Docstoc

Do Long-term Growth Forecasts Signal Analyst Quality - Do LTG

Document Sample
Do Long-term Growth Forecasts Signal Analyst Quality - Do LTG Powered By Docstoc
					Do Long-term Growth Forecasts Signal Analyst Quality or
                        Incentives?




                      Andreas Simon
                 Orfalea College of Business
                  Cal Poly San Luis Obispo
                    ansimon@calpoly.edu




                       John Nowland
               School of Economics & Finance
             Queensland University of Technology
                   j.nowland@qut.edu.au




                 This version: January 2009
       Do Long-term Growth Forecasts Signal Analyst Quality or
                                         Incentives?




Abstract
This study examines two explanations why not all analysts issue long-term earnings
growth (LTG) forecasts. First, analysts issue LTG forecasts to signal their superior
quality (forecasting ability, experience, private information). Second, analysts issue
LTG forecasts to establish and maintain other business relationships (e.g.
underwriting) with firms. Using I/B/E/S data from 1994 to 2005, we find that on
average only 57 percent of analysts issue LTG forecasts. Consistent with the first
explanation we find that analysts that issue LTG forecasts have superior forecasting
ability, more general experience and more private information. Prior to Regulation
FD, we find that analysts were more likely to issue LTG forecasts if their brokerage
also provides underwriting services. After Regulation FD, this relationship is no
longer significant. These results suggest that LTG forecasts are a good indicator of
analyst quality and that Regulation FD has helped to alleviate analysts’ incentives.


Keywords: Analysts’ forecasts; Analyst quality; Long-term earnings growth




* The authors gratefully acknowledge co mments and suggestions from Peter Clarkson, Irene Tutticci,
Jason Hall, Terry O’Keefe, Cyrus Ramezani, Beverly Walther and workshop participants at the UTS
2007 Research Su mmer School and USQ seminar series on previous versions of this paper.




2
1. Introduction
The objective of this study is to determine what influences a financial analyst’s
decision to issue a long-term earnings growth (henceforth, LTG) forecast for a firm
he/she covers. Prior research suggests that analysts’ LTG forecasts are more important
than their short-term earnings forecasts for the prediction of stock prices (Bradshaw
2004; Bradshaw 2002). However, only 35.8% of the 1,126 analyst reports (written by
56 different Institutional Investor All–American team members) examined by
Asquith, Mikhail and Au (2005) contain earnings growth information. This forecast
pattern is intriguing since the authors find that 99.1% of their sampled analysts use
earnings multiples as their valuation model and LTG is a key determinant of the price-
to-earnings (P/E) ratio. 1
        This study proposes two explanations as to why some analysts choose to issue
LTG forecasts and others do not. F irst, analysts issue LTG forecasts to signal to the
market that they are superior analysts. Due to the greater uncertainty analysts face
when making LTG forecasts (Chan, Karceski and Lakonishok 2003; Dechow, Hutton
and Sloan 2000; Frankel and Lee 1998) we expect only analysts that have superior
forecasting ability, greater experience and more private information will issue LTG
forecasts. Under this explanation, LTG forecasts could be used as an indicator of
analyst quality. Second, analysts issue LTG forecasts to establish and maintain other
business relationships with firms. As analysts are not generally evaluated on the
accuracy of their LTG forecasts (Dechow et al., 2000), there is an incentive for
analysts to use LTG forecasts to provide a positive outlook for the firm in the hope of
securing other business (e.g. underwriting). Under this explanation, LTG forecasts
could indicate a conflict of interest.
        This study uses LTG forecasts for US firms available from the I/B/E/S Detail
History Database between 1994 and 2005. Descriptive statistics indicate that only
57% of the 13,385 analysts listed on I/B/E/S provide at least one LTG, and that only
12% of the total 1,950,812 earnings forecasts over this period are LTG forecasts.
Consistent with the first explanation we find that analysts that issue LTG forecasts
have superior forecasting ability, more general experience and more private

1
 Expected earnings growth is widely used for valuation based on the Gordon Growth Model. Beaver
and Morse (1978) suggest that growth jointly determines P/ E ratios with risk and transitory earnings.
Penman (1996) shows that the P/E ratio indicates future earnings growth. Ohlson and Juettner-Nauroth
(2005) derive a theoretical model that takes into account two growth measures of earnings - short-term
growth in earnings and long-term earn ings growth - to explain the forward P/ E ratio.


3
information than other analysts. Prior to Regulation FD, we find that analysts were
more likely to issue LTG forecasts if their brokerage also provides underwriting
services. After Regulation FD, this relationship is no longer significant. In addition,
we find that analysts are more likely to issue LTG forecasts if they have less firm-
specific experience. This may indicate that analysts that are relatively newer to a firm
issue LTG forecasts to seek attention from firm management. These results are robust
to controlling for other analyst (resources, workload) and firm (size, volatility,
growth, intangibles) factors that could effect an analyst’s decision to issue a LTG
forecast. We also show that the result is not driven by the reporting of LTG forecasts
to I/B/E/S.
       This paper makes several contributions to the literature. First, it shows that
something as simple as the issuance of a LTG forecast can be used as a signal of
analyst quality. Second, it demonstrates that differences exist in analysts’ research
output, with not all analysts being able to issue more complex or uncertain LTG
forecasts. This is important to investors as the best equity analysis looks at the long-
term health of companies rather than merely suggesting shorter term earnings
expectations and trading opportunities. Third, this research provides evidence to
suggest that Regulation FD has been effective in allevia ting analysts’ incentives (or
conflicts of interest). In particular, the issuance of LTG forecasts to establish and
maintain other business relationships with firms.


