One share-one vote

Winner of the 2006 Barclays Global Investors Canada Research Award

Two themes have dominated North American capital markets in the last five years: corporate governance and a desire for a better and more transparent financial reporting regime for publicly listed companies. In this paper, we focus on Canadian restricted voting share firms (RV firms) that have deviated from the traditional “one share-one vote” structure. They provide a very interesting backdrop against which one can assess the tradeoff between the operating performance, governance regime, agency costs and the interests of external shareholders. This examination is also of interest in the light of the recent announcement by the Toronto Stock Exchange (TSX) that it will no longer show the voting structure associated with the shares it lists.

In this paper, we are interested in four questions.1 First, is there a systematic difference between the characteristics of RV firms and non-RV (one share-one vote) firms controlling for the industry factor? Second, is there a difference between operating and stock market performance across these two types of firms? Third, can we detect differences in governance environments and, fourth, is there any evidence of agency costs as proxied by excess CEO pay? To answer these questions, we focus on a sample of 263 TSX-listed RV firms over the 1993 - 2004 period that include some of Canada’s well-known firms.2

Background and Stylized Facts
There is a sufficiently large body of theoretical and empirical literature that documents and explains the emergence and existence of these RV firms as well as the value of the vote, performance differentials, governance issues, agency costs, the influence of differences in legal and capital market environments, and whether these types of structures are good for the economy and shareholders who hold these RV shares in their portfolio. Many have advanced arguments on philosophical ground, that any structure that deviates from the “one vote–one share rule” is inherently a bad structure as it allows for an entrenchment of controlling minority shareholders (CMS) from other external and internal governance mechanisms and gives them power to expropriate non-controlling shareholders.3

The stylized facts based on the current empirical evidence can be summarized as follows. The motives for adopting RV shares vary, but the two main motives seem to be a) the need to control a firm without proportional investment, and b) to resist a takeover attempt (SHARE, 2004). The empirical results about the stock market reaction to the announcement of a dual-class structure are mixed.4 The characteristics of dual-class firms vary across countries. For example, RV firms in Australia are smaller whereas firms in Canada and on NASDAQ are larger compared to their industry counterparts.5 For firms where two classes of shares exist, the price premium variesconsiderably across countries, ranging from 5% to over 100% in certain cases. The differences in premiums can be related to differences in dividends and liquidity, takeover protection and coattail provisions, and the role of controlling shareholders, and can be used to estimate the value of a “vote” which also varies considerably across countries.6

The interest in the RV firms is also a result of considerable interest in agency costs associated with these structures, as they might tend to weaken the disciplinary impact of capital markets and entrench management against various monitoring systems. This, in turn, may lead to management decisions to mitigate this impact. The existing empirical evidence seems to indicate that increasing agency costs in these structures leads to lower firm valuations and may also be associated with suboptimal investment decisions.7

Overall the existing empirical evidence suggests that the firms with RV structures are characteristically different from firms that follow the traditional “one vote–one share” structure in terms of both operating and governance variables. Also, there are differences across countries, RV firms have lower valuations than control firms, and the voting shares command a premium over RV shares.

Data and Research Methodology
As noted earlier, our sample consists of 263 Canadian firms listed on the TSX that had either dual-class or restricted voting share structure in any year between 1993 and 2004. We believe that this is a uniquely interesting sample for a number of reasons. First, the Canadian history of adopting dual-class share structure is quite long, dating back to 1927. Second, the TSX has a relatively large number of Canadian stocks with the dual-class share structure. In recent years, between 20% and 30% of TSX companies, including many of the large and wellknown companies in Canada, have used some form of dual-class or restricted voting share structure. In contrast, only about 3% of S&P 500 (U.S.) companies adopted dual-class mechanisms. Third, compared to other countries, Canada has fewer restrictions on dual-class shares. The Ontario Securities Commission (OSC) only requires that subordinated shareholders receive the same information received by those holding superior shares. Fourth, there is considerable regulatory oversight in terms of coattail provisions and the treatment of RV shareholders in a takeover context that alleviate some of the concerns in the literature arising from a relaxed regulatory regime.8

To examine these issues of interest and, given the evidence that the RV firms concentrate in certain sectors and have large firm size compared to others in their industry, we have adopted a set of coordinated techniques. In order to document the differentiating characteristics of RV firms, we compare them with similar characteristics of a control group of non-RV firms using a randomized selection of firms as per Palepu (1986). We also compare them to all other firms in their respective industries.9 We use both univariate and multivariate analysis to detect the differentiating characteristics. For multivariate analysis we use a model selection technique in the logistic regression analysis to control for the inherent multicollinearity among governance variables and operating variables.

