Peer Effects in Investment: Evidence from Early-tenure CEOs

Abstract

This study examines whether Chief Executive Officers’ (CEOs) incentives to rely on peers’ investments vary over their tenure at a firm. Newly appointed CEOs (hereinafter ‘early-tenure CEOs’) often encounter a lack of firm-specific information and are subject to rigorous evaluation. Therefore, when making investment decisions, they tend to seek more efficient and readily accessible external information, such as investment decisions made by peer firms. We find that the positive association between a focal firm’s investments and those of its peers (termed ‘peer effects’) is stronger when early-tenure CEOs manage the focal firm. Furthermore, this peer effect is stronger when managers possess greater incentives to rely on peers (i.e., lower ability and stronger monitoring) or when firms’ investment information quality is poorer (i.e., early stage of their life cycle, greater investment volatility and uncertainty). We also find that geographic proximity and the sharing of common board members or auditors may serve as mechanisms facilitating peer effects. Finally, we document improved future performance for early-tenure CEOs who align their investment decisions with those of peers. This study contributes to the existing literature by illustrating that manager-level characteristics can influence heterogeneity of peer effects and underscores the benefits of peer effects.

Keywords:

• Peer effects
• Learning from peers
• Investment
• CEO tenure
• Early-tenure CEO

JEL codes:

• G31
• M12
• M41

Acknowledgement

We thank Pepa Kraft (associate editor), Beatriz Garcia Osma (editor-in-chief), and two anonymous reviewers. We are grateful for the helpful comments of Bok Baik, Mark Clatworthy, Sunhwa Choi, Young-Soo Choi, Thierry Foucault, Iny Hwang, Lee-Seok Hwang, Sun-Moon Jung (discussant), Jeroen Koenraadt (discussant), Woo-Jong Lee, Tae Young Paik, Sebastian Richter, Jae Yong Shin, Kyle Welch (discussant), and seminar participants at Hong Kong Baptist University, Seoul National University, Sungkyunkwan University, 2018 European Accounting Association Annual Congress, 2018 Journal of Accounting, Auditing, and Finance Conference, 2018 American Accounting Association Annual Meeting, 2019 Korean Accounting Association and Asian Accounting Association Conference, and 2023 European Accounting Review Annual Conference. Ahrum Choi acknowledges financial support from Sungkyunkwan University.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data Availability

All data are publicly available from sources identified in the text.

Notes

1 Several studies suggest the existence of peer effects in various corporate policies such as capital structure, earnings management, tax avoidance, disclosure, or dividends (Adhikari & Agrawal, 2018; Bird et al., 2018; Kedia et al., 2015; Leary & Roberts, 2014; Seo, 2021).

2 In this paper, we use ‘early-tenure CEO’ to refer to a CEO who has been appointed recently at the firm. Please note that ‘early-tenure CEO’ does not indicate a CEO who works for a recently founded firm. Later, we use ‘early-stage firm’ to refer to a firm that is at the beginning phase of the firm life cycle.

3 Drawing on organizational legitimacy theory, De Franco et al. (2023) show that early-tenure CEOs attempt to gain legitimacy by mimicking the financial statements of firms with industry leaders.

4 For example, changes in growth opportunities or regulatory policies may affect the investment decisions of all firms in the same industry at the same time.

5 We examine investment peer effects after partitioning the sample into two groups based on the strength of industry shocks (i.e., strong versus weak shocks), where industry shock is measured as industry-average returns. We find that the investment peer effects among early-tenure CEOs are significant for both groups. This implies that early-tenured CEOs’ investment decisions are influenced by peers even when there is no significant industry shock, enhancing the importance of peer effects.

6 Our study provides additional contributions to Chen and Ma (2017) because we show how managerial characteristics affect investment peer effects, while they focus on the effect of firm characteristics in investment peer effects. In addition, we also show how our main results change with the factors mentioned in Chen and Ma (2017) (i.e., information advantage and information quality).

7 For example, Pan et al. (2016) show that CEO tenure is related to a cyclical pattern of investment. Cen and Doukas (2017) and Malmendier and Tate (2005) link CEO risk preference and overconfidence to investment. In addition, Kang and Kim (2020) show that founder CEOs invest more in employee relations.

8 Although not directly related to corporate investment, Hong et al. (2000) examine a similar topic among security analysts. They compare experienced analysts with inexperienced analysts and show that inexperienced analysts deviate less from consensus forecasts.

9 To be consistent with an indicator variable for early-tenure CEOs (Early), we include the CEO’s age, CEO duality, and the percentage of a CEO’s stock ownership in year t.

10 For the construction of the focal firms’ idiosyncratic returns (EquityShock) and the standard deviation of these returns (EquityRisk), see Section 3.2.

11 Since we focus on peers’ effects on a focal firm’s decision, one could argue that standard errors should be clustered at the group level (i.e., the SIC3 industry level). Our results are robust to the group-level clustering (untabulated).

12 Following Larcker and Rusticus (2010), we assess the appropriateness of our instrumental variables and report the test statistics at the bottom of Column (3). A weak identification test based on the Cragg–Donald Wald F-statistic rejects the null hypothesis that the instrument variables are weakly identified. Over-identification test using Hansen J-statistics does not reject the null hypothesis that all IVs are exogenous. Taken together, the test statistics suggest that our IVs satisfy both relevant and exclusion criteria.

13 Please note that 2SLS estimates local average treatment effects, rather than average treatment effects for the entire population.