2. Data and Methodology
2.1 Long-term earnings growth forecasts
Initially all US firms with analyst earnings forecasts (annual and LTG) are identified
from the I/B/E/S Detail History files for the period of January 1994 through
December 2005. Coverage of LTG data in I/B/E/S began in 1982. However, prior to
the early 1990s, LTG forecast dates recorded by I/B/E/S differed from the actual
forecast date (Cooper et al. 2001). We therefore restrict the start of our sample to
1994. LTG in the I/B/E/S US Detail History File are received directly from
contributing analysts, they are not calculated by I/B/E/S. While different analysts
apply different methodologies, the LTG forecast generally represents an expected




4
annual increase in operating earnings over the company’s next full business cycle. In
                                                                            2
general, these forecasts refer to a period of five years (I/B/E/S 2000).
        Table 1 shows the distribution of LTG and annual earnings forecasts over the
sample period and by year. The population consists of 1,950,812 earnings forecasts of
which 12 percent or 233,676 are LTG forecasts. As our analysis is focused on the
difference between analysts that do and don’t issue LTG forecasts, we exclude
multiple forecasts by the same analyst for the same firm in the same year (forecast
revisions).This leaves a total of 531,258 unique analyst- firm- year forecasts of which
30 percent or 157,512 are LTG forecasts. At the analyst level, a total of 13,385
analysts provided earnings forecasts to I/B/E/S during the period, of which 57 percent
or 7,660 analysts issued at least one LTG forecast. At the firm level, earnings
forecasts were provided to I/B/E/S for 13,778 firms during the sample period. Of
these, LTG forecasts were issued for 72% or 9,982 firms. The descriptive statistics by
year show that the proportion of LTG forecasts, analysts and firms has remained fairly
stable over the sample period.
        These descriptive statistics confirm the findings of Asquith, Mikhail and Au
(2005) in that not all analysts issue LTG forecasts and not all firms’ long-term growth
prospects are forecast by analysts. While the primary focus of this study is o n the
relationship between analyst characteristics and the issuance of LTG forecasts, we
cannot ignore the influence of industry and firm- level factors in the decision to issue
LTG forecasts. We first examine the prevalence of LTG forecasts across industries.
Table 2 presents the number and proportion of firms in each of the 48 industry groups
classified by Fama and French (1997). The need to match and categorize firms by
industry reduces the sample to 11,663 firms. The proportions of companies that have
LTG forecasts ranges from 57 percent in the non- metallic mines industry to 100
percent in the defence industry. Across the 48 industry groups the average is 80
percent and the standard deviation is 9 percent. This suggests there is minimal
industry bias in the proportion of firms receiving LTG forecasts. Firm- level factors
are examined in the following section.




2
  Long Term Growth estimates are in percentage terms. They represent the annual growth rate, in
percent, expected by analysts over the next few (3-5) years for the company
http://forum.wharton.upenn.edu/forums/index.php?showtopic=161&hl=long+term+growth .


5
2.2 Empirical Model
This study proposes two explanations for why some analysts choose to issue LTG
forecasts and others do not. First, analysts issue LTG forecasts to signal their superior
quality. Second, analysts issue LTG forecasts to establish and maintain other business
relationships with firms. To quantify analyst quality we follow the work of Barron et
al. (2002), Clement (1999) and Mikhail et al. (1997) and use a measure of analyst
ability (short term earnings forecasting accuracy), two measures of analyst experience
(general experience and firm-specific experience) and a measure of private
information. We predict that analysts with higher ability, more experience and more
private information are more likely to issue LTG forecasts to signal their superior
quality. To quantify analysts’ incentives, we utilise the Carter-Manaster underwriter
reputation index to identify if the brokerage the analyst is affiliated with also conducts
other business with firms (e.g. underwriting). We predict that analysts are more likely
to issue LTG forecasts if they are affiliated with a brokerage that has other business
relationships with firms.
        As the decision to issue LTG forecasts could be related to other analyst and
firm- level variables, we also include the following control variables. Following
Clement (1999) we include variables to control for the number of companies that an
analyst follows and the complexity of each analyst’s portfolio. Ana lysts that follow
more companies and forecast across more industries are expected to have less time to
devote to LTG forecasts. Consistent with Michaely and Womack (1999) we also
include brokerage size, which proxies for the resources available to analysts. Analysts
that have more resources are expected to issue more LTG forecasts. At the firm level,
previous research suggests that demand for analyst research increases with firm size,
growth opportunities and intangible assets (Lang and Lundholm 1996; Barth et al.,
2001; Frankel et al., 2006). However, analysts find it difficult to forecast firms with
higher earnings volatility. We therefore predict that the issuance of LTG forecasts is
positively related to firm size, growth and intangibles and negatively related to
earnings volatility.

        The full model and variable definitions are provided below. We use a logistic
regression to relate the likelihood of a LTG forecast being issued by analyst i, for firm
j in year t.




6
Prob[ LTG _ FORi , j ,t  1]  logit(  0
     n
   i , j ,t ,r Quality( 1 ABILITY   2 GEXP  3 FEXP   4 PRIVATE)
    r 1
     n
   i , j ,t ,r Incentives( 1 AFFIL)
    r 1
     n
  i , j ,t ,r Controls(1 NCOS   2 NSICS 3 BSIZE   4 EVOL   4 MV   4 M / B   4 INTAN)
    r 1

  i , j ,t


where:
LTG_FOR           is an indicator variable set to one if the analyst i issues a LTG forecasts
                  for firm j in year t and zero otherwise.
ABILITY           measures analyst’s ability, proxied by analyst short-term earnings
                  forecast accuracy, calculated as absolute one-year-ahead earnings
                  forecast error: AFEijt  Actual jt  Forecastijt where AFEijt is analyst

