The Effects of Table Versus Formula Presentation Formats on Investors’ Judgment about Executive Compensation

Abstract

We investigate how the data presentation format of executive compensation disclosures affects investors’ compensation- and investment-related judgment and decisions. We examine the effects of these formats and their underlying mechanisms via laboratory experiments. We compare bonus schemes presented in the table and formula formats and find an interaction effect between data presentation format and compensation favorability. When a manager receives unfavorable compensation due to missed performance targets, the table format is easier to understand than the formula format, whereas the latter appears more scientific than the former. These two effects work in opposite directions and therefore, result in an insignificant difference in approval ratings when the table, versus the formula, format is used. However, when the manager receives favorable compensation for beating performance targets, investors do not pay much attention to the disclosure’s understandability but still favor the formula format, as it has a scientific appearance. Therefore, we find that investors’ approval ratings are higher when using the formula format than the table format. We contribute to several research areas such as information understandability, the halo effect, compensation disclosures, and shareholder voting. Our findings also have practical implications for both managers and investors.

Keywords:

• Data presentation format
• Understandability effect
• Halo effect
• Say-on-pay

Acknowledgements

We appreciate the valuable comments and suggestions from Victor Maas (the Editor) and the two anonymous reviewers. We also thank Fei Du, Hun-Tong Tan, Steve Wu, Liu Zheng, Wen Zhou, seminar participants at Central University of Economics and Finance (China), Fudan University (China), Shanghai University of Finance and Economics (China), The University of Hong Kong, University of International Business and Economics (China), and participants of 39th EAA annual congress for helpful comments. We thank Xuejiao Liu for research assistance work. The work described in this paper was supported by grants from the National Natural Science Foundation of China [Project No. 71502096, 71972120] and the Research Grants Council of the Hong Kong Special Administrative Region, China [Project No. HKU 744009H].

Supplemental Data and Research Materials

Online Appendix A. Examples of bonus scheme presentation formats.

Online Appendix B. Cloze test example.

Data Availability

Contact the authors.

Notes

1 https://www.theinvestmentassociation.org/publicregister.html.

2 The new EU directive of compensation disclosure did not take effect until June 2019. Thus, detailed bonus information, such as on performance targets or payout structure, is not yet publicly available for most EU companies.

3 Tables and/or formulae are also used to present information on equity-based plans. We expect the predicted effect of the disclosure format to also be applicable to the context of equity-based plans or conditions wherever tables and formulae are present. We conduct our experiment in the setting of bonus disclosures, as the pay–performance relationship in a cash bonus scheme is more straightforward and suited to the scenario in laboratory experiments relative to other complicated long-term incentive plans (Conyon et al., 2000; Murphy & Jensen, 2011).

4 We focus on the table and formula formats, without considering the text-only format in the main analysis of this study. Because of lack of theory to formally predict the difference in understandability between the text-only format and the formula format or the difference in the halo effect between the text-only format and the table format, the aggregated effect of the text-only format on investors’ approval ratings is unclear (compared with the table and formula formats). In addition, psychology and accounting studies document that text is less effective in presenting numerical information than both tables and graphs (Feliciano et al., 1963; Kelly, 1993; Washburne, 1927). The US Securities and Exchange Commission (SEC) explicitly suggests that companies use tables, schedules, charts, and graphics rather than plain text to present financial data (SEC, 1998). Furthermore, a growing body of literature on information readability provides better methodologies to explore the effectiveness of textual disclosures. As such, it is less interesting to examine the text-only format effects in our experiments. Besides, the number of firms using the text-only format is decreasing and the number of firms using the table and formula formats is increasing in the Standard & Poor 500 firm-year sample (Table 1). Thus, we do not have a hypothesis regarding the effectiveness of the text-only format, but still include it in our analysis due to its high frequency in practice. According to Table 1, 321 of the 1,237 bonus schemes (25.95%) in our sample are presented in the text-only format.

5 We manually collect data on executive compensation disclosures in the United States (but not in the United Kingdom or other EU countries) due to the availability of compensation details and little room for manipulation under the strict US compensation disclosure rules (Baird & Stowasser, 2002).

6 In our manually collected sample, 74.11% of the table format bonus schemes present the actual performance together with the target performance, in contrast to 45.11% of the formula format bonus schemes (χ² = 65.82, p = 0.00). Of the table format bonus schemes, 91.96% have a ‘goal’ label. However, only 54.47% of the formula format bonus schemes have such a label (χ² = 167.67, p = 0.00).

7 The degree of information processing complexity increases with the number of steps executed (Bonner, 1994). According to the calculation formula of the formula format, basic information processing requires four steps: (1) translating corporate performance into payout factors with reference to the pre-set goals, (2) multiplying each payout factor by its weight, (3) adding the payout factors according to the calculation formula provided to obtain an overall achievement factor, and (4) multiplying the target bonus payment by the overall achievement factor. By contrast, calculating total compensation in the table format requires only two steps: (1) translating corporate performance into the percentage of the base salary payable for each performance measure and (2) adding them together. Thus, the pay–performance relationship in a bonus scheme is more straightforward in the table format than in the formula format.

