Summary of Comments and Responses: NI - 52-110 - Audit Committees

1.

General Comments

One commenter suggested that the document should include more details on some areas of the analysis that were just alluded to in the CBA. It was also noted that the discussion of the analytical techniques and variables used could be clearer.

The Office of the Chief Economist greatly appreciates the constructive suggestions that were made regarding the cost-benefit analysis for proposed Multilateral Instrument 52-110. To start, it should be noted that this document was not intended as an academic research article. It was directed at a more general level of knowledge and so some details and the results from alternate model specifications were excluded in the interests of clarity for a wider audience. In response to those looking for more technical detail:

A number of comments suggested that the entire CBA be written in plain language with fewer technical details.

•

Following the standard procedure for two stage least squares regressions, all of the variables in the second stage are also included in the first stage.

•

Net Income is calculated using the standard definition and so it is an after tax figure. However that does not mean that EVA® is calculated as Net Income minus WACC/ Assets. As stated on page 22, EVA® was derived as the difference between return on capital and the cost of capital multiplied by the total capital invested.

2.

Calculation of Economic Value Added

One commenter questioned the focus on Economic Value Added as a measure of performance.

In terms of the dependent variable, there are two main reasons why Economic Value Added was the focus. Firstly, while there are other measures available, the evidence did suggest that EVA® was most likely to generate a robust result. Also, time constraints did not allow for testing of a number of alternatives to EVA® as a dependent variable. Second, it was decided to avoid measures of performance incorporating market valuations. As discussed at length in the study, earnings manipulation can raise market valuation, artificially and at least temporarily, above that for other firms. There have been a substantial number of recent examples of this in cases under enforcement in Canada and very heavily covered in the U.S. We would not expect to see a positive relationship between governance and market value for those firms whose actions lead to the Sarbanes-Oxley Act.

3.

Analysis limited to a "bear" market

One commenter stated that the time period used in the analysis (March 1999 to March 2003) represents a bear market. To ensure that the results are not biased, a longer time period involving an entire business cycle should be used.

While the range does include a substantial bull market from March 1999 to March 2001, there is the possibility that the time period being used is influencing our results. However, it is our belief that a longer time period would not be appropriate for the examination of earnings smoothing. Accrual accounting tends to reverse itself over longer time periods and so increasing the time window may conceal aggressive accounting behaviour.

4.

Definition of Earnings Smoothing

One commenter suggested that the measure of earnings management used in the analysis was the "least appropriate" measure and that the cost benefit analysis erroneously suggested that this was inter-changeable with alternate earnings management measures.

Although time constraints prevented us from testing all possible measures of our earnings smoothing, the one we did use is well-established in economic literature (see references) and at no point did we suggest that the alternatives were interchangeable. Clearly the different measures capture the effects of different incentives for and pressures on companies. Testing additional forms of earnings manipulation would produce one of two possible results: either the results would be insignificant with no impact on the analysis, or they would be significant and add to total benefits.

Additional tests for robustness and other benefits, including increased liquidity, have been completed with positive results. Since these will only add to the total benefits estimated without altering the conclusion, the analysis has not been republished.

5.

Other variables to include

Commenters made a number of very constructive comments were made regarding improvements to our existing variable or alternate variable to use.

Having revisited the analysis we find that the definition of earnings does not make a substantive change to the impact of audit committee composition on the quality of earnings disclosure. We also found that using the correlation of accruals and cash flow produced results consistent with our initial analysis.

6.

Choice of variables

One commenter stated the cost benefit study suffers from a serious omitted correlated variable problem in the regression analysis which could lead to biased estimates.

It is agreed that omitted yet important variables can lead to bias in the estimated impact of the independent variables. Such issues are always a concern with econometric analysis. However, there is no specific evidence that this is a problem in the Cost-Benefit Analysis.

Generally researchers want to avoid using variables that are correlated with both the dependent variable and other independent variables as that introduces its own bias into the model. A large number of other variables were tested during the early stages of the analysis, but, following standard practices, those that added nothing to the model were excluded.

