## 1 Introduction

^{1}However, this assumption only partially matches real-world managers’ decision-making. For example, Graham (2022) recently provides survey evidence that irrational managers have optimistic views of their firms’ value and performance, which might lead managers to perceive their firms’ stock to be undervalued. Consistent with the findings of the survey, the literature suggests that overconfident CEOs, due to the overestimation of the profitability of existing projects, are less likely to timely disclose bad news (Ahmed and Duellman 2013; Kim et al. 2016; Pierk 2021).

^{2}These studies concentrate on CEOs, whereas other executives’ overconfidence has received less attention. This single focus might result in false attribution because it neglects the role of other managers who are accountable for particular areas, especially the firm’s accounting and finance (Malmendier et al. 2022).

^{3}However, much less is known about the association between CFO overconfidence and conditional accounting conservatism, and the motivations that drive overconfident CFOs to influence conservative reporting.

^{4}Furthermore, our findings suggest that overconfident CFOs would convey private information by reducing conditional accounting conservatism, complementing Ahmed and Duellman (2013) by investigating the motivation for overconfident managers to affect timely bad news disclosure.

^{5}

## 2 Theoretical framework, literature review, and hypothesis development

### 2.1 The effect of CFO overconfidence on conditional accounting conservatism

### 2.2 CFO power, CFO overconfidence, and conditional accounting conservatism

## 3 Research design

### 3.1 Sample selection and data sources

Steps | Observations |
---|---|

Total number of observations for US-listed firms with CCM and ExecuComp data | 41,153 |

Exclude: Financial firms (sic: 6000–6999) | (7,154) |

Exclude: Utility firms (sic: 4900–4999) | (2,151) |

Exclude: Missing value of the variable used to test the H1 and H2 | (10,222) |

Total number of observations to test the H1 | 21,626 |

Total number of observations to test the H2 | The different sample sizes depend on the number of observations of the independent variable |

### 3.2 Variable measurement

#### 3.2.1 The measurement of conditional accounting conservatism

^{6}The measures of Basu (1997) include the industry-year measurement which assumes that all firms in the same industry are homogeneous and the individual firm measurement which assumes that the firm’s operating traits are stable. However, Khan and Watts (2009) argue that both measures of Basu (1997) have limitations because firms’ conditional accounting conservatism is affected by time- and firm-specific factors.

^{7}Thus, Khan and Watts (2009) modify the method of Basu (1997) and develop a firm-specific estimation of the timeliness of bad news (C-Score) and good news (G-Score). The C-score refers to conditional accounting conservatism. The G-Score and C-Score are estimated as follows.

_{i,t}refers to net income before extraordinary items deflated by the market value of equity at the beginning of fiscal year t; RET

_{i,t}refers to the annual buy and hold return starting from the fourth month after the previous fiscal year-end; D

_{i,t}refers to an indicator variable that equals one if RET

_{i,t}is negative, and zero otherwise; MV

_{i,t}refers to the log of the market value of equity; MTB

_{i,t}refers to the market value of equity divided by the book value of equity; LEV

_{i,t}refers to total debt divided by total assets.

_{3}and β

_{4}from Eqs. (2) and (3) into regression Eq. (1) yields Eq. (4).

_{i,t}) is calculated by applying the estimates from Eq. (4) to Eq. (3). The higher value of C_score1

_{i,t}indicates more conditional accounting conservatism.

^{8}Thus, Banker et al. (2016) modify the measure of Khan and Watts (2009) by considering the potential confounding effect of sticky costs as follows.

_{i,t−1}refers to the log of the market value of equity in the fiscal year t−1; BM

_{i,t−1}refers to the book value of equity divided by the market value of equity in the fiscal year t−1; LEV

_{i,t−1}refers to total debt divided by total assets in the fiscal year t−1; S

_{i,t}/MKT

_{i,t−1}is the changes in sales divided by market value of equity in the fiscal year t−1; DS

_{i,t}equals one if S

_{i,t}is negative, and zero otherwise.

_{3}and β

_{4}from Eqs. (6) and (7) into regression Eq. (5) yields Eq. (8).

#### 3.2.2 The measurement of overconfidence

### 3.3 Baseline regression model design

_{i,t}is one of two conditional accounting conservatism proxies (C_score1

_{i,t}and C_score2

_{i,t}); Holder67CFO

_{i,t}, the proxy of overconfident CFOs, is the variable of interest; β

_{1}captures the relationship between CFO overconfidence and conditional accounting conservatism. If the β

_{1}is significantly positive, it indicates a positive association between CFO overconfidence and conditional accounting conservatism (hypothesis H1).

_{i,t}). In addition, we control for some other CFOs’ characteristics. We control for CFO gender (CFO_male

_{i,t}) since male CFOs tend to use fewer conservative accounting principles than their female counterparts (Francis et al. 2015). In addition to CFO gender, we control for the influence of CFO ownership (CFO_ownership

_{i,t}) as top managers’ ownership significantly affects conditional accounting conservatism (LaFond and Watts 2008). Given that firms are not static across time, we include firm performance in the regression. Prior studies suggest that firms’ growth opportunities, sale growth rates, operating uncertainty, and profitability affect conditional accounting conservatism (e.g., Smith and Watts 1992; Ahmed et al. 2002; Ahmed and Duellman 2007, 2013). As such, we use the market-to-book ratio (MTB

_{i,t}), sales growth (SaleGrowth

_{i,t}), sales volatility (VolSale

_{i,t}), and cash flows from operations (CashFlow

_{i,t}) to capture them, respectively. We also control for firm characteristics, including firm size (FirmSize

_{i,t}), leverage (Leverage

_{i,t}), R&D and advertising expenses (RDAD

_{i,t}), following Ahmed et al. (2002), Givoly et al. (2007), and Ahmed and Duellman (2013). Considering the monitoring effect, we include auditor type and litigation risk because a high-quality auditor (Big_four

_{i,t}) and high litigation risk (High_LIT

_{i,t}) affect firms’ conditional accounting conservatism decisions (Basu et al. 2001; Watts 2003; Cano-Rodríguez 2010). In addition, Beaver and Ryan (2005) suggest that unconditional accounting conservatism should be controlled for in the study of conditional accounting conservatism as conditional accounting conservatism can be preempted by unconditional accounting conservatism. In response to their call, we control for unconditional accounting conservatism (UnAC

_{i,t}). Finally, we use firm-fixed effect models and include year dummies to control for the potential omitted time-invariant effects of firms and years. Detailed variable measurements are provided in the Appendix.

