1. Introduction

1In this paper, we investigate the behavior of overconfident CEOs in competitive environments. We specifically examine if CEO overconfidence in competitive product markets affects the managerial bad news hoarding behavior, leading to stock price crash risk (SPCR, hereafter). Prior studies on the determinants of the SPCR are based on the traditional agency theory, suggesting that managers often have opportunistic incentives to hide poor performance and bad news for extended periods of time (Hutton et al., 2009; Bleck and Liu, 2007; Benmelech et al., 2010; Callen and Fang, 2017). However, the amount of bad news that a manager can hide is limited. Indeed, when a threshold is reached, bad news is released to investors all at once, leading to stock price crashes.

2The agency perspective is based on the idea that managers are rational but opportunistic. However, the behavioral finance theory argues that managers can make irrational decisions (Tversky and Kahneman, 1974; Kahneman and Tversky, 1977; Thaler and Sheffrin, 1981; Malmendier and Tate, 2005). Therefore, the agency explanation for SPCR is likely to be restrictive. In this paper, we rely on the behavioral explanation of the SPCR by investigating the impact of CEO overconfidence cognitive bias on future stock price crashes. Graham et al. (2013) define overconfidence as the tendency of individuals to overestimate their perceptions of firm performance. Overconfident CEOs attach too much importance to their private information when they believe that their knowledge is accurate.

4On the other hand, overconfidence is costly for the company. Overconfident CEOs are more likely to feel better than the average (Malmendier and Tate, 2005, 2008; Paquin, 2020). They overly estimate their ability to realize a good performance and have unrealistic optimism about the future. As a result, overconfident CEOs tend to misperceive ongoing negative NPV projects as value creating. Recent behavioral corporate finance literature links CEO overconfidence to lower investment efficiency. Roll (1986) and Ferris et al. (2013) argue that CEO overconfidence is associated with value-destroying mergers. In the same vein, Malmendier and Tate (2005, 2008) show that overconfidence leads to over-investment in negative NPV projects. Ben-David et al. (2010) document that overconfidence is associated with a less effective investment policy.

5The literature also suggests that overconfident CEOs suffer from the interpretational bias i.e. appreciating positive feedback and ignoring negative ones (Markus, 1977; Swann and Read, 1981; Wang et al., 2014). When overconfident CEOs evaluate a project, they are likely to disregard negative feedback and are convinced that the project has a promising future (Malmendier and Tate, 2005). This interpretational bias of overconfidence leads the CEO to persevere with negative NPV projects for extended periods. It also affects the informational environment for two reasons. First, unconsciously, overconfident managers naturally disregard bad news when they communicate with investors. Second, to prevent the involvement of impatient shareholders, overconfident CEOs may also consciously ignore negative news, and shift the reported information in an optimistic direction to convince investors. Consequently, overconfident CEOs conceal bad news for extended periods. After a threshold is reached, bad news is released to investors all at once, leading to stock price crashes.

6The objective of our paper is to investigate how overconfident managers behave in competitive product market. We choose to focus on product market competition as a major feature to influence overconfident CEOs’ decisions for two reasons. First, the “under-confidence hypothesis” argues that overconfident people are less likely to believe they are better than average on difficult tasks (Malmendier and Tate, 2008; Chen et al., 2014). People in competitive environments often face more difficult tasks than in less competitive ones (Ho, 2016). Under this hypothesis, we expect that in highly competitive market, overconfident CEOs may not overestimate their personal ability. In this perspective, Yu (2014) argues that the negative consequences of overconfidence lessen when competition increases in the product market. Indeed, we expect that the positive effect of overconfidence on crash risk is less prevalent in highly competitive markets. Second, the product market competition can exert pressure on managers and curb bad news hoarding, leading to sub-optimal decision-making. In this sense, Evans and Sridhar (2002) document that product market competition lessens the incentives of CEOs to inflate the share price by withholding bad news. Dhaliwal et al. (2014) show that firms facing high levels of pressure become more conservative in their financial reporting to prevent leakage of information to competitors.

