How accounting rule SFAS 123R change affects the
financial analysts’ projection?
1. Introduction ... 3
2. Literature Review ... 4
2.1 About the analysts’ forecast Literature ... 4
2.2 About SFAS 123R Literature ... 6
2.3 Literature contribution ... 7
3. Methodologies ... 7
3.1 Methodology set up ... 7
3.2 Interpretation of the 2SLS Estimation ... 11
4. Data ... 12
4.1. Match the database ... 12
4.2 Choose the variables and data description ... 13
5 Empirical results ... 18
5.1 First stage regression: Effect of Accounting rule SFAS 123R on accounting earnings . 18 5.2 empirical results: Main results on the second stage ... 23
6. Conclusion ... 27
1. Introduction
As the investors have less information than the analysts and it is hard for investors to get all the information about the stocks he wants to buy, most of the time the investors rely on the recommendation of analysts to make a decision whether to invest on the stocks. At the same time this recommendation affects the firm stock price and the ability to raise money in the financial market. In this way, it is important to understand how analysts predict the firm’s future earnings and how they form their recommendation.
Financial analysts’ recommendation is the expression of analysts’ belief about the stock value relative to the market price. In theory, intrinsic value of the stock price is the present value of future cash flow in the company. Market price should be around the intrinsic value. However, in accountants’ view, earnings also contain certain information about the stock price. Cash flow and earnings are not really same, but they often move together. When a firm suffers an accounting loss, its cash flows are probably decreasing. It is hard for us to distinguish how earnings and cash flow affect financial analysts’ projection separately.
Now, the accounting rule SFAS 123R offer an opportunity to prove how cash flows or accounting earnings affect financial analysts’ projection about stock price separately. SFAS 123R required all companies to expense the fair value of stock option. Companies have awarded the stock options for years, but they suddenly have to claim an accounting charge in the income statement, which will have a big influence on the firms’ accounting earning. But this accounting rule does not affect the cash flow of the
companies. In this way we can distinguish the cash flow effect from accounting earnings effects on the financial analysts’ projection. We can test how accounting earnings matters to analysts. We can prove how much accounting earnings will affect the analysts’ projection.
There are many literatures on the analysts’ forecast influence on the firm performance. Also much prior research has been done on the effect of SFAS 123R on the firm earnings and the incentive of how the managers and analysts’ take measure to deal with the accounting rule. Now we try to link the both side and explain how the change of accounting rule affect the analysts’ forecast precision and further we can discuss how the accounting earnings affect the projection quality and company performance.
The rest parts of the paper will structure as follows. Section 2 presents the literature review. Section 3 describes the data collection, data summary and simple background for SFAS 123R. Section 4 shows the methodology: how the SFAS 123R affects the firm performance. Section 5 presents the main results of regression and offers the explanation for this regression. Section 6 provides the robust test and Section 7 summarize the whole paper.
2. Literature Review
Jennifer Francis and Leonard Soffer (1997) try to prove that both analysts’ earnings forecast and their stock recommendations contain relative information but they do not include each other. They predict that the investors pay more attention on the earnings forecast with favorable recommendation than the revisions with unfavorable recommendations because the analysts’ working environment also encourages analysts to issue good information about the client firm. Jeffery S. Abarbanell and Victor L. Bernard (1992) present investors’ underreactions in the forecast earning can explain half of the magnitude for post earnings announcement drift while the overreaction forecast have no effect on the stock price overreaction. Terence Lim (2002) explains that financial analysts should trade off bias to get more precise forecast. A rational attribute of optimal earning forecast should include positive and predictable bias. However, if we use prior classical model to eliminate bias, it will prematurely dismiss analysts’ forecast and make the forecast inaccurate. Fried D. and Givoly D. (1982) indicate that the analysts’ forecast includes information and can be a better substitute for market expectation than forecasts generated by time-series model. They also identify that the broadness of information set and information reliance contribute to the accurate of analysts’ forecast. Jegadeesh N. and Kim W (2010) support that investors will recognize analysts tend to herd, especially when analysts from larger brokerages or analysts tracing stocks with small divergence. Park C. W. and Stice, E. K. (2000) empirically prove that after the superior analysts forecast one firm’s earnings, their subsequent forecast announcement have larger influence than the forecasts made by other analysts. Also this impact cannot spill over among the firms which the superior analysts follow. Gleason C. A. and Lee C. M. (2003) find that the process for adjusting stock price is faster and more complete in a firm with greater coverage analysis and a lot of delayed price
adjustment follows the subsequent forecast revision. The later news functions as catalysts in the price discovery process.
