An Investigation of the Impact of Publicly Available Accounting Data, Other Publicly Available Information

and Management Guidance on Buy-Side Analysts’ Forecasts

By Michael Newman, Wynne Chin, George Gamble and Michael Murray

Impetus for Study• This was an exploratory study

• There is a lot of information in the academic literature about sell-side analysts but little about buy-side analysts

• Question arose: what sources of information do buy-side analysts use to arrive at their stock recommendations and to what degree do they use it?

Academic Literature

• Givoly and Lakonishok (1979), Lys and Sohn (1990), Francis and Soffer (1997); and Barth and Hutton (2000) find that:– Financial (sell-side) analysts’ recommendations are

superior to other forecasts, such as those by time-series models, and

– Financial (sell-side) analysts help improve market efficiency since the information they provide on accruals and cash flows to their clients are quickly incorporated into stock prices.

Financial Analysts

• Financial analysts collect information about publicly traded companies from both public and private sources. – They then develop financial (earnings)

forecasts based on their evaluation of the current performance of firms that they follow, and make a stock recommendation of buy, hold or sell the stock.

• Source: Kothari (2001) and Fogarty & Rogers (2005)

Financial Analysts

• Sell-side Analysts produce research reports and/or perform analysis for customers of and are employed by securities brokerage and investment banking firms.

• Buy-side Analysts conduct analysis for their own firm that will be used for internal investment decisions, usually for institutional investors.

• Most academic studies have focused on sell-side analysts due to the availability of a large data set of individual analyst's actual decisions. – Little has been written about buy-side analysts because

they are very secretive. • Source: Kothari (2001) and Fogarty & Rogers (2005)

Limited Academic Literature on Buy-side Analysts

• Schipper (1991) finds that buy-side analysts’ stock recommendations are influenced by the recommendations made by sell-side analysts.

• Fogarty and Rogers (2005, p. 332) states that buy-side and sell-side analysts “have differing motivations and different information use motivations”.

• Question: what information is used and which sources of information are used the most by buy-side analysts?

Academic Literature

• To answer this question, we searched the literature to see what information sell-side analysts use in making their recommendations and to see if sell-side analysts base their stock recommendations on recommendations of other sell-side analysts.

• This information could then be used to form questions to ask of buy-side analysts.

Table 1 KEY LITERATURE FINDINGS

Source Construct Finding

Nofsinger & Sias (1996), Kothari (2001), and Fogarty & Rogers (2005)

Analysts Stock Recommendations, Forecast and Investment Advice

Analysts perform analysis, develop financial forecasts and make stock recommendation of buy, hold or sell the stock

Kothari (2001) and Fogarty & Rogers (2005)

Buy-Side Analysts Buy-side analysts work for institutional investors and their work is for the exclusive use of their employer (proprietary) and not released to the public.

Lewellen, Lease & Schlarbaum (1977, 1979) Francis & Soffer (1997), Fogarty & Rogers (2005)

Attitude Analysts' attitudes towards companies play a role in their stock purchase recommendations

Pankoff and Virgil (1970) Usefulness of information The extent to which information facilitates decision-making.

Fogarty & Rogers (2005) Information used by buy-side analysts Buy-side analysts have different motivations and uses of information than do sell side analysts

Schipper (1991) Other Analysts' Opinions Buy-side analysts are influenced by sell-side analysts

Previts, Bricker, Robinson & Young (1994)

Publicly Available Accounting Data Sell-side analysts use accounting data to get information about firms

Key Findings

Table 1 KEY LITERATURE FINDINGS

Source Construct Finding

Beneish (1999) and Kothari (2001) Evaluation Activity Detection of manipulation of earnings is of interest to analysts (Beneish, 1999, and Kothari, 2001) and use of analysis to identify hidden value (Kothari, 2001)

Tan, Libby & Hutton (2002) Trust of Management Refer to the fact that analysts sometimes have a "mistrust of management"

Bushee & Miller (2005) Find that “buy-side investors need to have faith in management.”

Kothari (2001) Management Guidance Analysts seek interaction with management to improve the flow of information and often issue optimistic forecasts to gain access to management (guidance).