8 A favorable (unfavorable) compensation result refers to when a company’s actual performance beats (misses) the performance targets and the manager receives a bonus award above (below) the target. In our experiment, the bonus scheme has two performance measures. Huang et al. (2013) report that 32% of the 7,550 bonus schemes for 1,626 non-financial firms from 2006 to 2011 use two performance measures, demonstrating a higher frequency than the use of one measure (26%) and three measures (26%). Within our manually collected sample, 154 (37%) of the 2013 bonus schemes use two performance measures. Furthermore, the actual performance beats both performance targets in 67 (44%) of the 154 bonus schemes, misses both performance targets in 38 (25%) of the schemes, and exceeds one target while missing the other target in 32% of the schemes. In our experiment, the performance targets of both measures are beaten (missed) under the favorable (unfavorable) outcome condition. We do not consider situations in which one target goal is beaten but the other is missed. As documented by Tan et al. (2015), the effect of textual information readability is contingent on benchmark performance consistency. Inconsistent relative performance across multiple benchmarks (beating one target goal but missing the other) leads to a greater readability effect than consistent benchmark performance (consistently beating or missing multiple benchmarks; Tan et al., 2015). Therefore, settings in which both targets are consistently beaten or missed may further eliminate the influence of textual information readability. This helps us focus on the effect of data understandability rather than text readability.

9 Ethical approval is obtained from the Human Research Ethics Committee for Non-Clinical Faculties at the university where the experiment is conducted (Reference No. EA160513).

10 The ANOVA tests show that the response means of working experience and education (i.e., completed courses) are not significantly different across the treatment conditions (p > 0.10). Although there are more participants with stock market investment experience in the formula format condition (mean = 0.91) than in the table format condition (mean = 0.63; F = 7.64, p = 0.01), the ANOVA test results for Hypotheses 1, 2, and 3 hold after including investment experience as a covariate.

11 Huang et al. (2013) indicate that earnings-based measures, such as earnings per share (EPS), earnings before interest and tax (EBIT), and operating income, are the most frequently used performance measures in executive bonus schemes (adopted by 74% of the bonus schemes in their sample), followed by sales (35%) and cash flow measures (16%). Therefore, we use two representative performance measures, namely EBIT and sales, in our setting. In the within-subjects test, we use EPS and operating cash flow as alternative performance measures in the bonus scheme to enhance the external validity of our results (Bracht & Glass, 1968).

12 Rennekamp (2012) and Tan et al. (2015) both manipulate news valence in their studies by beating or missing performance benchmarks by the same magnitude rather than the same percentage. Following the literature, we manipulate the outcome favorability by beating or missing the target goals of the performance measures by the same magnitude.

13 Studies validate the cloze test procedure as a superior measure of understandability (Smith & Taffler, 1992). A less understandable format should lead to a lower score than a more understandable format. We also use the cloze test to examine the information consistency across different formats by checking whether any of the participants complete the cloze test with 100% correct answers. If at least one participant receives full marks in the test, we can infer that the information is equivalent between the formats. This allows ruling out the alternative explanation that the investor reactions observed in the experiment are caused by information differences rather than format differences.

14 All the reported p-values are two-tailed unless otherwise stated.

15 The success rate of the manipulation check questions is relatively low. A probable reason is that the participants’ perceptions of bonus disclosures are confounded by the presentation of the background information. Including/excluding the participants who fail the manipulation check questions does not change the results. The following analysis is based on all the participants. Of the participants who receive the text-only format, 61% correctly indicate that they observe neither a table nor a formula. We include the text-only format condition in our analysis, but the 2 × 2 ANOVA tests conducted after excluding it do not change our main results regarding the table and formula formats.

16 The cloze test requires the participants to translate the presentation of the bonus scheme with key numerical information from the format they view in Envelope A into one of the other two formats. The participants in the table format condition are required to translate the bonus information into either (1) the formula format or (2) the text-only format. The cloze test scores do not differ significantly between the participants translating the table format into the formula format (mean = 4.44) and those translating the table format into the text-only format (mean = 5.12; t = 1.41, p = 0.17). The participants in the formula format condition are required to translate the bonus information into either the (1) table format or (2) text-only format. The cloze test scores do not differ significantly between the participants translating the formula format into the table format (mean = 3.50) and those translating the formula format into the text-only format (mean = 3.06; t = 0.62, p = 0.54). The results of the understandability effect hold when only the participants who translate the table format into the text-only format and the formula format into the text-only format (i.e., the target format of the format translation cloze test is the same) are included. The mean test score for the table format condition is significantly higher than that for the formula format condition (F = 12.56, p = 0.00).