One of the important contributions of this paper is to compare the CEO excess pay between RV firms and non-RV firms to allow for direct evidence of agency cost in RV firms. In order to detect the CEO excess pay, we have followed the methodology used by Core et al. (1999). They termed “CEO excess pay” as the portion of CEO pay (salary plus bonus) that is over and above the level determined by various governance variables such as board characteristics and ownership structure of a firm. The resulting two-step procedure is briefly explained in Appendix A.

Results
Our sample consists of 263 firms shown over the entire sample period. Fifty-seven percent of these firms have subordinated voting shares meaning there is a superior class of shares in existence, 33% have no votes and the rest have either limited or restricted voting rights. Three sectors as per the Global Industry Classification Standard 2003 (GICS 2002) comprise 60% of these firms: Consumer discretionary, Industrial and Financial (mostly real estate). Of the 143 firms that delisted these shares during this period, 50% were acquired, merged or privatized, 26% were delisted due to poor performance, and 17% because the shares were unified.

Table 1 shows some descriptive characteristics of the sample of 121 firms based on the base year of 2002 for both governance and performance variables.10 An explanation of the variables can be found in Appendix B. Our analysis of governance variables shows that RV firms have statistically significant relaxed governance characteristics using two different control firm sampling techniques and different time periods. In general, RV firms a) have higher directors’ ownership, b) insiders controlling the majority of voting rights, c) a much higher ratio of directors’ voting rights to cash flow (equity ownership), d) more inside directors, e) significantly larger board size, making it potentially less effective and more symbolic, and f) significantly higher CEO excess pay. It should also be noted that the ratio of directors’ voting right to cash flow and equity ownership in RV firms is 2.73, which is significantly different from the control group, where the ratio is close to 1. Therefore, it is empirically evident that, in RV firms, a disproportionate degree of control is achieved without owning equivalent cash flow equity.

With respect to operating characteristics, it can be seen that RV firms a) are significantly larger, b) maintain a statistically significant higher dividend payout ratio, c) have similar market-to-book ratios, and d) consistently higher average values for return on asset (ROA) and return on equity (ROE), indicating that they outperform benchmark firms on the operational basis. Two observations require further attention. First, RV firms maintain a statistically significant higher dividend payout ratio. This is not in line with general expectations that controlling shareholders in an RV firm would like to retain more cash, either for their benefits or to maintain control without resorting to external financing. On the contrary, higher dividend payout ratios are consistent with the hypothesis that, if the agency cost of a firm is higher, it may be in the best interest of insiders to minimize these costs through higher dividend payments. Second, the comparison of market-to-book value ratios between the RV firms and non-RV firms does not give us any conclusive evidence, as the difference is not statistically significant. However, we find that CEO excess pay is significantly higher for the RV firms, which suggests potentially high agency costs.

Our observations that market-to-book value ratios of RV firms and non-RV firms are similar, their systematic risk is lower, and their operating performance is significantly higher, lead us to investigate the overall longterm stock market performance of RV firms.11 We use two different samples and two different benchmarks to ensure robustness. First, we focus on 61 firms that adopted the RV structure between September 1993 and 2004. We investigate their long-term abnormal performance subsequent to the adoption of an RV structure using the BHAR methodology.12

Next, we focus on all firms with RV shares over the period December 1993 to December 2002 and calculate and compare their returns using the standard cumulative abnormal residual (CAR) methodology with a monthly rebalance approach, excluding the top 1% and bottom 1% of return data to avoid the effect of outliers. We use two benchmarks: the corresponding sector index and a matching controlling group sample. Although not shown here, the sample RV firm stocks underperform the sector index and the peer stocks by a wide margin. The difference in CARs between the two groups is also statistically significant at 1%.