7Based on a sample of 252 French companies from 2007 to 2016, we adopt the generalized least squares (GLS) and probit estimations for panel data. We find a strong positive relationship between CEO overconfidence and SPCR. These findings support the behavioral corporate finance theory that highlights the effect of cognitive biases, such as overconfidence, on various corporate policies. These results suggest that the irrational behavior of overconfident CEOs leads them to conceal bad news and poor performance. We show that CEO overconfidence can lead to a SPCR and this complements the traditional agency theory explanations of firm-level stock price crash. Furthermore, the results show that the effect of overconfidence on crash risk is less pronounced in highly competitive product markets. This means that under pressure from competitors, overconfident CEOs are less likely to hide bad news. Hence, the competition in the product market, designed to solve traditional agency problems, can also help constraining the managerial irrationality effect in relation to crash risk. In addition, we show that the positive effect of overconfidence on the SPCR is less prevalent in strongly governed firms and when the CEO is also the owner of the firm.

8This study contributes to the literature by combining the effect of overconfident CEOs and product market competition in relation to stock price crash risk. We show that the competitivity of product market is likely to influence the behavior of overconfident CEOs by reducing the bad news hoarding behavior, and then, the probability of SPCR. This study put emphasis on product market competition as an efficient corporate governance mechanism that protect the stock market stability. We then add to the corporate governance literature by documenting that governance devices, designed to mitigate the agency problems are also important in constraining managerial irrationality. On the other hand, our study complements recent studies on the harmful consequences of CEO overconfidence on many economic phenomena, including corporate financial and accounting decisions (Hackbarth, 2008), stock investments (Puri and Robinson, 2007), and entrepreneurial activities (Landier and Thesmar, 2008; Kim and Kim, 2019; Ramiah et al., 2016). Despite the economic importance of the stock market in raising capital, there has been little research on the role CEO overconfidence plays on stock market stability. This study extends the one by Kim et al. (2016) who focus on the relationship between CEO overconfidence and stock price crash risk in the US context by testing this relationship in a legal environment different from the American context i.e. in France a civil law country where shareholders’ rights are weakly protected. Our results have many practical implications. They may be useful to investors in their investment decisions, first, by shedding light on this type of risk, and second, by enabling them to identify the profile of companies with stock price crash risks.

9The remainder of the article is structured as follows: In section 2, we present the literature review and develop the hypotheses. In section 3, we describe the sample and research design. In section 4, we present and discuss the findings. In section 5, we conclude the paper.

2. Related literature and hypotheses

2.1. CEO overconfidence and stock price crash risk

10Based on psychology literature, overconfident people think they are better than average (Svenson, 1981; Alicke, 1985). The “better than average” effect is easily extended into economic decision-making process (Camerer and Lovallo, 1999). On the one hand, overconfident CEOs overly estimate their ability to bring about a good performance and have unrealistic optimism about the future. As a result, these CEOs may be likely to over-invest in negative NPV projects (Heaton, 2002; Malmendier et al., 2011). Ben-David et al. (2010) provide evidence that managerial overconfidence is associated with a less effective investment policy. In the same perspective, He et al. (2019) suggest that overconfident manager’s overestimate their own ability to bring about good performance, which negatively affects investment efficiency. Li et al. (2014) conclude that overconfident managers have unrealistic optimism about the future but underestimate the probability of negative events.

11On the other hand, overconfident CEOs, being highly committed to their projects, are likely to interpret ex post negative feedback badly. Taylor and Gollwitzer (1995) suggest that overconfident CEOs believe that they can control the outcome of their projects and underestimate the probability of project losses. They are also likely to disregard negative feedback because they are convinced about the promising future of their projects. As a result, they are more likely to keep money-losing projects that rational managers would terminate. This is likely to lead to a SPCR (Bleck and Liu, 2007).

12In addition, this cognitive bias also affects the corporate supply of financial information to the stock market. Overconfident CEOs unconsciously exclude negative news and ignore them when they communicate with investors on the stock market. Overconfident CEOs will then increase the firm opacity and accumulate bad news. Overconfident CEOs will then increase the firm opacity and accumulate bad news to convince impatient investors that their ongoing projects have a promising future. For instance, Hilary and Hsu (2011) and Hribar and Yang (2016), and Libby and Rennekamp (2012) argue that CEO overconfidence is associated with high earnings management and accounting fraud practices. Schrand and Zechman (2012) and Ahmed and Duellman (2013) show that overconfidence is associated with a less conservative accounting policy.