2.2 About SFAS 123R Literature
Aboody, David, Mary E. Barth, and Ron Kasznik (2004) present that the private incentive for executives, the level of asymmetry information, the extent to which the firm participate in capital markets and political cost are the four reasons why some firms recognize the stock-based compensation expense voluntarily. Barth M. E., Gow, I. D. and Taylor D. J. (2012) examine the incentive why two key market participants- managers and analysts-exclude the stock compensation expense from pro forma and street earnings. Managers do this because they want to increase earnings or smooth earnings or just want to meet earning benchmarks, which we regard as “opportunism.” However, the reason why analysts exclude the expense is they want to predict the firm performance more precisely. The authors give this behavior “predictive ability” explanation. Choudhary, Preeti, Shivaram Rajgopal, and Mohan Venkatachalam (2009) show that when the company anticipate the FAS 123R has great effect on the earning afterwards or when agency problem exists in the company, the executives prefer to accelerating vest of the employee stock option. Then this behavior will lead the stock price fall. Tomislav Ladika and Zacharias Sautner (2013) conclude that as FAS 123R cut the incentive horizon and then lead the managers decrease the investment for company.
2.3 Literature contribution
Most of the literature just talks about how the analysts’ projection affects the company performance. The literature gives us details how the projection affects stock price. For the literature about the FAS 123R, the authors want to find the motivation for the managers or the analysts accelerate the option expense execution. The authors also find when the corporate governance is poor, the companies are affected by FAS 123R largely. Still now, no literature has tried to distinguish the cash flow and the accounting earnings’ effect on the analysts’ projection quality. Now, the accounting rule FAS 123R offers an opportunity to separate the two effects. So this article focuses on the analysts’ projection quality- separating the effects of cash flow and the accounting earnings.
3. Methodologies
3.1 Methodology set upWe want to find out how the financial analysts form their forecast and evaluate the precision of their forecast. The quality of the analysts’ forecast will affect the companies’ performance or the ability to raise money in the financial market. Now we want to examine how exactly accounting earnings affect the financial analysts’ projection.
Financial analysts usually base their estimation on the accounting reports the firms released. They estimate the stock price by using the accounting earnings or using
the annual reports items to forecast the future cash flow the firm will generate. In this way, the accounting item net income includes both accounting component and financial component.
The basic model we want to use to test the hypothesis that accounting earnings affects the projections of the financial analysts is shown as below:
, 1 , 2 , 1 ,
' f t * f t * f t t f f t
financialanalysts forecast =β netincome +β x − + +m l n+ (1) Where financialanalysts forecast' f t, is a measure of financial analyst’s projection and netincomef t, is the accounting earnings extracted from the annual reports. xf t, 1− is the other firm characteristics while mt and lf are year and firm fixed effects. The other firm characteristics include firm size, stock return, book to market ratio and option granted.
Then we want to dig further to find out how different components affect financial analysts’ projection. Usually, it is hard to separate the effect of cash flows from the effect of accounting earnings on the financial analysts’ projection. The factors that increase the firm accounting earnings always increase the cash flow of the firm. When we regress directly on net income or EBITDA, we could not distinguish the effects of accounting and cash flow effects. Also, the omitted variables affect the net income and financial analysts’ forecast simultaneously. If any of the omitted variables are empirically unobservable, the coefficient of net income will be not precise and will mislead us to conclude the results incorrectly. So we introduce an exogenous variable FAS 123R. The coefficient for net income should be positive.
situation. The accounting rule SFAS 123R offer an opportunity to prove how cash flows and accounting earnings affect financial analysts’ projection about stock price separately and this is an exogenous variable that only regulated by the accounting policies. SFAS 123R required all companies to expense the fair value of stock option since June 15, 2005. In most cases, companies would be awarded the stock options for years before and would execute the option after the forbidden years. Executing the accounting rule FAS 123R suddenly push the firms to have to claim an accounting charge in the income statement. As when the firms expense their option, it would decrease their net income largely, especially for the new economy firms which grant a lot of option to their employees. Expensing the options does affect the accounting earnings for the companies but does not affect the cash flow in the companies. In this way we can distinguish the cash flow effect from accounting earnings effect on the financial analysts’ projection. We can test how accounting earnings matters to analysts. Another advantage for introducing the accounting rule is that accounting rule will affect the net income directly will have no influence on the financial analysts’ projection. Then we could make it clear for the causal effect between the dependent variable “analysts’ forecast” and independent variable “net income”. If we do not introduce the instrument variable, the dependent variable and independent variable interact with each other. For example, high net income shown on the annual report will initiate the financial analysts to give higher recommendation while at the same time the higher recommendation from analysts could push up the stock price and then promote the firm net income in certain extent. However, the instrument variable solves the
reversed causality.
