• Tidak ada hasil yang ditemukan

In response to Australia’s ‘if not why not?’ corporate governance disclosure requirements, are explanations for non-compliance

N/A
N/A
Protected

Academic year: 2023

Membagikan "In response to Australia’s ‘if not why not?’ corporate governance disclosure requirements, are explanations for non-compliance "

Copied!
37
0
0

Teks penuh

So far, not much has been published on the nature of “compliance” with (or rather, “legitimate response”) with respect to the two pillars of the “comply or explain” system. 4 When there is direct non-compliance, failure to provide an explanation (level 1 of the variable) is a failure to meet one of the requirements of the “comply or explain” system. An inadequate, uninformative explanation (Level 2) meets the letter but not the spirit of the “comply or explain” requirement and is effectively a Level 1 response disguised as a Level 3 response.

But none of the companies were outside the ASX's top 500 companies, reducing the research's ability to shed light on size-driven differences in company behavior in a "comply or explain" context. However, a report by the ASX Implementation Review Group (IRG, 2004) noted that feedback received from some companies indicated a perceived obligation to directly comply with the guidelines' recommendations, a perception previously noted by Fleming (2003) and later by Zadkovich (2007) . On the other hand, this would mean that smaller companies take advantage of the "comply or explain" flexibility to optimize their boards while staying within the rules.

Large institutional shareholders may prefer that a company's management be seen as directly compliant; and if it is not, then they will advocate for the company to provide adequate explanations in accordance with 'comply or explain'. Further, direct compliance may be seen as the fourth and highest level of value for the dependent variable.

DATA CHARACTERISTICS AND MODEL SPECIFICATION

12 However, no study of the quality of explanations for direct noncompliance can be undertaken without at least some coverage of the topic of direct compliance versus direct noncompliance, although that is not our focus. This provides grounds for assuming that the companies in the sample are relatively stable in their direct compliance choices. In terms of the overall non-compliance averages, it is 61 percent for recommendation 2.1, falling to 51 percent for recommendation 2.2 and 21 percent for recommendation 2.3.

Regarding the pattern of interpretation quality in Table 1, a clear trend is evident for recommendations 2.1 and 2.2 for most years (the exception being recommendation 2.2 in 2004). The evidence is that both the mean and standard deviation of LnTA increase monotonically from Level 2 to Level 4 under all three board structure recommendations. In the case of recommendation 2.3 (CEO separation) in Panel C, there is monotonicity at all four levels.

In Panel A and B, level 1 has the lowest standard deviation, but a mean slightly higher than level 2 although lower than any other level. In Panels A and B, the null response group (level 1) has the largest average PE ratio and also (but barely) in Panel C. Regarding the market-to-book ratio, a greater degree of compliance is associated with a increase in mean size and in the standard deviation.

The highest level of the dependent variable is four when direct compliance along with zero explanation, insufficient explanation and sufficient explanation (the first three levels) are included in the dependent variable. Since the ordinal form of the multinomial logistic regression assumes a uniform slope across all levels, its output provides a single set of coefficients for the independent variables along with multiple intercept coefficients (three if there are four levels, two if there are three levels) . In the case of the hierarchical form, an intercept coefficient and a set of coefficients for the independent variables are provided for each level (excluding the highest) of the dependent variable.

While the binary logistic regression code included the calculation of general summary measures (maximum likelihood test score and associated p-value, Wald estimate and p-value, Nagelkerke R2), these are not easily calculated in multinomial procedures.

RESULTS

In the case of recommendation 2.2 (independent director chairman) in panel A, the zero response coefficient breaks the pattern of monotonically decreasing coefficients, while the other two levels maintain it. In panel B, the difference between an insufficient response and a sufficient response is significant at the one-tenth of one percent level. With respect to age, there is a single strongly significant result in panel A for recommendation 2.1, which is supported by a weakly significant distinction in panel B.

In Panel A, the distinction between firms that comply directly and those that provide sufficient explanations for not doing so is significant at the one percent level and has a negative sign. In Panel B, insufficient explanations are distinguished from zero responses (five percent level of significance) and from adequate explanations at the one percent level of significance. However, Recommendation 2.3 yields inconclusive results because the strong distinctions in Panel A are weakly supported by one coefficient that is significant at the five percent level in Panel B.

