We thank Zhihong Chen, Sunhwa Choi, Simon Fung, Jeong-Bon Kim, Byungjin Kwok, Helen Hyejin Kwon, Sooyoung Kwon, Jay Junghun Lee, Charlie Sohn, Min-Sup Song, Catherine Sonu and other workshop participants at City University of Hong Kong, Seoul National University and conference participants at Korean Accounting Association annual meeting. Jong-Hag Choi gratefully acknowledges financial support from Samil PWC Research Fellowship. Woo-Jong Lee appreciates financial support by The Hong Kong Polytechnic University.

About

Sections

PDF

Tools

Share a link

Email
Wechat
Bluesky

Abstract

Prior studies document a negative association between Big 4 auditor choice and the implied cost of equity capital, suggesting that Big 4 auditors mitigate information asymmetry (IA) between shareholders and managers. This study extends this line of research and reports that the negative association is more pronounced in multiple-segment firms, where IA is more severe than in single-segment firms. We also find that the association between Big 4 auditor choice and the cost of equity capital becomes more negative as the number of segments increases. Taken together, our findings suggest that the role of Big 4 auditors in reducing the cost of equity capital becomes more significant when greater IA exists.

1. Introduction

This study examines whether the negative association between high-quality auditor choice and the implied cost of equity capital is more pronounced in multiple-segment firms than in single-segment firms. Prior studies document that Big 4 auditors provide higher quality auditing services than non-Big 4 auditors and that such high-quality service enables firms to decrease information asymmetry (IA) between shareholders and managers (e.g. Francis et al., 1999; Willenborg, 1999).1 In contrast, Lawrence et al. (2011) argue that the difference in audit quality documented in prior studies is mainly due to client characteristics that endogenously determine auditor choice; thus, they call for future research to further disentangle client characteristics from audit quality effects.

To respond to their call and to add on inconclusive prior literature, we examine whether or not the strong monitoring role of Big 4 auditors mitigates the diversification discount documented in prior studies. Lang and Stulz (1994), Berger and Ofek (1995) and Gomes and Livdan (2004) all report that multiple-segment firms are valued at a discount compared with single-segment firms. They describe this phenomenon as a ‘diversification discount’ or ‘negative excess value’. More specifically, previous studies show that diversified firms have higher expected returns, suggesting that the cost of equity capital is higher for the multiple-segment firms than for single-segment firms (Bens and Monahan, 2004; Demirkan et al., 2012). Investors in diversified firms require a premium, because they expect lower cash flow from inefficient resource allocation, such as over- or underinvestment, cross-subsidization among segments (Martin and Sayrak, 2003), self-interested suboptimal behaviours of managers (Morck et al., 1990; Denis et al., 1997) or mergers of already inefficient firms (Graham et al., 2002). Another line of research deals with information problems associated with the diversification discount, which suggests that information asymmetries can lead to higher expected returns for diversified firms (Krishnaswami and Subramaniam, 1999; Gilson et al., 2001). In addition, investors' ability to monitor the value-destroying activities of managers is limited due to the complexity of diversified firms. This exacerbates the IA problem between managers and shareholders in such firms (Denis et al., 1997; Bens and Monahan, 2004). Thus, investors value multiple-segments firms at a discount to compensate for the higher risk related to increased IA (Bens and Monahan, 2004; Demirkan et al., 2012). Hence, the cost of capital for multiple-segment firms becomes higher than that for single-segment firms.2

Big 4 auditors decrease the cost of equity capital by providing a high-quality audit service, which enables them to perform a stronger monitoring role than the non-Big 4 auditors (e.g. Fan and Wong, 2005). As a result, investors trust the financial information audited by Big 4 auditors, which is reflected in their decision-making (i.e. information role of audits). Consequently, Khurana and Raman (2004) and Chen et al. (2011) report a lower cost of equity capital for Big 4 clients compared with the clients of non-Big 4 firms. Therefore, we expect that the cost of equity capital is lower for both single-segment and multiple-segment firms audited by Big 4 auditors. Moreover, we expect that the association between Big 4 auditor choice and the decrease in the cost of equity capital is more pronounced in multiple-segment firms than in single-segment ones, because high-quality audits play a more important role in the presence of greater IA (Pittman and Fortin, 2004; Ahmed et al.,2008 Chen et al., 2011).

To empirically examine this prediction, we use 21,177 firm-year observations from 1987 to 2010, of which 12,460 firm-year observations come from single-segment firms and the remaining 8717 firm-year observations are from multiple-segment firms. The cost of equity capital is measured by the average value of four different implied cost of capital estimates based on Gebhardt et al. (2001), Claus and Thomas (2001), Gode and Mohanram (2003) and Easton (2004). We find that the negative association between Big 4 auditor choice and the cost of equity capital is more pronounced among multiple-segment firms. When we control other factors that may influence the cost of equity capital, Big 4 clients exhibit lower equity cost in multiple-segment firms only, but not in single-segment firms. Further tests reveal that the negative association between Big 4 and the cost of equity in multiple-segment firms is mainly driven by firms with more than four segments. When we employ Heckman's (1979) method, an instrumental variable approach or propensity score matching techniques (Lawrence et al., 2011), we find that the documented results are not sensitive to potential endogenous auditor choices.

This study contributes to firms, practitioners, regulators and academics in various ways. First, we confirm the beneficial effects of high-quality audits regarding the decrease in the cost of capital for the multiple-segment firms.3 Second, the findings in this study imply that investors respond more to the higher quality information audited by Big 4 auditors when IA is more severe. Thus, the present study complements the findings reported in prior studies regarding the stronger monitoring role of Big 4 audits (Fan and Wong, 2005; Choi and Wong, 2007), especially in situations where high IA exists (Pittman and Fortin, 2004; Chen et al., 2011). We also extend these findings to settings related to the diversification discount. Third, findings in this study are of interest to regulators and auditors. Maintaining proper audit quality is a great concern for regulators and auditors. Thus, this study suggests that auditors, especially small ones, should exert greater efforts when auditing multiple-segment firms to enhance audit quality. Both regulators and investors also need to be more vigilant of such firms. Finally, our findings provide academics with insights into the dynamics of multiple-segment firms, in terms of the effect of IA on the cost of equity capital and the influence of auditor choice on the effect. This study can serve as a starting point for academics, allowing them to explore the effect of multiple-segments in various accounting-related contexts that, in turn, complements the findings of Demirkan et al. (2012).4

2. Theory and hypotheses development

Prior studies document various valuation implications of firm diversifications. First, managers may channel the resources of one segment to others experiencing financial difficulties (Stulz, 1990; Lamont, 1997; Rajan et al., 2000; Ozbas and Scharfstein, 2010). They do so not to maximize firm value, but to build a managerial empire as this enables them to control more resources and increase their private wealth or perquisite consumption (Meyer et al., 1992; Denis et al., 1997; Rajan et al., 2000; Aggarwal and Samwick, 2003). Second, Graham et al. (2002) document that diversified firms engage in nonvalue maximizing merger and acquisition activities, which result in decreased expected future cash flow. Similarly, Morck et al. (1990) suggest that the maximization of the self-interest of managers is the main reason for the poor investment decisions of diversified firms. Third, Lamont and Polk (2001) document that diversified firms have a higher expected future return, which accounts for about half of the diversification discount, while the other half is attributable to future cash flow. Recently, Mitton and Vorkink (2010) document that the higher expected returns in diversified firms are attributable to smaller skewness exposure, which offers less upside potential to investors. Finally, there is evidence on the inherent inefficiency of diversified firms (Maksimovic and Phillips, 2002).