                  i’s absolute forecast error for firm j in year t, Actualjt is the actual
                  earnings per share for firm j in year t, and Forecast ijt is the most recent
                  forecast issued by analyst i prior the 30th June in year t. The measure is
                  adjusted to for differences in information environments across firms
                  following           Clement      (1999):           MAFEijt=         [(AFEijt-
                  Avg(AFEit))/Avg(AFEit]*-1 where Avg(AFEit) is the average absolute
                  error of all analysts covering the firm.
GEXP              measures analyst’s general experience, calculated as the number of
                  years for which analyst i supplied at least one forecast during year t.
FEXP              measures analyst’s firm-specific experience, calculated as the number
                  of years for which analyst i supplied at least one forecast during year t
                  for firm j.
PRIVATE           measures of the precision of idiosyncratic information, s, and measures
                  the extent to which analysts rely on private or idiosyncratic
                                                                       D
                  information; calculated as follows s 
                                                             [(1  1/ N ) D  SE ]2
                  where SE is the squared error of the consensus mean forecast (EPS actual
                  – EPSconsensus)2 , D is the dispersion among the forecasts (STDEV2
                  where STDEV is the standard deviation of I/B/E/S estimates) and N is



7
                the number of analysts making forecasts. We use one- year-ahead
                earnings forecasts made on the day of the first annual earnings
                announcement on Compustat and maximum 29 days thereafter (Barron
                et al. 1998)
AFFIL           measures analyst incentives, calculated as a dummy variable set to 1 if
                analyst i is affiliated with a brokerage that also conducts underwriting
                business (i.e. has a Carter-Manaster underwriter reputation index
                greater than zero) and set to 0 if analyst i is affiliated with a brokerage
                that has no Carter-Manaster index. 3
NCOS            measures analyst workload, calculated as the number of firms for
                which analyst i supplied at least one forecast during year t.
NSICS           measures analyst portfolio complexity, calculated as the number of
                four-digit SIC industry groups for which analyst i supplied at least one
                forecast during year t.
BSIZE           measures analyst resources, calculated as the number of analysts that
                work for the brokerage house analyst i works for during the year t.
EVOL            is the earnings volatility of the firm measured as the coefficient of
                variation of the firm’s earnings/price ratio, which is calculated as
                [standard deviation of E/P / mean of E/P]. E/P is defined as earnings
                per share before extraordinary items scaled by beginning stock price.
                The coefficient of variation of the firm’s E/P ratio uses information
                prior to the year the LTG forecast is issued and is based on a minimum
                (maximum) of 5 (10) annual E/P observations.
MV              is the natural logarithm of the firm’s market value of equity prior to the
                year of the analyst’s LTG forecast (Compustat#25*24).
M/B             is the firm’s market-to-book ratio in year t (Compustat#24*#25/#60).
INTAN           is the firm’s book value of balance sheet intangibles (including good-
                will), scaled by total assets, prior to the year of the analyst’s LTG
                forecast (Compustat #33/#6).


        As the model above includes the matching of data from Compustat and CRSP
and the calculation of additional analyst variables, the following analysis is conducted

3
  We thank Jay Ritter for making the modified Carter -Manaster ranks available on his website
http://bear.cba.ufl.edu/ritter/ipodata.htm.


8
on a sample of the population of 531,258 unique analyst-firm- year forecasts described
in Table 1. The sample comprises 198,897 observations once the control variables are
included and 181,437 with the inclusion of the private information variable. In all
cases, the proportion of LTG forecasts and analysts and firms with LTG forecasts are
comparable to those reported in Table 1. For example, the final sample of 181,437
forecasts is comprised 28 percent of LTG forecasts. The proportion of firms and
analysts with LTG forecasts are 86 percent and 40 percent respectively.


3. Results
3.1 Univariate Results
Table 3 presents descriptive statistics of the analyst quality, analyst incentive and
control variables partitioned on whether the analyst issued a LTG forecast. The far
right column shows the results of mean and median tests of the differences between
the two groups. The results of the analyst quality and incentive variables are
consistent with expectations, except for firm-specific experience. Analysts are more
likely to issue LTG forecasts if they have greater forecasting ability, more general
experience, more private information and if the brokerage they are affiliated with also
conducts underwriting business. In contrast to expectations, analysts are more likely
to issue LTG forecasts if they have less firm-specific experience. This result is
consistent with the Brown and Mohammed (2007) and may indicate that analysts that
are relatively newer to a firm issue LTG forecasts to seek attention from firm
management. The results for the control variables are generally as expected. Analysts
that follow more companies and forecast across more industries issue fewer LTG
forecasts. Analysts who work at larger brokerages are more likely to issue LTG
forecasts. Analysts are more likely to issue LTG forecasts for firms that are bigger
and have more intangible assets. The results for earnings volatility and market-to-
book (growth opportunities) are not significant across both tests.
        Table 4 reports Pearson (above the diagonal) and Spearman (below the
diagonal) correlation coefficients among all variables used in the analysis. Most
variables have modest correlations with each other. However, general experience and
firm-specific experience have a high correlation. This may be because they are both
measuring the same construct of forecasting experience or because the sample data is
left censored i.e. it is not possible to calculate analyst experience prior to the first year



9
of available data. To ensure this does not influence the results, the next section first
reports regression results on a variable-by-variable basis.