13Recently, in the American context, Kim et al. (2016) show that CEO overconfidence is positively associated with the SPCR, suggesting that overconfident CEOs hide bad news and accumulate poor performance. The authors argue that overconfident CEOs believe that they are increasing firm long-term value by engaging in bad projects and hiding the “interim” bad performance.

14The preceding discussion shows that CEO overconfidence can hoard bad news, which, in turn, leads to high SPCR. Therefore, we expect that:

H₁. The stock price crash risk increases with CEO overconfidence.

2.2. The moderating effect of product market competition

15Overconfident CEOs tend to exaggerate their own wisdom and think they are better than average. Yu (2014) suggests that overconfident CEOs feel better than average when facing easy tasks. Since competition in product market is associated with complex tasks, we expect that this competition mitigates “the better than average” feeling and the irrational decisions of overconfident CEOs. Jensen (1993) highlights the pressure the product market exerts on managers. Competitive pressures in the product market hinder managers’ discretionary decision-making and act as a monitoring device. In a highly competitive product market, managers have strong incentives to reduce the slack and maximize profits (Kim and Kim, 2019). A large amount of financial literature provides theory on how market competition can mitigate discretionary decision-making problems.

16Furthermore, as an external governance device, product market competition affects financial reporting quality and reduces managerial latitude to maintain a lock on the company. In highly competitive markets, shareholders can detect false statements by comparing the information reported by the company with its competitors. Holmström (1982) and Nalebuff and Stiglitz (1983) show that firms in highly competitive industries provide more information than firms in monopolistic ones. Evans and Sridhar (2002) show that managers, in highly competitive markets, are in a compromise either to declare good news to lower the cost of capital or to report bad news to avoid increased competition and theft of activity. This is likely to encourage managers to provide truthful information, and to avoid the accumulation of bad news when competition increases. Dhaliwal et al. (2014) document that firms facing high levels of pressure become more conservative in their financial reporting.

17Product market competition is considered a corporate governance device by putting pressure on managers and limiting their short-term behavior. As a result, overconfident CEOs do not have an interest in accumulating poor performance and are not likely to withhold bad news under market pressures. The preceding discussion leads then to the following hypothesis:

H₂. The positive effect of CEO overconfidence on the SPCR is less prevalent in highly competitive markets.

3. Data and research design

3.1. Data

3.2. Measures of variables

3.2.1. Dependent variable measures

19Following Kim et al. (2016), we first estimate firm-specific weekly returns using the following expanded index model regression:

20r_ij=α_i + β_1i r_m(j-2) + β_2i r_m(j-1) + β_3i r_mj+ β_4i r_m(j+1) + β_5i r_m(j+2) + β_6i r_s(j-2) + β_7i r_s(j-1) +β_8i r_sj+ β_9i r_s(j+1) + β_10i r_s(j+2) + ἐ_ij (1)

Where r_ij is the return on stock i in week j, r_mj is the return on the market index in week j, and r_sj is the return on the industry index in week j.

21We define the firm-specific weekly return for firm i in week j as the natural logarithm of one plus the residual of above regression (Equation 1) (W_ij= ln (1+ἐ_ij)). Following Kim et al. (2016), the first measure of crash risk, CRASH, is an indicator variable that equals 1 if a firm experiences one or more firm-specific weekly returns falling at least 3.2 standard deviations below its mean value each year, and 0 otherwise. The cut-off of 3.2 standard deviations is chosen to generate 0.1% of the distribution.

22Following Kim et al. (2011), our second measure of crash risk is the down-to-up volatility measure, DUVOL, of the crash risk. For each firm i over a fiscal-year period t, firm-specific weekly returns are divided into two groups. The first is down weeks when the returns are below the annual mean, and the second is up weeks when the returns are above the annual mean. The standard deviation of firm-specific returns is calculated apart for each of these two groups. DUVOL is the natural logarithm of the ratio of the standard deviation in the down weeks to the standard deviation in the up weeks:

23DUVOL_jt= log {(n_u-1) ∑ _downw²_jt / (n_d-1) ∑_upw²_jt} (2)

3.2.2. Independent variables measures

Measurement of managerial overconfidence

24To measure CEO overconfidence, we follow Schrand and Zechman (2012) and Kim et al. (2016). The first proxy is a dummy variable taking the value of 1 if the firm meets at least three of the following five criteria, and 0 otherwise (OC-SZ).

