Acciones digitales, cada vez más usadas en publicidad No es un secreto que las acciones digitales incrementan día con día y las marcas

4.3.3.1 The relationship of analysts recommendations and forecast management (H1)

The first hypothesis will be tested by running the regression of FM on AR, including moderators (Learning effect and Difficulty) and several control variables that the reason for using and measurement process of them are explained in the following paragraphs.

Prob(Down=1)= F( 0+ 1 AR+ 2AR×Difficulty + 3AR×FREQ+ 4 LMV+ 5 MB+ 6 Hightech

+ 7 Lag_Loss+ 8Year+ ) (4-13)

Where,

AR= the Analysts recommendations that takes the value of 1 to 5 (table 4-1) Difficulty to asses the credibility of managers forecasts.

FREQ Frequency of FM in the previous four years as index of learning effect

Logarithm of market value

Market to Book value

1 if the firm is in one of the high technology industries such as pharmaceuticals, aircraft and spacecraft, medical, precision and optical instruments, radio, television and communication equipment, office, accounting and computing machinery, or zero otherwise.

1 when a firm s quarterly earnings report preceding the forecast is negative and 0 otherwise.

= 1 if the firm-year is in 2010 and 0 otherwise.

Taking a page out of Rakow (2010), I converted LMV, MB and indicator variables that are set to one, if the value of the original variable is greater than or equal to the sample median, or zero otherwise.

Other control variable is the threat of litigation. Soffer et al. (2000) state that firms in a litigious environment want to prevent a large disappointment in the earnings announcement date (see Soffer et al. 2000), and this might be better accomplished by providing a less optimistic or even pessimistic forecast shortly before the earnings release date.

Kasznik and Lev (1995) posit that firms in high-tech industries face higher risk of litigation as they experience, larger price fluctuations, which might translate into potential losses to investors. Similarly, Baginski et al. (2002) uses high-tech industries to control potential firm- specific litigation risk. The earnings of high-tech firms are more volatile and inherently carry greater risks of inaccurate forecasts; all these factors could affect a firm s cost of capital. Therefore, a negative coefficient is predicted vis-à-vis high-tech, implying that high technology firms issue less optimistic forecasts.

Using dummy variables instead of continuous variable allows in equation (4-13) to be interpreted as the effect of independent variable when the dummy variable is equal to zero, while through can be interpreted as the effect of each variables when the dummy variable is equal to one.

For improving the robustness of the results, additional tests will be conducted, which includes tests such as ANOVA in order to compare difference of means of both buy and sell companies.

H2 is tested by running the following logit regression:

Prob(meet=1)= F( 0+ 1 AR+ 2 FREQ+ 3 Difficulty + 4 LMV+ 5 MB+ 6DA + 7 Hightech+ 8

Lag-loss+ 9 Year + ) (4-14)

Where,

is represented by variable, which equals 1 if a firm's actual earnings meets or exceeds the management s forecasts, or 0 otherwise.

is the firm's ability to manipulate earnings, as reflected by its discretionary accruals, which makes it ideal as a control. I use a version of the cross-sectional modified Jones model which is used by Bergstresser and Philippon (2006) to estimate discretionary accruals. Other independent and control variables are similar to what was explained for equation (4-13).

For testing H3 and H4, the ANOVA will be used to test the difference of mean value of FM between the companies that meet or miss forecasts in the buy and sell companies separately.

For testing H5 and H6, after removing (deducting) value of forecasts management from the issued forecast12, forecasts errors are calculated using nonbiased forecasts and reported forecasts. Then, chi-square test will be used to test the differences in the occurrences of positive FEs between FEs that are calculated by using unbiased forecasts (unmanaged FEs) and FEs that are calculated by using issued forecasts.

4.3.3.2 Analysis of Variance

One-way analysis of variance (ANOVA) is a technique used to compare means of two or more samples (Howell, 2012). The ANOVA tests the null hypothesis that samples in two or more groups are pulled from populations with the identical mean values. To do this, two estimates are got of the population variance. The ANOVA creates an F-statistic, the ratio of the variance calculated amongst the averages to the variance inside the samples. If the group means are pulled from populations with the same mean values, the variance among the group means would be lesser than the variance of the samples, following the central limit theorem. A upper ratio therefore suggests that the samples were drained from populations with dissimilar mean values (Howell, 2012). These estimates rely on various assumptions.

12

Following the methodology of (Gleason & Mills, 2008) the reported forecast is decomposed into the management and unmanaged component. The unmanaged forecast is defined as the difference between the reported forecast and the managed component (Felleg et al., 2012).

Response variable are normally distributed (or approximately normally distributed).

Samples are independent.

in order to obtain better comparison of forecasts management (H1) between buy and sell companies, the differences of means of forecasts management measure ( ), between buy and sell companies are tested in section 5.2.3.1.2.