## 4 Empirical results

### 4.1 Descriptive statistics

_{i,t}) proposed by Khan and Watts (2009) as our first conditional accounting conservatism measurement. The mean value of C_score1

_{i,t}is 0.062, which is similar to the result of Ahmed and Duellman (2013). Following Khalilov and Osma (2020), our second conditional accounting conservatism measurement (C_score2

_{i,t}) uses the approach of Banker et al. (2016). The mean value of C_score2

_{i,t}is 0.182, which is consistent with the findings of Khalilov and Osma (2020). Our overconfidence measure is consistent with Chen et al. (2022). Although our sample period differs from that used by Chen et al. (2022), we have comparable results: the mean values of overconfident CFOs and overconfident CEOs in their sample are 0.510 and 0.593, respectively, and the mean values of Holder67CFO

_{i,t}and Holder67CEO

_{i,t}are 0.548 and 0.650 in our sample. Over half of CFOs and CEOs are overconfident, which is in line with previous findings that overconfidence is a common trait among top managers (Goel and Thakor 2008; Malmendier et al. 2022), indicating that CFO overconfidence cannot be ignored.

N | Mean | SD | 10P | 25P | Median | 75P | |
---|---|---|---|---|---|---|---|

Panel A—Full sample | |||||||

C_score1 _{i,t} | 21,626 | 0.062 | 0.151 | − 0.130 | − 0.026 | 0.068 | 0.154 |

C_score2 _{i,t} | 21,626 | 0.182 | 0.187 | − 0.035 | 0.072 | 0.178 | 0.293 |

Holder67CFO _{i,t} | 21,626 | 0.548 | 0.498 | 0 | 0 | 1 | 1 |

Holder67CEO _{i,t} | 21,626 | 0.650 | 0.477 | 0 | 0 | 1 | 1 |

CFO_male _{i,t} | 21,626 | 0.921 | 0.269 | 1 | 1 | 1 | 1 |

CFO_ownership _{i,t} (%) | 21,626 | 0.087 | 0.234 | 0.000 | 0.000 | 0.000 | 0.059 |

MTB _{i,t} | 21,626 | 3.447 | 3.709 | 1.051 | 1.559 | 2.394 | 3.840 |

SaleGrowth _{i,t} | 21,626 | 0.106 | 0.239 | − 0.114 | − 0.004 | 0.075 | 0.177 |

VolSale _{i,t} | 21,626 | 0.144 | 0.124 | 0.038 | 0.063 | 0.106 | 0.182 |

CashFlow _{i,t} | 21,626 | 0.100 | 0.085 | 0.014 | 0.059 | 0.101 | 0.147 |

FirmSize _{i,t} | 21,626 | 7.318 | 1.582 | 5.378 | 6.172 | 7.200 | 8.348 |

Leverage _{i,t} | 21,626 | 0.213 | 0.169 | 0.000 | 0.056 | 0.205 | 0.328 |

RDAD _{i,t} | 21,626 | 0.067 | 0.126 | 0.000 | 0.000 | 0.022 | 0.080 |

Big_four _{i,t} | 21,626 | 0.864 | 0.343 | 0 | 1 | 1 | 1 |

High_LIT _{i,t} | 21,626 | 0.100 | 0.300 | 0 | 0 | 0 | 0 |

UnAC _{i,t} | 21,626 | 0.014 | 0.051 | − 0.034 | − 0.010 | 0.009 | 0.031 |

Panel B—Subsample | |||||||

Variable | Holder67CFO _{i,t} = 0 (N = 9,775) | Holder67CFO _{i,t} = 1 (N = 11,851) | T-statistics for tests of difference in means | ||||

Mean | Mean | (Non-overconfident CFOs-Overconfident CFOs) | |||||

C_score1 _{i,t} | 0.079 | 0.047 | 0.032*** | ||||

C_score2 _{i,t} | 0.210 | 0.159 | 0.051*** | ||||

Holder67CEO _{i,t} | 0.398 | 0.858 | − 0.459*** | ||||

CFO_male _{i,t} | 0.922 | 0.920 | 0.002 | ||||

CFO_ownership _{i,t} | 0.062 | 0.107 | − 0.045*** | ||||

MTB _{i,t} | 2.829 | 3.957 | − 1.129*** | ||||

SaleGrowth _{i,t} | 0.062 | 0.143 | − 0.081*** | ||||

VolSale _{i,t} | 0.143 | 0.145 | − 0.002 | ||||

CashFlow _{i,t} | 0.087 | 0.111 | − 0.025*** | ||||

FirmSize _{i,t} | 7.351 | 7.290 | 0.061** | ||||

Leverage _{i,t} | 0.225 | 0.203 | 0.021*** | ||||

RDAD _{i,t} | 0.069 | 0.066 | 0.004 | ||||

Big_four _{i,t} | 0.860 | 0.867 | − 0.007 | ||||

High_LIT _{i,t} | 0.109 | 0.092 | 0.017*** | ||||

UnAC _{i,t} | 0.018 | 0.011 | 0.007*** |

_{i,t}and C_score2

_{i,t}in the overconfident CFOs sample are significantly lower than the corresponding values for their non-overconfident CFOs sample.

### 4.2 Pairwise correlations

_{i,t}and Holder67CEO

_{i,t}have a significant association (a correlation coefficient of 0.479), indicating that both should be included in one regression to avoid the omitted variables problem. Holder67CFO

_{i,t}has significant negative correlations with C_score1

_{i,t}and C_score2

_{i,t}without controlling other variables, which is in line with our hypothesis H1 that overconfident CFOs tend to reduce conditional accounting conservatism compared with non-overconfident CFOs.

