Do Managers Trade on Public or Private Information? Evidence from Fundamental Valuations

Abstract

Using accounting-based (residual income) valuations, this study examines the extent to which abnormal returns after insider share trades are explained by private information versus mispricing of public information. For a sample of insider trades in the Netherlands (1999–2008), I find that managers' share purchase decisions are associated with positive future abnormal returns as well as equity undervaluation. Even though undervaluation results in predictable price increases, positive abnormal returns following purchases persist after controlling for fundamental valuations. Thus, this study provides evidence on the sources of managers' personal trading gains and suggests that positive abnormal returns after insider share purchases reflect both private information and managers' responses to market mispricing of public information.

Acknowledgements

This study is partly based on my dissertation completed at the University of Amsterdam. I appreciate the helpful comments and suggestions from an anonymous reviewer, Joost Impink, and Laurence van Lent (editor). I also thank seminar participants at the University of Amsterdam, the European Accounting Association annual meeting, and the Monash University PhD Accounting and Finance Symposium in Prato for helpful comments on earlier drafts.

Notes

Smaller firms are likely associated with greater mispricing of public information due to greater market uncertainty about the implications of firm-specific information for future earnings and current value. For example, Zhang (2006) shows that smaller firms are associated with greater price drifts after information events, suggesting greater under-reaction for firms with greater information uncertainty. Similarly, evidence in the post-earnings announcement drift literature suggests that smaller firms are associated with greater under-reaction (e.g., Bernard and Thomas, 1989) and, hence, more mispricing of public earnings information.

Jeng et al. (2003) find that, even though US insiders earn significant abnormal returns on their share purchases, the expected costs to shareholders of trading against informed insiders are relatively low at 10 cents per $10,000 transaction. They argue that outside investors (Jeng et al., 2003, p. 455) ‘have little to fear from these reported transactions, because insider trades make up but a tiny portion of the market.’ Hence, the benefits of allowing corporate insiders to trade their firms' shares, and convey their private information to the market, may outweigh the relatively low costs to outside investors.

See section 3 for further discussion on the alternative interpretations of variation in B/P ratios and prior returns.

Rozeff and Zaman (1998) and Jenter (2005) also examine the association of insider trades with prior returns and B/P ratios. Other studies have examined specific situations in which insider trades are driven by market mispricing. For example, Ali et al. (2011) find that US insiders respond to mispricing resulting from price pressure induced by mutual funds. Kolasinski and Li (2010) show that US insiders benefit from investor under-reaction to earnings announcements, while Core et al. (2006) find US insider trading activity to be associated with trading strategies based on the accrual anomaly. In contrast to these studies, I focus on the effect of mispricing on insider trading in general and further identify the implications of this effect for the abnormal returns after average insider trades.

The studies in this section focus on general associations between information and insider trades. Other studies have focused on specific settings such as earnings announcements (Sivakumar and Waymire, 1994), management forecasts (Cheng and Lo, 2006), annual/interim report filings (Huddart et al., 2007), restatements (Badertscher et al., 2011), bankruptcies (Seyhun and Bradley, 1997) or mergers and acquisitions (Seyhun, 1990). My study focuses on average insider trading behavior.

This comparison is analogous to comparisons between ‘active’ and ‘passive’ insider trading (e.g., Huddart et al., 2007). When in possession of private information, passively delaying purchases (sales) after bad (news) events is ex ante less costly for a manager compared with actively purchasing (selling) before good news (bad news) events, due to the lower risk of litigation.

For a comprehensive comparative legal analysis on insider trading in the UK, Germany, France, Spain and the Netherlands, and the role of European harmonization, see Welch et al. (2005).

For primary insiders, who by virtue of their relation with the issuer are privy to private information, the legal question of whether these persons knew or should have reasonably known that they possessed private information when trading is not relevant (AFM, 2010).

Prior to the Sarbanes-Oxley Act of 2002 (SOX), the reporting deadline for US insider transactions was also equal to the tenth day of the month following the trading.

Reporting of insider trades is timely in both settings, although transactions are slightly more timely in the US (two business days) versus the Netherlands (five business days). Another minor difference is that, since 2002, Dutch ‘top executives’ are required to report their transactions as soon as possible, ‘without delay’ (Degryse et al., 2009). Given the focus of my study on long-run returns after insider trades, this difference in reporting timeliness is unlikely to affect the generalizability of my results.

Degryse et al. (2009, p.11) also posit that ‘[t]he principles of law concerning insider trading are similar to those found in the U.S. or in the U.K.’

Clean surplus accounting requires that all changes in shareholders' equity, apart from transactions with shareholders (dividends, share issues, or repurchases) flow through the income statement. Other comprehensive income or ‘dirty surplus’ items, which are directly recognized in equity without affecting net income, violate this requirement.

Frankel and Lee (1998) and Ali et al. (2003) show that V/P ratios based on the RIV model predict future abnormal stock returns over and above B/P ratios. Given that B/P ratios are embedded in V/P ratios (Ali et al., 2003), this evidence can be interpreted as the superior ability of RIV model-based valuations to identify market mispricing.

In addition, one could argue that the practical power of RIV depends on the validity of the clean surplus assumption. However, it is unclear to what extent violations of clean surplus bias valuations in practice. Specifically, the randomness and transitory nature of most dirty surplus items (e.g., foreign currency translations) suggests these items should not matter in expectation (Penman, 2001). In this regard, Isidro et al. (2006) find only weak empirical evidence of an association between dirty surplus items and valuation errors based on RIV for the US, while they find no association for samples based on France, Germany, or UK firms. Hence, even though clean surplus is often violated, such violation appears to have limited effects on practical valuations based on RIV.

