From the latest report of Deutsche Bundesbank ( Germany's central bank ):
Securities analysts act, to a degree, as a link between the
companies they cover and potential investors or market
observers. They can therefore be regarded as information
intermediaries. Their task is to collect and evaluate a
wide range of information of varying quality and â in the
case of an earnings estimate â condense it into a single
figure. The result of their analysis is generally published
prominently, but not any information on the preceding
decision-making process. However, the quality and rationality
of the forecasts can be properly assessed only if
the factors influencing the decision are known. In the following,
we will analyse various determinants that could
have an impact on securities analystsâ decision-making
process and the quality of their forecasts. We will study,
first, what influence the individual environment has on
forecast quality and, second, to what extent publicly
available information is reflected in forecasts.1
Irrespectively of the determinants to be examined, any
empirical study on forecast quality must control for the
forecast horizon, which is defined as the period (generally
measured in months) between the time the forecast is
produced and the end of the business year for which the
forecast is made. As expected, there is a negative correlation:
forecast accuracy diminishes as the forecast horizon
grows longer. The business year in question can also
exert a specific influence, which has to be taken into account
in any analysis.
Analyst-specific and broker-specific factors
Differences in individual analystsâ forecast quality can, in
part, be explained by the individual environment or specific
analyst characteristics.2 For instance, various studies
show that long professional experience has a significant
positive impact on forecast quality. One explanation is
the âlearning by doingâ effect: the longer someone
works as an analyst, the greater his experience in the
field, which in turn leads to better forecast results. In
addition, Hong and Kubik (2003) state that, for analysts
working in the United States, continued employment in
the industry is closely linked to forecast accuracy.3 In
other words, analysts with comparatively many years of
experience have undergone a selection process in which
they were able to prevail over their rivals. However, the
positive correlation between professional experience and
forecast accuracy proved weak or even inexistent especially
for European analysts, which is explained, inter
alia, by differences in the incentive structure.4 The literature
therefore makes a further distinction between general
and company-specific professional experience. The
latter relates exclusively to the period over which an individual
analyst has covered a specific company. The positive
correlation generally proves robust in empirical analysis;
one possible reason is that communications between
the analyst and the management of the covered
company improve with years on the job.
A further analyst-specific determinant whose potential
impact is investigated in empirical studies is the number
of forecasts that an analyst makes for a firm in a business
year. If a large number of revisions are necessary, this
points to difficulties in establishing an adequate assessment,
which results in a negative correlation between
this variable and forecast accuracy, particularly at the beginning
of the business year. Conversely, at the end of
the business year, the number of revisions should have
ensured that the necessary adjustments have been made,
which would mean a statistically significant difference
can no longer be found.
Broker-specific factors include the size of the portfolio an
individual analyst covers. Here, a negative correlation is
assumed in theory: the more enterprises or sectors an
analyst covers, the less time he has to analyse a specific
company, which is reflected in a significantly greater
forecast error.
Various studies also show that the size of an analystâs employer
is statistically significant. US studies in particular demonstrate that analysts employed by larger brokerage
houses make better forecasts than their peers at smaller
houses. One possible explanation is that analysts with important
brokerage houses have better access to companiesâ
management. Similarly, they could have better resources
at their disposal. Larger brokerage houses are regarded
as the more attractive employers, partly for the
reasons outlined above, potentially leading to them employing
the better analysts. However, this argument does
not necessarily apply to the European market, as brokerage
houses do not hire staff based as exclusively on past
forecast accuracy as in the United States.5
Although the above-mentioned variables are used to try
to explain, as much as possible, the differences in forecast
quality based on analyst-specific and broker-specific behaviour,
a large part remains unexplained. As a result,
prior analyst-specific forecast quality generally proves
highly significant in addition to the above-mentioned
determinants. For instance, Brown (2001) shows that a
simple model containing only analystsâ individual prior
forecast quality as an explanatory variable performs just
as well as a model that contains the analyst characteristics
described above.6
Processing publicly available information
As mentioned above, a financial analystâs real achievement
is to collect, weight and compress existing information.