Variable | (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
---|---|---|---|---|---|---|---|---|---|

(1) | C_score1 _{i,t} | 1 | |||||||

(2) | C_score2 _{i,t} | 0.748*** | 1 | ||||||

(3) | Holder67CFO _{i,t} | − 0.107*** | − 0.136*** | 1 | |||||

(4) | Holder67CEO _{i,t} | − 0.078*** | − 0.117*** | 0.479*** | 1 | ||||

(5) | CFO_male _{i,t} | 0.021*** | 0.033*** | − 0.004 | − 0.030*** | 1 | |||

(6) | CFO_ownership _{i,t} | 0.184*** | 0.131*** | 0.095*** | 0.053*** | 0.031*** | 1 | ||

(7) | MTB _{i,t} | − 0.325*** | − 0.356*** | 0.151*** | 0.136*** | − 0.014** | − 0.034*** | 1 | |

(8) | SaleGrowth _{i,t} | − 0.074*** | − 0.144*** | 0.168*** | 0.159*** | 0.008 | − 0.018*** | 0.125*** | 1 |

(9) | VolSale _{i,t} | 0.147*** | 0.122*** | 0.009 | 0.021*** | 0.003 | 0.035*** | 0.007 | 0.065*** |

(10) | CashFlow _{i,t} | − 0.273*** | − 0.222*** | 0.145*** | 0.109*** | − 0.041*** | − 0.013* | 0.201*** | 0.056*** |

(11) | FirmSize _{i,t} | − 0.587*** | − 0.469*** | − 0.019*** | − 0.041*** | − 0.022*** | − 0.186*** | 0.051*** | − 0.061*** |

(12) | Leverage _{i,t} | − 0.011 | − 0.080*** | − 0.063*** | − 0.064*** | 0.050*** | − 0.040*** | 0.107*** | − 0.035*** |

(13) | RDAD _{i,t} | 0.025*** | − 0.013* | − 0.015** | 0.025*** | − 0.006 | − 0.029*** | 0.163*** | 0.049*** |

(14) | Big_four _{i,t} | − 0.096*** | − 0.082*** | 0.009 | 0.001 | − 0.032*** | − 0.066*** | 0.022*** | − 0.049*** |

(15) | High_LIT _{i,t} | − 0.015** | 0.047*** | − 0.027*** | − 0.015** | − 0.004 | − 0.048*** | − 0.054*** | − 0.061*** |

(16) | UnAC _{i,t} | 0.074*** | 0.057*** | − 0.067*** | − 0.049*** | − 0.017** | 0.008 | 0.089*** | − 0.005 |

Variable | (9) | (10) | (11) | (12) | (13) | (14) | (15) | (16) | |

(9) | VolSale _{i,t} | 1 | |||||||

(10) | CashFlow _{i,t} | − 0.088*** | 1 | ||||||

(11) | FirmSize _{i,t} | − 0.175*** | 0.123*** | 1 | |||||

(12) | Leverage _{i,t} | − 0.041*** | − 0.153*** | 0.355*** | 1 | ||||

(13) | RDAD _{i,t} | − 0.037*** | − 0.318*** | − 0.214*** | − 0.196*** | 1 | |||

(14) | Big_four _{i,t} | − 0.036*** | 0.040*** | 0.240*** | 0.070*** | − 0.014** | 1 | ||

(15) | High_LIT _{i,t} | 0.086*** | − 0.091*** | 0.184*** | 0.099*** | 0.036*** | 0.052*** | 1 | |

(16) | UnAC _{i,t} | 0.048*** | 0.003 | − 0.064*** | − 0.055*** | 0.186*** | 0.053*** | 0.042*** | 1 |

### 4.3 Baseline results—regression analyses of Hypothesis 1

^{10}Table 4 shows the regression results of the estimation of Eq. (12). The dependent variables are C_score1

_{i,t}measured using the method of Khan and Watts (2009) (columns (1) and (3)) and C_score2

_{i,t}measured using the approach of Banker et al. (2016) (columns (2) and (4)). The variable of interest is CFO overconfidence (Holder67CFO

_{i,t}) calculated following Campbell et al. (2011) and Chen et al. (2022).

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

Holder67CFO _{i,t} | − 0.016*** | − 0.025*** | − 0.012*** | − 0.018*** |

(0.002) | (0.002) | (0.002) | (0.003) | |

Holder67CEO _{i,t} | − 0.011*** | − 0.019*** | ||

(0.002) | (0.003) | |||

CFO_male _{i,t} | 0.006 | 0.008 | 0.006 | 0.008 |

(0.004) | (0.006) | (0.004) | (0.006) | |

CFO_ownership _{i,t} | 0.016*** | 0.031*** | 0.016*** | 0.032*** |

(0.005) | (0.008) | (0.005) | (0.008) | |

MTB _{i,t} | − 0.010*** | − 0.008*** | − 0.009*** | − 0.008*** |

(0.001) | (0.000) | (0.001) | (0.000) | |

SaleGrowth _{i,t} | − 0.014*** | − 0.024*** | − 0.013*** | − 0.023*** |

(0.003) | (0.004) | (0.003) | (0.004) | |

VolSale _{i,t} | − 0.014* | − 0.048*** | − 0.013* | − 0.045*** |

(0.008) | (0.012) | (0.008) | (0.012) | |

CashFlow _{i,t} | − 0.099*** | − 0.119*** | − 0.097*** | − 0.115*** |

(0.012) | (0.017) | (0.012) | (0.017) | |

FirmSize _{i,t} | − 0.070*** | − 0.060*** | − 0.069*** | − 0.058*** |

(0.002) | (0.003) | (0.002) | (0.003) | |

Leverage _{i,t} | 0.202*** | 0.090*** | 0.200*** | 0.086*** |

(0.009) | (0.012) | (0.009) | (0.012) | |

RDAD _{i,t} | − 0.048*** | − 0.088*** | − 0.047*** | − 0.086*** |

(0.018) | (0.030) | (0.018) | (0.030) | |

Big_four _{i,t} | − 0.014*** | − 0.018*** | − 0.014*** | − 0.018*** |

(0.003) | (0.005) | (0.003) | (0.005) | |

High_LIT _{i,t} | 0.020*** | 0.014*** | 0.020*** | 0.014*** |

(0.003) | (0.004) | (0.003) | (0.004) | |

UnAC _{i,t} | 0.074*** | 0.110*** | 0.071*** | 0.104*** |

(0.016) | (0.022) | (0.016) | (0.022) | |

Constant | 0.424*** | 0.517*** | 0.422*** | 0.514*** |

(0.015) | (0.028) | (0.015) | (0.028) | |

Observations | 21,626 | 21,626 | 21,626 | 21,626 |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.797 | 0.751 | 0.798 | 0.752 |