Accounting rule SFAS 123R also create a control group based on the compliance date across the firms. This variation causes some firms to take actions in 2005 while other firms may not. Then we modify the regression into the instrument regression by using the instrument variable “fas123r_take_effect” which equals 1 for firms whose fiscal year ending after May. Otherwise this dummy variable equals 0. We test our hypothesis using the two stage least square regression.
, 1
*
123 _
_
2*
, ,f t f t t f f t
netincome
=
π
lfas
r take effect
+
π
x
+ + +
m l
u
(1st stage) , 1 , 2 , ,
' f t * f t * f t t f f t
financialanalysts forecast =γ netincome +γ x + + +m l n (Reduced form) Where the netincome f t, is the fitted value from the first stage model.
, f t
netincome represents the accounting earnings such as net income or EBITDA.
123 _ _
lfas r take effectrepresents the lag of fas123 _r take effect_ . The coefficient we are interested in this model isπ and1 γ . A negative value for the coefficient of 1 π1 would indicate that the accounting rule SFAS 123R forces the firms to expense their option and then low down the firm’s net income on the annual report. A positive value for the coefficient of γ1 indicates that high net income often lead the financial analyst to give a higher recommendation for this firm’s stock.
If the instrument variable“fas123r_take_effect” is valid, it must follow two key assumption (See Stocks and Watson[2011, p.434]):
1. Instrument relevance: π1≠0. It means that instrument variable relates to the net income when we control other firm characteristicsxf t, .
means that cov(fas123r_take_effect, uf t, )=0.
In addition to discussing the accounting rule effect in year 2006, we also want to dig further to find out whether the accounting rule has effects before and after the execution point in a period. We choose the period 2005 to 2008. We modify the instrument variable to“fas123r_effective” which is equal to 1 when the time is June 2005 after. It equals to 0 if the time is before the June 2005. Then we get another two regression equation:
, 1
*
123 _
2*
, ,f t f t t f f t
netincome
=
π
lfas
r effective
+
π
x
+ + +
m l
u
(1st stage) , 1 , 2 , ,
' f t * f t * f t t f f t
financialanalysts forecast =γ netincome +γ x + + +m l n (Reduced form) Where the lfas123 _r effectiveis the lag of fas123 _r effective. Other variables keep the same meaning as above.
3.2 Interpretation of the 2SLS Estimation
From the first stage, we could indicate whether the accounting rule does affect the net income reported on the annual report by using the coefficientπ1. If the coefficient
1
π is significant, it means that the accounting rule has influence on the firm income when the rule takes effects and vice versa.
From the second stage, we pay attention to the coefficientγ1. If coefficient γ1 is significant, it means that the change of net income influences the analysts’ recommendation and vice versa.
4. Data
4.1. Match the database
The data are mainly obtained from two databases IBES and Compustat. We match the two databases by using the intermediate database CRSP. Firstly, we use the SAS program to match the IBES identifier to CRSP. When matching the databases, we prefer the identifier CUSIP, which is a unique identifier for all the firms even they are in the different databases. For database IBES we use the identifier CUSIP while for the database CRSP we use the identifier NCUSIP which is analogous to IBES’ CUSIP. After matching, we drop 37221 firms that cannot be matched between IBES and CRSP. Because CUSIP can change overtime with changes of company name, operations, or traded issues. We manually compare the name for all the firms in the two databases and got 876 firms name are exactly the same with each other while 6845 firms names are not exactly the same but they just the reasonable same company. For example, one firm is called 1st Century Bank while in the CRSP it is just called the first Century Bank INC. In this way we could hold 7721 firms in total which would offer us efficient base to begin the empirical research.
Secondly, after matching the IBES and CRSP, we continue match the CRSP and Compustat and finally match IBES firms’ CUSIP identifier to Compustat GVKEY. During the matching process, we meet that firms in the IBES database may match two CUSIP numbers in IBES. Often these will be firms that did mergers or changed their name. For example, GVKEY number 1164 matches two firms because MCI INC
changed to World.com. When meet such cases, we would search on the Google or Wikipedia to make sure whether the different firm names represent the same firm or not. At last we get there are 6741 firms match between the IBES and Compustat.
4.2 Choose the variables and data description
We get the data of financial analysts’ forecast through the IBES and use the proxy variables standard deviation from different analysts, mean recommendation from different analysts, buy percent, sell percent and the number of recommendation for the stocks. Table I shows summary statistics for the different proxy variables from different analysts.
Table I shows that the recommendation from the financial analysts are quite stable from 2004 to 2008. The mean for the standard deviation is 0.6557 for 2004 and just increased 0.0065 in year 2005. The mean for analysts’ recommendation is 2.3446 and almost no change when the accounting rule SFAS 123R took effect in 2005. The same situation happens to buy percent, sell percent and number of recommendation. All the mean and standard deviation for all five variables seem no obvious change during the period 2004 to 2008.