With the direct compliance level dropped in Panel B, the number of intercept terms drops to two. Whereas in Panel A of Table 3, Recommendation 2.1 was significant in seven of the nine procedures, in Panel A of Table 4 two of the three recommendations give a coefficient significant at the one-tenth of one percent error level, while the third gives a coefficient significant at the one percent level. Age's Table 3 significance was confirmed with respect to two recommendations in Panel A of Table 4.

Panel B of Table 4 confirms the significant findings in Table 3 for all three recommendations. For recommendation 2.3, this significance is at the level of one-tenth of one percent in both panels, while for recommendation 2.2 it falls to the level of ten percent in panel A and the level of five percent in panel B. The situation regarding recommendation 2.2 is not like that either. clearly in Panel B because the expected sign reversal in Table 4 does not occur.

The only other significant result (but only at the ten percent level) is liquidity with respect to Recommendation 2.1 in Panel A of Table 4. The effect of ownership concentration is also confirmed for two of the three levels in Panel A for Recommendations 2.1 and 2.2 and for one level of Recommendation 2.3. . In Panel A of Table 6 , LnTA continues to provide coefficients at tenths of one percent at the significance level, while the market-to-book ratio reaches significance at the five percent level in two cases.

CONCLUSIONS AND FUTURE DIRECTIONS

However, the economic effect of the age variable is very small; and both ownership concentration and age variables faded in significance in multinomial logistic regression procedures. While almost all of these produced sporadically significant results, none of them provided any reliable model from which to make predictions. But this does not diminish the effect size found in this sample that is significantly related to compliance choices made in the context of ASX corporate governance regulations.

So far research has been about 'comply or explain' or 'if not why?' corporate governance assessments, including those of the present study, conducted on relatively small samples. It is clear that there would be value to be gained in a future paper to expand the scope in several directions. A data set containing all companies listed on a stock exchange in a given period can be collected and used; the focus can be broadened to cover more than just the three measures examined.

We believe this is a promising direction for future research that will benefit regulators.

28 Australian Stock Exchange Corporate Governance Council (ASXCGC) 2003, Principles of good corporate governance and best practice recommendations. If not, why not?: in conversation with Karen Hamilton, Chair, ASX Corporate Governance Council, Keeping Good Companies (Monthly journal of the Governance Institute of Australia) April: 218. Use of 'follow-or-explain': compliance with codes of corporate governance in Great Britain and Germany.

This table shows the incidence of the three observation levels for all dependent variables, which are the company's response to ASX corporate governance compliance requirements. PANEL A: Binary Logistic Regressions Comparing Response Levels to Direct Compliance (i.e. Zero Response vs. Direct Compliance etc.) Response Level. PANEL B: Binary logistic regressions contrasting response levels (i.e., null response versus inadequate response, etc.) response pair.

Binary logistic regressions using Matlab's GLMFIT procedure are used to see if firm characteristics are associated with any systematic difference in firms' response choices when they choose to achieve full compliance with the three measures of Recommendation 2 of to avoid the ASX regarding the governance structure. In panel A, a certain level of quality of the response is tested against direct compliance, while in panel B, the three levels of quality of explanation for direct non-compliance are tested against each other. The total number of firm observations in the sample is 1032, but the number of observations used in each procedure depends on the sum of the two dependent variable subsets assembled for use in the given logistic regression.

In panel A, there are four levels for each dependent variable: the three levels of response quality when a company chooses not to directly comply with the given measure, and the fourth is direct compliance with the measure. Panel B: Only three response levels: null response, inadequate quality, adequate quality (two sets of coefficients per procedure). One set is an extended set containing four levels of the dependent variable, where the fourth level is "direct compliance".

This means that the dependent variable is 'Response to both pillars of the legal requirement', while the requirement is 'comply or explain'. The second set contains only the three response levels that are only possible for the 'explain' pillar. As in Table 5, each procedure has three sets of coefficients, based on the assumption of independent variables each having different intersections and slopes, that map the relationship between adjacent pairs of the four levels of the dependent variable.

Referensi

Dokumen terkait

2.2.3 Market to Book Value Market to book ratio of equity reflects that market assess the company’s investment return in the future can be seen from expected return of equity Smith