1. Excess investment of firm is in the high quartile within industry years. The excess investment is the residual from a regression of total asset growth on sales growth. Overconfident executives invest more with internal resources because they overestimate the future cash flows of their investments and underestimate the risk of mistrust (Malmendier and Tate, 2005; Ben-David et al., 2010).

2. The second component is the net acquisition obtained from the statement of cash flows. Net acquisitions from the statement of cash flows that are within the top quartile within industry-years suggest overconfidence.

3. The third component is the debt-to-equity ratio. CEO is considered overconfident about when the ratio of the firm’s debt-to-equity is in the top quartile within industry-years. Hackbarth (2008) suggests that overconfident managers underestimate the risk of investment project mistrust. They assess the firm’s cost of capital to fund the investment optimistically, so they rely more on debt.

4. The fourth component is a binary variable that takes the value of 1 if the firm uses convertible debt or the preferred stock is greater than 0. Ben-David et al. (2010) suggest that overconfident CEO will choose risky debt.

5. The last component is the dividend policy. Ben-David et al. (2010) and Kim and Kim (2019) suggest that overconfident managers are less likely to pay dividends to preserve liquidity for future investment opportunities. This component is measured by a dummy variable that takes the value of 1 if the dividend yield is 0, and 0 otherwise.

25Our second measure of CEO overconfidence is net buyer (NET-BUYER). Following Malmendier and Tate (2005), we capture CEO overconfidence through their behavior to buy company’s stocks rather than selling those stocks. Indeed, this behavior means that CEO overestimates the future performance of his company and is then considered overconfident (Malmendier and Tate, 2005). We then construct a dummy variable that takes the value of 1 if the CEO buys more of his own-company’s stocks than he sells annually and zero otherwise.

Measurement of product market competition

26Our first measure of product market competition is the Herfindahl-Hirschman Index (HHI), following Kim et al. (2011b). A high HHI indicates low product market competition. The HHI is calculated as follows:

where S_ijt is the market share of firm i in industry j in year t.

27Based on the HHI, we construct a binary variable (HHI-BIN) that takes the value of 1 if the value of the HHI of the firm i is lower than the sample median, and 0 otherwise.

28The second measure of product market competition is the speed of profit adjustment measure (SPA). According to Depoers and Jeanjean (2012), SPA measures the persistence of above-median earnings. SPA captures the speed at which abnormal earnings are driven towards zero. Consistent with Knechel et al. (2008), we measure SPA through the difference between the firm’s return on assets and the industry median. We then regress this difference and use (-1) * α_2j from the following regression as a measure of competitiveness:

29X_it = α₀ + α_1j (Dummy 1 X_it-1) + α_2j (Dummy 2 X_it-1) + ἐ_it (4)

30X_it is the difference between the firm’s return on assets and the industry median. Dummy 1 is a dummy variable taking the value 1 if X_it-1 < = 0 and 0 otherwise, and Dummy 2 is a dummy variable taking the value 1 if X_it-1 > 0 and 0 otherwise. This regression was run for each industry separately.

3.2.3. Control variables measures

31Following previous research (Hong and Stein, 2003; Chen et al., 2001; Kim et al., 2011a), we include a set of control variables. To capture the potential persistence of stock price crashes, we control for the ex-ante expected crash risk (CRASH_t-1). Based on Hong and Stein (2003) analysis, we also include the past volatility of firm-specific weekly returns (SIGMA_t–1) over the fiscal-year period and past average firm-specific weekly returns (RET_t-1). Firm size is the log of the total assets (SIZE_t-1). Hutton et al. (2009) find that firm size is positively related to crash risk. We include, again, the return on equity defined as income before extraordinary items to equity (ROA_t-1) and the market value of the equity to book value of equity (MTB_t-1). Finally, based on Hutton et al. (2009) analysis, we include (OPACITY_t-1) defined as the absolute value of discretionary accruals, where discretionary accruals are estimated by the Jones (1991) model.