Due to these suboptimal behaviours, the quality of accounting information in multiple-segment firms could be lower than that of single-segment firms. Consistent with this argument, Demirkan et al. (2012) find that the magnitude of discretionary accruals is higher in multiple-segment firms than in single-segment ones. Diversified firms are more likely to have greater agency problems between the top and divisional managers; thus, they tend to compensate the divisional managers based on observable performance. In contrast, single-segment firms are more directly monitored; thus, they are more likely to be compensated based on the level of managers' efforts.5 These differences lead to different incentives of managers to engage in earnings management in order to camouflage the observable level of performance, which results in lower quality financial information for multiple-segment firms (Demirkan et al., 2012). Based on prior evidence that high earnings lower the cost of capital (Bhattacharya et al., 2003; Francis et al., 2004), Demirkan et al. (2012) argue that investors and creditors require higher costs of capital for the same level of discretionary accruals in multiple-segment firms than in single-segment firms.6

One way to restore the credibility of accounting information is to hire auditors who can provide high-quality auditing service (Fan and Wong, 2005; Choi and Wong, 2007).7 Teoh and Wong (1993) find that investors respond more to the earnings news presented by Big 4 clients than by non-Big 4 clients. Willenborg (1999) finds that underpricing for IPO firms is less severe for Big 4 clients. Krishnan (2003a) also reports that market return reflects the discretionary accruals of the Big 4 clients more than those of non-Big 4 auditors. Similarly, Pittman and Fortin (2004) and Khurana and Raman (2004) find that the costs of debt and equity capital are lower for Big 4 clients.8 Kim et al. (2011) document that the cost of debt capital is lower for privately held firms that voluntarily hire auditors, especially when they are Big 4 clients. In summary, earlier findings suggest that Big 4 audits increase information quality and enhance the credibility of financial statements. As a result, IA between firms and shareholders is lower in Big 4 clients, thus resulting in lower costs of raised equity capital.9

However, none of these studies investigate the effect of high-quality audit services on the diversification discount. If the audits provided by Big 4 auditors can reduce IA via a more rigorous monitoring process, diversification discounts should decrease in Big 4 clients. In contrast, we expect single-segment firms to experience a less significant decrease in the cost of equity capital regardless of auditor choice, because they have relatively less IA than multiple-segment firms. If investors can easily obtain transparent information relevant to the value of the firm, which is the case for the single-segment firms in our context, then audit quality differences can have a relatively less impact on the cost of equity capital. As a result, we expect that the decrease in the cost of equity capital due to higher audit quality is more salient among multiple-segment firms than in single-segment ones. We propose the following hypothesis based on the discussions above:

H1: The magnitude of the decrease in the cost of equity capital for Big 4 clients is greater for multiple-segment firms than that for single-segment firms.

Consistent with our hypothesis, Brown and Hillegeist (2007) find that the negative association between disclosure quality and IA is stronger when the latter is more severe. This implies that in the presence of high IA, enhanced disclosure leads to reduced IA. In an auditing-related setting, Pittman and Fortin (2004) report that the choices of Big 4 auditors decrease the cost of debt capital as firms get older. In their study using Chinese firms, Chen et al. (2011) document the positive association between the cost of equity capital and the magnitude of earnings management. They also revealed that designating a large auditor decreases this association. The findings of Pittman and Fortin (2004) and Chen et al. (2011) imply that the role of Big 4 auditors becomes more important when greater IA exists. Choi and Wong (2007) also argue that the signalling role of Big 4 audits is more prominent when no other mechanism exists to guarantee the quality of financial reports. Similarly, Ahmed et al. (2008) report that the role of a high-quality audit is more important when corporate governance is weaker. In summary, prior studies conclude that Big 4 audits can be more effective when greater IA exists. This leads us to examine whether the effect of Big 4 audits on the cost of equity is more pronounced in multiple-segment firms.

3. Methodology

The research design and empirical evidence are provided in this section to answer the research question.

3.1. Empirical proxies for the implied cost of equity capital

Based on prior studies, we estimate the implied cost of equity capital using four methods developed by Gebhardt et al. (2001), Claus and Thomas (2001), Gode and Mohanram (2003) and Easton (2004), respectively. First, we estimate the implied cost of equity by deriving the discount rate that satisfies each of the four valuation models. The alternative estimates from different valuation methods are highly correlated with each other and similar within a reasonable range. Thus, by taking the average of these four estimates, we obtain a comprehensive measure of the cost of equity, which is denoted by CoE. This average estimate is used as the measure of the implied cost of equity throughout the article, because the averaging process mitigates the potential measurement errors associated with individual estimates.10

3.2. Empirical specification

To answer our research question, we posit regression Equation 1, which is based on prior studies that examine the determinants of the cost of equity capital (e.g. Gode and Mohanram, 2003; Dhaliwal et al., 2007).

$urn:x-wiley:08105391:media:acfi12011:acfi12011-math-0001$ (1)

where CoE_j,t is the average of the four costs of equity estimates for firm j at time t. Big is an indicator variable of audit quality that assumes a value of 1 when the firm retains a Big 4 auditor, and 0 otherwise. DiscAQ is the discretionary accruals quality estimates used in Francis et al. (2005).11 LnSize is the natural logarithm of total assets. LnBM is the natural logarithm of book-to-market ratio. LnDB is the natural logarithm of total liability-to-equity ratio. Beta(MKT), Beta(SMB) and Beta(HML) are the estimates of market, size and book-to-market premium, respectively; these come from the three-factor models derived from Fama and French (1996). These models are calculated over 36 months, ending in the month of the fiscal year-end. Bias is the difference in the actual and forecasted earnings scaled by the stock price of the fiscal year-end. Further, we construct industry dummies based on the 48 industry classifications proposed by Fama and French (1997). We also include year dummies to control time series variations in equity cost, which recognizes that the risk-free rate is a yearly constant in our firm-year analysis.