14 This claim is empirically demonstrated by Hilary et al. (2019), who show that a strong board prevents the CEO from deviating from the consensus when executing the capital budget.

15 We express our gratitude to Peter Demerjian for generously sharing the managerial ability data available on his website. Demerjian et al. (2012) estimate ManagerialAbility through a two-stage process. In the initial stage, they employ the data envelopment approach to assess the firm-level efficiency, which quantifies the extent that firms utilize economic resources efficiently to generate revenues. In the subsequent stage, non-managerial determinants of firm-level efficiency are isolated by regressing the first-stage estimates on firm-level determinants and time indicators. See Demerjian et al. (2012) for more detailed explanation.

16 We note that, while we divide the sample first and then perform 2SLS analyses on each subsample following prior studies, the validity of this method has not received much attention in the econometric literature.

17 Our focus is to examine whether the distinct peer effects hold for certain circumstances. That is, we are interested in whether the peer effects become stronger in the early years of the CEOs’ tenure for managers with high ability and under stricter monitoring. Hence, our analyses are not focused on whether the coefficients on IV(P_Investment*Early)_it are statistically different between the two subsamples. Nonetheless, we test the difference between subsamples using the bootstrapping approach following Shroff et al. (2014) and Amberger et al. (2021), and report the test statistics at the bottom of the table.

18 Additionally, we expect that peer effects become weaker in firms with CEO duality because strong CEO leadership makes it difficult for board members to monitor or discipline CEOs (Goyal & Park, 2002). Empirical results show that the coefficient on IV(P_Investment*Early) is insignificant in firms with CEO duality, but is positively significant in firms without CEO duality (untabulated).

19 We estimate GrowthUncertainty based on one-month-ahead expected idiosyncratic return volatility as of the beginning-of-the fiscal year, estimated by best-fitted EGARCH models using at least 60 previous monthly returns. We thank Seunghee Yang for sharing the data on the expected idiosyncratic return volatility.

20 In this analysis, we discard observations in the firm life cycles of shake-out and decline according to Dickinson (2011), which results in the decrease in the number of observations.

21 Foucault and Fresard (2014) show that investment sensitivity to peers’ stock prices is higher when peers’ stock prices are more informative. Thus, we test the cross-sectional effects of the information quality of peers and find that peer effects in investments are stronger when peers’ stock price is more informative. The results are available upon request.

22 The peer effects in investments may be different between internally and externally appointed CEOs because the information set available to the two CEOs may be different. Thus, we estimate Equation (1)(1) $\begin{array}{rl}\mathrm{Investmen}{t}_{i,t}& ={\text{β}}_0+{\text{β}}_1\mathrm{IV}{(P\mathrm{\_Investment}{)}_i}_{,t}+{\text{β}}_2\mathrm{IV}(P\mathrm{\_Investment}\ast \mathrm{Early}{)}_{i,t}\\ & +{\text{β}}_3\mathrm{Earl}{y}_{i,t}+\sum {\text{β}}_{i}P\mathrm{\_Control}{s}_{i,t-1}+\sum {\text{β}}_i\mathrm{Control}{s}_{i,t-1}+{ϵ}_{i,t},\end{array}$(1) after partitioning the sample into internally promoted and externally promoted CEOs. The results show that the investment peer effect exists in both subsamples, and the magnitude of peer effects is larger in the subsample where CEOs are externally hired, compared to the subsample where CEOs are internally promoted, although the difference in investment peer effects is not statistically significant. The results are available upon request.

23 We consider the role of various reasons for CEO turnover and how this affects our results. For example, we use data on forced CEO turnover from Professor Florian Peters (https://www.florianpeters.org/data) to compare how the CEO peer effect differs between forced and voluntary turnover of previous CEOs (Peters & Wagner, 2014; Jenter & Kanaan, 2015). Similarly, we use the CEO replacement reason variable from Execuomp to compare how the peer effect differs when the CEO resigns or retires, compared to when the CEO does not. The empirical analyses do not find significant differences based on the reasons for CEO replacement, but this may be due to the lack of data on CEO replacement reasons, which could be considered in future studies, as this discussion may provide important insights into the peer effect of CEOs. We thank Florian Peters for sharing CEO turnover data.

24 We obtain data on board directorship from the BoardEx database.

25 Alternatively, we divide the sample into quartiles or terciles and define firms in the bottom quartile or tercile as the ‘close’ group and firms in the top quartile and tercile as the ‘far’ group. Comparing the difference in peer effects between the close and far subsamples, peer effects are more pronounced in the close subsamples, supporting the robustness of our results.

26 Because of the missing values on future performance measures, the sample size slightly reduces in Table 7.

27 The monthly size factor (SMB) and the book-to-market factor (HML) are downloaded from Kenneth R. French’s online data library at https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html, and the MOM factor is constructed according to Carhart (1997). We thank Kenneth. R. French for sharing this data.

28 TNIC is an alternative industry classification developed by Hoberg and Phillips (2016). They classified the industry based on textual analysis of the product description in Item 1 or Item 1A of the annual 10-K disclosure. Reading the product description, they compute the product similarity score for every pair of firms by calculating the relative number of words that the two firms use in common in their product description. The more words shared by two firms, the more similar these firms are considered. Using this similarity score, they classify firms into the same industry if the score is above a certain threshold. The threshold is set at a value that makes the distribution of the TNIC industry as similar to that of the three-digit SIC distribution as possible.