3.2 Logistic analysis
Table 5 reports the results of the logistic regression model measuring the likelihood of
analyst LTG forecasts. The regressions are estimated using pooled data and include
all possible observations. Regression (1) presents the result for the analyst ability
variable. The coefficient is positive and significant indicating that analysts that are
more accurate at forecasting short-term earnings are also more likely to issue LTG
forecasts. Regression (2) examines the analyst general experience variable. The
coefficient is positive and significant indicating that analysts that have more
forecasting experience are more likely to issue LTG forecasts. Regression (3) reports
the result for analyst firm-specific experience. The coefficient is negative and
significant indicating that analysts are less likely to issue LTG forecasts the longer
they follow a company. This confirms the results of the univariate analysis and
suggests that the firm-specific experience variable may be more than just a measure of
analyst quality. It may be measuring another type of analyst incentive in that analysts
that are newer to a firm may issue LTG forecasts to seek attention from firm
management. Regression (4) identifies the relationship between private information
and the likelihood of analysts issuing LTG forecasts. The coefficient is positive and
significant, which suggests that analysts who have more private information are more
likely to issue LTG forecasts.
         Regression (5) examines the relationship between ana lyst incentives and the
issuance of LTG forecasts. The coefficient on AFFIL is positive and significant,
indicating that analysts affiliated with a brokerage that also provides underwriting
services to firms are more likely to issue LTG forecasts. This suggests that analysts in
these brokerages have more of an incentive to produce LTG forecasts in order to
establish and maintain other business relationships with firms. 4 To see if this
relationship is consistent before and after Regulation FD, we split the AFFIL
coefficient into AFFILpre and AFFILpost. If Regulation FD has had a material effect
on the incentive structure and business relationships between analysts and firms then
we expect the coefficient on AFFILpre to be positive and the coefficient on

4
  This finding is qualified by the same endogeneity problem as prior studies (e.g. Dechow et al., 2000)
as it does not explore causation between LTG forecasts and underwriting contracts.


10
AFFILpost to be insignificant. Regression (6) displays these results. The coefficient
on AFFILpre is significantly positive and the coefficient on AFFILpost is
insignificant, suggesting that after the introduction of Regulation FD in 2002, analysts
no longer have the same incentive to issue LTG forecasts.
       Regression (7) presents the results for the full model. The results are
consistent with the previous regressions. Analyst ability and private information are
positively related to the issuance of LTG forecasts. Analyst firm-specific experience
is negatively related to the issuance of LTG forecasts. The likelihood of issuing a
LTG forecast is higher for analysts affiliated to brokerages with underwriting business
only before Regulation FD. The only inconsistent result is that the coefficient on
analyst general experience is no longer significant. This is probably due to the high
correlation between analyst general and firm-specific experience. The coefficients on
the control variables are generally consistent with expectations. Negative relationships
are found between the number of companies followed and analyst portfolio
complexity and the issuance of LTG forecasts. Positive relationships are found
between brokerage size, firm size and intangible assets and the issuance of LTG
forecasts. The regressions were also run using the Fama-MacBeth (1973) annual
approach with consistent results.
       In summary, the results provide support for both explanations proposed in this
study. The issuance of LTG forecasts does provide a signal of analyst quality as we
find that analysts with greater forecasting ability, more general experience and more
private information issue more LTG forecasts. We also find that analysts affiliated
with a brokerage that also conducts underwriting b usiness are more likely to issue
LTG forecasts. However, this result is only significant before the introduction of
Regulation FD. In addition, we find that analysts are more likely to issue LTG
forecasts if they have less firm-specific experience. This may indicate that analysts
that are relatively newer to a firm issue LTG forecasts to seek attention from firm
management.


3.3 Further analysis
Prior research on analyst outputs (see Defond and Hung 2003) is vulnerable to the
issue of researchers not being able to observe these outputs if analysts choose not to
report them to I/B/E/S. To ensure the results of this study are not biased by the
forecasts reported to I/B/E/S, we conduct two checks. First we manually examine the


11
content of analysts’ research reports found on the Investext database for LTG-related
information (see for a similar approach Bradshaw 2002, Demirakos et al (2004).
Second, we reverse engineer LTG forecasts from price-to-earnings-to-LTG (PEG)
ratios for non-LTG forecasters.
        We randomly select one company that has both LTG forecasting and non-LTG
forecasting analysts. We then search for these analysts’ reports on Investext. We
obtain a total of 10 reports. Examining the content of these reports we find that
analysts that do not provide LTG data to I/B/E/S have no LTG data in their research
reports. Table 6 Panel A shows that none of the 5 reports from non-LTG forecasters
provide information about LTG prospects, whereas all 5 reports from LTG forecasters
do provide information. The reports from non-LTG forecasters also have less
information on growth trends at the firm and industry levels. Consistent with pr ior
research (e.g. Bradshaw 2002), the primary valuation model for both analyst groups
used to justify the stock recommendation is the P/E model. But LTG forecasters seem
to double check their valuation with other heuristics too (e.g. Price-to-Cash Flow).
Overall, based on the small sample of analyst reports collected, this study does not
suffer from bias due to unobserved LTG forecasts by non-forecasters.
        Alternatively the LTG forecast number can be reverse engineered from the
PEG ratio. The PEG ratio is equal to the price-to-earnings ratio divided by LTG. Prior
research has documented analysts use of the PEG ratio as a b asis for stock
recommendations (Asquith et al. 2005; Bradshaw 2004; Bradshaw 2002; Demirakos
et al. 2004). In addition, Bradshaw (2002) reports that a stock is fairly priced if it has
a PEG ratio of 1. Hence, if the analyst’s target price is used as the price, the PEG ratio
can be rearranged to solve for LTG. We therefore collect target price data for these
analysts from First Call and use one-year-ahead earnings forecasts made in the month
of the target price as expected earnings. Table 6 Panel B reports descriptive statistics
of the implied LTG rate. Non-LTG forecasting analysts have an implied LTG rate that
is 6.11% higher than LTG forecasting analysts. The difference between the means of
the two analyst groups is statistically significant.
        Next we construct a measure of the absolute LTG forecast accuracy of the two
groups of analysts, calculated as actual annualized growth in EPS for the past five
years minus forecasted LTG in absolute terms (Dechow et al. 2000). As LTG
forecasts and actual growth are already annualized and expressed in percentage terms



12
we do not scale the LTG forecast error further. 5 Analysts that are more (less) accurate
would thus have smaller (larger) absolute forecast errors. We find that non-LTG
forecasting analysts are on average 77.8% above the realized LTG forecast rate. This
is 67.7% more than LTG forecasting analysts. These find ings suggests that investors
would be better off following the investment recommendations of analysts providing
LTG forecasts, as their LTG predictions reflect more correctly the realized state. This
is important to investors as the best equity analysis looks at the long-term health of
companies rather than merely suggesting opportunities to trade the shares. Thus, the
results in Table 6 provide corroborating evidence for the logistic analysis results that
find that LTG forecasting analysts are of superior quality to non- LTG forecasting
analysts.