Hongren (1978) and Fogarty & Rogers (2005)

Firm’s management is an important source of information

Pankoff and Virgil (1970) Other Accounting Information Analysts prefer general economic and competitive data

FASB (1986) and Arthur Levitt (1998) Representational Faithfulness Importance of accounting data being free of errors and bias

Key Findings

Research Questions

• What sources of information do buy-side analysts use to make their stock recommendation? Do buy-side analysts use:

– publicly available accounting data in making their stock recommendation?

– other publicly available information in making their stock recommendation?

– information received from corporate managers in the form of management guidance in making their stock recommendation?

– information received from other analysts’ in making their stock recommendation?

Research Questions• Do buy-side analysts find the sources of

information they use to make their stock recommendation to be accurate and unbiased? Do buy-side analysts find:

– publicly available accounting data to be representationally faithful?

– other publicly available information to be representationally faithful?

– information received from corporate managers to be representationally faithful?

– information received from other analysts’ to be representationally faithful?

Research Questions• Is a buy-side analyst’s attitude towards a

company affected by the sources of information he or she uses? Is a buy-side analyst’s attitude towards a firm:

– affected by publicly available accounting data?– affected by other publicly available

information?– affected by information received from

corporate managers in the form of management guidance?

– heavily influenced by the opinions and recommendations of analysts they respect?

Research Questions• Does the degree to which buy-side

analysts’ trust management of a firm affect their attitude or their stock purchase recommendation?

– Are buy-side analysts’ attitude towards a company affected by how much they trust the management of a firm?

– Are buy-side analysts’ stock purchase decisions affected by how much they trust the management of a firm?

Research Questions• Does a buy-side analyst’s attitude

toward a firm influence his/her stock purchase recommendation for that firm?

– Are buy-side analysts’ stock purchase decisions affected by their attitude towards the firm?

Development of the Models

• Three main models were then developed from the existing literature using the SEM (structural equation modeling) based method:– The Stock Purchase Recommendation Model– The Attitude Model– The Full (Combined) Model

• The SEM (structural equation modeling) based method was used.

The Stock Purchase Recommendation Model (Model #1)• The Stock Purchase Recommendation Model is

developed based upon the findings in the academic literature that suggest that financial analysts use: – publicly available accounting data, – other publicly available information, – management guidance, – the opinions of other financial analysts and – their trust of management

as a basis for the recommendation (buy, hold or sell) they give on a stock purchase decision.

The Stock Purchase Recommendation Model

The Stock Purchase Recommendation Model Hypotheses

The Hypotheses we formulated for The Stock Purchase Recommendation Model are:

• H1: Buy-side analysts use publicly available accounting data in making their stock recommendation.

• H2: Buy-side analysts use other publicly available information in making their stock recommendation.

• H3: Buy-side analysts use information received from corporate managers in the form of “management guidance” in making their stock recommendation.

• H4: Buy-side analysts use information received from other analysts in making their stock recommendation.

• H5: Buy-side analysts’ stock recommendation for a firm is affected by how much they trust management of a firm.

The Stock Purchase Recommendation Model with Interaction Effects (Model #2)

• The Stock Purchase Recommendation Model is then modified to test if representational faithfulness and usefulness/degree of use of these sources is important in this process.

The Stock Purchase Recommendation Model with Interaction Effects

The Stock Purchase Recommendation Model with Interaction Effects Hypotheses

The additional hypotheses we formulated for The Stock Purchase Recommendation Model with Interaction Effects are:

• H6: Buy-side analysts find publicly available accounting data to be representationally faithful.

• H7: Buy-side analysts find other publicly available information to be representationally faithful.

• H8: Buy-side analysts find information received from corporate managers in the form of management guidance to be representationally faithful.

The Stock Purchase Recommendation Model with Interaction Effects Hypotheses

The additional hypotheses we formulated :

• H9: Buy-side analysts use publicly available accounting data heavily in determining their forecast and recommendation for a firm.

• H10: Buy-side analysts use other publicly available information in determining their forecast and recommendation for a firm.