_{i,t}) is significant and negative in column (1), suggesting a negative relationship between CFO overconfidence and conditional accounting conservatism. Given that the mean value of conditional accounting conservatism (C_score1

_{i,t}) is 0.062, there appears to be a sizeable difference (25.806%).

^{11}As shown in column (2), when we change the dependent variable from C_score1

_{i,t}to C_score2

_{i,t}, Holder67CFO

_{i,t}remains a significantly negative coefficient, indicating that the finding is robust under an alternative conditional accounting conservatism measure.

_{i,t}) and CEO overconfidence (Holder67CEO

_{i,t}) have a high correlation, and in response to the call made by Black and Gallemore (2013) and Malmendier et al. (2022) that CFO overconfidence and CEO overconfidence should be jointly considered in corporate decision-making, we control for the effect of Holder67CEO

_{i,t}in columns (3) and (4). Holder67CEO

_{i,t}has a significantly negative coefficient in columns (3) and (4), which is consistent with the finding of Ahmed and Duellman (2013). The coefficient on Holder67CFO

_{i,t}remains negative and statistically significant in columns (3) and (4). Besides, the coefficients on Holder67CFO

_{i,t}suggest that firms with overconfident CFOs cause a decrease in C_score1

_{i,t}by 19.355% (in comparison to the mean C_score1

_{i,t}) in column (3) and a decrease in C_score2

_{i,t}by 9.890% (in comparison to the mean C_score2

_{i,t}) in column (4), showing that these findings are also economically significant.

^{12}

_{i,t}) has a negative and significant coefficient in columns (1) to (4), which is consistent with the findings of Roychowdhury and Watts (2007) that firms with high growth opportunities adopt low conditional accounting conservatism. CashFlow

_{i,t}has a negative and significant coefficient in each column, showing that high-profit firms reduce the application of conditional accounting conservatism (Ahmed and Duellman 2013). The coefficient on firm size (FirmSize

_{i,t}) is negative and significant in each column, suggesting that larger firms are less likely to recognize losses timely (Givoly et al. 2007). Leverage

_{i,t}has a positive and significant coefficient in columns (1) to (4), indicating that high-leverage firms increase their accounting conservatism. This finding is in line with Ahmed et al. (2002) in that high-leverage firms have more debt-holder and equity-holder conflicts, which increase conservatism levels. The coefficient of high litigation risk (High_LIT

_{i,t}) is significantly positive in columns (1) to (4), indicating that firms under high litigation risk adopt more conditional accounting conservatism (Watts 2003).

### 4.4 Robustness tests

#### 4.4.1 Reverse causality concern

_{i,t}has a significantly negative coefficient in columns (1) and (2), proving that overconfident CFOs are able to reduce future conditional accounting conservatism.

(1) | (2) | |
---|---|---|

Variable | C_score1 _{i,t+1} | C_score2 _{i,t+1} |

Holder67CFO _{i,t} | − 0.010*** | − 0.021*** |

(0.002) | (0.003) | |

Constant | 0.359*** | 0.565*** |

(0.020) | (0.028) | |

Observations | 17,074 | 17,074 |

Controls in Eq. (12) | Yes | Yes |

Firm fixed effects | Yes | Yes |

Year fixed effects | Yes | Yes |

Adj. R ^{2} | 0.779 | 0.785 |

_{i}equals one if both outgoing CFO is non-overconfident and the incoming CFO is overconfident, and zero if both the outgoing CFO and incoming CFO are non-overconfident. The Post

_{i,t}equals one when observations occur in the first year after CFO turnover, and zero when observations occur in the last year before CFO turnover. The DID estimation model is as follows:

_{i}× Post

_{i,t}, is our variable of interest. We include firm fixed effects and year fixed effects, so we exclude Treat

_{i}and Post

_{i,t}to avoid multiple collinearities. Control variables are consistent with Eq. (12). Detailed variable information is shown in the Appendix.

_{i}× Post

_{i,t}is significantly negative in Panel B of Table 6, suggesting that the negative relationship between CFO overconfidence and conditional accounting conservatism remains under PSM-DID estimation.