4. Summary
This study investigates why some analysts choose to issue LTG forecasts and some do
not. We propose two explanations. First, analysts issue LTG forecasts to signal their
superior quality. Second, analysts issue LTG forecasts to establish and maintain other
business relationships with firms. Using I/B/E/S data from 1994 to 2005, we find that
on average only 57 percent of analysts issue LTG forecasts. Consistent with the first
explanation we find that analysts that issue LTG forecasts have superior forecasting
ability, more general experience and more private information. Prior to Regulation
FD, we find that analysts were more likely to issue LTG forecasts if their brokerage
also provides underwriting services. After Regulation FD, this relationship is no
longer significant.
           These results suggest that LTG forecasts are a good indicator of analyst
quality. In other words, something as simple as the issuance of a LTG forecast can be
used as a signal that an analyst has superior forecasting skills, more general
forecasting experience and more private information than other analysts. This is
important to investors as the best equity analysis looks at the long-term health of
companies rather than merely suggesting shorter term earnings expectations and
trading opportunities. In addition, this research provides evidence to suggest that
Regulation FD has been effective in alleviating analysts’ incentives (or conflicts of



5
    If an analyst has mult iple LTG forecasts we take the average over the period.


13
interest). In particular, the issuance of LTG forecasts to establish and maintain other
business relationships (e.g. underwriting services) with firms.




14
References

Asquith, P., M. B. Mikhail, and A. S. Au 2005, 'Information content of equity analyst
reports', Journal of Financial Economics, 75 (2), 245-282.

Barron, O. E., D. Byard, C. Kile, and E. J. Riedl 2002, 'High- Technology Intangibles
and Analysts’ Forecasts.' Journal of Accounting Research, 40 (2), 289.

Barron, O. E., O. Kim, S. C. Lim, and D. E. Stevens 1998, 'Using analysts' forecasts
to measure properties of analysts' information environment.' Accounting Review, 73
(4), 421.

Barth, M. E., R. Kasznik, and M. F. McNichols 2001, 'Analyst Coverage and
Intangible Assets.' Journal of Accounting Research, 39 (1), 1.

Beaver, W. and D. Morse 1978, 'What Determines Price-Earnings Ratios?' Financial
Analysts Journal, 34 (4), 65-76.

Bradshaw, M. T. 2004, 'How Do Analysts Use Their Earnings Forecasts in
Generating Stock Recommendations?' Accounting Review, 79 (1), 25-50.

Bradshaw, M. T. 2002, 'The use of target prices to justify sell-side analysts' stock
recommendations', Accounting Horizons, 16 (1), 27.

Brown, L. and E. Mohammed, 2007, 'Is analyst earnings forecast ability firm-
specific?', SSRN working paper.

Chan, L. K. C., J. Karceski, and J. Lakonishok 2003, 'The Level and Persistence of
Growth Rates', The Journal of Finance, 58 (2), 643-684.

Clement, M. B. 1999, 'Analyst forecast accuracy: Do ability, resources, and portfolio
complexity matter?' Journal of Accounting and Economics, 27 (3), 285-303.

Cooper, R. A., T. E. Day, and C. M. Lewis 2001, 'Following the leader: a study of
individual analysts' earnings forecasts', Journal of Financial Economics, 61 (3), 383-
416.

Dechow, P. M., A. P. Hutton, and R. G. Sloan 2000, ' The relation between analysts'
forecasts of long-term earnings growth and stock price performance following equity
offerings', Contemporary Accounting Research, 17 (1), 1.

DeFond, M. L. and M. Hung 2003, 'An empirical analysis of analysts' cash flow
forecasts', Journal of Accounting and Economics, 35 (1), 73-100.

Demirakos, E. G., N. C. Strong, and M. Walker 2004, 'What Valuation Models Do
Analysts Use?' Accounting Horizons, 18 (4), 221-240.

Fama, E. F. and K. R. French 1997, 'Industry costs of equity.' Journal of Financial
Economics, 43 (2), 153-193.



15
Fama, E. F. and J. D. MacBeth 1973, 'Risk, Return, and Equilibrium: Empirical
Tests.' Journal of Political Economy, 81 (3), 607.

Frankel, R., S. P. Kothari, and J. Weber 2006, 'Determinants of the informativeness of
analyst research', Journal of Accounting and Economics, 41 (1-2), 29-54.

Frankel, R. and C. M. C. Lee 1998, 'Accounting valuation, market expectation, and
cross-sectional stock returns', Journal of Accounting and Economics, 25 (3), 283-319.

I/B/E/S 2000, 'The I/B/E/S glossary, a guide to understanding I/B/E/S terms and
conventions', I/B/E/S International Inc.

Lang, M. H. and R. J. Lundholm 1996, 'Corporate Disclosure Policy and Analyst
Behavior.' Accounting Review, 71 (4), 467-492.

Michaely, R. and K. Womack 1999, 'Conflict of interest and the credibility of
underwriter analyst recommendations', Review of Financial Studies, 12 (4), 653-686.

Mikhail, M. B., B. R. Walther, and R. H. Willis 1997, 'Do security analysts improve
their performance with experience?' Journal of Accounting Research, 35 (3), 131.

Ohlson, J. A. and B. E. Juettner-Nauroth 2005, 'Expected EPS and EPS Growth as
Determinantsof Value', Review of Accounting Studies, 10 (2 - 3), 349-365.

