Mobile Edition
Print
Get the Mag
Weekly UpdatesTwo hypotheses result from this theoretical foundation (stated in the alternative form):
H1: Investors observing an optimistic analyst earnings forecast will rate the credibility of the forecast lower than those investors observing a pessimistic analyst earnings forecast.
H2: Higher expectations for analyst optimism, a priori, will result in lower analyst credibility assessments.
The Impact of Analyst Credibility Assessments
Whereas theory guided the construction of the preceding hypotheses, it is also important to ascertain whether investors' perceptions affect their behavior. Specifically, do investors' credibility perceptions affect the degree to which they rely upon analysts when constructing their own expectations of future earnings? If so, then analysts must also consider the impact on investors when they are faced with pressures to conform to the demands of management and investment banking colleagues. This leads to the following hypothesis (stated in the alternative form):
H3: Analyst credibility assessments are positively related to the degree to which investors rely upon their earnings forecasts.
RESEARCH METHOD
To test these hypotheses, an experiment was conducted. The experimental method permits certain variables to be controlled, such as analyst quality, while others are manipulated, such as analyst forecast type. The effect of equally extreme optimism and pessimism can thus be observed and measured, providing a clear test of the research questions.
Subjects
Subjects were restricted to individual investors to isolate the effects of analyst behavior on perceptions of credibility within this one constituent group. All of the 177 participants were members of the National Association of Investors Corporation (NAIC). They are active investors who regularly use fundamental analysis techniques to evaluate equity opportunities. Most NAIC members are skilled investors whose portfolio returns (unadjusted for risk) consistently equal or outperform the Standard & Poor's 500 stock index (Gottschalk, 1994). Half of the subjects had five or more years experience analyzing financial statements as part of their investment decision-making process.
Experimental Procedure
The experimental task required investors to forecast earnings for one of two companies at two intervals: before and after exposure to the analyst's forecast. This active involvement in the prediction task facilitated their assessment of the analyst's forecast credibility because they had personally formed a belief about the prospects of the stock. The stock-specific knowledge offered investors a grounded perspective from which they could evaluate the analyst's forecast.
The experiment proceeded in four stages. First, investors rated their beliefs that analysts in general were consistently optimistic in their earnings forecasts using a seven-point Likerttype scale (strongly agree/strongly disagree). Second, they were then provided with information on a real, but disguised, company's products and financial statements, industry statistics, and a brief description of management's plans. Using these data, they were asked to predict earnings one year ahead for the company. Third, the investors were given an analyst's earnings forecast along with similar financial information, but presented in ratio form. They were then offered the opportunity to revise their original forecast. Finally, subjects rated the credibility of the analyst. The entire task took an average of 30 minutes to complete.
The financial data were provided to avoid creating a demand effect whereby subjects would assume they were expected to react to the one new piece of information. A control group, consisting of 56 investors, was employed prior to the experiment. They received no analyst forecast in this stage of the experiment, only the additional financial data. The purpose of the control group was to determine whether the reorganized financial data had any influence on the subjects' forecasts.
The two companies subjects analyzed both had recently experienced poor financial performance. Prior research indicates that it is when firms are not performing well, they have the highest motivation to exert pressures on analysts (Schultz, 1990) and the optimistic forecast bias is prevalent (Ali et al., 1992). The goal in this experiment is to assess whether analyst credibility is a function of a perceived optimistic forecast bias. Thus, it is necessary to place the subjects in a setting where the prior literature has demonstrated the greatest probability for a systematic optimistic bias to occur.
Measures
Analyst's Forecast. Investors were provided with one of two types of analyst forecasts: optimistic or pessimistic. The control group was also used to set the forecast types at levels that would be perceived as optimistic and pessimistic given the data provided to subjects. The key was to construct optimistic and pessimistic forecasts that were perceived as unusual but still authentic. The control group had the same information available to it with the exception of the analyst's forecast. The mean control group forecast was judged to be equivalent to a neutral analyst forecast. The optimistic and pessimistic forecasts were set equidistant from the neutral forecast, approximately one standard deviation above and below the neutral forecast.
Analyst Credibility. Credibility was measured using a seven-point, bi-polar adjective scale that elicited the extent to which the investors perceived the analyst's forecast to be (1) dependable, (2) credible, (3), accurate, or (4) trustworthy. This scale had been developed and validated in prior research (Lichtenstein and Bearden, 1989).
Analyst Quality. Investors were told only that the analysts were professionals. The purpose of this explicit statement was to set a neutral tone for the analyst's abilities so that the subjects would focus on the content of the message rather than the qualifications of the analyst.
Investor Reliance. To determine the extent to which investors relied upon the analyst's forecast it is necessary to measure the magnitude of forecast revision relative to the analyst forecast provided. A relative measure is necessary to gauge reliance rather than simply the gross change in the investor's forecast. For instance, one investor might revise his/her forecast upwards by $0.20 and another might revise it by $0.05. It is impossible to determine which investor relied more upon the analyst forecast with-out judging their revisions relative to the first forecast and the analyst forecast provided. A Reliance Index was created to capture the degree to which the revision coincided with the observed analyst forecast. The Reliance Index is computed as:
Reliance Index = Forecast Revision - Initial Forecast/Analyst Forecast - Initial Forecast
The numerator represents the magnitude of each subject's forecast revision and the denominator measures the amount of forecast revision they would have needed to exhibit if they had completely relied upon and accepted the analyst forecast. For example, a Reliance Index of 0.40 indicates that the investor has revised his/her forecast towards the analyst's forecast and the revision magnitude was forty percent of the distance between the initial forecast and the analyst forecast. A Reliance Index of 1.0 indicates the subject's forecast revision matched the analyst forecast, indicating complete reliance on the analyst. If the Reliance Index score is greater than 1.0, the subject's forecast revision was greater than the distance towards the analyst forecast (i.e., the subject overshot the analyst forecast). When the Reliance Index is positive it indicates the subjects revised their forecasts towards the analyst forecast. The Reliance Index will be negative when subjects revise their initial forecast in the opposit e direction of the analyst forecast. If the numerator is zero then the investor made no change in his/her forecast after the analyst forecast was provided, thus resulting in a zero Reliance Index score. If the investor's initial forecast were equal to the analyst's forecast provided at a later stage in the experiment, the denominator would be zero and it would not be possible to determine the degree of reliance on the analyst forecast. This event did not occur for any of the subjects in this experiment.
Empirical Tests
The Analyst Credibility Model is best described using a linear representation of the independent and dependent variables:
Analyst Credibility = a + [b.sub.1] ([x.sub.1]) + [b.sub.2] ([x.sub.2])
Where: [b.sub.1] represents Expected Optimism [b.sub.2] represents Analyst Forecast Type
Analyst Credibility and Expected Optimism, were measured variables designed to elicit the perceptions of investors towards analysts' behavior. It was anticipated that greater Expected Optimism would lead to lower assessments of Analyst Credibility, or a negative coefficient associated with [x.sub.1]-Analyst Forecast Type was a manipulated variable where each subject received either an optimistic (1) or pessimistic (0) analyst forecast. Subjects receiving the optimistic forecast type were expected to perceive analyst credibility as lower than those subjects receiving the pessimistic forecast; hence a negative coefficient was also expected for [x.sub.2].
A simple linear model is sufficient to detect a relationship between the Reliance Index and Analyst Credibility:
Reliance Index = a + [b.sub.1] (x)
Where: [b.sub.1] represents Analyst Credibility
A positive relationship between Analyst Credibility and the Reliance Index was expected. A positive coefficient would be evidence that increases in analyst credibility assessments coincide with increased reliance on the analyst's forecast.
FEED