Summary of statistical findings

Table: Summary of statistical inferences

Click to view model

Table 6, Part 1: Performance regressions, EQ1: Market-to-book ratio, return on assets

Explanatory var.

Empirical findings and references

OWP_i,t,d

Managerial ownership

Managerial ownership appears to be able to explain about 0.7% of the total variation in financial performance (Table 2.3). Performance is significantly increasing for increasing managerial ownership in the 0 to 0.5% interval, but for all other intervals the evidence is not robust enough to conclude any particular pattern (Tables 2.1 & 2.2). In other words, there is some evidence of incentive alignment (Hypothesis 1, Chapter 2) for very low levels of managerial ownership. However, for higher levels of ownership it appears that firms have efficient ownership structures. This is perhaps a result of the economics of natural selection (Hypothesis 5, Chapter 2). Furthermore, the most significant results were produced with measures of officer and director ownership rather than by measures of insider ownership making the former the preferred measure. Get dissertation.

LnSALP_i,t,c

Officer and director salary

Combined salary for the two highest paid officers (executives):

Officer compensation appears to be able to explain about 0.8% of the total variation in financial performance (Table 2.3). From the 0 to $1 million interval there is evidence of a significant and negative relation for regressions on the market-to-book value (MTB), but it does not prevail when using return on assets. However, for the 1 to $6 million interval performance is significantly increasing for increasing manager compensation for the important weighted all-firms sample. This result holds for regressions on both MTB and return on assets and for the alternative definition of officer salary, but it does not hold for any of the equally weighted regressions (Tables 2.1 & 2.2 & 2.4). For the plus $6 million interval the evidence is mostly insignificant, although a significant negative relation is detected with regard to return on assets for the important weighted all-firms sample (Table 2.2).

Combined salary for the two highest paid directors (board members):

For the 0 to $0.5 million interval the evidence is weakly significant and negative. However, it is strongly significant and positive for the 0.5 to $3 million interval when considering the weighted all firm regressions. Interestingly the plus $3 million interval reveals a strongly significant and negative relation for all regressions (Table 2.4).

Hypotheses:

The argument, that managers at lower levels of compensation need to have higher salaries in order to attract the most capable and competent people appears to be supported. The evidence also supports that beyond a certain level ($6 million for officers and $3 million for directors) it is at best not improving corporate profit to pay more and at worst it is harmful to pay beyond these limits. The latter evidence is interesting because it cannot be explained by an alternative argument claiming that managers are paid more if the firm performs well. It is, however, consistent with the explanation that managers who get enormously rich very quickly may lose their incentives to continue to work hard for the company.

DSAL_i,t

Salary dummy

For officer salary as well as for director salary this dummy is mostly insignificant, thereby supporting the claim that firms with missing salary observations perform no differently than other firms (Tables 2.1 & 2.2 & 2.4).

LTDebt_i,t / Assets_i,t

Long-term debt to total assets

Leverage appears to explain about 1% of the total variation in financial performance (Table 2.3). It is strongly significant and negatively related to performance and the squared leverage is equally significant but positive. This is evidence of a U-shaped relation between performance and leverage. The evidence is remarkably robust and prevails for both MTB and for return on assets, regardless of sample used and regardless of whether the regressions are weighted or not (Tables 2.1 & 2.2). The explanation proposed here is that very high leverage is good for performance because it provides high-powered incentives for the management. By contrast, the higher financial performance for firms with low degrees of leverage could be explained by a pecking order argument stating that firms with temporarily high performance may prefer retained earnings to debt.

MarkCap_i,t

LN to market capitalization

Firm size is extremely significant and positively related to performance. This result is remarkably robust and holds independent of applied sample, performance measure and whether the regressions are weighted or not (Tables 2.1 & 2.2). The evidence is particularly strong for regressions on MTB, a finding that support the idea of strong liquidity effects because they are irrelevant for return on assets. In general, the evidence could also be seen as support of significant effects from monopoly pricing or survivor biases, although it is difficult to say which argument is the most important.

BETA_i,t

Stock market beta

The non-diversifiable risk is a significant and positive function of performance particularly, when the regressions are weighted with market value (Tables 2.1 & 2.2). This result may support the classic argument that investors demand higher performance for firms that are more risky in order to compensate them for their risk-aversion.

CapExp_i,t / PPECap_i,t

Capital expenditure to total property plant and equipment net

With regard to MTB this fraction is strongly significant and positive for all regressions, Table 2.1. However, when regressed on return on assets the evidence is inconclusive with different signs for the four regressions (Table 2.2). The MTB regressions could support both an argument for measurement bias and an argument for better investment opportunities. However, the regressions on return on assets cannot confirm the investment argument. Together this evidence supports the measurement bias for MTB, but not the investment argument.

OpeInc_i,t / Sales_i,t

Operating income to sales

The profit rate is strongly significant and positive for all regressions and to an extreme degree for the regressions on return on assets (Tables 2.1 & 2.2). There can be no doubt that a high profit rate is important for observing high financial performance. This should also be expected since a positive profit rate is a necessary, although not a sufficient, condition for high financial performance.

R&D_i,t / Assets_i,t

Research & development costs to assets

For all the MTB regressions the coefficient of this variable is strongly significant and positive (Table 2.1). However, for the return on assets it is strongly significant and negative for the equally weighted all-firms sample (Table 2.2). Together this evidence does not support an economic explanation of any particular relation between R&D expenditure and financial performance. However, it does lend support for a strong and predictable measurement bias with regard to MTB.

DR&D_i,t

R&D dummy

This dummy is insignificant on most regressions or it is weakly significant with varying signs. The evidence confirms that firms with missing observations of R&D perform no differently than firms reporting this value.

