Print Email Download Reference This Send to Kindle Reddit This
submit to reddit

Relative Performance Evaluation In Executive Compensation Contracts Finance Essay

This paper examines the presence of relative performance evaluation in executive compensation contracts. Based on a literature review I conclude that the findings of past studies are inconclusive about the presence of relative performance in executive compensation. Some possible explanations are offered by researchers, however, due to the lack of information these explanation could not be studied. However, a new Securities and Commission disclosure rule offers data to shed new insights into relative performance evaluation.

INTRODUCTION

XXXXX

What is Relative Performance Evaluation and what is the theoretical motivation behind?

The economic theory of relative performance was developed and introduced in 1982 by the researcher Holmström (1979; 1982). This theory, also referred as market indexing of compensation, suggests that the optimal executive compensation contracts and the rewards of managers should not be based on absolute performance, but based on relative performance to peers with similar characteristics (Murphy, 1990, p. 35).

In practice, this means that even if the industry is in a bear market an executive still could get rewarded because it performance relatively better than other firms in the same industry. However, this also imply that even if the industry is in a bull market and all the firms in that industry are doing well the executive could still be punished for bad performance relative to other firms in the same industry. The explanation is as follows: the positive performance of the firm is not due to superior managerial skills.

Theoretical motivation of relative performance evaluation

As mentioned earlier, the theoretical motivation and of Rpe can be found in the economic theory and is based on risk-sharing (Gibbons and Murphy, 1990, p.30). Executive compensation contracts that are based on absolute performance use, for example, the stock price of the firm as measure of the performance of the executive. Although these measures are proper measures for the absolute performance of a firm it is not the optimal measure for executive performance since many other factors, called systematic risk, beyond the power of the executive influences these measures. Therefore, an executive compensation contract that is solely based on absolute evaluation will be exposed to risks that are beyond the control of the executive. An executive compensation contract that is based on relative performance evaluation filters out these beyond the executive powers’ risks from the compensation of executives, whereas a strong form rpe filters out all the common risk the weak form of rpe filters out the common risk partially (Rajgopal, 2006, p.1817).

The formula below illustrates this. Suppose that the compensation contract of the executive is based on y, and y depends on an executives’ effort, see gibbons and Murphy, 1990

Executive compensation based on relative performance also has an incentive effect, especially for risk-averse executives, to increase shareholder wealth and put effort to achieve this since in rpe market/industry risks are filtered out. Executive will only be assessed on their own actions and not on common uncertainties, which provides incentives to perform better than its competitors (benchmarks). In practice, even if the measure used to assess the executive is negative rewards will still be granted to executives that perform better, less worse, than the benchmark. However, rpe also has its disadvantages for executives: it filters out cash windfalls (BRON).

Antle and Smith (1986, p.2) describes the gain in efficiency from rpe as follow:

“The gain in efficiency arises because information about the common uncertainty to be filtered out of the agent’s evaluation. This filtering permit a risk-averse agent’s to exposure to “systematic risk” to be reduced without a reduction in the agent’s incentives”

In general, an executive compensation package contains four components: a base salary, an annual bonus, stock options and long-term incentive plans (Murphy, 1999, p.3). One of the most direct way to reward of punish executives in their performance is by their compensation package.

According to Murphy (1999) rpe can be distinguished into two types: implicit rpe and explicit rpe. Gong et al. (2010) also makes a distinction among executive compensation contracts. Relative performance in executive compensation contracts can be either an implicit part or explicit part of the contract. Relative performance as implicit part of an executive’s compensation contract implies that the relation between the performance of the executive and her relative performance is not specified in the form of a contract. However, although the compensation level is not officially specified in a contract in practice the level of compensation is related to the performance of this executive relative to a benchmark as firms with the same two digit-SIC code (Meulbroek, 2001, p.4). On the other hand, relative performance as explicit part of executive compensation contracts explicitly state the relative performance based level of compensation in the executive compensation contract, which implies that the benchmark used in relative performance evaluation is explicitly stated.