Penman, S. H. 1996, 'The Articulation of Price- Earnings Ratios and Market-to-Book
Ratios and the Evaluation of Growth.' Journal of Accounting Research, 34 (2), 235-
259.




16
Table 1
Descriptive analysis of all US firms with earnings forecasts contained in the I/B/E/S Detail Histo ry database from 1994 through 2005
                                             Analyst-                                                          Observations per year
                                   Total     firm-year     1994      1995        1996       1997        1998      1999      2000      2001     2002      2003      2004      2005
No. forecasts on I/B/E/S
 1 No. LTG forecasts             233,676      157,512     13,954    15,534      18,966     21,821      21,375    19,336    15,407    20,770   20,959    21,383    22,238    21,933
 2 No. EA RN forecasts          1,717,136     373,746    113,280 124,228 132,013 138,716 150,650 147,876 140,808 145,178                      137,465   147,075   166,228   173,619
 3 Total no. of forecasts       1,950,812     531,258    127,234 139,762 150,979 160,537 172,025 167,212 156,215 165,948                      158,424   168,458   188,466   195,552
 4 % LTG of total                  12%          30%        11%       11%         13%        14%         12%        12%      10%       13%      13%       13%       12%       11%
No. analysts on I/B/E/S
5 No. of unique LTG
    analysts                      7,660        7,660      1,461      1,508       1,847     2,130       2,158      2,260     2,159    2,368    2,516     2,456     2,383     2,330
6 No. of unique EA RN
    analysts                      5,725        5,725      1,409      1,625       1,659     1,838       2,209      2,266     2,474    2,198    2,213     2,244     2,159     2,248
7 Total no. of unique
    analysts                      13,385       13,385     2,870      3,133       3,506     3,968       4,367      4,526     4,633    4,566    4,729     4,700     4,542     4,578
8 % total of LTG analysts          57%          57%        51%       48%         53%        54%         49%        50%      47%       52%     53%        52%      52%       51%
No. of firms on I/B/E/S
 9 Firms with LTG                 9,982        9,982      4,607      5,105       5,835     6,163       6,108      6,033     5,624    4,980    4,747     4,665     4,859     4,835
10 Firms without LTG              3,796        3,796       877        851         985      1,208       1,194       948       821       489     451       558       813      1,092
11 Total no. of firms             13,778       13,778     5,484      5,956       6,820     7,371       7,302      6,981     6,445    5,469    5,198     5,223     5,672     5,927
12 % LTG of total                  72%          72%        84%       86%         86%        84%         84%        86%      87%       91%     91%        89%      86%       82%




17
Table 2
Distribution of number and proportion of firms with LTG forecasts by industry groups
Fama-French (1997) Industry                        Nu mber of firms              Proportion of firms
Classifications                              Total       with LTG forecasts      with LTG forecasts
Agriculture                                   24                    22                  92%
Aircraft                                      33                    32                  97%
Alcoholic Beverages                           35                    34                  97%
Apparel                                       81                    70                  86%
Automobiles and Trucks                        124                  111                  90%
Banking                                      1098                  765                  70%
Business Services                            1644                 1352                  82%
Business Services                             135                  103                  76%
Candy and Soda                                25                    24                  96%
Chemicals                                     153                  130                  85%
Coal                                          23                    22                  96%
Co mputers                                    364                  292                  80%
Construction                                  128                  107                  84%
Construction Materials                        165                  138                  84%
Consumer Goods                                153                  121                  79%
Defence                                       18                    18                 100%
Drugs                                         511                  341                  67%
Electrical Equ ip ment                        286                  227                  79%
Electronic Equip ment                         547                  474                  87%
Entertain ment                                90                    77                  86%
Fabricated Products                           32                    26                  81%
Food Products                                 119                  102                  86%
Healthcare                                    261                  225                  86%
Insurance                                     341                  288                  84%
Machinery                                     258                  217                  84%
Medical Equip ment                            316                  256                  81%
Miscellaneous                                 73                    54                  74%
Nonmetallic M ines                            67                    38                  57%
Personal Services                             127                  109                  86%
Personal Services                             501                  431                  86%
Petroleu m and Natural Gas                    409                  305                  75%
Pharmaceutical Products                       171                  143                  84%
Precious Metals                               95                    47                  49%
Printing and Publishing                       103                   90                  87%
Real Estate                                   80                    61                  76%
Recreational Products                         84                    68                  81%
Restaurants, Hotel, Motel                     210                  192                  91%
Rubber and Plastic Products                   78                    59                  76%
Shipbuild ing, Railroad Eq                    19                    17                  89%
Shipping Containers                           36                    31                  86%
Steel Works, Etc.                             131                  109                  83%
Teleco mmunications                           474                  388                  82%
Textiles                                      50                    43                  86%
Tobacco Products                              15                    14                  93%
Trading                                      1032                  761                  74%
Transportation                                253                  212                  84%
Utilit ies                                    289                  263                  91%
Wholesale                                     402                  334                  83%
Total                                       11,663               9,343                  80%