These results indicate that there are statistically positive and significant differences in accounting and governance variables between RV and non-RV firms as well as significantly worse stock market performance. Since these are somewhat puzzling results, we have performed various multivariate and out of sample analyses as well as some robustness tests to ensure that the results are not sensitive to the controlling sample methodology and time period we have chosen.13 Given the fact that many of the operating variables and governance variables are correlated, our multivariate analyses show that the differential governance factor dominates all other factors.

Conclusions and Caveats
Overall, we find that the accounting-based performance of RV firms is significantly better than their counterparts using two different matching techniques but that they demonstrate a significant long-term stock market underperformance. Note the fact that the RV firms have a weaker governance regime and higher agency costs, as proxied by excess CEO pay, and that, in mixed models, the differential governance factor dominates all other factors. We term this contrasting result as “governance discount” associated with RV firms.

We do recognize the challenges associated with conducting such a study. These relate to the size of the sample, the short time period, and the robustness of the methodologies and of the controlling sample. It also relates to the fact that some of the firms in our sample have been in existence for a long time period and that, by now, the stock market should have understood the governance nature of the firm and adjusted expectations accordingly. We also need to better recognize the interdependence of operating and governance variables and the fact that more work may be required to ensure that these results are extremely robust. However, we believe that we have taken care of these issues to the best of our ability, using the best methodological techniques available to us, and that our results and conclusions are robust.

In one sense, our results are somewhat surprising. The efficient market proponents would suggest that firms which consistently outperform operationally must be valued by capital markets in terms of either higher marketto-book ratios (and) or in terms of better stock market performance. This should be the case, especially in a somewhat vigilant and sophisticated regulatory regime that exists in Canada and the existence of coattail provisions even if there are some questions about the governance structure of these firms. In addition, the RV firms do pay higher dividends, potentially to mitigate the higher agency costs, and may thereby forgo attractive investment opportunities that could generate higher operating returns since they cannot readily raise equity capital without diluting control. Our evidence of higher excess CEO pay also indicates higher agency costs for RV firms. Our evidence shows that, at least during the study time period, differential governance matters and is priced by market participants by according a governance discount to RV firms. It is also clear that these valuation and performance differentials displayed by the RV firms have implications for their investment potential. Given these results, we find it surprising that the TSX has announced it will no longer show the voting structure associated with the shares it lists. We believe that this decision may be detrimental to an investor who may unknowingly buy shares of a firm with restricted voting structure as opposed to the one with a traditional one vote-one share structure.

Appendix A
Excess CEO Pay

Essentially, the Core et al. methodology results in a two-step procedure. First, we use the following regression to estimate the determinants of CEO pay (cash salary plus bonus): CEO Pay = Fn (Revenue, Investment Opportunity, Return on Asset (ROA), Stock Return (RET), Standard Deviation of ROA, Beta, CEO Duality, Board Size, Percentage of Inside Directors, Directors’ Ownership, Blockholder Ownership). Then the estimated coefficients of the governance variables (i.e. board and ownership variables used in the regression equation above) are used to estimate the CEO excess pay of each firm in the sample as follows: CEO Excess Pay = Coefficient*Board variables + Coefficient*Ownership variables We use the CEO excess pay as one of the variables in the logistics regression.

Appendix B

References

Endnotes
1. Many studies term these shares as ‘dual-class’ shares. We feel more comfortable using ‘restricted voting share’ (RV) terminology, as many firms do not list both categories of shares on the stock exchange. This also implies that comparing our results with any previous pure dual-class share literature should be done with caution, since some of the firms have only one class of share—restricted voting share alone. In this paper, we use both terms interchangeably.

2. For example: Quebecor Inc., Four Seasons Hotels Inc., Canadian Tire Corp., Magna International Inc., Onex Corporation, Rogers Communications, Shaw Communications, Bombardier Inc., Molson Inc., and Jean Coutu Group Inc. etc.

3. For example, see Amoako-Adu, B., and Smith, B.F (2001), Bennedsen, M, and Nielsen, K.M. (2004); Cronqvist, H., and Nilsson, M. (2003); Jog, V., and Riding, A. (1986); Nenova (2003); SHARE (2004); and Smith, B.F., and Amoako-Adu, B. (1995) just to name a few of these papers.

4. Partch (1987) finds positive announcement impacts for U.S. firms whereas Jarrell and Poulson (1988) report significantly negative impact. In Canada, Jog and Riding (1986) report no impact whereas Maynes (1992) reports negative impact.