In general, the implicit approach use the following Ordinary Least Squares regression for testing the presence of relative performance evaluation in executive compensation contracts (Rajgopal et al., 2006, p.1817)

Compensation = β0 + β1 Firm returns + β2 Market/ Industry returns+β3 Controls + error. (1)

Whereas:

Compensation = CEO salary + bonuses

Firm returns = performance measure (for example RET, ROE or ROA) of the firm

Industry returns = peer performance measure

Controls = controlling variables

Error = residual of the regression

The betas in this equation can be seen as the pay-for-performance sensitivity coefficients. As mentioned previously, the rpe attempts to filter out the systematic risks. Rpe attempts to filter out industry effects, and reward executives based on managerial performance. Therefore, a negative pay-for-performance sensitivity coefficient for market/industry coefficient is interpreted as the proof that relative performance evaluation is present in executive compensation contracts, whereas (β1+ β2=0) implies the presence of the strong form of relative performance evaluation. Proof for the weak form of relative performance evaluation is present if β2 is negative and less than β1in absolute magnitude.

While prior implicit approach studies make their own assumptions for the best peer group in the OLS regression (for example firms with the same two digit-SIC code) the explicit approach use the peer performance of the explicitly stated peer in the regression.

This paper will investigate whether there is rpe in executive compensation contracts.

Evidence empirical studies on implicit/explicit rpe

Implicit relative performance evaluation

Studies that support the presence of relative performance evaluation

One of the first studies that tested the presence of rpe in executive compensation contracts is a study by Antle and Smith (1986). The authors investigated the presence of rpe based on a sample of 39 firms in the aerospace, chemical and electronic industries. The data of this study covers the period from 1947-1977. The firm performance measures chosen by the authors to measure performance are the return on common stocks (RET) and the return on assets (ROA) measure. The empirical procedure in this study is based on the following two regressions for all 39 firms. The results of this study are mixed. Out of the 39 firms this study investigated 16 firms seems to use relative performance evaluation. STATE PEER GROUP USED

Another study that investigates the relationship between relative performance and executive compensation is a study by Gibbons and Murphy (1990). This study is based on data of 1668 CEOs compensation contracts from 1049 different firms. CEO compensation in this study consists of the sum of salary and bonus. The measure of firm performance in this study is represented by the change in the logarithm of shareholder wealth. This change is equal to the continuously accrued rate of return received by the firms’ shareholders, including price appreciation and dividends. The benchmark firms in this study can be divided into four different SIC industry groups: one-digit, two-digit, three-digit and a four-digit SIC industry group. The industry return in this study is represented by the continuously accrued rate of return value weighted portfolio of all the firms that belongs to the same SIC industry as the sample firm. The variable market return is defined as the continuously accrued return on the value-weighted portfolio of all the firms listed in Compustat.

Results of this regression show that rpe is present in the sample of this study. In general, results [1] shows that rpe is present in all of the one to four SIC digit industry groups. Systematic risks are partially filtered out. However, pay-for-performance sensitivity coefficients are not equal for the four different industry groups. CEO pay seems to be more strongly related to broad industry definition (1-digit SIC) than to narrow industry definitions (4-digit SIC). Another explanatory variable in the regression was the market-return. Result shows that all the market return pay-for-performance coefficients are negative and significant. Anyhow, since, holding market returns constant, CEO pay growth is not significantly related to the three- and four-digit industry return while it is significant related to both the three- and four-digit market return, it can be implied as prove that the explanatory variable market-return is more likely than industry returns to be used as relative performance measure.

Janakiraman et al. (1992) also studies the presence of rpe in executive compensation. The executive compensation package applied in this study consists of the sum of the following two cash compensation components: Salary and annual bonus. The compensation data comes from 609 firms and covers the period from 1970 to 1988. This study also uses two measures that represent firm performance. The firms’ accounting performance is measured by the return on equity while the security price performance is measured by the sum of capital gains and dividends divided by the stock price at the beginning of the year (RET). The benchmark group accounting performance measure is represented by an industry index which consists of the value-weighted ROE for all other firms than the sample firms with the same two-digit SIC code on the 1989 annual industrial and research Compustat data files. Furthermore, the benchmark security price performance index for the sample firm is estimated by the value-weighted RET for all other firms in the same two-digit SIC code on the monthly CRSP data file (Janakiraman et. al. 1992, p.58). In order to investigate the presence of relative performance evaluation the authors regressed the change in compensation on the change in accounting ROE and the change in industry accounting ROE. The same was done for the second measure RET. Regression results show that there is weak evidence for the relative performance evaluation hypothesis in the accounting performance measure ROE. According to the results of this study CEOs cash compensation is positively related to the firm’s accounting ROE and not or slightly related to the industry. If we look at the results of the security price return (RET) we can see that the strong form of relative performance evaluation does not hold. This implies that not all the industry component is filtered out in the CEOs compensation package. Nevertheless, there is support for the weak form of relative performance evaluation. Table 2 Appendix [2] show that the cash compensation of a CEO is positively related with the security price return of the firm but negatively related with the industry security price return. In short, there is no evidence of the strong form of relative performance evaluation and some support of the weak form of relative performance evaluation.