Equation 1 is separately estimated for single-segment and multiple-segment firms. We expect the mean effect of Big to be negative (c₁ < 0), because a higher audit quality reduces the cost of equity capital in the United States (Khurana and Raman, 2004). Such impact is expected to be more pronounced in multiple-segment firms where investors are likely to face greater IA than in single-segment firms.

In addition, we include several variables to control other factors that are likely to affect the cost of equity capital. Poor accounting quality (DiscAQ) is expected to be positively related to the cost of equity, which yields a positive sign for c₂ (Francis et al., 2005; Demirkan et al., 2012). Based on existing literature (e.g. Gode and Mohanram, 2003; Dhaliwal et al., 2007), we expect that the cost of equity is negatively related to firm size (LnSize) and positively related to book-to-market ratio (LnBM), financial leverage (LnDB) and market risk premia. To adjust risk premia, we include Beta(MKT), Beta(SMB) and Beta(HML). Finally, optimism in forecasts of analysts produces an upwardly biased cost of capital estimates (Easton and Sommers, 2007). Thus, we control forecast bias to reduce potential measurement errors in the cost of equity proxies. The formal variable definitions used for the regression analysesb are reported in Table 1.

Table 1. Variable definitions

Variable	Description
Regression variables
CoE	The average value of four cost of equity estimates based on Gode and Mohanram (2003), Easton (2004), Claus and Thomas (2001) and Gebhardt et al. (2001)
Multi	Equals 1 if a firm operates at least two business segments, and 0 otherwise
#Seg	The number of business segments
Big	Equals 1 if a firm is audited by one of the Big 4 audit firms, and 0 otherwise
Specialist	Equals 1 if a firm is audited by an industry specialist auditor, and 0 otherwise. We classify an auditor as an industry specialist when he earns the greatest market share in terms of audit fee in each 2-digit SIC industry
DiscAQ	Discretionary accrual quality as defined by Francis et al. (2005)
LnSize	The natural logarithm of total assets
LnBM	The natural logarithm of book-to-market ratio
LnDB	The natural logarithm of total debt-to-equity ratio
Beta(MKT)	Estimates of market premium from the three-factor model by Fama and French (1996) are calculated for more than 36 months until the end month of the fiscal year
Beta(SMB)	Estimates of size premium from the three-factor model by Fama and French (1996) are calculated for more than 36 months until the end month of the fiscal year
Beta(HML)	Estimates of book-to-market premium from the three-factor model by Fama and French (1996) are calculated for more than 36 months until the end month of the fiscal year
Bias	Actual EPS minus the mean analyst forecast, deflated by stock price at the end of the fiscal year
Partitioning variables
Size	The natural logarithm of total assets
Age	The number of years between the event year and the first year when a stock appears in CRSP
Volatility	The standard deviation of daily returns for the fiscal year
Volume	The average daily turnover in percentage over the fiscal year, where daily turnover is the ratio of the number of shares traded each day to the number of outstanding shares at the end of the day
Following	The number of analysts following a firm
Dispersion	Forecast dispersion of analysts measured when the number of analysts following a firm is more than two

4. Empirical results

4.1. Sample selection and descriptive statistics

Our sample selection process begins with all the firms listed on NYSE, AMEX or NASDAQ for any year from 1987 through 2010. Financial statement data, stock price and monthly return data, as well as analyst earnings and long-term growth forecasts data are extracted from Compustat, CRSP and I/B/E/S, respectively. Multiple-segment firms are those that report at least two segments operating in four-digit SIC codes obtained from the Compustat NAIC file.12 Similar to Berger and Ofek (1995), Lamont and Polk (2001) and Bens and Monahan (2004), we eliminate firm-years with the following characteristics: firm-years that lack at least one segment on the Compustat Industry Segment database, firm-years with one or more segments in the financial services industry (SIC 6000–6999) or in regulated industries (SIC 4900–4999), firm-years with less than $20 million total sales and firm-years in which the absolute value of the difference between the aggregate segment sales and the total reported sales exceeds 1 per cent of the total reported sales. Finally, we ensure that all observations have no missing variables, ultimately yielding 21,177 firm-year observations. Observations that fall within the top and bottom 1 per cent of the major research variables included in Equation 1 are winsorized to avoid potential outlier influence.

Table 2 presents descriptive statistics of the key variables. Panel A of Table 2 shows the mean and median of the implied cost of equity measure (CoE variable), which are 13.4 per cent and 10.5 per cent, respectively.13 The number of multiple-segment firms is 8717, which covers 41.2 per cent of the entire sample, as evidenced by the mean value of the Multi variable. On average, about two segments operate within a firm (#Seg variable). Notably, about 92.1 per cent of the sample observations are Big 4 clients (Big).14 With respect to the fundamental determinants of the cost of equity, the sample mean values are 6.324 for LnSize, −0.792 for LnBM and −0.151 for LnDB.15 The mean values of the three-factor loadings [Beta(MKT), Beta(SMB) and Beta(HML)] also fall within reasonable ranges. The mean value of Bias is −0.193, which is consistent with the optimistic bias of analysts in their earnings forecasts.

Table 2. Descriptive statistics. Panel A and B presents descriptive statistics of the key variables (the Pearson correlation coefficients and their p-values in italics). The sample period is from 1987 to 2010, yielding 21,177 firm-year observations. See Table 1 for variable definitions

Variable	Mean	SD	5%	Q1.	Median	Q3.	95%
(Panel A) Summary statistics (N = 21,177)
CoE	0.134	0.092	0.051	0.079	0.105	0.151	0.337
Multi	0.412	0.492	0.000	0.000	0.000	1.000	1.000
#Seg	1.903	1.344	1.000	1.000	1.000	3.000	5.000
Big	0.921	0.270	0.000	1.000	1.000	1.000	1.000
DiscAQ	0.000	0.021	−0.029	−0.013	−0.003	0.009	0.041
LnSize	6.324	1.696	3.743	5.080	6.161	7.429	9.517
LnBM	−0.792	0.681	−1.985	−1.212	−0.756	−0.336	0.290
LnDB	−0.151	0.911	−1.740	−0.753	−0.094	0.459	1.290
Beta(MKT)	1.081	0.650	0.069	0.673	1.028	1.439	2.290
Beta(SMB)	0.752	0.937	−0.620	0.119	0.637	1.275	2.514
Beta(HML)	0.031	1.115	−1.956	−0.617	0.094	0.734	1.805
Bias	−0.193	0.388	−0.894	−0.208	−0.054	−0.001	0.022

