Market Reactions to the Regulation of Executive Compensation

Abstract

This paper investigates equity market reactions to the regulation of executive compensation. We exploit a natural experimental setting in Germany, where recent legislation introduces restrictions on the amount and on the components of board executive compensation packages, and invokes liability for the supervisory board in case of inappropriate remuneration arrangements. We use this exogenous shock to the contracting environment to infer market perceptions of the usefulness of the regulation. Using event study methodology, we investigate market reactions for the first-time announcement of regulatory intent and for a pooled sample of seven events leading to the adoption of the law act. We find weak evidence of an average negative market reaction to the proposed regulation. Multivariate analyses reveal that firms which were particularly affected by the regulation because board members received high abnormal remuneration experienced larger stock price discounts on average. Consistent with this, we find a positive relation between pay-performance sensitivity and the equity market reaction. Taken together, these findings indicate that the regulation was not considered beneficial from a shareholder perspective. This result is consistent with the market perceiving the regulation of executive compensation to impose potentially inefficient contractual arrangements for some firms.

Acknowledgements

For helpful comments and suggestions, we thank Miles B. Gietzmann, Philip P. M. Joos, Nico Lehmann, Serena Morricone, Bharat Sarath, participants at the HEC Lausanne Accounting Seminar, at the Workshop on Accounting and Regulation in Siena 2013, at the Annual Congress of the European Accounting Association in Tallinn 2014, at the JIAR Conference in Hong Kong 2014, and at the Annual Meeting of the German Academic Association for Business Research in Leipzig 2014, two anonymous reviewers, and Robert F. Göx, the guest editor. Also, we are indebted to Christian Drefahl for his assistance in collecting the compensation data. All remaining errors are ours.

Notes

1 While provisions (1) and (2) apply to all incorporated firms, (3) is restricted to listed corporations.

2 Firms disclose in their annual reports how they implemented the VorstAG. For example, in the 2009 annual report of Volkswagen, a German car manufacturer, it is outlined that ‘the remuneration structure is focused on ensuring sustainable business growth in accordance with the [VorstAG]’ (Volkswagen, 2010, p. 112). In addition to fixed compensation and a business performance-based bonus, Volkswagen implemented the VorstAG by adopting a new long-term incentive programme starting in 2010, the so-called Strategy 2018. Under this programme, long-term performance-related compensation of executive board members is based on the four-year average of four criteria, a customer satisfaction index, an employee index (including employment rate, productivity, and employee satisfaction), sales growth, and increases in return on sales (Volkswagen, 2010, p. 113).

3 For events that decrease the likelihood/scope of the regulation, MAR is multiplied by negative one.

4 Joos and Leung (2013) and Larcker et al. (2011) also cluster standard errors at the event date to deal with the correlation of MARs in the multi-event analysis. Because we do not have a sufficient number of clusters (e.g. Gow et al., 2010; Petersen, 2009), we bootstrap the clustered standard errors (Cameron et al., 2008; Gow et al., 2010). Results remain unchanged when using bootstrapped standard errors clustered by event and firm.

5 In addition, we also apply the procedure used by Armstrong et al. (2010) and Joos and Leung (2013) and test whether the coefficient estimates based on the event-date regression differ from coefficient estimates obtained using non-event date returns for the pooled event analysis. Our results are similar to those obtained from the second test statistic.

6 Similar approaches are applied by Core, Holthausen, and Larcker (1999) and Ferri and Maber (2013). We use the two-year average for years 2006 and 2007 to estimate the level of ABNORMAL PAY, since detailed disclosure of executive compensation only became mandatory in 2006, when the VorstAG was introduced. This restriction also applies to the estimation of PAY SENSITIVITY. Also, we do not use 2008 compensation data for two reasons. First, the 2008 financial crisis is likely to be reflected in the level of executive compensation. Second, at least for events 1–3, we need to rule out any look-ahead bias (Larcker et al., 2011), that is, we use only data in our tests that were available to the market participants when regulatory discussions occured. Still, we note that inferences are similar when using a two-year average for 2007 and 2008 instead of 2006 and 2007 when 2008 data are available to market participants and the re-estimated ABNORMAL PAY is significantly negative for the single event and pooled analysis (5% level and 1% level, respectively).