7.

Presentation of results

One commenter made the assertion that the relationship between earnings smoothing and EVA is only weakly supported by the evidence and that the other variables have a larger impact on performance.

The most important variables must always be included to ensure the model constructed is robust and avoids any problems associated with omitted variables. Major economic determinants are always expected to have the highest weight in a valid model. This is why, for this study and others that were reviewed, the earnings management variable had the lowest impact. Those familiar with econometric analysis will be aware that the key factor is that earnings smoothing did show a high probability of being statistically significantly in determining EVA®. If net income was not significantly more important in determining value-added for the majority of firms, that could be evidence of an incorrectly specified model.

8.

Calculation errors

One commenter pointed out that there is an error in the present value calculations in Table 9.

The suggested change has only minor implications for the top end of the range of estimated benefits. The primary concern was that the lower end of the benefit range was greater than the high-end cost estimate. This relationship was unaffected by the change.

9.

Calculation of benefits

One commenter suggested that there were errors in the process used to translate the estimated coefficients in to estimated dollar value benefits. These included errors in the estimated impact of audit committee independence and earnings smoothing, as well as calculating benefits for an incorrect number of firms.

The benefit calculation used the estimated impact of an independent audit committee and earnings smoothing (along with measures of the reliability of those estimates) to calculate the benefits that would accrue to firms. This was done for firms that would have to alter their audit committee composition as a result of the proposed multilateral instrument. To err on the side of caution we also used a measure of the fit of our analysis, R2, to scale the estimated benefits. While this is not a standard econometric practice, staff reduced the projected benefits by half based on the R2 to ensure that the estimates were as conservative as possible. Restoring the estimate to double the reported amount would show a stronger outcome, but would not affect the conclusions.

10.

Implications of Canada's market size

One commenter suggested that the calculation of benefits must more explicitly incorporate the distribution between small and large companies in Canada's markets.

It is true that smaller Canadian corporations are less likely to have an independent audit committee. But given that, smaller firms are more likely to reap the benefits of the proposed rule. The cost estimates were based on the average director costs for larger companies (approximately the 300 largest companies) and therefore are almost certain to overstate costs for smaller firms. Attempts to control for the effects of firm size were made, to the fullest extent possible, throughout the analysis and the authors are confident that are results are not biased by the composition of Canada's equity market.

11.

Alternatives to full independence

One commenter noted that the CBA focuses on audit committee independence while ignoring the possibility that a lesser standard might be sufficient.

Although some have found evidence that a majority of independent directors is as effective as full independence, other authors have found that even one management representative on the audit committee produces the same result as no independent members. Examples of such authors are Tuffano or Bédard (including his own response to the proposed instrument). Given that audit committees are generally quite small it is easy to see how even one individual related to management could influence outcomes. The preponderance of published research was relied upon to effectively use the limited time available for our research. If data and time allow, the analysis will be expanded to include other governance structures.

12.

Costs of a Financial Expert

One commenter suggested that forcing disclosure of whether or not a "financial expert" sits on the audit committee is the same as requiring that such an individual be on the committee. Therefore it should be included in the cost-benefit analysis. It was also noted that in research done on U.S. companies the presence of a financial expert did decrease manipulative accounting.

The additional costs of including this requirement were included in the cost-benefit analysis. There was no way, a priori, to determine how many companies would include a financial expert purely because disclosure was required. Canadian companies are subject to a number of guidelines and disclosure requirements and a significant number of issuers do not follow them. The empirical record in these instances does not support the claim that disclosure is a de facto requirement. The costs for having a financial expert were included in the CBA report, but not in the total costs. Adding the range for financial expertise to the total does not change the outcome with benefits still exceeding costs by a significant multiple.

It should also be noted that even with an extreme estimate of all audit committees retaining a financial expert, the total high-end cost estimate would still only represent 0.014% of 2002 operating expenses.