Adver_i,t / Assets_i,t

Marketing costs to assets

For all MTB regressions this variable is strongly significant and positive (Table 2.1). However, this is not true for regressions on the return on assets for which it is either insignificant or strongly significant and negative as in the case of the weighted all-firms sample (Table 2.2). This evidence is similar to the evidence for R&D and, although it does support an explanation of measurement bias in the MTB regressions, it does not support any particular economic theory of this relation.

DAdver_i,t

Marketing dummy

This dummy is extremely significant and positive for the weighted all-firms sample, both with regard to MTB and return on assets (Tables 2.1 & 2.2). However, for all other regressions the evidence is insignificant. The evidence cannot reject the hypothesis that firms with missing observations on marketing perform better than firms without missing observations. This possible bias is perhaps caused by the high fraction of missing observations for the marketing variable (Chapter 6, Table 3, Part 2). Get dissertation.

Dummies for industry, exchange and incorporation

The three-digit industry dummies contribute significantly to the adjusted R² (6 to 15%) in all samples. However, the contribution from exchange dummies and location of incorporation is limited to about 0.5% and 0.2% respectively (Table 2.3). The evidence indicates that institutional factors and in particular those associated with industry are of major importance for the measurement and/or the level of financial performance.

Table 6, Part 2: Ownership regressions, EQ2: Off & dir ownership, insider ownership

Explanatory var.

Empirical findings and references

Expected performance

Consensus stock recommendation:

The coefficient is strongly significant and negative for the all-firms sample regressed on officer and managerial ownership (Table 3.1). The likely biased NYSE sample produces weaker results (Table 3.1). Both of these findings are similar when substituting officer and director ownership with insider ownership (Table 3.2).

Next five year average EPS growth:

Although this measure is only available for the NYSE sample, it was possible to get the data for two years. Both for year 2000 and for 1999 the weighted samples produce highly significant and positive coefficients. However, the equally weighted regressions produce an insignificant result for year 2000 and a significant and positive result for 1999 (Table 3.4).

Hypotheses:

There is fairly strong evidence that expected performance is an important and positive determinant of managerial ownership. This evidence supports the insider investment argument that insiders increase their ownership when they expect performance to improve and decrease it when they expect it to deteriorate. The evidence is also consistent with the insider reward argument that managers are able to increase their equity rewards when they expect better performance and that they actually do it. The details of the arguments which belong to Hypothesis 4 are discussed in Chapter 2, Section 2.4. Get dissertation.

P_i,t-1

Past performance

Three-year average return on assets:

The coefficient in the officer and director regressions is strongly significant and positive for the weighted regressions. Furthermore, it is significant and positive for the equally weighted regressions (Table 3.1). Although less significant these, results prevail when the ownership measure is substituted by insider ownership (Table 3.2).

Two-year average return on assets:

The findings are practically the same as those reported above when using the three-year average return on assets (Table 3.5).

Hypothesis:

The evidence is supportive of the reward argument in Hypothesis 4 stating that firms reward their managers with equity stakes if they have delivered high financial performance in the previous periods. For details, see Chapter 2, Section 2.4. Get dissertation.

MarkCap_i,t

Market capitalization

Size is extremely significant and negative for all regressions. This result holds for both measures of manager ownership, whether the sample is weighted or not, and for both the all-firms sample and the NYSE sample (Tables 3.1 & 3.2). The evidence convincingly supports the argument that higher market value makes it more difficult for managers to afford a large ownership stake. Alternatively, it also supports the idea that the managers of large firms are satisfied by smaller ownership stakes than the managers of small firms because it is easier to control a large firm for a given fraction of ownership than it is to control a small firm with the same fraction of ownership. Finally, it could also support a risk argument claiming that managers of large firms prefer less ownership in order to be more able to keep a fully diversified portfolio.

Dummies for industry, exchange and incorporation

The one-digit industry dummies contribute significantly to the adjusted R² (3 to 5%). Perhaps more surprising is the finding that the dummies for the location of incorporation seem to be even more important for the determination of managerial ownership than the industry dummies (4 to 5%). The stock exchange dummies are also important, although they are far less important than the two other categories of dummies (0.3 to 1.2%) (Table 3.3). All in all, the evidence indicates that institutional factors are of major importance for the determination of managerial ownership.

Table 6, Part 3: Expectation regressions, EQ3: Consensus stock recom., EPS growth

Explanatory var.

Empirical findings and references

P_i,t

Performance

Both MTB and return on assets are strongly significant and negative when regressed on the stock consensus recommendation and this result holds for all samples, whether weighted or not (Table 4.1). Substituting the consensus stock recommendation with the alternative measure of the next five-year average growth in earnings per share further confirms these findings. In this case both MTB and returns on assets are strongly significant and positive on this measure for all regressions, with exception of the weighted NYSE 2000 sample on return on assets, which is insignificant (Table 4.2). Finally, the evidence shows that the performance variables account for an important portion of the total variation in the performance expectations (Table 4.3). All together, the evidence seems to supports the claim that adaptive expectations play a significant role in the formation of performance expectations, although they are only believed to be a small part of a vastly unknown story about the formation of performance expectations. This evidence is important for the primary model in the sense that, it also supports the claim that adaptive expectations may be an important source of endogeneity between managerial ownership and financial performance.

Dummies for industry, exchange and incorporation

The three-digit industry dummies contribute very significantly to the adjusted R² (4.3% to 15.6%). Dummies for country of incorporation contribute more than 10 times more to the adjusted R² when the regression is weighted (it increases from 0.16 to 2.5%). The stock exchange dummies also contribute significantly to the adjusted R² (with about 1.5%) (Table 4.3). Together the evidence clearly indicates that institutional factors are of major importance as determinants of expected performance.