	(b)	(c)	(d)	(e)	(f)	(g)	(h)	(i)	(j)	(k)	Bias
(Panel B) Pearson correlation matrix
(a) CoE	0.076	0.120	−0.044	0.034	−0.044	0.220	0.102	0.071	0.062	0.144	−0.195
(a) CoE	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
(b) Multi		0.803	0.053	0.002	0.263	0.041	0.179	−0.014	−0.105	0.125	0.083
(b) Multi		0.00	0.00	0.76	0.00	0.00	0.00	0.04	0.00	0.00	0.00
(c) #Segment			0.067	0.012	0.333	0.015	0.197	−0.012	−0.147	0.121	0.082
(c) #Segment			0.00	0.08	0.00	0.03	0.00	0.09	0.00	0.00	0.00
(d) Big				−0.024	0.202	−0.053	0.136	−0.003	−0.074	−0.016	0.102
(d) Big				0.00	0.00	0.00	0.00	0.68	0.00	0.02	0.00
(e) DiscAQ					0.003	−0.101	−0.004	0.037	0.033	−0.045	−0.034
(e) DiscAQ					0.68	0.00	0.54	0.00	0.00	0.00	0.00
(f) LnSize						−0.129	0.355	0.020	−0.368	0.035	0.232
(f) LnSize						0.00	0.00	0.00	0.00	0.00	0.00
(g) lnBM							0.000	0.022	0.139	0.147	−0.198
(g) lnBM							0.99	0.00	0.00	0.00	0.00
(h) LnDB								−0.028	−0.094	0.177	0.080
(h) LnDB								0.00	0.00	0.00	0.00
(i) Beta(MKT)									0.010	0.142	−0.075
(i) Beta(MKT)									0.13	0.00	0.00
(j) Beta(SMB)										0.135	−0.164
(j) Beta(SMB)										0.00	0.00
(k) Beta(HML)											0.047
(k) Beta(HML)											0.00

Panel B of Table 2 reports the Pearson correlations between the key variables. Both the multiple-segment indicator (Multi) and the number of segments (#Seg) are positively associated with CoE. This means that the cost of raised equity capital is higher in multiple-segment firms than in single-segment ones. Consistent with Khurana and Raman (2004), Big is negatively associated with CoE, which suggests that higher audit quality mitigates IA within a firm, thereby reducing the cost of equity capital. DiscAQ is positively associated with CoE. This result is consistent with Francis et al. (2005) who suggest that market participants impose a higher cost of capital on firms with poor accruals quality. CoE is negatively associated with firm size (LnSize), but positively associated with book-to-market ratio (LnBM), financial leverage (LnDB) and market risk premia [Beta(MKT), Beta(SMB) and Beta(HML)]. Correlations between control variables are not high. The largest correlation is between LnSize and Beta(SMB) at −0.368. Thus, it is unlikely that multicollinearity can significantly affect the empirical findings in the multivariate analyses.

4.2. Univariate analyses

Table 3 presents the average values of the implied cost of capital (CoE) by the number of segments. The sample firms are partitioned into four groups: single-segment firms, firms with two or three segments, firms with four or five segments and firms with six or more segments. Column (3) shows that the proportion of Big 4 clients increases with the number of segments, whereas Column (4) shows that the cost of equity capital monotonically increases with the number of segments. For example, the average implied cost of equity capital for single-segment firms is 13.2 per cent, whereas it increases to 17.2 per cent for firms with at least six segments. This trend is observed in both Big 4 clients [Column (6)] and non-Big 4 clients [Column (5)]. Furthermore, the mean values of the cost of equity capital are consistently greater for non-Big clients, as reported in Khurana and Raman (2004). In addition, the difference in the cost of equity capital between Big 4 and non-Big 4 clients [Column (7)] increases with the number of segments. The difference is −0.012 in single-segment firms but increases to −0.134 in firms with six or more segments. A difference of −0.134 for firms with six or more segments implies that the cost of equity capital for non-Big 4 clients is 13.4 per cent higher than that of Big 4 clients. This highlights the dramatic effect of Big 4 audits; it also implies that the informative role of audit quality becomes more important as IA increases. The above analyses, however, are not conclusive given that other factors potentially influencing the cost of capital are not controlled. This control seems more critical for multiple-segment firms, because of the huge difference in the cost of capital between Big 4 and non-Big 4 clients. Results of multivariate analyses are reported in the next section.

Table 3. Univariate analyses. The table presents the average cost of equity (CoE) by the number of segments (#Seg). The sample period is from 1987 to 2010, yielding 21,177 firm-year observations. Column (3) shows the proportion of Big 4 clients in each segment group. Columns (4) to (6) show CoE for the full sample and for non-Big 4 and Big 4 clients. The last column shows the difference between CoE reported in columns (5) and (6)

(1) #Seg	(2) Sample size	(3) Big 4 (%)	(4) Full sample	Big 4 versus non-Big 4
(1) #Seg	(2) Sample size	(3) Big 4 (%)	(4) Full sample	(5) Non-Big 4	(6) Big 4	(7) Diff. [(6)–(5)]
Total	21,177	0.921	0.134	0.148	0.133	−0.015***
Least
1	12,460	0.909	0.132	0.139	0.127	−0.012***
2–3	5859	0.928	0.139	0.151	0.133	−0.018***
4–5	2391	0.955	0.166	0.218	0.155	−0.063***
Most
6	467	0.979	0.172	0.305	0.171	−0.134*
Diff. (most to least)		0.070***	0.040***	0.166***	0.044***	−0.122***

* and *** represent statistical significances at the 10% and 1% levels, respectively. See Table 1 for variable definitions.

4.3. Regression analyses

Multivariate regression analyses are performed to ensure that the results are not driven by other firm characteristics. The report t-statistics are estimated based on robust standard errors adjusted for clustering by each firm. For every regression result reported below, t-statistics are based on robust standard errors adjusted for clustering by each firm.

Equation 1 is applied separately for each firm group in consideration of the underlying factors that may have caused the reported differences between single- and multiple-segment firms. In Panel A of Table 4, the Big coefficient for the single-segment firms is not statistically different from zero, whereas it is significantly negative for multiple-segment firms. This implies that the earlier findings of Khurana and Raman (2004) may be driven by multiple-segment firms. The coefficient for the multiple-segment group is −0.009, which indicates that Big 4 clients enjoy an approximately 0.9 per cent lower implied cost of equity capital compared with non-Big 4 clients. The difference between the two coefficients (−0.010) is significant at the 5 per cent level as reported in the rightmost column of Panel A. This finding is consistent with the notion that the informative role of big auditors is greater under greater information asymmetry. Consistent with Francis et al. (2005), the coefficients on DiscAQ are significantly positive in both groups, although this difference is not significant. Other coefficients on control variables also exhibit expected signs even though some coefficients are not statistically significant.