Penman and Sougiannis (1998) and Francis et al. (2000) show that practical implementations of RIV outperform DCF and DDM valuations. Assuming an efficient market, they find that RIV valuations are associated with lower valuation errors and explain more variation in prices than valuations based on the other models. Jiang and Lee (2005) test the dynamic stock price implications of RIV versus DDM and find that book values and earnings in RIV contain more useful information for equity valuation than dividends in DDM alone.

While relaxing the market efficiency assumption, Frankel and Lee (1998) find that V/P ratios based on RIV are cross-sectionally associated with future abnormal stock returns. This finding suggests that the relative magnitude of a V/P ratio based on RIV is indicative of the extent to which a firm is under- or overpriced by the market at a specific point in time. Barniv et al. (2010) present evidence suggesting RIV can also be used to identify mispriced securities in international settings. Lee et al. (1999) show that the time series of aggregate Dow 30 RIV valuations is cointegrated with index levels and that value-to-price ratios based on RIV are associated with future market returns. Overall, these findings suggest valuations based on RIV can be used to detect instances of mispricing.

See Easton (2004) for an intuitive derivation of this model. See Brief (2007), Ohlson (2005), Ohlson and Juettner-Nauroth (2005) and Penman (2005) for more theory and discussions on the benefits and drawbacks of the model.

Penman (2005, p. 376) argues that ‘despite decades of endeavor in research in finance, we do not know how to estimate the cost of capital that features in both RIV and AEG formulas. To be honest, it is a speculation and fundamental analysts warn of building speculation into a valuation. A method that puts less weight on this speculative component is to be preferred, all else equal. Of the two, the AEG valuation is more subject to this criticism.’

Penman (2005) presents descriptive evidence suggesting that valuations based on AEG are less accurate and more volatile compared with valuations based on RIV. Jorgenson et al. (2011) find similar results while varying assumptions and forecast horizons.

All results presented in this study are qualitatively highly similar when using alternative rates for cost of equity and/or growth and using firm-specific discount rates based on CAPM.

Although the book value of equity data may not be available to the market at the time of the earnings announcement, results are qualitatively similar when using ‘synthetic’ book values as in Lee and Swaminathan (1999) or when book value is assumed to be available to the market four months after fiscal year end.

For example, a manager that exercises one stock option and subsequently sells the acquired share has three records on one day in the register reflecting the same disposition trade. First, the manager ‘sells’ an option. Second, the manager buys a share at exercise price. Third, the manager sells the share at a price greater than the exercise price. This example illustrates the importance of carefully examining the records in the AFM register. A failure to do so would result in the purchase of the share at exercise price being treated as purchase, whereas in fact this transaction relates to a sale.

Prior to October 2005, issuing companies were treated as corporate insiders under Wte 46b similar to officers, directors, and large shareholders. Hence, the register includes public disclosures of repurchase and share issue transactions by firms. Although these transactions are initiated by firms' management, they are not conducted for managers' personal accounts and hence are eliminated for the purpose of this study.

However, the relation between the two variables is positive when focusing on Q2 through Q5. Untabulated tests suggest that the Pearson (Spearman) correlation equals 0.350 (0.353) when observations in Q1 are excluded. Nevertheless, these correlations are modest and suggest the two variables capture different underlying constructs.

Although the main interest of this study is in the six- and 12-month holding periods due to the focus on insider trades, I also examined the abnormal returns for 18- and 24-month holding periods (not tabulated). For the 18-month holding period, the abnormal return differential increases to 14.9%, while it increases to 19.2% for the 24-month holding period.

The 2.1% is calculated as follows. The marginal effect of 0.00524 (computed in STATA using the ‘margins’ command) indicates the predicted change in BUY when VPQ changes by one unit. Given that VPQ takes on values between 1 and 5, the marginal effect of switching from the smallest to the largest group is 4 × 0.00524 = 0.02095 (=2.1%).

(2/4 × 0.02095)/0.027 = 0.388.

Note that in model 1 the estimated coefficient on BUY is zero in 12 cases (=75–42–21). This is because the sample is restricted to firms having any buying or selling over the sample period. Restricting the analyses to the 63 firms with insider buying or 64 firms with insider selling does not affect the presented results. The same applies to the results based on the sample of smaller firms.

Again, as with the RIV implementation, the discount rate is set equal to 10% for all firms. Results are unaffected by changing the constant discount rate and growth rate assumptions, or using firm-specific discount rates based on CAPM.

Untabulated analysis reveals that the (Spearman rank) correlation between VP and VP_AEG for the sample of 8199 firm-months equals 0.37 (0.31).

Untabulated tests suggest that increasing the holding period to 18 or 24 months provides similar insights. The return differentials (Q5-Q1) equal 1.2% and 0.3% for the 18- and 24-month holding periods, respectively.

Jorgenson et al. (2011) find that AEG valuations implemented using longer forecast horizons improve the model's ability to explain observed stock prices, but still RIV estimations using similar forecast horizons outperform AEG estimations. In untabulated tests, I find that increasing the forecast horizon for AEG to three or five years does not improve the model's ability to detect mispriced securities.

Additional information

Notes on contributors

David Veenman

Paper accepted by Laurence van Lent.