/(l + waccj ]x (EVIEBITDA), (1)

    where F, is the fundamental value ofthe firm at date /, EBITDA^^^ is earnings before

    interest, taxes, depreciation, and amortization in period / + x, wacc is the firm's weighted

    average cost of capital, and (JEVIEBITDA)Jis (enterprise value)/(eamings before interest,

    taxes, depreciation and amortization) for comparable firms at date /. Financial analysts

    project forward to the period when the firm is expected to reach a sustainable level of

    performance and discount the relevant future earnings to the

    present using the firm's weighted average cost of capital. Multiplying by a current

    benchmark value of EVIEBITDA for a set of comparable firms yields the fundamental

    value ofthe firm.

    APPENDIX B

    RATING TO ECONOMIC PROFIT

    The rating to economic profit (REP) is based on the relation between the market-to-book

    ratio at the enterprise level and the ratio of the retum on invested capital to the weighted

    average cost of capital:

    REP = (EV,/IC,)/{ROIC,^J wacc) (2)

    whereEV^is the market value ofthe firm's equity plus the book value ofthe firm's debt at

    date t, IC, is the book value ofthe capital invested in the firm at t, ROIC,^^is the expected

    retum on investedcapital in period / + 1, and wacc is the firm's weighted average cost of

    capital. Valuation theorysuggests that the book-to-market ratio is an increasing function

    ofthe finn's cost of capital, and that

    a relatively high spread between the expected return on invested capital and the weighted

    average cost of capital should lead to a high market-to-book multiple. If the latter relation

    does not hold, then the market does not impound properly all the available information

    about a firm's future performance in its current market value, and, hence, the firm is

    undervalued. Similarly, a low expected economic performance leads to a relatively low

    market-to-book ratio, otherwise the firm is overvalued.

    Although analysts consider REP to be a sophisticated form of price-to-book ratio, it can

    be shown that REP equals EVINOPLAT (where the denominator is one-year ahead netoperating profit less adjusted tax) multiplied by the weighted average cost of capital.

  • 7/31/2019 Validade Dos Modelos Us Ado Singles

    15/15

    Assuming costs of capital and leverage are constant within sectors, the profitability of an

    investment strategy based on REP should be similar to one based on one-year-ahead PE

    ratios.