LA ESTRUCTURACION DE LAS BASES DE DATOS Y EL WINDOW DRESSING
EVENTOS DE IMPACTO EN LA ARMONIZACIÓN CONTABLE INTERNACIONAL
The second model relates to the determination of purely voluntary items. The definition of purely voluntary disclosure is items that are not explicitly required in any of the years covered in this study (see appendix C). However, every regulation contains a general statement that firms should disclose any extraordinary events affecting their operations. This means that some aspects of disclosure are not explicitly named, but are nonetheless required. This anomaly might be expected to affect the purity of DIVOLPUR. This shortcoming must remain, however, because the information available does not permit a more precise distinction of the mandatory- voluntary dichotomy. This calls for a separate study. The most refined delineation between mandatory and voluntary disclosures would be achieved via the use of a case study approach, focusing on just a few, highly selected firms.
Table 2 below gives additional support for the importance of market maturity in the development of disclosure. Despite the fact that the items in the index applied are not explicitly mandated during the research period, the influence of mandatory disclosure elements at a given time is very strong. A joint test, adding the yearly dichotomy variables to the model yielded F(7, 263) = 4.11, p=.000. The high significance of the test value indicates that the yearly dichotomy variables should be included in the model. The results confirm anecdotal evidence that interim reports have established their position in corporate communication. The high statistical significance of the F- test value, 10.02, for the whole model indicates that overall the model provides
Table 2. Regression results of all interim reports including purely voluntary disclosures only
Dependent variable: DIVOLPUR
Independent Parameter Standard t value variables estimate error standard White a
D85 -.110 .061 D86 -.058 .057 D87 -.044 .058 D88 -.007 .059 D89 .018 .059 D90 .006 .057 D91 .000 .056 D92 -.018 .058 LHOLNU .012 .005 2.321 2.363 FIRMS -.001 .000 -2.165 -2.303 SCGNETS .087 .047 1.869 1.803 SNIQKPOP .003 .001 2.822 2.624 ANNBETA -.014 .019 -.721 -.758 ANNDEBTS .003 .003 .970 1.060 POSBCAR -.002 .021 -.097 -.094 NIQKPOP .001 .001 1.041 1.010 PROFNETP -.390 .112 -3.478 -3.701 LANPERSO .029 .006 4.968 4.774 Adjusted R2=35.40% F(17, 263)=10.02, p=.000
White: Jarque-Bera: RESET: 22(139)=141.02, p=.436 22(2)=1.92, p=.383 F(3, 277)=.35, p=.789
a = White refers to heteroscedasticity-corrected t values (White, 1980), DIVOLPUR = index of all interim reports containing purely voluntary disclosures
only,
D to D85 92 = yearly dichotomy variables,
LHOLNU = natural logarithm of the number of shareholders, FIRMS = percentage of corporate ownership,
Table 2 (continued)
SCGNETS = standard deviation of percentage change in net sales, SNIQKPOP = standard deviation of net investments/total assets ratio, ANNBETA = market model beta,
ANNDEBTS = debt/equity ratio,
POSBCAR = post-event cumulative abnormal return (CAR) at business day 140. Expected returns are based on the market model.
NIQKPOP = net investments/total assets ratio, PROFNETP = profit/net sales ratio, and
LANPERSO = natural logarithm of the number of personnel.
Boldface (italic boldface) designates statistical significance at the 5% (0.1%) level.
evidence of a linear relationship between DIVOLPUR and the explanatory variables (see table 2).
The same set of tests as those used for the DIALL model were applied for the DIVOLPUR model. This allows the reader to see how well the OLS model corresponds to alternative data. The tests are detailed in table 2. The heteroscedasticity test (White, 1980) of the error terms indicates that the residual variance cannot be forecasted for the DIVOLPUR model. The error terms from the DIVOLPUR regression give 22(139)=141.02, p=.436. The test value indicates that no heteroscedasticity is found. Another heteroscedasticity test was performed, following the procedure established with the DIALL regression (Breusch & Pagan, 1979). This tested whether the squared error terms and the moments of the predicted values were correlated. Contrary to the DIALL model results, no significant correlation existed with the DIVOLPUR model. This contrast gives further evidence that the data in the DIVOLPUR model are not heteroscedastic.
The normality of the residual terms was investigated by means of the Jarque-Bera test (Jarque & Bera, 1987). The test value for the DIVOLPUR regression error term yielded 22(2) =1.92, p=.383. This indicates no deviation from normality. Therefore the DIVOLPUR model should be acceptable for predictive use in chapter 8.
Finally, the RESET test for identifying nonlinearity (Ramsey, 1969) was performed. The model containing error terms from the DIVOLPUR regression yielded F(3, 277) =.35, p=.789. The test value indicates that the linear model should be used in preference to a nonlinear model.
Besides the impact of regulation, there are several other interesting conclusions that may be drawn from table 2 above. In general: (1) governance structure, (2) business risk, (3) growth potential, and (4) firm size have variables with significant coefficients in the DIVOLPUR model. The governance structure exercises two types of influence over voluntary disclosure. One, the variable number of shareholders has a positive and significant coefficient. In the DIVOLPUR model a greater number of shareholders also causes extended voluntary disclosure in interim reports. In the DIALL model the coefficient of the number of shareholders is not statistically significant. One reason for the different outcome might be that the DIVOLPUR index is more sensitive to firm-specific factors than the DIALL index. Managers have fewer choices to make when disclosing mandatory items. This could cause the sensitivity of DIALL to explanatory variables be lower than that of DIVOLPUR. Two, the coefficient of the percentage of corporate ownership variable is negative and statistically significant. The interpretation of this result is that where a large proportion of a firm’s shares are owned by other firms, that firm will tend to disclose less than a firm with a broad base
of shareholders. Firms, as owners, can be classified as institutional investors. They are in a position to demand high quality information and communication from the firms whose shares they own. This leads to the conclusion that other communication channels are used in cases where other corporates own a significant portion of a firm. One of these other communication channels is a seat on the board. Bradbury (1992, p. 144) suggests that if a capital market is thin, indirect channels of corporate disclosure might become more widely used. This argument is also valid in the current study.
One of the business risk variables, SNIQKPOP, has a positive and significant coefficient. In the DIVOLPUR model a high standard deviation in net investments is related to extended disclosure. The reasons for extended disclosure are likely to be very similar to those already discussed with the DIALL results.
One of the growth potential variables, PROFNETP, has a negative and highly significant coefficient. The sign of the coefficient is contrary to that expected. This result suggests that a high profit/net sales ratio does not in itself result in any voluntary disclosure in interim reports. This view is similar to that previously discussed in connection with the DIALL results. One possible interpretation of this outcome is that a firm’s managers view good profit generation as sufficiently strong communication to outside interest groups. No extension of the formal disclosure process to the capital markets and other interest groups appears to be viewed as warranted.
The firm size variable, LANPERSO, has a positive and significant coefficient. This is an identical outcome as with the DIALL model. The coefficients of the variables
approximating the firm’s market risk (ANNBETA), capital structure (ANNDEBTS), mispricing of the firm (POSBCAR), and the firm’s growth (NIQKPOP), are not statistically significant.
A comparison of the results of the DIALL and DIVOLPUR models (tables 1 and 2) indicates that, besides the year in which an interim report is published, the variables for: (1) business risk, (2) growth potential, and (3) firm size have significant coefficients in both of the models. In addition, the variables for governance structure have significant coefficients in the DIVOLPUR model but not in the DIALL model. The next chapter investigates more of the implications in the market resulting from different levels of disclosure in interim reports.