• H11: Buy-side analysts predominantly use information received from corporate managers in the form of management guidance in determining their forecast and recommendation for a firm.

The Attitude Model (Model #3)

• The Attitude Model is developed based upon the findings in the academic literature that suggests that financial analysts’ attitudes may influence their stock purchase recommendation.

• This model tests the effect that publicly available accounting data, other publicly available information, management guidance, the opinions of other financial analysts and their trust of management has on their attitude towards the organization they are making a stock purchase recommendation about.

The Attitude Model

The Attitude Model Hypotheses

The relationships of the buy-side analysts’ attitude towards the firm they are making a stock purchase recommendation for and their stock purchase recommendation are captured in the following hypotheses:

• H12: Buy-side analysts’ attitude towards a firm is affected by publicly available accounting data..

• H13: Buy-side analysts’ attitude towards a firm is affected by other publicly available information.

• H14: Buy-side analysts’ attitude towards a firm is affected by information received from corporate managers in the form of management guidance.

• H15: Buy-side analysts’ attitude towards a company is heavily influenced by the opinions and recommendations of analysts they respect.

• H16: Buy-side analysts’ attitude towards a company is affected by how much they trust management of a firm.

The Attitude Model with Interaction Effects (Model #4)

• The Attitude Model is then modified to test if representational faithfulness and usefulness/degree of use of these sources is important in this process.

• By examining both models, we hope to see if the information buy-side analysts receive from these sources (publicly available accounting data, other publicly available information, management guidance, the opinions of other financial analysts and the analyst’s trust of management) has similar or different influences on their stock purchase recommendation or their attitude.

The Attitude Model with Interaction Effects

The Full (Combined) Model • The Full (Combined) Model, tests to see if attitude

has an effect on an analyst’s stock purchase recommendation.

• It is tested in three ways. – First, we add attitude as a construct to the Stock

Purchase Recommendation Model (the Combined Model).

– Next we test it by adding attitude as a construct to the Stock Purchase Recommendation Model with Interaction Effects (the Combined Model with Interaction Effects).

– Finally, we take the Stock Purchase Recommendation Model and add the Attitude Model as a construct to see the extent of the effect that the constructs the academic literature identified have on the stock purchase recommendation decision (the Full (Combined) Model).

The Full (Combined) Model (Model #5)

• The Full (Combined) Model integrates “Attitude” into the Stock Recommendation Model based on the findings in the literature that analysts' attitudes towards companies play a role in their stock purchase recommendations (Lewellen, Lease & Schlarbaum,1977 and 1979, Francis & Soffer, 1997), and Fogarty & Rogers, 2005)

The Full (Combined) Model

These relationship between Attitude and a buy-side analyst’s stock purchase recommendation is captured in the final hypothesis:

• H17: Buy-side analysts’ attitude towards a company influences their forecast and stock purchase recommendation for a firm

The Combined Model

• The Combined Model is developed by taking the Stock Purchase Recommendation Model and adding Attitude as a construct

The Combined Model

The Combined Model with Interaction Effects (Model #6)

• Next we take the Stock Purchase Recommendation Model with Interaction Effects and add Attitude. – We refer to this model as the Combined

Model with Interaction Effects

The Combined Model with Interaction Effects

The Full (Combined) Model (Model #7)

• The Full (Combined) Model is a combination of the Stock Purchase Recommendation Model and the Attitude Model. – Attitude is added as a construct to see the

extent of the effect that the constructs the academic literature identified have on the stock purchase recommendation decision and on the buy-side analyst’s attitude.

The Full (Combined) Model

Selection of Subjects• The selection of the subjects for this study (i.e.,

buy-side analysts currently employed by institutional investors) was guided by the academic literature.

• The literature finds that institutional investors (e.g., employers of buy-side analysts) prefer to own stock in large firms with high visibility. – O’Brien and Bhushan (1990), Hessel and Norman (1992), Del

Guercio (1996), Falkenstein (1996), Gompers and Metrick (2001), and Bushee (2001) find that institutional investors prefer to invest in larger public firms.

– Bushee and Noe (2000), and Bradshaw, Bushee and Miller (2004) find that disclosure quality and visibility also makes a difference in how they choose to invest their funds.