18
Table 3
Descriptive statistics of variables partitioned by analysts with and without long-term growth forecasts
                                                                           Standard                          t-test p-value
Variable                                    n        Mean       Median     Deviation      Q1       Q3
                                                                                                          Wilco xon p-value
Analyst quality
ABILITY
with long-term growth forecasts        74,409      0.025      0.033          0.624      -0.11       0.18        <0.0001
without long-term growth forecasts     122,488 0.00004           0           0.466      -0.16       0.33        <0.0001
GEXP
with long-term growth forecasts        74,409       4.28       4.00           2.71       2.00       6.00        <0.0001
without long-term growth forecasts 122,488          4.11       3.00           2.68       2.00       6.00        <0.0001
FEXP
with long-term growth forecasts        74,409       2.53       2.00           1.93       1.00       3.00        <0.0001
without long-term growth forecasts 122,488          2.60       2.00           1.91       1.00       3.00        <0.0001
PRIVATE
with long-term growth forecasts        70,383      995.8       3.39        32375.15      0.15      49.02         0.0092
without long-term growth forecasts 113,512        649.05       2.74        22158.93      0.14      38.11        <0.0001
Analyst incentives
AFFIL
with long-term growth forecasts        74,409       0.38         1            0.44         0          1         <0.0001
without long-term growth forecasts 122,488          0.24         0            0.43         0          1         <0.0001
Control variables
NCOS
with long-term growth forecasts        74,409      17.55      15.00          12.36      11.00      21.00        <0.0001
without long-term growth forecasts 122,488         19.02      16.00          15.23      11.00      23.00        <0.0001
NSICS
with long-term growth forecasts        74,409       5.15       5.00           3.32       3.00       7.00        <0.0001
without long-term growth forecasts 122,488          5.49       5.00           4.05       3.00       7.00        <0.0001
BSIZE
with long-term growth forecasts        74,409      66.31      48.00          65.00      22.00      88.00        <0.0001
without long-term growth forecasts 122,488         47.93      29.00          56.00      10.00      63.00        <0.0001
EVOL
with long-term growth forecasts        74,409      15.56       0.13         2177.84      0.00       0.35          0.734
without long-term growth forecasts 122,488         12.76       0.13         1985.40      0.00       0.37        <0.0001
MV
with long-term growth forecasts        74,409       7.60       7.54           1.87       6.30       8.88        <0.0001
without long-term growth forecasts 122,488          7.42       7.40           1.90       6.11       8.70        <0.0001
M/B
with long-term growth forecasts        74,409       3.95       2.60          29.68       1.67       4.37          0.913
without long-term growth forecasts 122,488          3.89       2.32          60.45       1.52       3.83        <0.0001
INTAN
with long-term growth forecasts        74,409       0.13       0.05           0.17       0.00       0.20        <0.0001
without long-term growth forecasts 122,488          0.11       0.04           0.15       0.00       0.15        <0.0001
The t-tests test the null hypothesis that the mean difference between observations with and without long -term earnings
 growth forecast is zero. The Wilco xon test, a non-parametric statistical method, tests the null hypothesis that the median
 difference between observations with and without long-term earnings growth forecasts is zero.




     19
Table 4
Correlation coefficients with two-tailed p-values (total observations = 181,437)
  Variable    LTG_FOR ABILITY GEXP                   FEXP PRIVATE            AFFIL    NCOS     NSICS    BSIZE    EVOL       MV       M/B    INTAN
LTG_FOR            1.00          0.05       0.02      -0.02       0.01        0.02     -0.05    -0.04     0.11     0.00     0.03     0.00     0.06
                               <.0001     <.0001 <.0001           0.01      <.0001    <.0001   <.0001   <.0001     0.85   <.0001     0.56   <.0001
 ABILITY           0.05          1.00       0.00       0.00       0.00        0.01     -0.01    -0.02     0.03     0.00     0.01     0.00     0.00
                <.0001                      0.18       0.64       0.61        0.02    <.0001   <.0001   <.0001     0.92     0.00     0.73     1.00
   GEXP            0.02          0.00       1.00       0.66      -0.01        0.06      0.04     0.03     0.19     0.00     0.10     0.00     0.10
                <.0001           0.09                <.0001     <.0001      <.0001    <.0001   <.0001   <.0001     0.28   <.0001     0.34   <.0001
   FEXP           -0.03          0.00       0.65       1.00      -0.01        0.07      0.05     0.00     0.16     0.00     0.16     0.00     0.02
                <.0001           0.10     <.0001                <.0001      <.0001    <.0001     0.56   <.0001     0.11   <.0001     0.53   <.0001
 PRIVATE           0.02          0.04      -0.06      -0.05       1.00        0.06      0.00    -0.01    -0.02     0.05     0.14     0.20     0.05
                <.0001         <.0001     <.0001 <.0001                      0.004      0.61     0.00   <.0001   <.0001   <.0001   <.0001   <.0001
  AFFIL            0.02          0.03       0.05       0.05       0.07        1.00     -0.04    -0.01     0.07    -0.01    -0.04    -0.01     0.01
                <.0001           0.00     <.0001 <.0001         <.0001                <.0001   <.0001   <.0001    <.001   <.0001   <.0001   <.0001
   NCOS           -0.05          0.00       0.16       0.15       0.00       -0.04      1.00     0.44    -0.06     0.02    -0.02     0.00    -0.11
                <.0001           0.55     <.0001 <.0001           0.94      <.0001             <.0001   <.0001   <.0001   <.0001     0.08   <.0001
  NSICS           -0.02         -0.01       0.08       0.03       0.01       -0.01      0.39     1.00    -0.10     0.00    -0.12     0.00     0.04
                <.0001         <.0001     <.0001 <.0001           0.00      <.0001    <.0001            <.0001     0.57   <.0001     1.00   <.0001
  BSIZE            0.16          0.04       0.20       0.15      -0.01        0.08      0.03    -0.12     1.00     0.00     0.13     0.00     0.07
                <.0001         <.0001     <.0001 <.0001           0.00      <.0001    <.0001   <.0001              0.27   <.0001     0.49   <.0001
   EVOL            0.03         -0.01      -0.08      -0.01       0.00       -0.02      0.05    -0.02    0.00      1.00     0.00     0.00     0.01
                <.0001           0.02     <.0001       0.02       0.95       <.001    <.0001   <.0001    0.37               0.46     0.89     0.04
    MV             0.03          0.04       0.10       0.15       0.00       -0.04     -0.02    -0.14    0.16      0.11     1.00     0.02     0.04
                <.0001         <.0001     <.0001 <.0001           0.06      <.0001    <.0001   <.0001   <.0001   <.0001            <.0001   <.0001
    M/B            0.07          0.01      -0.02      -0.03       0.00       -..001    -0.06     0.02    0.02      0.06    0.37      1.00     0.00
                <.0001         <.0001     <.0001 <.0001           0.73       <.001    <.0001   <.0001   <.0001   <.0001   <.0001              0.41
  INTAN            0.06          0.01       0.13       0.06       0.00        0.01     -0.14     0.10    0.09     -0.10    0.11     0.02      1.00
                <.0001           0.06     <.0001 <.0001           0.27      <.0001    <.0001   <.0001   <.0001   <.0001   <.0001   <.0001