Table 4. Regression results. Panel A and B presents the regression results where CoE is regressed on Big with other control variables for firms with single or multiple segments (by the number of segments) using Equation 1. The sample period is from 1987 to 2010, yielding 21,177 firm-year observations. Industry and Year dummies are included, but their coefficients are not reported for brevity

Parameter	Single segment (Multi = 0)		Multiple segments (Multi = 1)		Diff (multi-single)
Parameter	Estimate	t-stat.	Estimate	t-stat.	Diff (multi-single)
(Panel A) Regression results for single-segment versus multiple-segment firms
Intercept	0.175***	24.59	0.148***	11.96	−0.027***
Big	0.001	0.36	−0.009**	−2.15	−0.010**
DiscAQ	0.123***	3.79	0.033	0.65	−0.090**
LnSize	−0.007***	−13.22	−0.003***	−3.94	0.004***
LnBM	0.030***	28.15	0.031***	19.94	0.001
LnDB	0.017***	19.82	0.019***	14.66	0.002
Beta(MKT)	0.004***	3.82	0.004***	2.49	0.000
Beta(SMB)	0.002***	2.75	0.005***	4.82	0.004**
Beta(HML)	0.001	1.22	0.001*	1.39	0.001*
Bias	−0.025***	−12.80	−0.029***	−9.01	−0.004
Industry/year	Yes		Yes
No. observations	12,460		8,717
Adj. R²	0.371		0.402

Parameter	(1) #Seg = 1		(2) #Seg = 2 or 3		(3) #Seg = 4 or 5		(4) #Seg ≥ 6
Parameter	Estimate	t-stat.	Estimate	t-stat.	Estimate	t-stat.	Estimate	t-stat.
(Panel B) Regression results by the number of segments
Intercept	0.175***	24.59	0.126***	12.68	0.206***	5.29	0.123*	1.56
Big	0.001	0.36	−0.003	−0.73	−0.024**	−2.04	−0.101**	−1.67
DiscAQ	0.123***	3.79	0.018	0.33	0.114	1.08	−0.170	−0.45
LnSize	−0.007***	−13.22	−0.004***	−4.69	−0.001	−0.94	0.003	0.63
LnBM	0.030***	28.15	0.028***	16.02	0.038***	11.36	0.035***	3.73
LnDB	0.017***	19.82	0.018***	12.93	0.023***	7.33	0.023**	2.29
Beta(MKT)	0.004***	3.82	0.005***	2.62	0.004	0.97	−0.006	−0.39
Beta(SMB)	0.002***	2.75	0.006***	5.29	0.003	1.07	0.011	1.04
Beta(HML)	0.001	1.22	0.002**	1.66	0.000	−0.12	−0.007	−0.74
Bias	−0.025***	−12.80	−0.028***	−8.01	−0.031***	−4.22	−0.028	−0.62
Industry/year	Yes		Yes		Yes		Yes
No. observ-ations	12,460		5,859		2,391		467
Adj. R²	0.371		0.382		0.415		0.415

*, ** and *** represent statistical significances at the 10%, 5% and 1% levels, respectively (one-tailed). The t-statistics are computed using robust standard errors that are adjusted for clustering by each firm. See Table 1 for variable definitions.

The full sample in Panel B of Table 4 is divided into four subsamples based on the number of segments. Regression analyses are performed in each subsample. Column (1) shows the results for single-segment firms, which are consistent with the results in Panel A. Panel B clearly demonstrates a monotonic increase in Big coefficients with the number of segments. The coefficients on Big are -0.003 in firms with two or three segments [Column (2)], −0.024 in firms with four or five segments [Column (3)] and −0.101 in firms with the most segments [Column (4)]. Interestingly, only the coefficients on Big in Columns (3) and (4) are statistically significant, suggesting that the findings in Panel A are driven by firms with at least four segments. Thus, the findings in Panel B corroborate those in Panel A, providing strong evidence of the negative association between Big 4 auditor choice and the cost of capital in multiple-segment firms.16

The above argument focuses on the role of auditors in mitigating IA in multiple-segment firms. However, it can be argued that the findings may be driven by operational complexity rather than IA in multiple-segment firms. One way to reaffirm the information-based explanation for the previously reported findings is to examine whether or not the effect of auditors in multiple-segment firms varies with other proxies for information asymmetry. If the role of Big 4 firms in decreasing the cost of equity capital for multiple-segment firms is more significant when the direct measure of IA is relatively high (i.e. when IA is more severe), the results support our main findings. On the other hand, if our previous findings are only driven by the operational complexity inherent in multiple-segment firms, the monitoring role of Big 4 firms in multiple-segment firms should be observed regardless of the degree of IA measured by other proxies.

We employ six measures of firm-level IA in addition to our original measure (i.e. the number of segments). The six measures represent the level of IA from different and more direct perspectives than the number of segments. Based on prior studies, we classify smaller firms, young firms, those followed by fewer analysts, as well as those with higher trading volatility, with lower liquidity and with higher forecast dispersion as firms with higher IA.17 The median values of each variable are used to divide the sample into high- versus low-IA groups. The relation of the six variables to information symmetry is summarized in Panel A of Table 5.

Table 5. Regression results for subsamples. Panel A presents various firm characteristics associated with IA. Panel B reports regression coefficients on #Seg and their corresponding interaction terms when we use Equation 2. In the specification, other control variables are included, but their coefficients are not reported for brevity. Size is the natural logarithm of total assets. Age is the number of years between the event year and the first year when a stock appears in CRSP. Volatility is the standard deviation of daily returns for the fiscal year. Volume is the average daily turnover in percentage over the fiscal year, where daily turnover is the ratio of the number of shares traded each day to the number of outstanding shares at the end of the day. Following is the number of analysts following a stock. Dispersion is the dispersion of analysts' forecasts

	Size	Age	Volatility	Volume	Following	Dispersion
(Panel A) Firm characteristics associated with information asymmetry
High IA	Small	Young	High	Low	Low	High
Low IA	Large	Old	Low	High	High	Low

Firm characteristics	IA	#Seg		Big * #Seg		N	Adj. R²
Firm characteristics	IA	Estimate	t-stat.	Estimate	t-stat.	N	Adj. R²
(Panel B) Regression results by the degree of information asymmetry
Size	High	0.005	1.57	−0.006*	−1.71	10,588	0.314
Size	Low	0.007	1.48	−0.007	−1.37	10,589	0.472
Age	High	0.008**	2.16	−0.006*	−1.71	10,631	0.389
Age	Low	0.004	0.98	−0.003	−0.80	10,546	0.387
Volatility	High	0.008**	2.39	−0.009**	−2.61	10,495	0.342
Volatility	Low	0.001	0.34	−0.005	−1.00	10,494	0.448
Volume	High	0.008*	1.85	−0.006	−1.52	10,494	0.365
Volume	Low	0.004	1.26	−0.004	−1.23	10,495	0.421
Following	High	0.007**	1.96	−0.006*	−1.68	10,396	0.314
Following	Low	0.002	0.86	−0.002	−0.89	10,781	0.376
Dispersion	High	0.008**	2.28	−0.008**	−2.17	8,495	0.476
Dispersion	Low	0.001	0.43	−0.001	−0.28	8,501	0.331

* and ** and&!ast;&!ast;&!ast; represent statistical significances at the 10%, 5% and 1% levels, respectively (one-tailed). The t-statistics are computed using robust standard errors that are adjusted for clustering by each firm. See Table 1 for other variable definitions.