Selection of Subjects• Based on this, a list of 703 buy-side analysts was

compiled from a number of sources that met this criteria, primarily from analyst meetings sponsored by public companies and from referrals (from Investor Relations Departments and other buy-side analysts). – The list was then culled to eliminate bad email addresses

and each buy-side analyst was then contacted to verify if they were still buy-side analysts and asked if they would be willing to participate in a survey.

• We had responses from 227 analysts. – One hundred forty analysts participated in the survey. – Of these, one hundred seventeen identified themselves

as “buy-side” and gave complete sets of data.

Selection of Subjects

• The sample size needed for this experiment, calculated to be 60, was determined based on the amount of power needed given the relationships that were described earlier in this section.

• The actual sample of 117 “buy-side” analysts surpassed this minimum requirement.

Method of Analysis

• The partial least squares (PLS) approach is used in this investigation because it can be used to model more complex situations (compared to covariance methods). – It is also consistent with Multiple Regression

Analysis in that it gives Beta and R2 statistics. – And allows the researcher to put greater

reliance on theory in analyzing data when it is strong (Chin, 1998).

Accounting Literature using PLS• Ittner, Larcker and Rajan (1997, p. 243-245) used PLS

to estimate their latent variables (the underlying theoretical constructs) to minimize the impact of their observable (manifest) indicators in their paper which examines the factors influencing the relative weights placed on financial and non-financial performance measures in CEO bonus contracts.

• Ittner and Larcker (1998, p. 5) used PLS to construct their customer-satisfaction index, stating “the index is constructed using Partial Least Squares (PLS) to weight the three items such that the resulting index has the maximum correlation with expected economic consequences (customers' self-reports of recommendations, repurchase intentions, and price tolerance).”

Analysis of Data• The analysis of data was completed by first assessing the

measurement model and then assessing the structural model. – In this case, significant tests were assessed using boot strap

analysis with 1000 re-samples. – No assumptions of normality were made in accordance with the

partial least squares (PLS) approach used in this study.

• Individual item loadings, internal consistency and discriminant validity were assessed for the measurement model using PLS.

• The structural model and the 17 hypotheses were tested by examining the path coefficients and their respective statistical significance.

• The predictive power of the model was based on the explained variance in the dependent constructs.

Survey Design• As discussed earlier, the structural model was developed

after an extensive review of the existing academic literature. – Key research areas were identified, constructs were established,

and key questions were developed based on the key facets of each construct.

• The study includes the appropriate use of negation in some cases, and asks questions several ways to better measure each construct. – The items were measured using mostly seven-point Likert-type

scales (with anchors such as “strongly disagree” to “strongly agree” and measurements ranging from “-3” to “+3” where “0” was neutral) and some eleven-point Likert-type scales (with anchors such as “pessimistic” to “optimistic” with a range of “0” to “+10”).

– A pilot study involving subject matter experts (including buy-side analysts) was conducted to clarify questions and to measure consistency and validate these items.

– Some changes in wording were made and some questions were eliminated as a result of the pilot study in order to refine the survey instrument.

Data Collection• The methodological approach used to test the relationships

involved a survey that was accessed on the Web. – A number of papers have been written about the use of Web based

surveys. – Recommendations made in this academic literature were

incorporated into the survey design, especially at it pertained to increasing survey response and the development of the survey tool.

– Kaplowitz, Hadlock and Levine (2004) states “For special populations that regularly use the Internet, the Web has been found to be a useful means of conducting research.”

• Selection of the relationships that were tested was appropriate since each was based on evidence of such relationships in the academic literature among financial analysts. – In addition, the survey questionnaire was administered to buy-side

analysts, a group whose use of various sources of information (the constructs in this study) is not public information.

Development of the Survey Instrument

• A facet based approach was used to develop the questions to be asked in the survey. – As key areas for research were identified:

• Definitions were developed, • The definition was decomposed into component parts (the

key facets of the question) in order to understand the underlying meaning, and

• Several questions were developed in order to select those that had the most congruence/meaning.