An important source of information is doubtless the
current consensus forecast among other analysts, which
is the subject of intense debate in the literature. The individual
analyst starts out in the same situation as other uninvolved
market players. While he is familiar with the result
of the consensus estimate, he does not know what
factors may have played a role in his peersâ decision-making
process. Unlike other market players, however, the
analystâs own forecast gives him an idea of how the consensus
estimate could change.
According to Banerjeeâs definition (1992), individual analystsâ
behaviour is classified as non-rational herd behaviour
if they base their forecast exclusively on other analystsâ
consensus estimate and neglect their own information.
7 Such behaviour is, however, difficult to prove empirically.
As different analysts usually respond to similar
information signals, they will likely arrive at similar recommendations.
In this case, it is therefore not clear
whether synchronised analystsâ earnings revisions are
due to herd behaviour or merely to the fact that they
base their decisions on the same information. Clement,
Hales and Xue (2007) demonstrate for the United States
that the consensus forecast is used in a rational manner.
They find that analysts are more likely to incorporate information
from the consensus forecast into their own
forecast the greater the number of analysts involved in
the consensus forecast.8 In this case â if analysts use the
consensus forecast as one of several sources of information
â their own forecast accuracy may improve.9 It would
therefore be premature to describe proof that the consensus
forecast has an influence on analyst decisions as irrational
behaviour. For Germany, Naujoks et al (2009)
even show that analysts systematically go against the
consensus forecast in order to raise their profile (antiherding).
10
Past stock market performance is similar to the consensus
forecast. Even though this is publicly known and available
information, taking share prices into account may
well help improve the quality of earnings forecasts.11
Another source of information for which one would expect
similar analyst behaviour is the macroeconomic outlook.
Unexpected changes in and increased uncertainty
about future macroeconomic developments are both
likely to impact analyst-specific earnings forecasts.
http://www.bundesbank.de/download/volkswirtschaft/monatsberichte/2009/200907mb_en.pdf
Securities analysts act, to a degree, as a link between the
companies they cover and potential investors or market
observers. They can therefore be regarded as information
intermediaries. Their task is to collect and evaluate a
wide range of information of varying quality and â in the
case of an earnings estimate â condense it into a single
figure. The result of their analysis is generally published
prominently, but not any information on the preceding
decision-making process. However, the quality and rationality
of the forecasts can be properly assessed only if
the factors influencing the decision are known. In the following,
we will analyse various determinants that could
have an impact on securities analystsâ decision-making
process and the quality of their forecasts. We will study,
first, what influence the individual environment has on
forecast quality and, second, to what extent publicly
available information is reflected in forecasts.1
Irrespectively of the determinants to be examined, any
empirical study on forecast quality must control for the
forecast horizon, which is defined as the period (generally
measured in months) between the time the forecast is
produced and the end of the business year for which the
forecast is made. As expected, there is a negative correlation:
forecast accuracy diminishes as the forecast horizon
grows longer. The business year in question can also
exert a specific influence, which has to be taken into account
in any analysis.
Analyst-specific and broker-specific factors
Differences in individual analystsâ forecast quality can, in
part, be explained by the individual environment or specific
analyst characteristics.2 For instance, various studies
show that long professional experience has a significant
positive impact on forecast quality. One explanation is
the âlearning by doingâ effect: the longer someone
works as an analyst, the greater his experience in the
field, which in turn leads to better forecast results. In
addition, Hong and Kubik (2003) state that, for analysts
working in the United States, continued employment in
the industry is closely linked to forecast accuracy.3 In
other words, analysts with comparatively many years of
experience have undergone a selection process in which
they were able to prevail over their rivals. However, the
positive correlation between professional experience and
forecast accuracy proved weak or even inexistent especially
for European analysts, which is explained, inter
alia, by differences in the incentive structure.4 The literature
therefore makes a further distinction between general
and company-specific professional experience. The
latter relates exclusively to the period over which an individual
analyst has covered a specific company. The positive
correlation generally proves robust in empirical analysis;
one possible reason is that communications between
the analyst and the management of the covered
company improve with years on the job.