7 We also computed abnormal pay using an augmented version of model (2) that also includes FREE FLOAT and blue chip index membership (DAX) as explanatory variables. Using this augmented measure yielded essentially the same results for all our multivariate analyses as did the original measure based on model (2).

8 Results remain unchanged when restricting the sample for estimating ABNORMAL PAY to the 203 firms included in our final sample.

9 Rapp, Schaller, and Wolff (2011) observe that executive compensation components based on stock performance were only used by 37.4% of firms listed in the Prime Standard.

10 In our robustness Section 5.3, we address potential concerns with the inclusion of the fair value of stock options granted by calculating the pay-performance sensitivity based on cash compensation only.

11 To generate reasonable estimates for the industry fixed effects, six of these firms are excluded because they operate in industries with no more than three firms. For example, Core et al. (1999, p. 377) also exclude firms from industries with less than 10 firms in order to calculate reasonable estimates for their industry effect variables.

12 Additionally, we exclude another firm (Conergy) that was in financial distress from November 2007 on, requiring substantial amounts of liquidity, followed by accusations of balance sheet fraud in 2009, and accompanied by substantial losses in market value from late 2007 on.

13 Multicollinearity tests are performed for all explanatory variables in regression model (1) for the pooled sample (N = 1353) without bootstrapping. VIFs for variables other than SIZE and DAX are below 2.00 and mean VIF is 1.39. Thus, inferences are not affected by multicollinearity concerns (e.g. Gujarati, 2003, p. 362).

14 When we exclude financial firms from the sample, the aggregated market reaction remains negative and is statistically insignificant. We also re-estimate the overall market reaction using a one-day event window. The overall market reaction is significantly negative (10% level).

15 To shed further light on the market reactions conditional on the level of executive compensation, we form five portfolios based on quintiles for our test variable ABNORMAL PAY. In this analysis, which is conducted both for the first-time announcement and for the pooled sample, we find that market reactions are not significant for the first three portfolios. For the fourth portfolio, market reactions are significantly different from zero on the first-time announcement. For the portfolio with presumably highly overpaid executives (portfolio 5), market reactions are significantly negative for the first-time announcement and the pooled event analysis. This indicates that market reactions are mainly driven by firms with presumably overpaid executive board members. Also, this finding might explain why we fail to find significant market reactions based on the weighted bootstrap p-value in Table 4. The multivariate analyses in the following sections shed more light on this.

16 When estimating ABNORMAL PAY for model (3), financial firms are also excluded.

17 Larcker et al. (2011) do not report R²s.

18 We also estimate model (1) without control variables for ABNORMAL PAY and PAY SENSITIVITY separately, and for both test variables. For all re-estimations, results remain basically unchanged.

19 When comparing coefficients from model (1) to coefficients from 21 non-overlapping non-events in 2008 results are confirmed. The test statistic indicates that ABNORMAL PAY is different from ABNORMAL PAY on non-events at the 5% level.

20 In additional tests, we re-estimate model (1) without control variables for ABNORMAL PAY and PAY SENSITIVITY separately, and for both test variables. For all re-estimations, results remain basically unchanged.

21 Untabulated pooled regression estimates for the sample of financial firms confirm this result, yielding coefficient estimates in line with our full sample analyses. We also estimated models (1) and (3) for event Nos. 2–7 to rule out that results are driven by event 1, which may be confounded due to the financial crisis. We find that the coefficient on ABNORMAL PAY is significantly negative (5% level) and the coefficient on PAY SENSITIVITY is significantly positive at the 10% level (5% level) in model (1) (model (3)). These slightly weaker results are not surprising since the first-time announcement is our main event where the regulatory intent was announced for the first time.

22 Larcker et al. (2011) use a substantially larger sample and subtract from the natural logarithm of CEO compensation the natural logarithm of median annual CEO compensation for all firms in the same industry group and size quintile.

23 Similarly, our data show that for firms with positive values for ABNORMAL PAY, and thus presumably overpaid executives, mean ABNORMAL PAY is significantly higher (5% level) for firms not included in this robustness test than for firms for which the relevant CEO compensation data are available.