#	Theme	Comments		Responses

1.	General Comments	One commenter suggested that the document should include more details on some areas of the analysis that were just alluded to in the CBA. It was also noted that the discussion of the analytical techniques and variables used could be clearer.	The Office of the Chief Economist greatly appreciates the constructive suggestions that were made regarding the cost-benefit analysis for proposed Multilateral Instrument 52-110. To start, it should be noted that this document was not intended as an academic research article. It was directed at a more general level of knowledge and so some details and the results from alternate model specifications were excluded in the interests of clarity for a wider audience. In response to those looking for more technical detail:

		A number of comments suggested that the entire CBA be written in plain language with fewer technical details.	•	Following the standard procedure for two stage least squares regressions, all of the variables in the second stage are also included in the first stage.

			•	Net Income is calculated using the standard definition and so it is an after tax figure. However that does not mean that EVA® is calculated as Net Income minus WACC/ Assets. As stated on page 22, EVA® was derived as the difference between return on capital and the cost of capital multiplied by the total capital invested.

2.	Calculation of Economic Value Added	One commenter questioned the focus on Economic Value Added as a measure of performance.	In terms of the dependent variable, there are two main reasons why Economic Value Added was the focus. Firstly, while there are other measures available, the evidence did suggest that EVA® was most likely to generate a robust result. Also, time constraints did not allow for testing of a number of alternatives to EVA® as a dependent variable. Second, it was decided to avoid measures of performance incorporating market valuations. As discussed at length in the study, earnings manipulation can raise market valuation, artificially and at least temporarily, above that for other firms. There have been a substantial number of recent examples of this in cases under enforcement in Canada and very heavily covered in the U.S. We would not expect to see a positive relationship between governance and market value for those firms whose actions lead to the Sarbanes-Oxley Act.

3.	Analysis limited to a "bear" market	One commenter stated that the time period used in the analysis (March 1999 to March 2003) represents a bear market. To ensure that the results are not biased, a longer time period involving an entire business cycle should be used.	While the range does include a substantial bull market from March 1999 to March 2001, there is the possibility that the time period being used is influencing our results. However, it is our belief that a longer time period would not be appropriate for the examination of earnings smoothing. Accrual accounting tends to reverse itself over longer time periods and so increasing the time window may conceal aggressive accounting behaviour.

4.	Definition of Earnings Smoothing	One commenter suggested that the measure of earnings management used in the analysis was the "least appropriate" measure and that the cost benefit analysis erroneously suggested that this was inter-changeable with alternate earnings management measures.	Although time constraints prevented us from testing all possible measures of our earnings smoothing, the one we did use is well-established in economic literature (see references) and at no point did we suggest that the alternatives were interchangeable. Clearly the different measures capture the effects of different incentives for and pressures on companies. Testing additional forms of earnings manipulation would produce one of two possible results: either the results would be insignificant with no impact on the analysis, or they would be significant and add to total benefits.

			Additional tests for robustness and other benefits, including increased liquidity, have been completed with positive results. Since these will only add to the total benefits estimated without altering the conclusion, the analysis has not been republished.

5.	Other variables to include	Commenters made a number of very constructive comments were made regarding improvements to our existing variable or alternate variable to use.	Having revisited the analysis we find that the definition of earnings does not make a substantive change to the impact of audit committee composition on the quality of earnings disclosure. We also found that using the correlation of accruals and cash flow produced results consistent with our initial analysis.

6.	Choice of variables	One commenter stated the cost benefit study suffers from a serious omitted correlated variable problem in the regression analysis which could lead to biased estimates.	It is agreed that omitted yet important variables can lead to bias in the estimated impact of the independent variables. Such issues are always a concern with econometric analysis. However, there is no specific evidence that this is a problem in the Cost-Benefit Analysis.

			Generally researchers want to avoid using variables that are correlated with both the dependent variable and other independent variables as that introduces its own bias into the model. A large number of other variables were tested during the early stages of the analysis, but, following standard practices, those that added nothing to the model were excluded.