3.3. Model specification

32We use panel data regressions to test the hypotheses. The dependent variable Crash, measured in year t, is one of two stock price crash risk measures: CRASH and DUVOL. Following prior research, all independent variables are measured in year t-1. The key variable CEO overconfidence is one of the two proxies for CEO Overconfidence (OC-SZ and NET-BUYER). We use GLS regressions on panel data when using a continuous dependent variable DUVOL and correct the problems of heteroscedasticity and autocorrelation of the residuals. When the dependent variable is CRASH, a binary variable, we apply a Probit regression on the panel data and correct for the heteroscedasticity and autocorrelation of the residuals.

33We model the relationship between CEO overconfidence and crash risk as follows:

34Crash _it = β₀ + β₁CEO overconfidence_it-1 + β₂CRASH_t-1 + β₃SIGMA_t–1 + β₄RET_t-1 + β₅SIZE_t-1 + β₆ ROA_t-1 + β₇MTB_t-1 + β₈OPACITY_t-1 + ∑Year fixed effect + ἐ_it (5)

35We use an interaction term between CEO overconfidence (OC-SZ or NET-BUYER) and product market competition (HHI-BIN or SPA) to explore the moderating effect of product market competition on the relationship between overconfidence and crash risk:

36Crash _it = β₀ + β₁CEO overconfidence_t-1 + β₂Product market competition _t-1 + β₃CEO overconfidence _t-1* Product market competition_t-1 + β₄CRASH_t-1 + β₅SIGMA_t–1 + β₆RET_t-1 + β₇SIZe_t-1 + β₈ROA_t-1+ β₉ MTB_t-1 + β₁₀ OPACITY_t-1 + ∑ Year fixed effect + ἐ_it (6)

4. Empirical results

4.1. Descriptive statistics

37Table 1 presents descriptive statistics. The mean (median) value of DUVOL is - 0.38 (-0.052), lesser than the mean reported by Kim et al. (2016) in the US context (0.027). In the entire period of study, 16.8% of French companies set go through at least one stock price crash each year, based on our definition of a crash event. This proportion is almost identical to the 17.2% reported by Kim et al. (2016) and is higher than the proportion (10.25%) reported by Wen et al. (2019) in the Chinese context. As for CEO overconfidence measures, the average of OC-SZ and NET-BUYER are 0.301 and 0.265, respectively. These averages are same as the value founded by Kim et al. (2016). Finally, the mean of HHI-BIN is 0.496 and the mean of SPA is -0.712.

38Table 2 displays the Pearson correlation matrix. It shows that the correlation between both proxies of CEO overconfidence and the other independent variables are relatively small, below the critical value of 0.8 reported by Gujarati (2004). In additional, the values of the variance inflation factor (VIF) for each variable vary between 1.01 and 1.94 far below the critical value of 10 (Neter, 1986). Therefore, multicollinearity is not a serious problem in our multivariate analyses. In addition, both CEO overconfidence measures are positively correlated and significant at the 1% level with DUVOL and CRASH variables, which is intuitionally consistent with our first hypothesis.

4.2. Main regression analysis

39Table 3 presents the empirical results of the regression testing the effect of CEO overconfidence on crash risk (H₁). The first two columns of Table 3 present the results by the probit model regressions with CRASH as the dependent variable. As reported, the coefficients of both proxy of overconfidence are positive and statistically significant at the 1% level. These results show that CEO overconfidence positively affect the likelihood of a stock price crash, measured by CRASH. These results also hold for the magnitude of the price fall, measured by DUVOL. This means that there is a positive relationship between both measures of CEO overconfidence and the likelihood and extent of the stock price crash in the French market. Our first hypothesis (H₁) is then supported. These findings support the behavioral corporate finance perspective that links the overconfidence bias to many sub-optimal decisions. These results confirm those of Kim et al. (2016) in the US context and suggest that overconfident CEOs are more likely to keep money-losing projects because they are convinced that their projects have a promising future. In addition, overconfident CEOs unconsciously ignore negative news which will accumulate and lead to a crash of the stock price. This finding complements recent studies on the harmful consequences of managerial overconfidence, such as excessive investment (He et al., 2019; Ben-David et al., 2010), accounting manipulation (Hilary and Hsu, 2011; Hribar et Yang, 2016; Libby and Rennekamp, 2012), and less conservative earnings (Schrand and Zechman, 2012; Ahmed and Duellman, 2013).