Table I
Summary Statistics of Financial analysts’ recommendation
This table shows that the mean and standard deviation for the financial analysts’ recommendation. Stdev is the standard deviation for all the financial analysts estimating the stock price. Meanrec is the mean for all the financial analysts’ recommendation of stock price. Buypct is the percentage which financial analysts’ recommending buying occupies the entire analysts offer. Sellpct is the percentage which financial analysts’ recommending selling occupies the entire analysts offer. Numrec is the number of recommendation the financial analysts’ give.
Year N
stdev meanrec buypct sellpct numrec
Mean S.D Mean S.D Mean S.D Mean S.D Mean S.D
2004 2570 0.6557 0.4257 2.3464 0.627 48.1032 34.2307 6.2095 10.6319 7.0117 5.9455
2005 2562 0.6622 0.4219 2.3419 0.6124 48.1515 33.4600 5.6824 10.1352 7.1959 5.9905
2006 2458 0.65415 0.4158 2.3497 0.6073 48.0218 33.8724 5.6392 10.1349 7.4951 6.0554
2007 2299 0.6735 0.4067 2.3269 0.5875 48.8077 33.2597 4.9693 9.5277 7.5341 5.8296
2008 2144 0.6590 0.4171 2.3340 0.6053 48.1895 33.9467 5.4806 10.047 7.0601 5.5683
We get the data about the accounting items in the Balance Sheet from the database Compustat. We extract the data such as net income, EBITDA, stock price, total asset, market to book ratio and options granted during the period 2000 to 2012. Table II shows the summary statistics for the accounting items in the Balance Sheet.
Because most articles use net income and EBITDA as a measure of the accounting earnings for a firm, we just regard percent of net income and percent of EBITDA as the proxy variables. The mean for percent of net income from the 2004 to 2003 did not change too much but the standard deviation changed quite a lot. The mean for percent of EBITDA is 0.0673 in 2005 while in 2006 it decreased to 0.0604 and the standard deviation for the PEBITDA volatile quite heavily. For other variables, the standard deviation in the period 2005 and 2006 changed a lot at the same time.
Table II
Summary Statistics for firm characteristics
Table II give the summary statistics for the firm characteristics. PNI is the percentage of net income in the total asset. PEBITDA is the Earnings before Interest divide the total assets. We could get the relative value for net income and EBITDA for all the firms and decrease the size effects. MBV is the firm market to book ratio while the LMBV is the lag one period for market to book ratio. Market to book ratio is the firm market value divide the book value of the firm. At is the firm total asset and LAT is the lag one period for the total asset. Optgr is the firm total granted option which is accumulated by year. It measures how much option the firm own for the current year. As the optgr is quite large we use the logarithm for optgr. We extract the close stock price for the fiscal year on the database Compustat and use the close stock price to calculate the stock return. Then we calculate the lag one period for the stock return. The samples are the firms which exist both in the IBES and Compustat databases.
Year N
PNI PEBITDA LMBV LAT INOPTGR LStockreturn
Mean S.D Mean S.D Mean S.D Mean S.D Mean S.D Mean S.D
2004 2522 -0.01278 0.4148 0.0675 0.3304 1.5038 2.3747 5317.097 18540.96 -0.4876 1.6622 0.4279 0.6319
2005 2512 -0.01219 0.4466 0.0673 0.3447 1.6387 2.4457 5689.911 19413.42 -0.5503 1.6572 .02564 0.5200
2006 2398 -0.01663 0.4448 0.0604 0.2980 1.5320 1.9072 5861.865 19474.75 -0.7492 1.7219 0.0548 0.4306
2007 2243 -0.0096 0.3362 0.0571 0.34403 1.5610 2.1532 6563.853 21106.23 -.07270 1.6825 0.1438 0.4431
2008 2077 -0.06025 0.4931 0.0524 0.2975 1.4513 1.7546 6941.835 21265.58 -0.5534 1.6562 0.0279 0.4427
We take PNI and PEBITDA as the dependent variables for the first regression while take the LMBV, LAT, Lstockreturn and INOPTGR as the control variables. Because accounting rule SFAS 123R regulate how to expense the option, we include the option granted as the control variable. For other firm characteristics, we all take one period lag because accounting conditions last year would have effects on the accounting earnings this year.
Using the date which SFAS 123 regulates, we create the instrument variable “fas123r_take_effect” and “fas123r_effective”. In calendar year 2004, this variable equals to 0 for all the firms. In calendar year 2005, “fas123r_take_effect” equals 1 if
the firm fiscal year ends in June through December and equals 0 if the firm fiscal year ends in January through May. In calendar year 2006, the instrument variable “fas123r_take_effect” is vice versa. After calendar year 2006, “fas123r_take_effect” equals to 0 for all the firms again. “Fas123r_effective” functions as the measure to distinguish the period before and after the accounting rule took effects. “Fas123r_effective” equals to 1 after June 2005 and equals to 0 before June 2005 when the accounting rule executed for all firms in US. For this reason we separate our study period as 2006 and from 2005 to 2008.