A further analyst-specific determinant whose potential
impact is investigated in empirical studies is the number
of forecasts that an analyst makes for a firm in a business
year. If a large number of revisions are necessary, this
points to difficulties in establishing an adequate assessment,
which results in a negative correlation between
this variable and forecast accuracy, particularly at the beginning
of the business year. Conversely, at the end of
the business year, the number of revisions should have
ensured that the necessary adjustments have been made,
which would mean a statistically significant difference
can no longer be found.
Broker-specific factors include the size of the portfolio an
individual analyst covers. Here, a negative correlation is
assumed in theory: the more enterprises or sectors an
analyst covers, the less time he has to analyse a specific
company, which is reflected in a significantly greater
forecast error.
Various studies also show that the size of an analystâs employer
is statistically significant. US studies in particular demonstrate that analysts employed by larger brokerage
houses make better forecasts than their peers at smaller
houses. One possible explanation is that analysts with important
brokerage houses have better access to companiesâ
management. Similarly, they could have better resources
at their disposal. Larger brokerage houses are regarded
as the more attractive employers, partly for the
reasons outlined above, potentially leading to them employing
the better analysts. However, this argument does
not necessarily apply to the European market, as brokerage
houses do not hire staff based as exclusively on past
forecast accuracy as in the United States.5
Although the above-mentioned variables are used to try
to explain, as much as possible, the differences in forecast
quality based on analyst-specific and broker-specific behaviour,
a large part remains unexplained. As a result,
prior analyst-specific forecast quality generally proves
highly significant in addition to the above-mentioned
determinants. For instance, Brown (2001) shows that a
simple model containing only analystsâ individual prior
forecast quality as an explanatory variable performs just
as well as a model that contains the analyst characteristics
described above.6
Processing publicly available information
As mentioned above, a financial analystâs real achievement
is to collect, weight and compress existing information.
An important source of information is doubtless the
current consensus forecast among other analysts, which
is the subject of intense debate in the literature. The individual
analyst starts out in the same situation as other uninvolved
market players. While he is familiar with the result
of the consensus estimate, he does not know what
factors may have played a role in his peersâ decision-making
process. Unlike other market players, however, the
analystâs own forecast gives him an idea of how the consensus
estimate could change.
According to Banerjeeâs definition (1992), individual analystsâ
behaviour is classified as non-rational herd behaviour
if they base their forecast exclusively on other analystsâ
consensus estimate and neglect their own information.
7 Such behaviour is, however, difficult to prove empirically.
As different analysts usually respond to similar
information signals, they will likely arrive at similar recommendations.
In this case, it is therefore not clear
whether synchronised analystsâ earnings revisions are
due to herd behaviour or merely to the fact that they
base their decisions on the same information. Clement,
Hales and Xue (2007) demonstrate for the United States
that the consensus forecast is used in a rational manner.
They find that analysts are more likely to incorporate information
from the consensus forecast into their own
forecast the greater the number of analysts involved in
the consensus forecast.8 In this case â if analysts use the
consensus forecast as one of several sources of information
â their own forecast accuracy may improve.9 It would
therefore be premature to describe proof that the consensus
forecast has an influence on analyst decisions as irrational
behaviour. For Germany, Naujoks et al (2009)
even show that analysts systematically go against the
consensus forecast in order to raise their profile (antiherding).
10
Past stock market performance is similar to the consensus
forecast. Even though this is publicly known and available
information, taking share prices into account may
well help improve the quality of earnings forecasts.11
Another source of information for which one would expect
similar analyst behaviour is the macroeconomic outlook.
Unexpected changes in and increased uncertainty
about future macroeconomic developments are both
likely to impact analyst-specific earnings forecasts.
http://www.bundesbank.de/download/volkswirtschaft/monatsberichte/2009/200907mb_en.pdf