40Regarding control variables, we find that future crash is positively related to the volatility of firm-specific returns (SIGMA), similar to Chen et al. (2001), and negatively related to firm-specific returns (RET). The results show that firm size positively affects crash risk, as large companies are more likely to attract investors’ attention when disclosing hidden bad news. This result confirms the findings of Hutton et al. (2009) and Chen et al. (2001). In line with our expectation, a company’s performance measured by the ROA ratio negatively affects the crash risk. Finally, the coefficient of OPACITY is statistically significantly positive, suggesting that firms engaging in earnings manipulation are more likely to incur stock price crashes in the future. This result suggests that earnings management practices are used by managers to hide bad news and poor performance.

41We now examine the CEO overconfidence effect on stock price crash risk in competitive product markets. To test this effect, we used two measures of product market competition, HHI-BIN and SPA. Table 4 presents the results of the effect of product market competition on SPCR, and on the relationship between overconfidence and SPCR (H₂). First, our results show that the coefficients of our two measures of product market competition are negative and significant. These results suggest that the pressure exerted by competitors limits the crash risk. We confirm the results of Andreou et al. (2016) and Kim et al. (2011b) that increased competition in the product market encourages managers to make better investment decisions, disregard losing projects and avoid poor performance. Additionally, they suggest that firms facing high levels of pressure are constrained to enhance their informational environment as their competitors do. In such a rich informational environment and under product market pressures, managers are less likely to accumulate bad news, leading to decrease the crash risk.

42Regarding how CEOs overconfident behave in competitive product market, we find results that support the hypothesis H₂. The coefficients of our interaction terms (OC-SZ*HHI-BIN and NET-BUYER*HHI-BIN) are negative and statistically significant. Also, Table 4 shows the same results using our second measure of product market competition i.e SPA. These results suggest that in presence of high market competition, the effect of overconfidence on crash risk is less pronounced. Competition in product markets is then a good corporate governance device likely to influence the behavior of overconfident CEOs. The latter become more aware in their investment decision-making and more realistic in their forecast, under pressure from competitors. This result supports the “under-confidence hypothesis” suggesting that because the overconfidence bias is stronger for simple tasks and competition makes management more difficult, the competition mitigates the perverse effect of overconfidence (Ho, 2016).

43These results highlight the importance of product market competition as a disciplinary mechanism that limits managerial decision-making and protects the interests of minority shareholders especially in a less shareholder-oriented market i.e. in France. In addition, these findings have implications regarding the importance of corporate governance mechanisms including product market competition not only to mitigate agency conflicts between opportunistic managers and external shareholders but also to constrain the adverse consequences of managerial irrationality.

4.3. Further evidence

4.3.1. The effect of managerial ownership

44We further examine the effect of managerial overconfidence on crash risk when the CEO holds large factions in the capital. The agency theory suggests managerial ownership helps aligning the interests of the CEO with those of external shareholders (Korczak and Korczak, 2009). We then examine if managerial ownership can attenuate the positive effect of CEO overconfidence on the stock price crash risk.

45We group our sample into firms according to managerial ownership. We designate firms with managerial ownership if the manager or her family holds the largest percentage of votes in the company. Table 5 reports the results of the impact of CEO overconfidence on both measures of stock price crash risk (Panels A and B). The results show that the positive effect of CEO overconfidence on crash risk is less prevalent when the overconfident CEO is also the owner of the company. These results mean that the commitment of overconfident managers with their wealth leading them to be more sensitive to losses and more rational in their decision-making. This is likely to lessen the crash risk. This finding supports the alignment hypothesis of agency theory assuming that the owner-manager becomes more concerned about the risk of losing her wealth (Chen 2006).

4.3.2. The effect of corporate governance quality

46Corporate governance quality has an important role in enhancing financial disclosure and reporting quality (Larcker et al., 2007) and hence mitigates the bad news hoarding behavior by overconfident CEOs. We expect that the effect of managerial overconfidence on crash risk will be less pronounced in strongly governed firms. We use the corporate governance score retrieved from the DataStream database. This score ranges from 0 to 100. Then, we group our sample into firms with high corporate governance score (above median) and low corporate governance score (below median).