Panel B of Table 5 reports the regression results using Equation 2, where the interaction variables of Big and #Seg (Big*#Seg) are added into Equation 1. The number of observations used in these analyses varies across partitioning variables, because not all of the IA-related variables are available for all of the sample observations and all the available observations are used for each analysis. Regression analyses are performed separately for firms with high IA and low IA. The coefficients on the two variables of interest (i.e. #Seg and Big*#Seg) are only used for simplicity. For illustration purposes, we use bold characters to represent the results of subsamples with high IA and normal characters to represent subsamples with low IA.

Interestingly, a similar pattern is observed in all partitioning IA variables. Most of the coefficients on Big*#Seg are significant and negative when IA is more severe. For example, when a firm is relatively young and IA is high, the coefficient on Big*#Seg is −0.006, which is significant at the 10 per cent level (t = −1.71). On the other hand, when IA is less severe in low-IA groups, their coefficients are all negative but generally insignificant. For example, when a firm is relatively old and IA is low, the coefficient on Big*#Seg is −0.003, which is insignificant at any conventional level (t = −0.80).

In summary, the negative association between the choice of high-quality auditors and the cost of capital is more pronounced among firms with greater information asymmetry. In contrast, we fail to find significant results in most cases for firms with low information asymmetry. These results indicate that the greater influence of audit quality in multiple-segment firms is related to the information role of the audit. Moreover, these findings do not support the claims of a number of finance articles that the diversification discount is caused by factors other than agency problem (e.g. Lang and Stulz, 1994).

4.4. Tests on the potential endogeneity of auditor choice

Fan and Wong (2005) and Choi and Wong (2007) report that firms with greater agency problems tend to hire high-quality auditors to mitigate investors' concerns. They also observe that this tendency is more pronounced in firms with a higher demand for equity capital, implying that such firms demand high-quality auditors to decrease their equity capital costs. Therefore, controlling for auditor choice decision is critical to the extent that the results of Table 4 could be driven by the intentions of multiple-segment firms to lower their costs of raising equity capital by hiring high-quality auditors.18

In Table 6, we perform an additional test to address potential endogeneity concerns in relation to auditor choice. First, we utilize the instrumental variable approach and report first-stage results in Panel A. Based on prior studies including Chaney et al. (2004) and Behn et al. (2008), we adopt the probit model to estimate the probability of selecting a Big 4 auditor and include several firm characteristics. The determinant variables include firm size (Size, natural logarithm of total assets), asset turnover (Aturn, sales divided by total assets), financial leverage (Lev, long-term debt divided by average total assets), current ratio (Curr, current assets divided by total assets), quick ratio (Quick, current assets minus inventory divided by current liabilities), return-on-assets ratio (ROA, net income divided by average total assets), loss indicator (Loss, having a value of 1 if a firm incurred a loss, and 0 otherwise), an indicator for long-term debt issuance (ISSUE, having a value of 1 if a firm issued more than 20 per cent of existing long-term debt for the past 2 years) and year- and industry-indicator variables. The model fits reasonably well as pseudo R² reaches 0.163.

Table 6. Addressing potential endogeneity of auditor choice. Panel A presents the first-stage probit regression results for the auditor choice model. In Panel B, we present the regression results, where CoE is regressed on the fitted value of Big with other control variables for firms with single versus multiple segments. Panel C presents the regression results with the addition of the inverse Mills ratio (IMR) to the Equation 2. In Panel B and C, the fitted value of Big (IMR) is estimated from the regression where Big 4 is regressed on its potential determinants and is included in the main regression. The sample period is from 1987 to 2010, yielding 21,177 firm-year observations. Industry and year dummies are included, but their coefficients are not reported for brevity

	Estimate	p-Value
(Panel A) First-stage regression
Intercept	−3.996***	0.00
Size	0.818***	0.00
Aturn	0.257***	0.00
Lev	0.918***	0.00
Curr	0.402*	0.07
Quick	−0.006	0.72
ROA	0.487	0.47
ROALoss*	−1.590*	0.05
Loss	0.012	0.90
Issue	−0.306***	0.00
Accruals	−1.520***	0.00
Year	Yes
Industry	Yes
No. observations	17,488
Pseudo R²	0.163

Parameter	Single segment (Multi = 0)		Multiple segments (Multi = 1)		Diff (multi-single)
Parameter	Estimate	t-stat.	Estimate	t-stat.	Diff (multi-single)
(Panel B) Instrumental variable approach
Intercept	0.161***	9.11	0.195***	18.34	0.034*
Fitted value of Big	−0.026*	−1.42	−0.040***	−3.34	−0.014**
DiscAQ	−0.011	−0.20	0.109***	2.97	0.120
LnSize	−0.002***	−2.33	−0.004***	−5.61	−0.002*
LnBM	0.031***	18.31	0.030***	26.22	−0.001
LnDB	0.019***	13.78	0.017***	17.95	−0.002
Beta(MKT)	0.004**	2.06	0.004***	3.62	0.000
Beta(SMB)	0.006***	4.58	0.002***	2.65	−0.004
Beta(HML)	0.001	0.82	0.001**	1.98	0.000***
Bias	−0.026***	−7.43	−0.025***	−11.61	0.001
Year	Yes		Yes
Industry	Yes		Yes
No. observations	10,221		7,267
Adj. R²	0.370		0.400

Parameter	Single segment (Multi = 0)		Multiple segments (Multi = 1)		Diff (multi-single)
Parameter	Estimate	t-stat.	Estimate	t-stat.	Diff (multi-single)
(Panel C) Heckman's approach
Intercept	0.153***	9.10	0.184***	18.13	0.031*
Big	−0.015	−0.92	−0.022**	−1.99	−0.006*
DiscAQ	−0.011	−0.20	0.111***	3.03	0.122
LnSize	−0.002***	−2.80	−0.005***	−6.88	−0.002*
LnBM	0.031***	18.19	0.030***	26.13	0.000
LnDB	0.019***	13.65	0.017***	17.76	−0.002
Beta(MKT)	0.004**	1.97	0.004***	3.54	0.000
Beta(SMB)	0.006***	4.54	0.002***	2.49	−0.004
Beta(HML)	0.001	0.84	0.001**	2.09	0.000
Bias	−0.027***	−7.53	−0.026***	−11.76	0.001***
IMR	0.003	0.35	0.013**	2.18	0.010**
Year	Yes		Yes
Industry	Yes		Yes
No. observations	10,221		7,267
Adj. R²	0.369		0.400

*, ** and *** represent statistical significances at the 10%, 5% and 1% levels, respectively (two-tailed for Panel A and one-tailed for Panels B and C). The t-statistics are computed using robust standard errors that are adjusted for clustering by each firm. ROA, return-on-assets.