Even though the accounting rules SFAS 123R took effect since June 2005, the accounting items was reported in 2006. So in fact we use the accounting items in the year 2006 while the instrument variable we use is 2005. In order to match all the variables, we use the lag one period for the “fas123r_take_effect” to match the accounting items in 2006. We applied the same rule to the instrument variable “fas123r_effective”. It means if we want to test how accounting rule affect the accounting earnings we will just test in year 2006. However, if we want to examine whether accounting rules have different effects before or after the rules executed, we would like to test in the period 2005 to 2008. If the results are significant, it means that FAS 123-R remains in effect in 2007 and 2008 for all firms.
As the SFAS rules was issued by FASB, it just constraints the US firms or the firm which list in US. So we study two different subsamples to find out how these effects relate to each other. The first subsample is for all firms in US. We could use FIC to separate US firms from the whole sample. We generate the FIC equals to 1 if the
firm is a US firm. Not surprisingly, we pay more attention to the new economy firms which grant more option to their employees.
Specially, we define the new economy firms as firms competing in the computer, software, internet, telecommunications, or networking fields. (Anderson et al. (2000) and ILL ) We select the new economy firms with the SIC code and filter out the SIC code equal to 3570 (Computer and Office Equipment), 3571 (Electronic Computers),3572 (Computer Storage Devices), 3576(Computer Communication Equipment), 3577 (Computer Peripheral Equipment), 3661 (Telephone &Telegraph Apparatus), 3674 (Semiconductor and Related Devices), 4812 (Wireless Telecommunication),4813 (Telecommunication), 5045 (Computers and Software Wholesalers), 5961 (Electronic Mail-Order Houses), 7370 (Computer Programming, Data Processing), 7371 (Computer Programming Service), 7372(Prepackaged Software), and 7373 (Computer Integrated Systems Design). Old economy firms are firms with SIC codes less than 4000 not otherwise categorized as new economy (Murphy 2003 P.132 Note 5). Then we generate a new series called newfirm equal to 1 if the SIC is the same as the SIC code mentioned above.
Before we regress on all the variables, we must wisorize all the variables. Because we run the distribution for all the variables and find out that the smallest and largest values are thousands of times bigger than even the 99th percentile, which are most cases for ratio variables. To address this, winsorize all variables at the 1st and 99th percentile expect for stock return, which we winsorize by using 5th and 95th percentile
5 Empirical results
5.1 First stage regression: Effect of Accounting rule SFAS 123R on accounting earnings
From the Table II we could see that the regression results strongly support the conclusion that the accounting rule SFAS 123R do affect the accounting earnings for the firm. Table II shows that this result changes quite a lot when we add the control variables into the regression in 2006. This table presents estimates from cross-sectional regression for the calendar year 2006. The sample is the firms from Compustat and IBES firms. The specification control variables for firm characteristics are log of option granted, lag one period of market to book ratio, lag one period total asset and lag one period of stock return and all specifications include industry and all specifications include industry fixed effects. When we see the F statistics test for “fas123r_take_effect”, we find the values are all above ten which mean that the “fas123r_take_effect” is a strong instrument variable. The stock return is annual stock return by calculating using the close stock price for fiscal year.
The coefficient of “fas123r_take_effect” is negative which matches our expectation that when the accounting rules force the firms to expense their options, the accounting earnings of that firm will decrease. For Panel (A) no matter the dependent variable is percentage of net income or the percentage of EBITDA, the coefficients are all significant at the significance level 1%. For column (2) and column (4), we add control variables when we regress. The coefficients for the control variables are also significant at the significance level 1%. The coefficient of option granted is negative. It
indicates that the more option granted for the firm, the more will accounting earning change in 2006. The coefficient of stock return is positive. It shows that the stock return correlates with the accounting earnings. The higher the stock returns, the more net income will be next year. Perhaps when the stock return is higher, investors would rather buy this firm stock and believe this firm will perform better in the future. In this way, this firm acquires a higher reputation which benefits its business. Then the stock returns just promote the accounting earnings next year.
For the Panel (B), we expand the regression and mainly prove the relation for the new economy firms in US. The regression results are similar to the US firms. Usually, the new economy firms focus more on technology and release more option to encourage their employees. The options take a quite large part of the firm value. So the new economy receives more impact when the accounting rule SFAS 123R takes effect. The coefficient of lag one period “fas123r_take_effect” is significant. It also proves that the accounting rules SFAS 123R affects the accounting earnings. Not all the firms would accelerate to expense the option in the firms before the accounting rule take effects. Why this situation could happen. It contributes to our next discussion- would the change of accounting earnings affect the financial analysts’ projection?