• We began with the main concept (e.g., the final outcome) -- how buy-side analysts determine their stock recommendation…..

Stock Purchase Recommendation• The literature confirms that analysts make financial

forecasts, provide investment advice and make stock purchase recommendations.

• We next defined the variable and developed the questions for the main variable as well as the reflective and formative variables

Stock Purchase Recommendation

DEFINITION: Extent or degree to which an analyst would make a recommendation involving the purchase or sale of a company's common stock.

OVERALL STOCK RECOMMENDATION Overall, all things considered, my feelings towards this

firm's stock is (please circle one rating for each descriptor):

NeutralBad -3 -2 -1 0 1 2 3 GoodUnsound -3 -2 -1 0 1 2 3 SoundBelow Standard -3 -2 -1 0 1 2 3 Above StandardWeak -3 -2 -1 0 1 2 3 StrongNegative -3 -2 -1 0 1 2 3

PositivePessimistic -3 -2 -1 0 1 2 3

Optimistic

Stock Purchase Recommendation

OVERALL STOCK RECOMMENDATION My recommendation for this firm's stock is

very positive.-3 -2 -1 0 1 2 3Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree

I base my stock investment recommendations on my financial forecasts for this firm.

-3 -2 -1 0 1 2 3Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree

Overall, I really like this firm's stock and recommend its purchase.

-3 -2 -1 0 1 2 3Strongly Disagree Somewhat Disagree Somewhat Agree Strongly Agree

Attitude• The literature next led to another potential direct

relationship based on the question of whether or not the buy-side analyst formulates his or her stock recommendation based on these variables or if the relationship is one between the buy-side analyst and his/her attitude towards the firm he/she is analyzing.

• DEFINITION: Affective evaluation an analyst has towards this firm’s financial performance.

Sources of Information• Next we determined the sources of information

analysts use to make their decisions. • The literature identifies several sources of

information that sell-side analysts use to arrive at their stock recommendation:– Publicly Available Accounting Data (Previts, Bricker,

Robinson and Young, 1994), – Other Publicly Available Information (Pankoff and

Virgil, 1970), – Management Guidance (Hongren, 1978; and Fogarty

& Rogers, 2005),– Opinions of Other Analysts (Schipper, 1991), and – Trust of Management (Tan, Libby and Hutton, 2002)

Representational Faithfulness and Degree of Use of Information

• The question of Representational Faithfulness (FASB, 1986 and Arthur Levitt, 1998) and Usefulness/Use of Information (Pankoff and Virgil, 1970), were also areas for contribution according to the literature.

• Connecting lines were then drawn between :– The two facets of representational faithfulness (Accuracy

and Bias), Representational Faithfulness and four of the sources of information (Publicly Available Accounting Data, Other Publicly Available Information, Management Guidance and Other Analysts Opinions); and,

– The Degree of Use of Information and these same four sources.

Representational FaithfulnessREPRESENTATIONAL FAITHFULNESS • Analysts evaluate the financial statements issued by

firms to determine what they feel the value of a firm should be. They make these decisions based on how much they trust the representational faithfulness of the information and whether or not they believe that the accounting information is unbiased and accurately reported based on GAAP.

DEFINITION Representational faithfulness is the extent to which the accounting data/financial statements truthfully represent the company's financial picture. Representational faithfulness is characterized by accuracy in GAAP reporting and the absence of reporting bias.

Degree of Usage

We also looked at “Degree of Use” or how often ananysts used each source of information.

Evaluation Activity• Detection of manipulation of earnings is of interest to

analysts (Beneish, 1999, and Kothari, 2001) as is the use of analysis to identify hidden value (Kothari, 2001)

• Based on the academic literature, a direct relationship was hypothesized to exist between “attitude” and “evaluation activity.”

Evaluation ActivityDEFINITION Extent to which an analyst is very diligent in evaluating

and understanding a firm.

Evaluation methods are used to:1) Get a better understanding of the firm; and,2) Check for bias and accuracy in GAAP reporting.