5. See Taylor and Whittred (1998) for Australian results, Bohmer et al. (1996) for NASDAQ, and Amoako-Adu and Smith (2001) for Canadian results.

6. See Megginson (1990), Nenova (2003), Zingales (1994, 1995).

7. For some of the empirical evidence, see Cronqvist and Nilsson (2003) and Bebchuck et al. (2003), La Porta et al. (2002), Lins (2003) and Taylor and Whittred (1998). These results are also consistent with theoretical conjectures that “one vote-one share” structure is optimal. See Grossman and Hart (1988) and Harris and Raviv (1988).

8. However, it should also be noted that due to the ‘exempt’ offer provision of the Ontario Securities Act and Regulation, acquirers do not require to extend their offer to all shareholders of a target company. This means a change of control could occur without triggering the coattail if an exempt offer is used.

9. Essentially, all the stocks on TSX (as listed in 2002) were organized in alphabetical order. Then, RV shares, preferred shares, financial companies—essentially real estate companies and trust units are deleted from this list. Finally, every fifth company from the list is selected, which gives a list of 121 control firms comprising the non-RV firms. This is the sample of controlling firms in Table 2.

10. We also conducted a robustness test for this set of results by identifying a non-RV firm for each RV firm stratified by industry by choosing a firm as closest in size (i.e. three-year average total assets) as possible. The results are similar and for the sake of brevity not reported here. These are available from the authors upon request.

11. This is actually a troubling issue. First, in an efficient market, one expects no relation between governance (e.g. RVS structure vs. non RVS structure) and future stock returns. In order to justify significant return differences, we have to make some argument about market inefficiency or some behavioural assumption about investors’ feeling about RV firms. The second issue is that assessing long-run abnormal performance is very sensitive to the model and the control firm approach that is used. We have attempted to account for the second issue by using two advanced methods of controlling sample techniques and using two different time periods to ensure very careful matching.

12. We have used a two-year period to detect Buy and Hold Abnormal Returns (BHAR). This is rather a medium-term time frame. We have chosen this time frame to increase the sample size as many recent restricted voting shares delisted or get reunified after a few years. We also conducted the BHAR analysis for a sample of 61 RV firms that went public within this time period and found similar results—minus 15 to 20 percent negative returns. To account for the small sample, we used the non-parametric bootstrap method (Efron and Tibshrani, 1993) and bootstrap the mean statistics of the 1-Year and 2-Year BHAR based on sector index return benchmark. We specified the bootstrap sample as 500, and the number of replications to be 10,000. Our results show significant difference at 10% even with a small sample size. We do not show the results for this sample here due to space limitations but note that RV share firms underperform in the post-adoption period.

13. More specifically, we used the principal component analysis method to reduce the number of variables listed in Table 2 to six key factors and used these factor scores as the independent variables in the subsequent logistic regression where we coded RV firms as 1 and coded the control sample non-RV firms as 0 and used this variable as the dependent variable in the logistic regression. We used three models: model 1 included only the operating performance-related factors, the second model included governance factors only and the third model included all factors. We found that the “differential governance” factor dominates all other factors. In addition, we performed a robustness test to ensure that the results are not sensitive to our controlling sample methodology and time period. Due to the fact that RV firms are concentrated in certain industries and are large within their industry, we compared the RV firm sample with the entire non-RV firm sample within their respective industry sectors and in a different time period instead of using the traditional matching firm technique. Next, we conducted a robustness test for the governance variable by identifying a non-RV firm for each RV firm stratified by industry by choosing the firm as closest in size (i.e. three-year average total assets) as possible. We collected the corporate board information from annual reports and the ownership/voting right information from Stock Guide database. All the governance information is for the year 2002. These “out of sample” and “out of time period” matching techniques confirm the previous results that RV firms have superior operating performance and a weaker corporate governance regime and are accompanied by an inferior stock market performance. This confirms the presence of what we term as a “governance discount”.

—Vijay Jog, Chancellor Professor at Eric Sprott School of Business, Carleton University, Ottawa, ON.; PengCheng Zhu, Ph.D. Student at Eric Sprott School of Business; and Shantanu Dutta, Assistant Professor, Department of Business Administration, St. Francis Xavier University, Antigonish, NS.

For a PDF version of this article, click here.