17 We use the readability measures implemented by Tan et al. (2015) to assess the readability of the textual information in the bonus disclosure. These measures are based on the SEC’s plain English principles (SEC, 1998) for textual disclosures. In the experiment, the participants are asked to indicate the extent to which they think the bonus scheme information is difficult to read/understand/process on 11-point scales from 0 (‘not at all difficult’) to 10 (‘extremely difficult’). We conduct a factor analysis and the three readability measures load onto one factor (Cronbach’s alpha = 0.81). We do not find a significant difference in information readability between the formats (F = 0.83, p = 0.37). We test the readability of the textual information to exclude the alternative explanation that the difference in understandability between formats is due to the variance in text readability rather than the data presentation format.

18 The participants also complete a within-subjects test in which they assess a company’s bonus scheme, where the compensation result is either favorable or unfavorable. EPS and operating cash flow are the two performance measures. Each participant reads identical bonus information presented in the table, formula, and text-only formats. The order in which the formats are presented is counterbalanced, and the participants are required to assess all three formats using 11-point scales. Analysis shows that under the favorable compensation condition, the approval rating of bonus awards is significantly higher for the formula format (mean = 0.29) than for the table format (mean = 0.02; t = 1.98, p = 0.05). Under the unfavorable compensation condition, the approval rating does not differ significantly when using the table (mean = -0.02), versus the formula (mean = 0.15; t = 1.51, p = 0.14), format. The within-subjects test results replicate the results of the between-subjects test, providing complementary evidence of how bonus scheme presentation formats affect investors’ judgment through the understandability and halo effects.

19 The finding that the understandability of the text-only format is not significantly different from that of the table format is inconsistent with our prediction, probably because the understandability effect of the data presentation is confounded with the readability effect of the textual descriptions in the text-only format (Bloomfield, 2008; Tan et al., 2015).

20 In the experiment, the participants are asked to indicate the extent to which they understand the bonus payment decision, on a scale from 0 (‘do not understand at all’) to 10 (‘completely understand’), and how confident they are in their understanding, on a scale from 0 (‘not at all confident’) to 10 (‘extremely confident’).

21 We use the factor score of the scientific appearance and the effectiveness of the bonus scheme to keep the measurement scale of the variables consistent in the model. We reperform the SEM analysis using the average rating of the investors’ assessments of scientific appearance and motivation effectiveness to measure the halo effect. The results are qualitatively the same as the reported results.

22 We predict that the formula format is less understandable than, and has a positive halo effect relative to, the table format. In the favorable compensation condition, the results of our model support these predictions (Link 1, coefficient = -0.58, p = 0.04; Link 2, coefficient = 0.88, p = 0.00; Figure 3, Panel A). However, the coefficient of the relationship between the format and understandability or the halo effect is not significant in the unfavorable compensation condition (p > 0.30). This is probably due to a variable measurement problem. A well-known rule of thumb in SEM analysis is that every variable should be measured with at least two indicators to identify a model (Kenny et al., 1998). Our use of only one indicator for the presentation format probably decreases the degree of model fit and accounts for the insignificant coefficient results. Alternatively, both effects may lose efficacy, such that we observe no difference in investors’ judgment between the formats in the main experiment. To further examine the underlying mechanisms of the format effect in the context of an unfavorable compensation result, we conduct a supplementary experiment that provides evidence that is complementary to our predictions (Section V provides a detailed description).

23 AMT is an increasingly popular source of experimental data for social scientists because the subject pool is large, readily accessible, and at least as representative as more traditional subject pools (Paolacci et al., 2010). Furthermore, studies run on AMT reliably replicate a wide range of judgment and decision-making findings, with accounting research increasingly turning to AMT to recruit participants when the experimental task does not require specialized accounting knowledge (Asay et al., 2017; Paolacci et al., 2010; Rennekamp, 2012). In this study, the supplementary experiment replicates some of the findings of the main experiment (in which the participants are Master’s students), thereby demonstrating the validity of AMT participants.

24 The participants are asked to assess the vividness of the bonus disclosure in the experiment. Analysis shows that the table format receives a higher evaluation than either the formula format (t = 2.81, one-tailed p = 0.00) or the hybrid format (t = 3.27, one-tailed p = 0.00), both of which have a positive halo effect. Vividness does not have a significant effect on the investors’ approval ratings (F = 2.34, p = 0.13). Meanwhile, the format effect on the investors’ approval ratings remains significant after including vividness as a covariate (F = 2.84, p = 0.06). The results indicate that vividness is not confounded with the theorized halo effect and does not mediate the format effect on investor judgment.

Additional information

Funding

The work described in this paper was supported by grants from the National Natural Science Foundation of China [project numbers 71502096, 71972120] and the Research Grants Council of the Hong Kong Special Administrative Region, China [project number HKU 744009H].