Panel A—PSM | |||||||
---|---|---|---|---|---|---|---|

Variable | Unmatched | Mean | Bias | T-test | |||

Matched | Treated | Control | %bias | %reduct |bias| | t | p >|t| | |

(1) | (2) | (3) | (4) | (5) | (6) | (7) | |

Holder67CEO _{i,t} | U | 0.610 | 0.326 | 59.100 | 4.820 | 0.000 | |

M | 0.610 | 0.598 | 2.500 | 95.700 | 0.160 | 0.874 | |

CFO_male _{i,t} | U | 0.890 | 0.909 | − 6.100 | − 0.510 | 0.612 | |

M | 0.890 | 0.890 | 0.000 | 100.000 | 0.000 | 1.000 | |

CFO_ownership _{i,t} | U | 0.056 | 0.032 | 22.600 | 2.110 | 0.036 | |

M | 0.056 | 0.042 | 13.300 | 41.100 | 0.930 | 0.354 | |

MTB _{i,t} | U | 2.717 | 2.449 | 16.800 | 1.250 | 0.213 | |

M | 2.717 | 2.789 | − 4.500 | 73.300 | − 0.220 | 0.824 | |

SaleGrowth _{i,t} | U | 0.107 | 0.035 | 31.300 | 2.210 | 0.028 | |

M | 0.107 | 0.108 | − 0.400 | 98.800 | − 0.020 | 0.984 | |

VolSale _{i,t} | U | 0.134 | 0.136 | − 2.000 | − 0.180 | 0.860 | |

M | 0.134 | 0.149 | − 11.300 | − 477.200 | − 0.740 | 0.461 | |

CashFlow _{i,t} | U | 0.099 | 0.083 | 22.300 | 1.580 | 0.114 | |

M | 0.099 | 0.102 | − 3.100 | 85.900 | − 0.250 | 0.803 | |

FirmSize _{i,t} | U | 7.679 | 7.795 | − 7.700 | − 0.620 | 0.538 | |

M | 7.679 | 7.689 | − 0.700 | 91.100 | − 0.050 | 0.962 | |

Leverage _{i,t} | U | 0.208 | 0.231 | − 15.500 | − 1.260 | 0.209 | |

M | 0.208 | 0.202 | 3.600 | 76.800 | 0.240 | 0.814 | |

RDAD _{i,t} | U | 0.060 | 0.089 | − 11.900 | − 0.780 | 0.437 | |

M | 0.060 | 0.049 | 4.200 | 64.800 | 0.760 | 0.449 | |

Big_four _{i,t} | U | 0.927 | 0.919 | 3.100 | 0.240 | 0.807 | |

M | 0.927 | 0.963 | − 13.600 | − 342.800 | − 1.030 | 0.307 | |

High_LIT _{i,t} | U | 0.037 | 0.085 | − 20.200 | − 1.470 | 0.141 | |

M | 0.037 | 0.049 | − 5.100 | 74.600 | − 0.380 | 0.701 | |

UnAC _{i,t} | U | 0.011 | 0.016 | − 12.500 | − 0.960 | 0.338 | |

M | 0.011 | 0.005 | 12.300 | 2.000 | 0.950 | 0.342 | |

Panel B—DID | |||||||

(1) | (2) | ||||||

Variable | C_score1 _{i,t} | C_score2 _{i,t} | |||||

Treat _{i} × Post_{i,t} | − 0.022* | − 0.025** | |||||

(0.012) | (0.010) | ||||||

Constant | 0.919*** | 0.826*** | |||||

(0.174) | (0.139) | ||||||

Observations | 120 | 120 | |||||

Controls in Eq. (13) | Yes | Yes | |||||

Firm fixed effects | Yes | Yes | |||||

Year fixed effects | Yes | Yes | |||||

Adj. R ^{2} | 0.922 | 0.949 |

#### 4.4.2 Measurement error

_{i,t}(Holder100CEO

_{i,t}) equals one from the first time that CFOs (CEOs) hold vested options that are at least 100% in the money to the end of their tenure, and zero otherwise. In addition, to rule out the concern that CFOs and CEOs exercise their options late because they are risk-tolerant rather than overconfident, our study, following Huang et al. (2016), controls for CFO and CEO risk tolerance (CFO_vege

_{i,t}and CEO_vega

_{i,t}). As shown in Table 7, the Holder100CFO

_{i,t}has a significant negative coefficient, suggesting that the negative relationship still holds when using the new proxy of overconfidence and controlling for CFOs’ and CEOs’ risk tolerance.

^{13}

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

Holder100CFO _{i,t} | − 0.010*** | − 0.019*** | − 0.011*** | − 0.021*** |

(0.002) | (0.003) | (0.002) | (0.003) | |

Holder100CEO _{i,t} | − 0.011*** | − 0.017*** | − 0.010*** | − 0.018*** |

(0.002) | (0.003) | (0.002) | (0.003) | |

CFO_vega _{i,t} | 0.001 | − 0.003** | ||

(0.001) | (0.001) | |||

CEO_vega _{i,t} | − 0.001 | − 0.003** | ||

(0.001) | (0.001) | |||

Constant | 0.416*** | 0.510*** | 0.490*** | 0.595*** |

(0.015) | (0.028) | (0.019) | (0.033) | |

Observations | 21,388 | 21,388 | 18,601 | 18,601 |

Other controls in Eq. (12) | Yes | Yes | Yes | Yes |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.797 | 0.752 | 0.799 | 0.760 |

#### 4.4.3 The effect of CEO characteristics

_{i,t}× Holder67CEO

_{i,t}, into Eq. (12). As shown in Table 8, the coefficient on Holder67CFO

_{i,t}× Holder67CEO

_{i,t}is not significant, suggesting that we do not find a statistically significant joint effect of CFO and CEO overconfidence on conditional accounting conservatism. The coefficients on Holder67CFO

_{i,t}and Holder67CFO

_{i,t}+ Holder67CFO

_{i,t}× Holder67CEO

_{i,t}are significantly negative, indicating that overconfident CFOs have an independent effect on conditional accounting conservatism no matter whether CEOs are overconfident or not. These findings are held after controlling for CEO gender (CEO_male

_{i,t}) and CEO ownership (CEO_ownership

_{i,t}), which are shown in columns (3) and (4).

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

Holder67CFO _{i,t} | − 0.015*** | − 0.024*** | − 0.015*** | − 0.024*** |

(0.003) | (0.005) | (0.003) | (0.005) | |

Holder67CEO _{i,t} | − 0.011*** | − 0.021*** | − 0.012*** | − 0.021*** |

(0.003) | (0.004) | (0.003) | (0.004) | |

Holder67CFO _{i,t} × Holder67CEO_{i,t} | 0.002 | 0.005 | 0.002 | 0.005 |

(0.004) | (0.005) | (0.004) | (0.005) | |

CFO_vega _{i,t} | 0.001 | − 0.003** | 0.001 | − 0.003** |

(0.001) | (0.001) | (0.001) | (0.001) | |

CEO_vega _{i,t} | − 0.001 | − 0.003** | − 0.001 | − 0.003** |

(0.001) | (0.001) | (0.001) | (0.001) | |

CEO_male _{i,t} | 0.013 | 0.004 | ||

(0.008) | (0.011) | |||

CEO_ownership _{i,t} | 0.000 | 0.000 | ||

(0.000) | (0.001) | |||

Constant | 0.496*** | 0.599*** | 0.482*** | 0.594*** |

(0.019) | (0.033) | (0.021) | (0.034) | |

Holder67CFO _{i,t} + Holder67CFO_{i,t} × Holder67CEO_{i,t} | − 0.013*** | − 0.019*** | − 0.012*** | − 0.019*** |