7.	Presentation of results	One commenter made the assertion that the relationship between earnings smoothing and EVA is only weakly supported by the evidence and that the other variables have a larger impact on performance.	The most important variables must always be included to ensure the model constructed is robust and avoids any problems associated with omitted variables. Major economic determinants are always expected to have the highest weight in a valid model. This is why, for this study and others that were reviewed, the earnings management variable had the lowest impact. Those familiar with econometric analysis will be aware that the key factor is that earnings smoothing did show a high probability of being statistically significantly in determining EVA®. If net income was not significantly more important in determining value-added for the majority of firms, that could be evidence of an incorrectly specified model.

8.	Calculation errors	One commenter pointed out that there is an error in the present value calculations in Table 9.	The suggested change has only minor implications for the top end of the range of estimated benefits. The primary concern was that the lower end of the benefit range was greater than the high-end cost estimate. This relationship was unaffected by the change.

9.	Calculation of benefits	One commenter suggested that there were errors in the process used to translate the estimated coefficients in to estimated dollar value benefits. These included errors in the estimated impact of audit committee independence and earnings smoothing, as well as calculating benefits for an incorrect number of firms.	The benefit calculation used the estimated impact of an independent audit committee and earnings smoothing (along with measures of the reliability of those estimates) to calculate the benefits that would accrue to firms. This was done for firms that would have to alter their audit committee composition as a result of the proposed multilateral instrument. To err on the side of caution we also used a measure of the fit of our analysis, R2, to scale the estimated benefits. While this is not a standard econometric practice, staff reduced the projected benefits by half based on the R2 to ensure that the estimates were as conservative as possible. Restoring the estimate to double the reported amount would show a stronger outcome, but would not affect the conclusions.

10.	Implications of Canada's market size	One commenter suggested that the calculation of benefits must more explicitly incorporate the distribution between small and large companies in Canada's markets.	It is true that smaller Canadian corporations are less likely to have an independent audit committee. But given that, smaller firms are more likely to reap the benefits of the proposed rule. The cost estimates were based on the average director costs for larger companies (approximately the 300 largest companies) and therefore are almost certain to overstate costs for smaller firms. Attempts to control for the effects of firm size were made, to the fullest extent possible, throughout the analysis and the authors are confident that are results are not biased by the composition of Canada's equity market.

11.	Alternatives to full independence	One commenter noted that the CBA focuses on audit committee independence while ignoring the possibility that a lesser standard might be sufficient.	Although some have found evidence that a majority of independent directors is as effective as full independence, other authors have found that even one management representative on the audit committee produces the same result as no independent members. Examples of such authors are Tuffano or Bédard (including his own response to the proposed instrument). Given that audit committees are generally quite small it is easy to see how even one individual related to management could influence outcomes. The preponderance of published research was relied upon to effectively use the limited time available for our research. If data and time allow, the analysis will be expanded to include other governance structures.

12.	Costs of a Financial Expert	One commenter suggested that forcing disclosure of whether or not a "financial expert" sits on the audit committee is the same as requiring that such an individual be on the committee. Therefore it should be included in the cost-benefit analysis. It was also noted that in research done on U.S. companies the presence of a financial expert did decrease manipulative accounting.	The additional costs of including this requirement were included in the cost-benefit analysis. There was no way, a priori, to determine how many companies would include a financial expert purely because disclosure was required. Canadian companies are subject to a number of guidelines and disclosure requirements and a significant number of issuers do not follow them. The empirical record in these instances does not support the claim that disclosure is a de facto requirement. The costs for having a financial expert were included in the CBA report, but not in the total costs. Adding the range for financial expertise to the total does not change the outcome with benefits still exceeding costs by a significant multiple.

			It should also be noted that even with an extreme estimate of all audit committees retaining a financial expert, the total high-end cost estimate would still only represent 0.014% of 2002 operating expenses.