Evaluation may lead to more or less bias checking. Analysts use various analytical methods (such as fundamental analysis, technical analysis, and ratio analysis) and accounting measures to evaluate a firm's accounting performance. Measures designed to provide information useful for earnings/cash flow evaluation give an indication of the firm’s future economic income while managerial performance measures indicate the value added by the manager’s efforts or actions in a period.

Model Evaluation

• The 117 completed questionnaires were examined for outliers.

• A frequency Test was run using SPSS that examined the raw data file for skewness, kurtosis and outliers.

• No significant outliers were identified.

Reliability Assessment • Individual item loadings and internal consistency of

reliabilities were examined as a test of reliability. • Internal consistency was first assessed by

examining Cronbach’s alphas which is influenced by the loadings of the items. – Consistent with the academic literature, the reliability of the

constructs exceeded .7 in all but one of the constructs (use of other publicly available information which had a .678).

• A second test of reliability was then performed nusing the Composite Reliability Index. – All of the composite reliabilities exceeded the minimum acceptable

value of 0.7 including other publicly available information with a .862 Composite Reliability (Hair, Anderson, Tatham and Black, 1995).

Convergent Validity Assessment

• Convergent validity is shown when the t-values of the Outer Model Loadings are above 1.96. – The t-values in all the convergent validity

assessments are much greater than 1.96 indicating high convergent validity.

Discriminant Validity Assessment• Discriminant validity was assessed in two

ways: – One based on how well the items load on their own

construct as compared to other constructs in the model• The individual cross-loadings of the items all

exceeded .82 and the factor cross-loadings of all the items shared by the constructs are low, showing clear discriminant validity.

– A second based on the fact that the average variance shared between the constructs and their measures is greater than the variance shared between the constructs themselves.

Discriminant Validity Assessment

• The second test of Discriminant Validity is the AVE (Average Variance Extracted) Analysis can be performed two ways. – One approach requires that the square root of the AVE of each

construct be much larger, although there are no guidelines about how much larger, than any correlation between this construct and any other construct.

– A second approach to AVE (Average Variance Extracted) Analysis, suggested by Chin (1998), involves the squaring of the correlations and comparing those to the AVEs.

– Both approaches indicate Discriminant Validity. • Results of the other models tested give similar results.

Structural Model Evaluation• Each structural model was tested by examining the path

coefficients (and their respective statistical significance) and the R2 to determine the predictive power of the model, based on the explained variance in the dependent constructs.

• Each hypothesis was tested using PLS– Path coefficients, along with their degree of significance based on

T-values, are calculated using a boot-strapping procedure. – The latent scores derived from PLS Graph were then loaded into

SPSS to further test the significance of the path coefficients. – SPSS was then used to calculate the interaction terms.

• The PLS coefficients are different from the SPSS coefficient combination of the interactions, rounding and the fact that the results of the boot-strapping is based on an average of 1,000 re-samples.

Structural Model Evaluation• Results of the study are broken down into the following

models: • The Stock Purchase Recommendation Model

– The Stock Purchase Recommendation Model– The Stock Purchase Recommendation Model with Interaction Effects

• The Attitude Model– The Attitude Model– The Attitude Model with Interaction Effects

• The Full (Combined) Model– The Combined (the Stock Purchase Recommendation Model with

Attitude Added) Model – The Combined (the Stock Purchase Recommendation Model with

Attitude Added) Model with Interaction Effects– The Full (Combined) Model

Structural Model Evaluation

• The significance of the path (betas) and the adjusted R2 was used to measure significance in each model.

• Each was then evaluated by first looking at the main effect model without interactions and then by looking at it with interactions included.

Stock Purchase Recommendation Model (Model #1)

Stock Purchase Recommendation Model

• The R2 of the structural model used to measure the main effects of the Stock Purchase Recommendation Model was .398 with an Adjusted R2 of .371.

• Two significant effects were found with this model:– Use of publicly available accounting data had

a path of .393 with a significance of .000; and, – The opinions of other analysts had a path

of .258 with a significance of .002.

Stock Purchase Recommendation Model with Interaction Effects

(Model #2)

Stock Purchase Recommendation Model with Interaction Effects

• The R2 was .527 with an Adjusted R2 of .434.