(0.002) | (0.004) | (0.002) | (0.004) | |

Observations | 18,824 | 18,824 | 18,824 | 18,824 |

Other controls in Eq. (12) | Yes | Yes | Yes | Yes |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.800 | 0.761 | 0.800 | 0.761 |

#### 4.4.4 The effect of governance

_{i,t}) if they meet three of the four criteria listed below: First, the CEO is not simultaneously the Chairman. Second, the percentage of outside directors is greater than the sample’s median figure. Third, the institutional ownership percentage is higher than the sample’s median figure as institutional investors significantly affect conditional accounting conservatism (Lin 2016). Fourth, the percentage of inside directors is lower than the median value of the sample. As shown in columns (1) and (2) of Table 9, we additionally control the StrongCG

_{i,t}in Eq. (12). Holder67CFO

_{i,t}remains a significantly negative coefficient. Besides, we add the interaction term, Holder67CFO

_{i,t}× StrongCG

_{i,t}, into Eq. (12). As shown in columns (3) and (4), the coefficient on the interaction term is not significant. However, the coefficients on Holder67CFO

_{i,t}and Holder67CFO

_{i,t}+ Holder67CFO

_{i,t}× StrongCG

_{i,t}are significantly negative, suggesting that strong governance cannot moderate the association between CFO overconfidence and conditional accounting conservatism. These findings are consistent with the findings of Schrand and Zechman (2012) and Ahmed and Duellman (2013) that governance mechanisms do not mitigate the effect of overconfident managers.

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

Holder67CFO _{i,t} | − 0.016*** | − 0.025*** | − 0.016*** | − 0.025*** |

(0.003) | (0.004) | (0.003) | (0.004) | |

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

StrongCG _{i,t} | 0.002 | 0.007 | 0.002 | 0.008 |

(0.003) | (0.004) | (0.004) | (0.006) | |

Holder67CFO _{i,t} × StrongCG_{i,t} | 0.000 | − 0.003 | ||

(0.004) | (0.007) | |||

CFO_vega _{i,t} | 0.001 | − 0.002 | 0.001 | − 0.002 |

(0.001) | (0.002) | (0.001) | (0.002) | |

CEO_vega _{i,t} | − 0.000 | − 0.001 | − 0.000 | − 0.001 |

(0.001) | (0.001) | (0.001) | (0.001) | |

CEO_male _{i,t} | 0.009 | 0.012 | 0.009 | 0.012 |

(0.011) | (0.017) | (0.011) | (0.017) | |

CEO_ownership _{i,t} | 0.000 | 0.000 | 0.000 | 0.000 |

(0.001) | (0.001) | (0.001) | (0.001) | |

Constant | 0.577*** | 0.509*** | 0.577*** | 0.508*** |

(0.028) | (0.053) | (0.028) | (0.053) | |

Holder67CFO _{i,t} + Holder67CFO_{i,t} × StrongCG_{i,t} | − 0.016*** | − 0.028*** | ||

(0.005) | (0.007) | |||

Observations | 11,361 | 11,361 | 11,361 | 11,361 |

Other controls in Eq. (12) | Yes | Yes | Yes | Yes |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.807 | 0.769 | 0.807 | 0.769 |

## 5 Analysis of mechanisms

### 5.1 CFO overconfidence, conditional accounting conservatism, and conveying information

_{i,t}). We include last year’s earnings per share (Earn

_{i,t-1}), earnings per share (Earn

_{i,t}), the sum of earnings per share for fiscal years t + 1 to t + 3 (Earn

_{i,t+3}), the sum of stock returns for fiscal years t + 1 to t + 3 (R

_{i,t+3}) and CFO overconfidence (Hold67CFO

_{i,t}) in the regressions. To easily interpret our results, we multiply C_score

_{i,t}(including C_score1

_{i,t}and C_score2

_{i,t}) by a negative one to generate RevC_score

_{i,t}(including RevC_score1

_{i,t}and RevC_score2

_{i,t}). The greater the value of RevC_score

_{i,t}the lower the conditional accounting conservatism. We also include interaction terms among them. The variable of interest is the triple interaction term, Hold67CFO

_{i,t}× RevC_score

_{i,t}× Earn

_{i,t+3}. In addition, following prior studies (e.g., Shu 2021) control CEO overconfidence (Hold67CEO

_{i,t}), CFO gender (CFO_male

_{i,t}), CFO ownership (CFO_ownership

_{i,t}), CEO gender (CEO_male

_{i,t}), CEO ownership (CEO_ownership

_{i,t}), Market to book ratio (MTB

_{i,t}), Sales growth (SaleGrowth

_{i,t}), Sales volatility (VolSale

_{i,t}), Firm Size (FirmSize

_{i,t}), Firm age (FirmAge

_{i,t}), Leverage (Leverage

_{i,t}), Investment (Invest

_{i,t}), Property, Plant & Equipment (PPE

_{i,t}), Negative earnings per share (Loss

_{i,t}), Number of the analyst (Analyst_number

_{i,t}), CFO risk tolerance (CFO_vega

_{i,t}), CEO risk tolerance (CEO_vega

_{i,t}), and Strong governance (StrongCG

_{i,t}). Detailed variable measurements are provided in the Appendix.

_{i,t}× RevC_score1

_{i,t}× Earn

_{i,t+3}and Holder67CFO

_{i,t}× RevC_score2

_{i,t}× Earn

_{i,t+3}are significantly positive in columns (1) and (2), respectively. These findings show that overconfident CFOs who reduce conditional accounting conservatism increase earnings informativeness, which is consistent with our prediction that overconfident CFOs tend to convey information by reducing conditional accounting conservatism.