In Panel B of Table 6, we present the results of using the instrumental variables (IV) approach by incorporating the fitted value of the Big 4 auditor indicator from the first stage of regression. In Panel C, we adopt the Heckman's 2-stage least squares approach by calculating the inverse Mills ratio (IMR) for the Big 4 auditor choice. As shown in Panels B and C (Table 6), the results corroborate our previous findings that the association between audit quality and cost of equity is stronger in multiple-segment firms than in single-segment firms even after controlling the potential endogeneity of auditor choice. The IMR coefficient is also only statistically significant in the multiple-segment group.

Although these approaches are common in accounting literature, we are cautious in drawing clear inferences from these tests. For the IV approach, tests of instrument power reflect the relevance of our specification. The overidentification test using the Hansen's J statistic marginally fails to reject the null with p-value of 0.1053, whereas the underidentification test of Kleibergen and Paap (2006) that uses the rk LM statistic rejects the null with p-value of 0.0091. In addition, the p-value (<0.01) of the endogeneity test shows that the selection of Big 4 auditors is endogenous. In other words, no correlation exists between the explanatory variable and the error terms in the failure of OLS. Therefore, the IV application, which collectively supports the validity of our approach, is necessary.19 However, if the instrument is only weakly correlated with the endogenous regressor, instrumental variable estimates are more likely to be biased and produce an incorrect statistical inference than simple OLS estimates (Larcker and Rusticus, 2010). The weak identification test of Kleibergen and Paap (2006) with a yield of 3.855 for the Wald F statistic, which does not exceed the heuristic threshold value of 10 by Staiger and Stock (1997) for weak instruments, indicates the relative weakness of selected instruments. Moreover, for the Heckman approach, identifying a valid exogenous variable is vital, but we are unable to perfectly identify the exogenous variable in our selection model. These caveats to our specifications necessitate additional specification, which will be discussed in the next section.

4.5. Controlling for client characteristics

Propensity score matching has several advantages over IV methods as a solution to the endogeneity problem (Core, 2010). First, it does not require the identification of reliable instrument variables that can be very challenging in many empirical settings. Second, it is robust to misspecification of the functional form because it is not subject to parametric assumptions of IV methods (Armstrong et al., 2010). Thus, we believe that the propensity score matching technique supplements the weak instrument problem discussed in Table 6. Another reason for the use of propensity score matching is that it matches observations based on the probability of selecting a Big 4 auditor. For example, if significant size differences are present between the clients of Big 4 and of non-Big 4 auditors in each segment grouping, our previous results may simply capture the size effect on the cost of equity. Lawrence et al. (2011) also argue that differences between Big 4 and non-Big 4 auditors are a reflection of their clients' characteristics.

To address these concerns, we estimate the propensity of selecting Big 4 auditors from the auditor choice model in Table 6 and then match a non-Big 4 client with a Big 4 client with the closest predicted value.20 Based on the method by Lawrence et al. (2011), we match the pairs with replacements, which yields 2788 matched pairs. Panel A in Table 7 presents the mean values of the variables for both full and propensity score-matched samples. Before matching, all variables except Beta(HML) exhibit significant differences between Big 4 and non-Big 4 clients. However, after matching, differences of most variables become statistically insignificant, implying effective propensity score matching.

Table 7. Regression results using propensity score-matched sample. Panel A presents the average values of variables for full and propensity score-matched samples. Propensity scores are calculated using the first-stage probit model in Table 6. Panel B presents the regression results using the propensity score-matched pairs. The sample period is from 1987 to 2010, yielding 21,177 firm-year observations. Industry and year dummies are included, but their coefficients are not reported for brevity

	Non-Big 4	Big 4	Diff.	Non-Big 4	Big 4	Diff.
	Full sample			Propensity-matched sample
(Panel A) Mean values of variables for full and propensity score-matched samples
CoE	0.148	0.133	−0.015***	0.138	0.125	−0.012
DiscAQ	0.002	0.000	−0.002***	0.003	−0.003	−0.006*
LnSize	5.156	6.424	1.269***	5.219	5.522	0.303
LnBM	−0.668	−0.803	−0.135***	−0.638	−0.749	−0.111
LnDB	−0.574	−0.114	0.460***	−0.615	−0.409	0.206
Beta(MKT)	1.087	1.080	−0.007	1.189	0.983	−0.205*
Beta(SMB)	0.990	0.732	−0.258***	0.879	0.914	0.035
Beta(HML)	0.091	0.026	−0.065**	0.159	0.093	−0.066
Bias	−0.329	−0.182	0.147***	−0.241	−0.192	0.049
No. observations	1,672	19,505		1,394	1,394

Parameter	Single segment (Multi = 0)		Multiple segments (Multi = 1)		Diff (multi-single)
Parameter	Estimate	t-stat.	Estimate	t-stat.	Diff (multi-single)
(Panel B) Regression results for single-segment versus multiple-segment firms
Intercept	0.188***	9.89	0.109***	4.66	−0.079
Big	0.003	0.92	−0.013**	−1.99	−0.016***
DiscAQ	0.232***	2.72	−0.013	−0.09	−0.245
LnSize	−0.010***	−5.99	−0.004*	−1.62	0.005*
LnBM	0.025***	8.50	0.033***	6.38	0.008
LnDB	0.017***	7.52	0.016***	4.20	−0.001**
Beta(MKT)	0.005**	2.04	0.005	1.14	0.000
Beta(SMB)	0.001	0.93	0.009***	2.85	0.007
Beta(HML)	0.000	0.18	−0.001	−0.32	−0.001
Bias	−0.030***	−7.70	−0.018***	−2.71	0.012
Year	Yes		Yes
Industry	Yes		Yes
No. observations	1,878		910
Adj. R²	0.358		0.345

*, ** and *** represent statistical significances at the 10%, 5% and 1% levels, respectively (one-tailed). The t-statistics are computed using robust standard errors that are adjusted for clustering by each firm. See Table 1 for variable definitions.

In Panel B of Table 7, we confirm our previous results using the matched sample. For single-segment firms, the proportion of Big audit clients is 50.18 per cent, while that of multiple-segment firms is 49.67 per cent, suggesting effective control of the potentially endogenous relation between auditor and diversification choice. Even after propensity score matching, a significant difference is still found in the Big coefficients between single-segment and multiple-segment firms. Consistent with the result in Table 4, the negative association between Big and CoE is higher in multiple-segment firms for the matched sample. In sum, the greater impact of a number of auditors in mitigating IA for multiple-segment firms is robust in several treatments that address endogeneity concerns.