Studies that rejects the presence of relative performance evaluation

While these studies do detect some presence of rpe there are also studies that don’t observe relative performance evaluation.

Barro and Barro (1990) studied the pay of executive’s in banks. The authors had data on CEOs pay in 83 large commercial banks. Based on this information Barro and Barro (1990) investigated whether relative performance evaluation is present in the executive compensation packages in these banks. The data covers the period of 1983 to 1987. The performance measures in this study are stock returns and changes in earning yields. An interesting detail about the peer group in this study is that the peer group is not based on SIC digit codes but based on banks in other geographical regions. Results show that peer –group returns are not filtered out in the sensitivity of pay to performance. In other words, the cash compensation of executives investigated in their study is not adjusted for the performance of the peer group which rejects the presence of relative performance evaluation.

Bertrand and Mullainathan (2001) investigated in their study whether CEOs are rewarded for luck. In this study luck is defined as systematic risk, (positive) shocks to performance that is beyond the control of the CEO. If the result of the studies show that changes in CEO rewards are sensitive to luck than this would be evidence that relative performance evaluation is not present, because according to the relative performance evaluation hypothesis this luck component should be filtered out and equal to zero. The data cover the period 1977 to 1994 and consists of the 51 largest American oil companies. Results of this study show that the pay of a CEO is associated with dollar earned by luck as well as a dollar earned due to superior executive performance. This result suggests that CEOs are rewarded for luck, also called cash windfall, and rejects the presence of relative performance evaluation.

Another study that rejects the presence of relative performance evaluation in executive compensation and goes even further is a study by Garvey and Milbourn (2006). Principle-agent theory suggest that systematic risk should not affect executive’s pay and that this risk can be filtered out by paying an executive relative to a benchmark it is assumed that under this theory an executive’s pay is unaffected by market movements. However, the results of Bertrand and Mullainathan (2001) reveal that CEOs are paid for luck and that positive market movements are not filtered out. Since cash windfall is present in the pay of executives it would imply that executives’ pay should also be affected by negative market movements. However, Garvey and Milbourn (2006) documents asymmetric benchmarking in executive pay, which implies that executives are rewarded for positive market movements but sheltered for negative market movements. Therefore, the result of this study is that asymmetric benchmarking is present in executive compensation. Executive’s benefit from positive market movement but do not suffer from market movements.

Table 1 Prior empirical studies

Empirical Study

Peer group

relative performance evaluation

Antle and Smith (1986)

Two-digit SIC

Results are mixed, observed relative performance evaluation in 16 out of 39 firms

Gibbons and Murphy (1990)

One to four-digit SIC

Yes

Janakiraman et al. (1992)

Two-digit SIC

Did observe relative performance evaluation in one out of the two performance measures.

Barro and Barro (1990)

Geographical region: banks in the same region

No

Bertrand and Mullaintain (2001)

Two-digit SIC

No

Garvey and Milbourn (2006)

Two-digit SIC

No

Other studies which also investigated relative performance evaluation include Jensen and Murphy (1990), Aggarwal and Samwick (1999), Himmelberg and Hubbard (2000), Garvey and Milbourn (2003) and Rajgopal et al. (2006) among others. Put these studies in table.

Most of the prior empirical studies adopted the implicit RELATIVE PERFORMANCE EVALUATION approach because of the limited disclosures with respect to relative performance evaluation contracts. Therefore most studies attempt to observe relative performance evaluation by benchmarking a firms’ performance measure on the average of firms with the same industry two digit-SIC Code. Although the amount of studies using the explicit approach is scarce there are a few studies which investigated the presence of relative performance evaluation using the explicit approach.

Murphy (1999) NOG BESCHRIJVEN 2 EXPLICIT STUDIES

Possible explanations

Reviewing prior studies on relative performance evaluation reveals that empirical results are mixed. However, researchers are still inconclusive about the reason behind why the results are mixed. Nevertheless, some possible explanations are offered. The first possible explanation is based on factors that could influence a firms´ decision to use relative performance evaluation in executive compensation. The other possible explanation is that some researchers believe that the implicit test has incorrect assumptions and model misspecifications (Murphy, 1999; Bannister and Newman, 2003; Albuquerque, 2009).