Table III: Effect of SFAS 123R on accounting earnings: First stage results (Year 2006) This table examine whether accounting rule SFAS 123R would affect the accounting earnings when the accounting rule took effect in 2005. Even though the accounting rule became effective in June 2005, the reporting accounting items were exactly in 2006.In this way we just use the lag one period for fas123R_take_effect which is based on the fiscal year ending month for all the US firms in 2005 while other accounting items we still use is in 2006. In panel (A) the dependent variable is percent of net income which is equal to net income divide by total asset while in Panel (B) the dependent variable is the percent of EBITDA which is equal to EBITDA divided by total asset. Fas123R_take_effect is equal to 1 when the fiscal year ending month is after May and is equal to 0 if the fiscal year ending month is from January through May. LfAS123R_Take_Effect is one period lag for Fas123R_take_effect. Inoptgr_w is the log of the grant options for all the firms. Lat_w is the lag one period total asset. Lmbv_w is the lag one period market to book ratio. LstockReturn_w is the lag one period for stock return. Percentage of net income, lat, lmbv and inoptgr are all winsorized at 99% while the stockreturn is winsorized at 95%. In Panel (B) the sample is shown US firms and new economy firms. In Panel (A) the sample is extracted from US firms.Column (1) and (3) show the regression without control variables while Column (2) and (4) show the regression withcontrol variables
Panel A: Sample: US firms
Dependent variable: Percent of net income Percent of EBITDA Sample: US Firms Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) LfAS123R_Take_Effect -.07740 *** (-3.95) -0.10828*** (-3.42) -0.08879*** (-5.84) -0.1167*** (-4.49) inoptgr_w -0.0212*** (-2.70) -0.0132** (-2.54) Lat_w 1.50e-06*** (3.42) 1.02e-06*** (3.15) Lmbv_w -0.0708*** (-4.94) -0.0618*** (-5.27) LstockReturn_w 0.1733*** (3.73) 0.1242*** (2.80) Observations 4089 3134 3974 3073
Industry fixed effects YES YES YES YES
F-statistics (FAS
Panel B: Sample: US Firms & New Economy firms
Dependent variable: Percent of net income Percent of EBITDA Sample:
US Firms & New Economy firms
Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) LfAS123R_Take_Effect -0.11570*** (-3.91) -0.16780*** (-3.67) -0.0733*** (-3.87) -0.1152*** (-4.45) inoptgr_w -0.02616 (-1.37) -0.01141 (-1.12) Lat_w 2.11e-06 (1.51) 2.11e-06** (2.45) Lmbv_w -0.05598*** (-2.10) -0.04642** (-2.49) LstockReturn_w 0.2854*** (2.96) 0.1982*** (4.09) Observations 345 329 345 329
Industry fixed effects YES YES YES YES
F-statistics (FAS
123R_Take_effect) 15.42 14.3 14.87 18.9
Then we go further to expand the period to 2005 to 2008 to compare the effects before and after the accounting rule takes effect. The instrument we use is “fas123r_effective”. The coefficients of this variable are all significant at the significance level 5%. It means as the accounting rules SFAS 123R take effect, the firms’ accounting earnings change. There is quite a log difference between the before and the after effects. The coefficient is negative, which means the accounting rules have a negative effect on the accounting earnings. It indicates that not all the firms try to execute their options before the effective date of accounting rule to avoid expensing the options.
Table IV: Effect of SFAS 123R on accounting earnings: First stage results (2005-2008) This table shows how FAS 123R has effect on the accounting earnings during the period 2005 to 2008. FAS123R_effective is an instrument variable and it is equal to 1 since the calendar time is after June 2005 while it is equal to 0 when the calendar time is before June 2005. LfAS123R_Take_Effect is the lag one period for the variable fAS123R_Take_Effect. Other variables are similar to the variables in Table III.