• Three significant main effects and one interaction effect are found with this model:– Publicly available accounting data’s path increased

from .393 to .497 with a significance of .000; – Degree of usage of publicly available accounting data

had a path of .175 with a significance of .035;– The other analysts’ opinions path coefficient

decreased from .258 to .197 with a significance of .040 (compared to .002 in the main effects model); and,

– Trust of management became significant with a path coefficient of .172 and a significance of .097.

Attitude Model (Model #3)

Attitude Model• The R2 of the structural model used to

measure the Attitude Model is .642 with an Adjusted R2 of .626.

• Three significant main effects are found in the Attitude Model (main effects model), the first two of which are similar to the Recommendation Model (main effects model):– Use of publicly available accounting data had a path

of .450 with a significance of .000; – The opinions of other analysts had a path of .195 with

a significance of .002; and,– Trust of management was significant in this model

with a path coefficient of .383 and a significance of .000.

Attitude Model with Interaction Effects (Modeln #4)

Attitude Model with Interaction Effects

• The R2 of the structural model used to measure the Attitude Model with Interaction Effects indicates that it has slightly more explanatory power than the Attitude Model (main effects model) with an R2 of .700 (compared to .642) and an Adjusted R2 of .641 (compared to .626).

Attitude Model with Interaction Effects

• The three significant main effects found in the Attitude Model (main effects model) remain significant although the path coefficients have changed:– Use of publicly available accounting data has

a path of .437 with a significance of .000; – The opinions of other analysts has a path

of .194 with a significance of .011 (versus .002); and,

– Trust of Management was significant in this model with a path coefficient of .334 and a significance of .000.

Attitude Model with Interaction Effects • Three additional interaction effects are

also found that are significant:– The interaction of degree of usage of publicly

available accounting data and use of publicly available accounting data has a path coefficient of .192 and a significance of .005;

– The interaction of degree of usage of management guidance and management guidance has a path coefficient of .151 and a significance of .072; and,

– The interaction of representational faithfulness of management guidance and management guidance has a path coefficient of -.186 and a significance of .015.

The Combined Model (Model #5)

The Combined Model

• The R2 of the Combined Model is .569 (compared to .398 for the Stock Purchase Recommendation Model) with an Adjusted R2 of .546 (vs. .371), indicating that the model has better explanatory power than the Stock Purchase Recommendation Model.

The Combined Model• Interestingly enough, only one of two main effects

found in the Stock Purchase Recommendation Model is still present in the Combined Model albeit at a lower significance level -- the opinions of other analysts has a path of .124 with a significance of .089. – The beta coefficient of the other main effect that was

present, publicly available accounting data, dropped from .383 to .083 and its significance level decreased from a significant .000 to .380.

• Two new significant main effects are found:– Trust of management has a negative path coefficient of

-.175 with a significance of .032; and, – The main effect with the largest significance is attitude

with a path coefficient of .685 and a significance level of .000.

Combined Model with Interaction Effects (Model #6)

Combined Model with Interaction Effects

• The R2 of the structural model used to measure the Combined Model with Interaction Effects (the Stock Purchase Recommendation Model with Interaction Effects and Attitude added) is considerably higher (e.g., had much better explanatory power) than that of the original Stock Purchase Recommendation Model with Interaction Effects -- .659 (compared to .527) with an Adjusted R2 of .587 (vs. .434).

Combined Model with Interaction Effects• There is one similarity and two differences in the significant

main effects and interaction effects found between the Combined Model and the Combined Model with Interaction Effects:– Attitude remained as the main effect with the largest significance with

a path coefficient of .679 and a significance level of .000;– Publicly available accounting data became significant again (it has

statistical significance in both of the Stock Purchase Recommendation Models and the two Attitude Models, but not the Combination Model) with a path coefficient of .210 and a significance of .057; and,

– Degree of usage of publicly available accounting data became significant for the first time in any of our models with a path coefficient of .133 and a significance of .041.

• It is interesting to note that trust of management lost significance with a negative path coefficient of -.082 and a significance level of .401 in the Combination Model with Interactions Effects after being statistically significant in the Stock Purchase Recommendation Model with Interaction Effects, both Attitude Models and the Combined Model.