Factors that may affect a firms´ decision to use relative performance evaluation in executive compensation

The majority of the prior studies did not only test the presence of relative performance evaluation. Most of them also tried to identify other factor that could influence the probability of a firms´ decision to include relative performance evaluation in executive compensation. One possible factor is the common risk, Holmstrom (1982) predicts that when common risk has a greater chance to influence the performance of a firm that firm is more likely to use relative performance evaluation in their executive compensation. Another factor that seems to have influence is the degree of industry competition. Aggarwal and Samwick (1999) predict that firms in a more competitive industry are less likely to use relative performance evaluation. Murphy (2001) argues that the growth opportunity of a firm has affect on the probability of a firms’ decision to use relative performance evaluation. High growth firms are more likely to use relative performance evaluation in setting executive compensation. Leone and Rock (2002) argue that setting executive pay in high growth firms using internal standards will provide executives the incentive to smooth performance, which reduces the fears of budget ratcheting. The ownership structure and board attributes are also factors that may influence the use of relative performance evaluation. Several studies documented that executive pay-for-performance sensitivity is higher for firms with a high quality governance structure (Core et al., 1998; Bertrand and Mullainathan, 2001; Sautner and Weber, 2005; Garvey and Milbourn, 2006). Therefore, the better a firm is governed the more likely relative performance evaluation is used in executive compensation. Garvey and Milbourn (2003) predict that firms with wealthier and older executives show weaker evidence of relative performance evaluation. These authors predict that the wealthier and older the CEO the greater the self-hedging ability. Therefore, the wealthier and older the CEO, which implies the greater the ability to hedge for market risk, the weaker the benefits of relative performance evaluation.

Rajgopal et al. (2006) predict that the lack of relative performance evaluation support in empirical studies could be attributed to a covariation between and the economy’s fortunes. The prediction is that the compensation of talented CEOs is less likely filtered for market or industry performance. For that reason relative performance evaluation is less frequent observed in large firms. Rajgopal et al. (2006) also predict that firm performance is related to the use of relative performance evaluation.

A more recent study by Murphy and Sandino (2009) propose that the availability of compensation consultants can also influence a firms’ decision to use relative performance evaluation in their executive compensation. The motivation is that compensation consultants have access to exclusive information on how competitors compensate their executives and industry-wide compensation schemes. Therefore, the availability of professional advices from compensation consultants may influence firms’ decision to use relative performance evaluation. Albuquerque (2009) argues that the availability of similar peers is also one of the potential factors that may influence the use of relative performance evaluation in executive compensation. Albuquerque (2009) argues that including peers with dissimilar abilities could increase the probability of avoiding/investing in risky investments. Hence, it is important to benchmark a firm relative to other firms with similar abilities.

Table 2 Summary Factors that may affect a firms’ decision to use relative performance evaluation

Introduced by

Factor

Prediction

Holmstrom 1982

Common risk

Firms are more likely to use relative performance evaluation when their performance is more heavily influenced by common risk

Aggarwal and Samwick (1999)

Industry competition

Firms that are facing a more competitive environment are less likely to use relative performance evaluation due to the concern that relative performance evaluation may encourage destructive competition

Murphy (2001)

Growth opportunity

Firms with high growth opportunities are more likely to adopt relative performance evaluation

Core et al. (1998)

Bertrand and Mullainathan (2001)

Sautner and Weber (2005)

Garvey andMilbourn (2006)

Ownership structure and board attributes

Relative performance evaluation is positively associated with the quality of corporate governance. The higher the quality the more likely relative performance evaluation is observed

Garvey and Milbourn (2003)

Executive attributes

Theoretical benefit of relative performance evaluation is diminished to the extent that executives can hedge market risk. Firms with healthier and older CEOs (greater self-hedging ability) exhibit weaker evidence of relative performance evaluation.

Rajgopal et al.(2006)

Albuquerque (2009)

Firm size

Firms are less likely to filter out industry and market-wide performance when compensating more talented CEOs. Hence, less frequent use of relative performance evaluation is observed among larger firms

Rajgopal et al. (2006)

Firm performance

Better performing firms exhibit less evidence of relative performance evaluation use.

