The effect of earnings forecast precision on firm value and insider trading under voluntary disclosure in Taiwan
38
4.2 Return structure of forecast revision
In the empirical analysis of forecast revision on stock returns, I first investigate the Abnormal Return AR and Cumulative Abnormal Return CAR. Market adjusted return is used as the abnormal return. As is
documented in Figure 1, the abnormal returns are negative around the announcement date of the first downward revision. The cumulative abnormal return is about -7 during four days after announcement. However, the
abnormal returns are first positive and then turn negative later on after the announcement date of upward revision in Figure 2.
-8 -7
-6 -5
-4 -3
-2 -1
1 2
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 2
4 6
8 10 12 14 16 18 20
the date of the first downward revision re
tu rn ra
te
AR CAR
Figure 1 Abnormal Return AR and Cumulative Abnormal Return CAR around the announcement date of the first downward revision.
-3 -2
-1 1
2 3
4 5
6
-20 -18 -16 -14 -12 -10 -8 -6
-4 -2
2 4
6 8
10 12 14 16 18 20
the date of the upward revision ret
u rn
rat e
AR CAR
Figure 2 Abnormal Return AR and Cumulative Abnormal Return CAR around the announcement date of the upward revision.
4.3 Firm value and earnings forecast error
Listed companies in Taiwan are required to release quarterly financial statements, in addition to annual reports in April of each year. Semiannual reports are released in August. In order to test the model prediction that
forecast error under voluntary disclosure reduces firm value, I estimate the relationship between firm value and
The effect of earnings forecast precision on firm value and insider trading under voluntary disclosure in Taiwan
39
forecast error. Tobin’s Q is used to measure firm value. The market value of the stock is based on the market price at the end of the month following the release of the financial statements. Dynamic panel datasets play an
increasingly significant role in corporate finance research because researchers frequently require models that include lagged dependent variables as well as fixed effects to control for unobserved heterogeneity. This paper
specifies a simple dynamic model, equation 6, in which the firm value Q
it
depends on its past value Q
it-1
and on the absolute forecast error and the firm size. Hossain, et al. 1995 find a positive relationship between size and
the level of information disclosed. The following dynamic fixed effects model is estimated using quarterly panel data.
, 1
, 1 2
, 3
, ,
i t i
i t i t
i t i t
Q a
b Q b AFE
b Size ε
−
= + ⋅ + ⋅
+ ⋅ +
6 Where, Q denotes the Tobin’s Q ratio, AFE represents the absolute value of quarterly forecast error, Size is the
natural logarithms of firm market value and a
i
is a fixed effect. Table 2 lists the estimated results of equation 6. The estimated coefficient on AFE is negative at the 1 level, consistent with the prediction that there is a negative
relationship between firm value and earnings forecast error. Tobin’s Q is positively associated with the size of the company. The Sargan test J-statistic suggests that a dynamic specification of model for the sample of voluntary
disclosure firms is valid.
Table 2 GMM estimates of a Tobin’s Q model for dynamic panel data
Predicted sign Model 1 One-step
Model 2 Two-step Intercept
Coefficient -2.3274
-2.2596 p-Value
0.0000 0.0000 Q
t-1
Coefficient + 0.2828
0.2858 p-Value
0.0000 0.0000 AFE
Coefficient - -0.0408
-0.0417 p-Value
0.0005 0.0003 SIZE
Coefficient + 0.2012
0.1968 p-Value
0.0000 0.0000 Adjusted-R square
0.9880 0.9881
Sum squared resid. 22.5024
22.4561 J-statistic
1.2550 1.1377
Notes: Q denotes the measurement of the firm value, AFE represents the absolute value of quarterly forecast error, and Size is the log value of firm market value. Equation 6 is estimated by dynamic fixed effects model with instrumental variables, using the
following as instruments: lagged earnings forecast error, lagged firm size, ratio of net income to net sales, growth rate of revenue and industry. Industry is a dummy variable that expresses the sector of operation of the company. The received value =1 if the company is
in the electronics sector. The panel data consists of 117 cross-sections and 4 observations in a cross-section. Significance levels p-values of each independent variable are reported.
, , and
denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively.
4.4 Correlation between trading profit and forecast error in prior period
The stock returns around the announcement of forecast revisions are calculated to measure the potential trading profit of insiders.
4
For example, insiders can make profit by selling short prior to the announcement of a
4
For example, one of the return rate series around the announcement of downward revision among the sample firms as follows: 2.29, -0.81, -2.26, -0.21, 0, -0.63, -0.42, -1.28, 0, -4.31, -6.98, -6.78, -6.75, -2.79, -1.72, 0. Where -4.31 is the rate of return on the event
date. The return rate R during this period is -28.5. Insiders can make profit by selling short prior to the announcement of a downward revision in earnings forecast, and buy back after the price fall.