DATA DESCRIPTIONS 1 Sample Selection

We collect data on social ties from BoardEx for the years 2004-2012. After excluding financial institutions (SIC codes 6000-6999) due to their special regulations and operating practices, we obtain a sample of 29,225 firm-year observations. We then exclude observations that do not have the requisite data in COMPUSTAT and Audit Analytics, giving rise to our initial sample of 20,965 firm-year observations from 3,546 firms. For our audit effort models, we exclude missing adjusted report lag, cases where the difference between a firm’s fiscal year-end and the audit report date is greater than 90 days from our initial sample, and missing control variables, resulting in 18,932 observations. For our conservatism test, we follow Kothari et al. (2005) to compute performance-adjusted discretionary accruals. After excluding observations with missing data to calculate these accruals and control variables, we have 10,079 observations. We use adjusted discretionary accruals to proxy for auditor conservatism. Finally, for our auditor change model, we restrict our sample to firms with high probability of committing fraudulent financial reporting. Excluding observations with missing control variables and data to compute firms’ fraud risk gives rise to 7,364 observations. To control for the

13 

potential effects of extreme values, we winsorize observations that fall in the top and bottom 1 percent of the sample distributions of the dependent and the continuous independent variables. 

Table 1 shows the distribution of sample firms by year and industry. Panel A reports that the sample is evenly distributed across 2004-2012, except in the first year when there are fewer observations. Panel B reports that, across our four samples, the top three industry categories are, respectively, manufacturingPart 2 (ranging from 25.83%~40.03%), services (ranging from 19.84%~26.52%), and manufacturingPart 1 (ranging from 17.45%~19.58%). Overall, our sample distribution by industry is very similar to that reported in Bruynseels and Cardinaels (2014).

[Insert Table 1 here]

3.2 CEO-AC Social Ties

For comparison with Bruynseels and Cardinaels (2014), we follow their definitions of the social ties measures. Education ties arise when the CEO and AC member(s) earn their bachelor or graduate degrees from the same educational institution. Employment ties arise when the CEO and AC member(s) have current or past overlapping employment as employees or directors in any firms. Other non-professional ties arise when CEO and AC member(s) have present or past membership in the same charities, clubs, foundations, and non-profit organizations. To test H1, we use an aggregate tie measure ALLTIES, which is defined as the proportion of AC members who have at least one type of ties with the CEO. To test H2, we use EDU, EMPLOY, and OTHERACT to denote the proportions of AC members who have connections to the CEO based on education, employment, and other non-professional ties, respectively. To test H3, we estimate all the regressions for our S&P 500 and non-S&P 500 sub-samples separately. Using the initial sample of 20965 firm-observations, Table 2 reports a complete picture of the average distributions of these tie measures over our sample period.

[Insert Table 2 here]

14 

Panel A of Table 2 shows a sharp increase in ALLTIES, from 16% in 2004 to 21% in 2012.

Turning to the three types of social ties, the proportion of EDU remains stable at 4% while the proportion of OTHERACT decreases from 7% in 2004 to 4% in 2012. Notably, the proportion of EMPLOY doubles, from 8% in 2004 to 16% in 2012. These 9-year distributions suggest that, on average, CEOs substantially increase the appointment of AC members that have employment ties with them and reduce AC members that are connected to them through other non-professional activities. If we separate the full sample into S&P 500 and non-S&P 500 groups, Panels B and C report that the S&P 500 firms tend to reduce AC members who are socially tied with the CEOs, mainly through a notable decrease, from 14% in 2004 to 9% in 2012, of AC members with other non-professional ties.

In contrast, the non-S&P 500 firms tend to increase the appointment of socially-tied individuals to their ACs, mainly through a substantial increase, from 8% in 2004 to 17% in 2012, of AC members with employment ties. These results suggest that S&P 500 and non-S&P 500 firms appear to adopt entirely different policies in hiring AC members. Particularly, CEOs in non-S&P 500 firms may have changed their AC selection decision by appointing more individuals who have previously worked with them and reducing individuals who are connected to them through non-professional activities. One possible reason underlying this phenomenon is that the CEOs tend to focus more on AC members’ professional expertise and competence rather than their friendship. This notion is consistent with Cohen et al.’s (2014) result that industry expertise together with accounting expertise can improve AC oversight effectiveness. In contrast, CEOs in the S&P 500 firms tend to decrease individuals who have connections with them from non-professional activities. Given Bruynseels and Cardinaels’ (2014) finding that it is non-professional ties that jeopardize AC oversight quality, Table 2 implies that the impairment of AC effectiveness appears to decrease in S&P 500 firms.

Note that our social ties distributions are similar to those documented in Panel A of Table 2 in

15 

Bruynseels and Cardinaels (2014). Looking at years 2004~2008, the average proportion of ALLTIES is between 15%~17% in our study and 16%~17% in theirs. In addition, we both report that the average proportion of EDU is 4%. Finally, the average proportions of EMPLOY and OTHERACT in our study (8%~10% and 5%~7%, respectively) are almost the same as those in their study (i.e., 9%~10% and 4%~6%, respectively). These basic statistics provide evidence that our identifications of ALLTIES, EDU, EMPLOY, and OTHERACT are consistent with those in Bruynseels and Cardinaels (2014).