47Table 6 presents the subsample analyses for strongly and weakly governed firms for both measures of stock price crash risk (Panels A and B). The results show that the effect of overconfidence on SPCR is less prevalent for strongly governed firms. These findings highlight the importance of corporate governance as a disciplinary controlling device (Paquin, 2020) able to enrich the firm’s informational environment. It is then likely to limit bad news hoarding behavior by the overconfident manager and protect the stability in the financial market.

4.4. Robustness checks

4.4.1. Alternative measure of crash risk

48We use an alternative measure of stock price crash risk i.e. the negative skewness of firm-specific returns (NCSKEW). Specifically, NCSKEW is the negative of the third moment of firm specific weekly returns for each sample year and dividing it by the standard deviation of firm-specific weekly returns raised to the third power.

49NCSKEW = - [n (n-1) ^3/2 ∑w³_jt] / [(n-1) (n-2) (∑w²_jt) ^3/2] (7)

50The results reported in Table 7 remain qualitatively unchanged and show that CEO overconfidence has a positive effect on the stock price crash risk and that product market competition can curb this effect.

4.4.2. Endogeneity issue

51GMM regression: one concern with the empirical findings is the reverse causality. Overconfident CEOs may self-select into companies with high growth or business risk (Graham et al., 2013), and high-growth (risk) firms may have a higher crash risk than low-growth firms. This overconfidence–growth matching issue is addressed by including instrumental variables of growth and risk (the market-to-book ratio and the Herfindahl-Hirschman Index). We perform robustness analysis in this study using the generalized method of moments (GMM). Table 8 shows that there is a positive relationship between managerial overconfidence and crash risk. This relationship is less pronounced when competition increases in the product market supporting our previous findings.

52Propensity score matching regression: we use the propensity score matching (PSM) procedure to cross-check the results reported in Table 3. We use OC-SZ and NET-BUYER variables as dummy variables that take the value of one if the CEO is overconfident, and 0 otherwise. Using propensity score, we match without replacement a firm-year observation with OC-SZN or NRT-BUYER equal to one (treatment observation) against another firm-year observation with OC-SZ or NET-BUYER equal to zero (control observation). We use PSM regression to address omitted variables bias, endogeneity issues and self-selection bias. We first estimate the likelihood (i.e., the propensity score) that a firm with a set of firm-level characteristics is run by an overconfident manager, using logit model where OC-SZ or NET-BUYER is regressed on all control variables in equation (5). Using the propensity score estimated in the first stage, we one-to-one match each firm in our treatment observations with firm in our control observations that the closest propensity score within 3% width without replacement. The results reported in table 9 show a positive relationship between CEO overconfidence and the crash risk measures. These results are similar to those found in the main analysis.

5. Discussion and conclusion

53This paper examines the effect of overconfident CEOs on the SPCR in competitive product markets. Starting from the existing literature on overconfidence, we consider CEO overconfidence as the ability of CEOs to overestimate their personal capacity to create wealth for shareholders, and to master all the risks associated with the investments. This irrational behavior affects investment and financing choices, and the quality of financial reporting. This study is among the few studies that relate stock crash risk to executive personal characteristics or psychological traits, which is an important complement of the traditional agency explanations of crash risk.

54Using a sample of French companies over the period 2007-2016, we find that CEO overconfidence is positively associated with the stock price crash risk, suggesting that overconfident managers believe that they are maximizing long-term value of firm by hiding the “temporary” bad performance and bad news, leading to large stock price falls. We also show that the positive effect of CEO overconfidence on stock price crash risk is less pronounced in highly competitive product markets. Hence, the competition in the product market, designed to solve traditional agency problems, can also help constraining the managerial irrationality effect in relation to stock crash risk. These findings suggest the importance of corporate governance, and particularly, external governance devices such as product market competition, able to exert pressure on managers, protect shareholders' interests and to strengthen confidence in the financial market.

55Future research may explain the crash risk by other cognitive bias (optimism, mimicry, etc.) and may focus on other moderating mechanisms which can limit the negative effects of CEO overconfidence.