(1) | (2) | |
---|---|---|

Variable | R _{i,t} | R _{i,t} |

Earn _{i,t−1} | − 0.170 | − 0.074 |

(0.223) | (0.335) | |

Earn _{i,t} | − 0.319 | 0.535** |

(0.222) | (0.233) | |

Earn _{i,t+3} | 0.551*** | 0.674*** |

(0.094) | (0.115) | |

R _{i,t+3} | − 0.154*** | − 0.150*** |

(0.020) | (0.028) | |

Holder67CFO _{i,t} | − 0.060** | − 0.025 |

(0.030) | (0.043) | |

Holder67CFO _{i,t} × Earn_{i,t−1} | − 0.023 | − 0.502 |

(0.390) | (0.488) | |

Holder67CFO _{i,t} × Earn_{i,t} | 0.650** | 0.598 |

(0.299) | (0.456) | |

Holder67CFO _{i,t} × Earn_{i,t+3} | 0.169 | 0.238* |

(0.118) | (0.142) | |

Holder67CFO _{i,t} × R_{i,t+3} | 0.031 | 0.005 |

(0.023) | (0.031) | |

RevC_score1 _{i,t} | 0.727*** | |

(0.144) | ||

RevC_score1 _{i,t} × Earn_{i,t−1} | 1.625* | |

(0.886) | ||

RevC_score1 _{i,t} × Earn_{i,t} | − 1.226 | |

(0.754) | ||

RevC_score1 _{i,t} × Earn_{i,t+3} | 0.116 | |

(0.406) | ||

RevC_score1 _{i,t} × R_{i,t+3} | − 0.028 | |

(0.100) | ||

Holder67CFO _{i,t} × RevC_score1_{i,t} | − 0.360*** | |

(0.135) | ||

Holder67CFO _{i,t} × RevC_score1_{i,t} × Earn_{i,t−1} | − 0.499 | |

(1.880) | ||

Holder67CFO _{i,t} × RevC_score1_{i,t} × Earn_{i,t} | − 1.096 | |

(1.645) | ||

Holder67CFO _{i,t} × RevC_score1_{i,t} × Earn_{i,t+3} | 1.106* | |

(0.623) | ||

Holder67CFO _{i,t} × RevC_score1_{i,t} × F_{i,t+3} | 0.107 | |

(0.139) | ||

RevC_score2 _{i,t} | − 1.114*** | |

(0.123) | ||

RevC_score2 _{i,t} × Earn_{i,t−1} | 0.448 | |

(0.929) | ||

RevC_score2 _{i,t} × Earn_{i,t} | 1.529* | |

(0.780) | ||

RevC_score2 _{i,t} × Earn_{i,t+3} | 0.348 | |

(0.379) | ||

RevC_score2 _{i,t} × R_{i,t+3} | 0.037 | |

(0.128) | ||

Holder67CFO _{i,t} × RevC_score2_{i,t} | − 0.248* | |

(0.135) | ||

Holder67CFO _{i,t} × RevC_score2_{i,t} × Earn_{i,t−1} | − 1.039 | |

(1.313) | ||

Holder67CFO _{i,t} × RevC_score2_{i,t} × Earn_{i,t} | − 0.498 | |

(1.397) | ||

Holder67CFO _{i,t} × RevC_score2_{i,t} × Earn_{i,t+3} | 0.854* | |

(0.496) | ||

Holder67CFO _{i,t} × RevC_score2_{i,t} × R_{i,t+3} | 0.013 | |

(0.146) | ||

Constant | 1.551*** | 1.181*** |

(0.272) | (0.274) | |

Observations | 5,264 | 5,264 |

Controls in Eq. (14) | Yes | Yes |

Firm fixed effects | Yes | Yes |

Year fixed effects | Yes | Yes |

Adj. R ^{2} | 0.340 | 0.378 |

_{i,t}). The variable of interest is the interaction term, Hold67CFO

_{i,t}× RevC_score

_{i,t}. We follow prior studies (e.g., Lee 2010) to control CEO overconfidence (Hold67CEO

_{i,t}), CFO gender (CFO_male

_{i,t}), CFO ownership (CFO_ownership

_{i,t}), CEO gender (CEO_male

_{i,t}), CEO ownership (CEO_ownership

_{i,t}), Market to book ratio (MTB

_{i,t}), Volatility of cash flow (VolCashFlow

_{i,t}), Value-weighted 12-month market-adjusted returns (Vwretd

_{i,t}), Returns volatility (SdVwretd

_{i,t}), Net working capital (NWC

_{i,t}), Capital expenditure (CAPEX

_{i,t}), Acquisition expenditure (ACQ

_{i,t}), Research and development expenditure (R&D

_{i,t}), Firm Size (FirmSize

_{i,t}), Leverage (Leverage

_{i,t}), Dividend payer (Dividend

_{i,t}), CFO risk tolerance (CFO_vega

_{i,t}), CEO risk tolerance (CEO_vega

_{i,t}) and Strong governance (StrongCG

_{i,t}). Detailed variable measurements are provided in the Appendix.

_{i,t}and Holder67CFO

_{i,t}have a significantly positive coefficient, respectively. These findings suggest that overconfident managers have more incentives to hold cash, which is consistent with the finding of Chen et al. (2020). The coefficient of Holder67CFO

_{i,t}× RevC_score1

_{i,t}(Holder67CFO

_{i,t}× RevC_score2

_{i,t}) is significantly negative in column (3) (column (4)), suggesting that overconfident CFOs reduce the incentive to hold cash by reducing conditional accounting conservatism. Overall, this evidence indicates that overconfident CFOs reduce precautionary motives to save cash, as less conditional accounting conservatism conveys more primary information and increases firms’ financial flexibility.