The Full (Combined) Model (Model #7)

The Full (Combined) Model

• The R2 of the Full (Combined) Model is .565 (compared to .569 for the Combined Model and .659 for the Combined Model with Interaction Effects) with an Adjusted R2 of .541 (versus .546 and .587, respectively). Although its “goodness of fit” is less that the two Combination Models, it still has high explanatory power.

The Full (Combined) Model• Three significant main effects and three significant

moderating effects are found in the Full (Combined) Model:

• The largest significant effect is attitude with a .684 path coefficient and a significance of .000 (this was consistent with the earlier Combination Models which has .beta coefficients of .685 and .679, respectively, both at a .0000 significance level);

• Trust of management has two significant effects:– a main effect negative path coefficient of -.179 with a

significance of .030 when regressed against stock purchase recommendation; and,

– a moderating effect positive path coefficient of .383 with a significance of .000 when regressed against attitude;

The Full (Combined) Model• Other analysts’ opinions also has two significant

effects:– a main effect positive path coefficient of .124 with a

significance of .091 when regressed against stock purchase recommendation; and,

– a moderating effect positive path coefficient of .195 with a significance of .002 when regressed against attitude; and,

• Publicly available accounting data does not have a significant main effect when regressed against the stock purchase recommendation dependent variable (.083 with a .383 level of significance) but does have a moderating effect when regressed with attitude (with a path coefficient of .450 and a significance level of .000).

Summary and Conclusions

• Overall, we find that the three groups of models that were developed to test our hypotheses have good explanatory power:

Summary and Conclusions• A summary of key (significant) results for all of the models tested is

shown below. – It should be noted that the constant term (alpha) does not have an effect of

any magnitude in any of the seven models, which is generally the case.

Conclusions• The best interpretation of these findings,

especially when coupled with the findings of the model, is that buy-side analysts – base their decisions to recommend

purchase of a stock largely based on their attitude towards a company and

– that that attitude is influenced largely by publicly available accounting data and trust of management,

– as well as what other analysts they respect say about the company.

Conclusions• Our most interesting finding was that trust of

management plays an important role in both a buy-side analyst’s attitude towards an organization and in their decisions regarding purchasing, holding or selling a security. Specifically,– there is evidence that there is a positive, significant relationship

between a buy-side analyst’s attitude towards a stock and trust of management, and

– a lesser weighted but still significant negative relationship between a buy-side analyst’s stock purchase recommendation and trust of management.

• This implies that while buy-side analysts may feel good about companies with management they trust, they are cautious about investing in firms run by trustworthy management.– An interesting side note is that while buy-side analysts do not

consider management guidance in their decisions, their contact with management does seem to influence their trust of management.

Conclusions• Our second most interesting finding is that

the independent variable that has the greatest single effect on a buy-side analyst’s stock purchase recommendation is the buy-side analyst’s attitude towards that company.– In other words, based on the findings of the models,

all of which were found to have good explanatory power, it appears that buy-side analysts base their buy or sell decisions of a stock largely on their “gut feeling” (“attitude”) towards a company and that this attitude is largely influenced by the publicly available accounting data, their trust of management and the opinions of other analysts they respect about the firm they are making a stock purchase recommendation about.

Conclusions• We also find that publicly available accounting

data – plays an important role in the buy or sell recommendation a buy-

side analysts makes when attitude is not part of the equation but loses some of its effect when attitude is added; and,

– has a huge effect on the buy-side analyst’s attitude towards a company.

• This leads us to conclude that while publicly available accounting data is considered in the stock purchase recommendation decision (direct impact), its largest influence is an indirect one (on a buy-side analyst’s attitude toward the firm).– It should be noted that the degree of use/usefulness of publicly

available accounting data was also significant, leading us to conclude that buy-side analysts do use this information.

Conclusions

Finally, we find that other analysts’ opinions affects both the buy or sell recommendation and the attitude of buy-side analysts, and does so in a positive way, implying that buy-side analysts are more likely to recommend purchase of a stock if (other) analysts they trust are making similar recommendations (herd behavior).

THANK YOU