Panel A: Sample: US firms
Dependent variable: Percent of net income Percent of EBITDA Sample: US Firms Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) LfAS123R_Take_Effect -0.0344*** (-3.37) -0.0237** (-3.05) -0.0219*** (-3.17) -0.0164** (-3.01) inoptgr_w 0.0086 (1.38) 0.0056 (1.29) Lat_w -5.37e-07 (-0.25) -7.72e-08 (-0.05) Lmbv_w 0.0198*** (6.27) 0.0135*** (6.09) LstockReturn_w 0.0379*** (3.24) 0.0068 (0.83) Observations 15845 12162 15393 11924
Industry fixed effects YES YES YES YES
F-statistics (FAS
Panel B: Sample: US Firms & New Economy firms
Dependent variable: Percent of net income Percent of EBITDA Sample:
US Firms & New Economy firms
Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) LfAS123R_Take_Effect -0.1157427*** (-3.91) -0.1678*** (-2.67) -.0733*** (-3.87) -0.1152*** (-3.45) inoptgr_w -0.0262 (-1.37) -0.0114099 (-1.12) Lat_w 2.11e-06 (1.51) 2.11e-06** (2.45) Lmbv_w -0.0560*** (-2.10) -0.0464** (-2.49) LstockReturn_w 0.2854468*** (2.96) 0.1982*** (4.09) Observations 2623 2242 2616 2242
Industry fixed effects YES YES YES YES
F-statistics (FAS
123R_Take_effect) 17.84 8.85 14.87 13.52
5.2 empirical results: Main results on the second stage
From the Table V we could see that the regression results do not strongly support the conclusion that the analysts’ recommendation is affected by the accounting earnings of the firm. Table V shows that this result changes quite a lot when we add the control variables into the regression in 2006. This table presents estimates from cross-sectional regression for the calendar year 2006. The sample is the firms from Compustat and IBES firms. The specification control variables for firm characteristics are log of option granted, lag one period of market to book ratio, lag one period total
asset and lag one period of stock return and all specifications include industry and all specifications include industry fixed effects. The stock return is annual stock return by calculating using the close stock price for fiscal year.
The coefficient of “net income” is positive which matches our expectation that when the net income increases, the financial analyst will recommendation it and the disagreement among all the analysts will decrease. For Panel (A) no matter the dependent variable is standard deviation or the mean of recommendation, the coefficients are all not significant even at the significance level 10%. For column (2) and column (4), we add control variables when we regress. The coefficients for the control variables are also not significant at the significance level 10%. It means that the change of net income because of the change of accounting rule does not affect the analysts’ projection of the firm stock.
For the Panel (B), we expand the regression and mainly prove the relation for the new economy firms in US. The regression results are similar to the US firms. Usually, the new economy firms focus more on technology and release more option to encourage their employees. The options take a quite large part of the firm value. So the new economy receives more impact when the accounting rule SFAS 123R takes effect. The coefficients of net income are not significant either. It means even in the new economy in US the effects are not obvious. We may want to ask why this situation would happen. When the firm decreases its net income, the financial analysts do not adjust their recommendation. The reason is that financial analysts know how the accounting rules SFAS 123R works and they think this only affect the net income for
just one time. The effects will not last long and in fact the business of the firms does not change. The decrease of the net income is because the firm has to expense its option. As the financial analysts have their expectation before they offer the recommendation, they do not adjust the recommendation even though the accounting earnings of the firm decrease. In this situation, the mean of recommendation stays almost the same and the standard deviation does not change a lot, which means the disagreement among the analysts does not increase. The financial analysts all give the similar expectation and recommendation for the US firms and new economy firms.
Table V: Effect of Analysts’ recommendation on accounting earnings: Second stage results (Year 2006)
This table shows how net income affects the financial analysts’ recommendation when we take FAS123R_take_effect as the instrument variable. Standard deviation represents the disagreement among all the financial analysts’ recommendation. The mean of recommendation represents the mean recommendation. Other variables are similar to the variables in Table III.
Panel A: Sample: US Firms
Dependent variable: Standard deviation Mean of recommendation Sample: US Firms Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) Pni_w 0.5416 (1.29) 0.2164 0.78 0.6746 -1.12 0.2642 -0.64 inoptgr_w 0.0590*** 8.92 -0.01175 -1.19 Lat_w 1.48e-07 0.29 2.15e-06*** 2.78 Lmbv_w -0.0016 -0.12 -0.0477** -2.37 LstockReturn_w 0.01883 0.35 -0.1788** -2.25 Observations 2311 1889 2311 1889
Industry fixed effects YES YES YES YES
F-statistics (FAS
Panel B: Sample: US Firms & New Economy firms
Dependent variable: Standard deviation Mean of recommendation Sample:
US Firms & New Economy firms
Without control variable With control variable Without control variable With control variable (1) (2) (3) (4) Pni_w 0.5307 0.90 0.3671 0.95 .8007 0.97 0.3042 0.66 inoptgr_w 0.0814*** 4.73 0.0318 1.49 Lat_w 2.56e-06 1.52 2.78e-08 0.01 Lmbv_w -0.0086 -0.65 -0.0460* -1.81 LstockReturn_w 0.0295 0.30 -0.1803 -1.50 Observations 306 281 306 281
Industry fixed effects YES YES YES YES
F-statistics (FAS
123R_Take_effect) 1.81 1.91 1.94 1.43
6. Conclusion
As we all know, there was quite a lot disagreement for the accounting rule SFAS 123R since 1972 and the accounting rule about the option changes all the time. Now most people have study how this accounting rule affects the firm performance or the investment etc. Most study focus on the firm itself. Seldom people study how the accounting rule SFAS 123R affect the analysts’ projection. In this paper, we discuss how the accounting rule SFAS 123R affect the analysts’ projection and we also
distinguish the different effect of accounting and finance aspects. From the empirical study, we find out that the accounting rule SFAS 123R does have a large influence on the firms’ accounting earnings while this accounting rule does not affect the financial analysts’ recommendation that much. The reason is that the financial analysts know how the accounting rules SFAS 123R works and they think this accounting rule only affects the net income for just one time. The effects will not last long and in fact the business of the firms does not change. The decrease of the net income is because the firm has to expense its option. As the financial analysts have their expectation before they offer the recommendation, they do not adjust the recommendation even though the accounting earnings of the firm decrease. In this situation, the mean of recommendation stays almost the same and the standard deviation does not change a lot, which means the disagreement among the analysts does not increase. The financial analysts all give the similar expectation and recommendation for the US firms and new economy firms.