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | Cash _{i,t} | Cash _{i,t} | Cash _{i,t} | Cash _{i,t} |

Holder67CFO _{i,t} | 0.011*** | 0.007** | 0.005 | |

(0.003) | (0.003) | (0.003) | ||

RevC_score1 _{i,t} | 0.106*** | |||

(0.015) | ||||

Holder67CFO _{i,t} × RevC_score1_{i,t} | − 0.036*** | |||

(0.013) | ||||

RevC_score2 _{i,t} | 0.080*** | |||

(0.011) | ||||

Holder67CFO _{i,t} × RevC_score2_{i,t} | − 0.022** | |||

(0.010) | ||||

Holder67CEO _{i,t} | 0.005* | 0.001 | 0.000 | 0.000 |

(0.003) | (0.003) | (0.003) | (0.003) | |

Constant | 0.466*** | 0.463*** | 0.516*** | 0.505*** |

(0.058) | (0.058) | (0.059) | (0.060) | |

Observations | 11,568 | 11,568 | 11,568 | 11,568 |

Controls in Eq. (15) | Yes | Yes | Yes | Yes |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.817 | 0.817 | 0.819 | 0.819 |

### 5.2 Alternative explanation

_{i,t}is the CFO equity incentive proxy; Holder67CFO

_{i,t}× CFO_equityincentive

_{i,t}is the variable of interest; β

_{3}captures the moderating effect of CFO equity incentive on the relationship between CFO overconfidence and conditional accounting conservatism. If the β

_{3}is significantly negative, overconfident CFOs with high equity incentives are more likely to reduce conditional accounting conservatism. Detailed variable measurements are provided in the Appendix.

_{i,t}× CFO_equityincentive

_{i,t}is not significant in columns (1) and (2). There is no evidence to suggest that the CFO equity incentive stimulates the negative relationship between CFO overconfidence and conditional accounting conservatism. Overall, we rule out this alternative explanation and further consolidate our main mechanism.

(1) | (2) | |
---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} |

Holder67_CFO _{i,t} | − 0.017*** | − 0.025*** |

(0.004) | (0.005) | |

CFO_equityincentive _{i,t} | − 0.016*** | − 0.011*** |

(0.003) | (0.004) | |

Holder67CFO _{i,t} × CFO_equityincentive_{i,t} | 0.006 | 0.003 |

(0.004) | (0.005) | |

Constant | 0.567*** | 0.492*** |

(0.029) | (0.052) | |

Observations | 11,710 | 11,710 |

Controls in Eq. (16) | Yes | Yes |

Firm fixed effects | Yes | Yes |

Year fixed effects | Yes | Yes |

Adj. R ^{2} | 0.801 | 0.765 |

## 6 CFO power, CFO overconfidence, and conditional accounting conservatism–regression analyses of Hypothesis 2

_{i,t}, is an indicator variable that equals one if the CFO is among the three highest-paid managers, and zero otherwise (Florackis and Sainani 2021). This rank represents the value the board assigns to the CFO and the CFO’s authority and responsibility within the firm (Bebchuk et al. 2011; Feng et al. 2011; Baker et al. 2019).

^{14}In addition to using CFO compensation, we also use a measure that reflects the CFO’s position in the management hierarchy to capture CFO power. In particular, earlier research has noted that managers who serve on boards are more likely to influence board and management team decisions (e.g., Beck and Mauldin 2014; Baker et al. 2019). Thus, the CFO board membership is widely used to measure CFO power (e.g., Caglio et al. 2018; Baker et al. 2019; Florackis and Sainani 2021). Besides, since the CEO sets the tone at the top, the close relationship between the CFO and the CEO increases the possibility and power of the CFO to participate in decision-making (e.g., Hsieh et al. 2018). The close relationship between the CEO and CFO is reflected in the long overlap in their tenure (Zenger and Lawrence 1989; Ancona and Caldwell 1992; Zhang 2019; Bowen et al. 2022). To reflect the CFO’s position in the management team, we create a dummy variable called CFO_PowerAlter

_{i,t}. It equals one if the CFO meets at least one of the following two conditions: being a member of the board of directors; working together with the CEO for a long time; otherwise, it equals zero.

_{i,t}is one of two CFO power proxies (CFO_rank

_{i,t}and CFO_PowerAlter

_{i,t}); Holder67CFO

_{i,t}× CFO_power

_{i,t}is the variable of interest; β

_{3}captures the moderating effect of CFO power on the relationship between CFO overconfidence and conditional accounting conservatism. If the β

_{3}is significantly negative, overconfident CFOs with more power are more likely to reduce conditional accounting conservatism. Detailed variable measurements are provided in the Appendix.

_{i,t}× CFO_rank

_{i,t}is significantly negative, suggesting that the CFO power enhances the negative relationship between CFO overconfidence and conditional accounting conservatism, which is in line with our second hypothesis (H2). The results are broadly similar when we use CFO_PowerAlter

_{i,t}to measure CFO power.

(1) | (2) | (3) | (4) | |
---|---|---|---|---|

Variable | C_score1 _{i,t} | C_score2 _{i,t} | C_score1 _{i,t} | C_score2 _{i,t} |

Holder67CFO _{i,t} | − 0.013*** | − 0.018*** | − 0.021*** | − 0.020*** |

(0.004) | (0.005) | (0.004) | (0.004) | |

CFO_rank _{i,t} | 0.006** | 0.014*** | ||

(0.003) | (0.004) | |||

Holder67CFO _{i,t} × CFO_rank_{i,t} | − 0.007* | − 0.015*** | ||

(0.004) | (0.005) | |||

CFO_PowerAlter _{i,t} | 0.007** | 0.013*** | ||

(0.003) | (0.003) | |||

Holder67CFO _{i,t} × CFO_PowerAlter_{i,t} | 0.005 | − 0.010** | ||

(0.004) | (0.005) | |||

Constant | 0.566*** | 0.484*** | 0.614*** | 0.483*** |

(0.030) | (0.053) | (0.032) | (0.051) | |

Observations | 10,679 | 10,679 | 11,890 | 11,890 |

Controls in Eq. (17) | Yes | Yes | Yes | Yes |

Firm fixed effects | Yes | Yes | Yes | Yes |

Year fixed effects | Yes | Yes | Yes | Yes |

Adj. R ^{2} | 0.804 | 0.764 | 0.743 | 0.764 |