20
Table 5
Logit analysis of the likelihood of analyst long-term growth forecasts
                              Regression (1)       Regression (2)        Regression (3)   Regression (4)   Regression (5)   Regression (6)   Regression (7)
      Variable     pred.          Analyst         Analyst General        Firm-specific       Private          Incentive      Regulatory            All
                   sign           Ability          Experience             Experience       Information        Structure     Environment        Variables
     intercept                  -0.418***           -0.734***              -0.698***       -1.1642***        -0.784***       -0.7873***        -0.453***
                                 <0.0001             <0.0001                <0.0001           0.001             0.001           0.001           <0.0001
     ABILITY         +           0.102***                                                                                                       0.046***
                                 <0.0001                                                                                                        <0.0001
      GEXP           +                               0.102***                                                                                     0.04
                                                     <0.0001                                                                                      0.127
       FEXP          +                                                     -0.048***                                                           -0.063***
                                                                            <0.0001                                                             <0.0001
     PRIVATE         +                                                                     4.78E-7***                                          4.07E-6**
                                                                                              0.003                                               0.05
      AFFIL          +                                                                                       0.075***
                                                                                                             <0.0001
     AFFILpre        +                                                                                                         0.249***         0.225***
                                                                                                                               <0.0001          <0.0001
     AFFILpost       0                                                                                                           0.018           -0.025
                                                                                                                                 0.134            0.124
      NCOS           -          -0.004***            -0.005***             -0.004***        -0.001***        -0.005***        -0.005***        -0.003***
                                 <0.0001              <0.0001               <0.0001          <0.0001          <0.0001          <0.0001          <0.0001
      NSICS          -          -0.017***            -0.014***             -0.001***         -0.02***          -0.014         -0.013***        -0.015***
                                 <0.0001              <0.0001               <0.0001          <0.0001          <0.0001          <0.0001          <0.0001
      BSIZE          +           0.010***             0.003***              0.004***         0.003***         0.003***         0.004***         0.003***
                                 <0.0001              <0.0001               <0.0001          <0.0001          <0.0001          <0.0001          <0.0001
      EVOL           -            -0.004             -0.003***               -0.002            -0.002          -0.003            0.003          -0.00001
                                   0.167                0.243                  0.28             0.249           0.223            0.246             0.25
        MV           +           0.023***             0.02***               0.003***         0.014***         0.019***         0.022***         0.037***
                                 <0.0001              <0.0001               <0.0001          <0.0001          <0.0001          <0.0001           0.0001
       M/B           +            -0.001               -0.004                -0.001            -0.002          -0.001           -0.001           0.0008
                                    0.38                0.063                -0.632           0.7183            0.616             0.63            0.396


21
      INTAN            +           0.508***              0.575***              0.574***            0.557***             0.561***             0.558***             0.52***
                                   <0.0001               <0.0001               <0.0001             <0.0001              <0.0001              <0.0001              <0.0001
     Pseudo-R2                        0.10                  0.11                  0.10                0.06                 0.14                 0.15                0.13
         n                          198,897              198,897               198,897              181,437             198,897              198,897              181,437

Long-term earnings growth indicator (LTG_FOR) is an indicator variable set to one if the analyst issues a LTG for firm i in year t and zero otherwise.
*** significantly different fro m zero at the 1% level, ** significantly d ifferent fro m zero at the 5% level, and * significantly different fro m zero at the 10% level.




22
Table 6
Content analysis and implied long-term growth rate
Panel A: Investext report
                           Non-LTG             LTG
                           forecaster       forecaster         All reports
Growth prospects
LTG                            0                 5                 5
M&A                            1                 2                 3
Industry trends                0                 2                 2
Position in Industry           1                 3                 4
Stock price data
Target price                   2                 4                  6
EPS 1                          5                 5                 10
EPS 2                          1                 4                  5
P/E                            5                 5                 10
Price/cash                     0                 3                  3
Div idend rate                 0                 3                  3
Income Statement
Revenue                        0                 3                 3
ROE                            2                 4                 6

Panel B: Implied long-term growth rate and accuracy
                   non-LTG forecaster        LTG forecaster                   Difference
                     Mean (Median)           Mean (Median)
Characteristic             [Std]                  [Std]           Mean          t-stat    Significance
                      24.96 (17.29)             18.85 (15)
LTG forecast             [191.43]                 [19.1]           6.11          3.99        <0.001
                      78.81 (25.49)            11.15 (8.01)
LTG accuracy             [355.57]                [27.55]          67.66          6.11        <0.001
This table summarizes the justifications citied by a sample of analysts separated into LTG forecasting
and non-LTG forecasting analysts. Panel A presents the distribution of the number of justifications
contained in each report. The reports are collected fro m Investext. Panel B reports the implied LTG
rate fro m non-LTG forecasting analysts derived by reverse engineering the PEG model and the
reported LTG on I/B/ E/S for forecasting analysts. The last row reports LTG forecast accuracy
measured as the realized LTG rate minus the forecasted/ imp lied LTG.




23

				
DOCUMENT INFO