This finding may lead to another discussion for financial analysts’ recommendation- Finance behavior, which we do not discuss in this paper to dig further. Through this empirical study, we know more about how financial analysts form their recommendation. Not all the change of accounting earnings would lead the financial analysts change their recommendation. This point is also proven in the article of Taylor etc. (2012) that the analysts would exclude the option expense from the earnings when exclusion increases earnings’ predictive ability for future performance but not for the reason of opportunism to smooth the profits of the firm. As the analysts
exclude the option expense, then the accounting rule would not affect the recommendation of the financial analysts in this way.
7. Reference
Abarbanell, J. S., and Bernard, V. L. 1992. Tests of analysts' overreaction/underreaction to earnings information as an explanation for anomalous stock price behavior. The Journal of Finance, 47(3), 1181-1207.
Aboody, David, Mary E. Barth, and Ron Kasznik, 2004, Firms’ voluntary recognition of stock-based compensation expense, Journal of Accounting and Economics 42, 123-150.
Barth, M. E., Gow, I. D., and Taylor, D. J. 2012. Why do pro forma and Street earnings not reflect changes in GAAP? Evidence from SFAS 123R. Review of Accounting Studies, 17(3), 526-562.
BANDYOPADHYAY, S. P.; L. D. BROWN; AND G. D. RICHARDSON. "Analysts' Use of Earnings Forecasts in Predicting Stock Returns: Forecast Horizon Effects." InternationalJournal of Forecasting (1995): 429- 45
Choudhary, Preeti, Shivaram Rajgopal, and Mohan Venkatachalam, 2009, Accelerated vesting of employee stock options in anticipation of FAS 123-R, Journal of Accounting Research 47, 105-146. D. Eric Hirst, Patrick E. Hopkins, and James M. Wahlen (2004) Fair Values, Income Measurement, and Bank Analysts' Risk and Valuation Judgments. The Accounting Review: April 2004, Vol. 79, No. 2, pp. 453-472.
Brown, L. D., & Y.-J. Lee. (2006). The impact of SFAS 123R on changes in option-based compensation. Working paper, Georgia State University.
Choudhary, P., Rajgopal, S., & Venkatachalam, M. (2009). Accelerated vesting of employee stock options in anticipation of FAS 123-R. Journal of Accounting Research, 47(1), 1–42.
Fried, D., and Givoly, D. 1982. Financial analysts' forecasts of earnings: A better surrogate for market expectations. Journal of Accounting and Economics, 4(2), 85-107.
Gleason, C. A., and Lee, C. M. 2003. Analyst forecast revisions and market price discovery. The Accounting Review, 78(1), 193-225.
Hirst, D. E., Koonce, L., and Simko, P. J. 1995. Investor reactions to financial analysts' research reports. Journal of Accounting Research, 33(2), 335-351.
Jegadeesh, N., and Kim, W. (2010). Do analysts herd? An analysis of recommendations and market reactions. Review of Financial Studies, 23(2), 901-937.
Landsman, W. R., Miller, B. L., & Yeh, S. (2007). Implications of components of income excluded from pro forma earnings for future profitability and equity valuation. Journal of Business Finance and Accounting, 34(3–4), 650–675.
Lim, T. (2001). Rationality and analysts' forecast bias.The Journal of Finance,56(1), 369-385.
Murphy, Kevin J. "Stock-based pay in new economy firms."Journal of Accounting and Economics34.1 (2003): 129-147.
Park, C. W., and Stice, E. K. 2000. Analyst forecasting ability and the stock price reaction to forecast revisions. Review of Accounting Studies, 5(3), 259-272.