Murphy and Sandino (2009)

Compensation consultants

The availability of professional advices from compensation consultants may influence firms’ decision to use relative performance evaluation. Compensation consultants have access to proprietary information on industry-wide compensation practices and hence can facilitate the structuring and execution of relative performance evaluation contracts

Albuquerque(2009)

Availability of similar peers

Firms are less likely to use relative performance evaluation if potential peers exhibit dissimilar ability

Misspecifications and incorrect assumptions

Beside the explanation of factors that could influence the use of relative performance evaluation there is also another explanation. The other possible explanation for the mixed results, lack of support for relative performance evaluation in executive compensation, is proposed by Murphy (1999), Bannister and Newman (2003) and (Albuquerque 2009). According to them the reason behind the mixed results in observing relative performance evaluation in executive pay is due to the implicit approach used by prior studies to detect relative performance evaluation. The proposition is that the implicit test, which is mostly used in prior empirical studies to detect relative performance evaluation, induces model misspecifications and incorrect assumptions that consequently lead to measurement errors.

Bannister and Newman (2003) show that assumptions about performance measures are for a high percentage incorrect in relative performance evaluation studies, which consequently lead to measurement errors. They find that 28% of the firms in their study make use of relative performance evaluation. Out of this percentage 44% uses stock return as its performance measure. Hence, in a random sample only 12% (0.28*0.44=0.12) would contribute to the detection of relative performance evaluation in a study which uses stock return as performance measure. The other 16% (0.28*0.56=0.16) of firms which uses another performance measure would not contribute anything to the probability in observing relative performance evaluation in a study which uses stock returns as performance measure. Another study which argues that the implicit test induces measurement errors is a study by Albuquerque (2005, p.6) propose that the evidence on the use of relative performance evaluation is mixed because prior studies specified the wrong peer group as benchmark. Most prior relative performance evaluation formed the peer group by combining all the firms with the same two-digit SIC industry code. However, Albuquerque (2009) argues that firms with the same two-digit SIC code may not be the most suitable peer group.

Hence, researchers (Murphy, 1999; Bannister and Newman, 2003; Albuquerque, 2005; Albuquerque, 2009 among others) argue that relative performance evaluation studies should include details of explicit relative performance evaluation contracts.

SEC’s new disclosure rule can help to shed new light on relative performance evaluation research

A new disclosure rule by the Securities and Exchange Commission can help to shed new light on relative performance evaluation research. This new disclosure rule, which became effective on 15th of December 2006, obliged firms to provide a “Compensation Discussion and Analysis” report in their proxy statements. In practice, the SEC required all firms to provide details about the process of selecting performance targets. Firms are also obligated to provide an evaluation of how objective compensation was determined based on these performance targets. As a result, firms which claim to use relative performance evaluation in their compensation schemes are required to disclose relative performance evaluation contract details in the “Compensation Discussion and Analysis” report. This new rule makes it possible for future research to explicitly test the proposed explanations of mixed results in prior empirical studies on relative performance evaluation.

Discussion and Future recommendations

The question remains: how can the new disclosure rule help researchers to shed new light on relative performance evaluation in executive compensation?

As mentioned previously, the probability of the use of relative performance evaluation in executive compensation may be affected by different factors. Although these factors have been examined in prior empirical studies and results indicated the affective power of these factors a lot is still unknown about these factors. Prior studies most of the time examined the influence of these factors on relative performance evaluation once at the time due to the limitation of the implicit test. In addition, prior studies did not control for other factors that could also affect a firms’ decision in using relative performance evaluation, which could impose endogeneity problems, in their study results (Gong et al., 2010. p.19). The new disclosure rule obligates firms to provide information which can be used to test the affective power of each of these factors simultaneously.

Future studies should make great use of the new data and closely examine how these economic factors, independently and collectively, affect the decision of a firm to use relative performance evaluation in their executive compensation schemes. In addition, future studies should examine the affect of these economic factors on each of the different types (base salary, an annual bonus, stock options and long-term incentive plans) of executive compensation that is based on relative performance. An economic factor might be highly influential in determining the use of relative performance evaluation in for example an annual bonus that is based on relative performance while the same factor might be less influential in for example indexed stock options.

A study by Gong et al. (2010) tested the affective power of these factors simultaneously and got some interesting results. MISSCHIEN ERBIJ ZETTEN

Another way in which the new disclosure rule offers researchers the possibility to enrich existing relative performance evaluation literature with new insights is that due to the new rule researchers can get access to information with respect to executive compensation which could be used to examine before and after tests.

As mentioned previously, some researchers argue that the implicit test and its constraints is the reason behind the mixed results of studies on relative performance evaluation (Murphy, 1999; Bannister and Newman, 2003; Albuquerque, 2005; Albuquerque, 2009 among others).

The new disclosure rule obligates firms to offer data, like the peer group used in relative performance evaluation, to investigate this claim. Future studies can examine this claim by performing before the new SEC disclosure rule and after the new SEC disclosure rule tests.

For example, most of the prior studies, see table 1, used firms with the same two-digit SIC code as peer group in their studies due to the lack of details provided by the firms. However, since the new disclosure rule obligates firms to provide explicit details with respect to the peer group in relative performance evaluation future studies can use this new data to test whether including explicit details about the peer group really does increase the power of the results in prior studies. Researchers can replicate the prior studies and replace the two-sic digit firms used in the tests before the new disclosure rule by the actual peer group and compare the results of the implicit approach with the explicit approach.

Also, Bannister and Newman (2003) documented that the chosen performance measure in prior studies may lead to measurement errors. Since the new disclosure rule provide explicit performance measure details future studies can examine whether including explicit performance measure details increases the power of relative performance evaluation studies. Researchers can replicate past studies and replicate the assumed measure by the actual explicitly stated measure and compare the results of these studies to see whether including explicit details increases the pay-for-performance sensitivity.

Figure one illustrates this.

CREATE TIMELINE

Conclusion

This paper examined the question: Is there relative performance evaluation in executive compensation. While this question seems straightforward and easy to answer the opposite is true.

Based on a thorough literature review I conclude that researchers are still inconclusive and unclear about the existence or absence of relative performance evaluation in executive compensation. Results of prior studies that examined the presence of relative performance evaluation in executive compensation are mixed.

Researches offered several possible explanations for the mixed results in prior studies. One of them is that the probability of using relative performance is dependent on several factors. The more these factors are in favor of relative performance the higher the probability of observing relative performance in executive compensation. Past studies have made an attempt to identify these factors. However, the true affective power of these factors are still unclear because prior studies tested the affect of these factors one at the time and did not control for other factors that also might have an influence. Another possible explanation is because of the implicit approach in prior studies. According to some researchers this approach induces misspecifications and wrong assumptions in prior empirical studies.

A new SEC disclosure which became effective on 15th of December 2006 rule may offer researchers the possibility to examine these claims and gain further insights in relative performance evaluation. This new disclosure rule obligate firms which claim to use relative performance to provide explicit relative performance details. This new data offer researchers the possibility to examine the affective power of the factors that influence a firm’s decision to use relative performance independently and as a whole. In addition, researchers replicate the prior implicit studies and incorporate explicit relative performance details to compare the power of the implicit and explicit approach.

EXPLICIT relative performance evaluation

Explicit: see Gong et al.

Only a few studies have examined explicit relative performance evaluation use in executive compensation contracts

(Murphy 1999, 2001; Bannister and Newman 2003; Carter et al. 2009). Murphy (1999) examines

177 large U.S. firms included in the 1997 Towers Perrin survey, and reports that 28.8 percent of his sample use relative performance evaluation in annual incentive plans. Bannister and Newman (2003) examine the proxy disclosures of 160 Fortune 250 firms in fiscal years 1992 and 1993, and provide descriptive statistics about relative performance evaluation contracts used by 45 firms in their sample.

Gong p2 We provide evidence that the traditional implicit test is unable to detect relative performance evaluation use, even among firms that claim to use relative performance evaluation in setting executive pay.

In particular, our finding of a biased selection of relative performance evaluation peers

supports the conjecture that self-serving behavior of management could compromise the theoretical benefits of relative performance evaluation (e.g., Gibbons and Murphy 1990; Murphy 2001).

Gong. p4

We demonstrate that using the disclosed relative performance evaluation peer groups, instead of firms with similar size and industry, can significantly improve the power of the implicit test to detect relative performance evaluation use. This finding highlights the importance of incorporating explicit relative performance evaluation contract details in testing the use of relative performance evaluation.

Differences in study:

Different Samples,

Different industries,

See Gibbons and Murphy (1990) for differences between Antle and Smith (1986)

Print Email Download Reference This Send to Kindle Reddit This

Share This Essay

To share this essay on Reddit, Facebook, Twitter, or Google+ just click on the buttons below:

Request Removal

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please click on the link below to request removal:

Request the removal of this essay.


More from UK Essays

Doing your resits? We can help!