4.6. Alternative measures of audit quality

In this section, we discuss sensitivity analyses using alternative measures of audit quality. First, we consider auditor tenure as a proxy for audit quality, because understanding client-specific knowledge, such as operations, accounting system and internal control structure, is crucial for auditors to enhance audit quality. Despite earlier debates on how the tenure of the auditor affects auditor independence, recent studies conclude that auditor tenure allows auditors to rely less on managerial estimates and become more independent of management (e.g. Johnson et al., 2002; Myers et al., 2003; Mansi et al., 2004; Ghosh and Moon, 2005).21 Thus, we include Tenure instead of Big as an alternate for audit quality in Equation 1. We count auditor tenure from 1978, while the first year observation is counted as 1 by construction. We define Tenure as an indicator variable with a value of 1 if auditor tenure is 10 years or more, and 0 otherwise.22 We use 10 years to define Tenure, because the sample mean (median) of auditor tenure is 10.10 (9.00) years. We also restrict the sample to Big 4 clients (N = 19,505) to draw clearer inferences on auditor tenure as a proxy of audit quality. We then replace Big with Tenure in Equation 1.

This specification resolves a few potential statistical problems inherent in Equations 1 and 2. As reported in Table 2, only 8 per cent of our samples are non-Big 4 clients, and this proportion decreases as the number of segments increases. Thus, firms in this study are rarely audited by non-Big 4 auditors due to the lengthy requirements in sample selection criteria that allow most large firms to be included in the data set. This small percentage may cause a lack of statistical vigour in Big 4 versus non-Big 4 tests, although our earlier findings remain robust. Nevertheless, to address potential problems in relation to the small proportion of non-Big 4 clients, we employ empirical analyses using only the Tenure variable for Big 4 clients. Untabulated results reveal that the Tenure coefficient for multiple-segment firms is significantly negative, although it is not statistically different from that of single-segment firms. The coefficient on Tenure in single-segment firms is −0.003 (t = −1.41), while that of multiple-segment firms is −0.003 (t = −2.39). This result corroborates our earlier findings that are documented in Panel B, Table 4.23

Thus far, the discussion relies on the relationship between auditor tenure and audit quality based on extensive empirical evidence. However, other evidences also suggest that long auditor tenure does not have a negative impact on audit quality (e.g., Geiger and Raghunandan, 2002). Thus, we choose auditor industry expertise as another alternative proxy for audit quality. Prior studies report that auditor expertise plays an important role in audit quality (e.g. Balsam et al., 2003; Krishnan, 2003b). We follow these studies and examine whether the choice of an industry expert auditor decreases the cost of equity capital, unlike in single- and multiple-segment firms that are audited by Big 4 auditors. Then, we determine auditor expertise by considering the individual market shares of different auditor industries in a specific year. We construct an indicator variable (Specialist) that is equal to 1 if the firm is audited by an industry specialist that reports the highest market share in terms of audit fees in each 2-digit SIC industry. We then replace the Big variable in Equation 1 with Specialist. Accordingly, we restrict our sample to firms that are audited by one of the Big 4 auditors with audit fee data. The resulting sample consists of 5642 firm-year observations.24 Untabulated results suggest that the Specialist coefficient is significantly negative when we use the entire sample. If we divide our sample into single- and multiple-segment firms, only the Specialist coefficient for multiple-segment firms (−0.015, t = −2.00) is significantly negative. Such coefficient is also significantly higher than that of single-segment firms (p-value of the difference = 0.02), which is consistent with the results reported in Table 4.25

5. Conclusion

This study examines whether the role of Big 4 auditors in decreasing the cost of equity capital is more pronounced in multiple-segment firms than in single-segment firms. Prior studies document that multiple-segment firms are valued at a more discounted price than single-segment firms are, and describe the finding as a ‘diversification discount’ or ‘negative excess value’. A diversification discount implies that the cost of equity capital is higher for multiple-segment firms than for single-segment firms. This phenomenon occurs due to higher IA between managers and shareholders in multiple-segment firms. Thus, investors discount the value of multiple-segment firms to compensate for the increased agency risk.

We use 21,177 firm-year observations from 1987 to 2010 to empirically test this prediction. We use the implied cost of equity capital measures to proxy for the cost of equity capital. The empirical results support our prediction that the association between the Big 4 auditor choice and the decrease in cost of equity capital is more pronounced among multiple-segment firms. As the number of segments increases, the association becomes stronger.

Although the findings in this study clearly reveal the significant effect of Big 4 audits on the cost of capital, the scope of this research still has limitations. First, a firm's diversified decision may possibly be endogenous. We do not formally incorporate this issue; rather, we perform subsample analyses classified by the degree of diversifications (i.e. number of segments) to show the significant effect of high-quality audits in each subsample. Second, although we attempt to solve the endogenous auditor choice issue, some unknown factors, such as governance mechanism, may simultaneously influence both auditor choice and the cost of capital. Third, the measure of implied cost of equity capital in this study can be correlated with unknown factors that, in turn, may also be correlated with auditor choice. Third, this study does not directly observe the effort of auditors (e.g. audit hours), but uses the choice of Big 4 auditors as a proxy for audit quality. Fourth, although our theory highlights the role of IA as a determinant of the cost of capital differential, IA may possibly not be priced in the market (Hughes et al., 2007; Lambert et al., 2007). Instead, the quality of accounting information can directly and indirectly influence the cost of capital (Lambert et al., 2007). We leave these caveats for future research.

1 Big 4 auditors include all larger auditors. In other words, the term refers not only to the current Big 4 auditors, but also to the former Big 5, Big 6 or Big 8 auditors.

2 Note that we specifically focus on the cost of equity capital among other proxies of the diversification discount. In so doing, we aim to demonstrate the effect of Big 4 auditor choices in a more comparable way to most prior studies, including that of Lawrence et al. (2011).

3 This contradicts a recent counterargument by Lawrence et al. (2011). Note that they investigate the existence of superior audit quality between Big 4 and non-Big 4 auditors using various proxies, including cost of capital, but they did not tackle the relation between audit quality and the cost of capital with respect to the differential level of information asymmetry.

4 Furthermore, this study sheds some light on the recent controversy in accounting and finance literature on whether or not information risk is priced (e.g. Easley and O'Hara, 2004; Hughes et al., 2007; Lambert et al., 2007). Rather than directly focusing on the controversy, we provide insights into the controversy by showing the monitoring role of Big 4 audits in settings where information asymmetry is relatively high. We admit, however, that this study is not conclusive regarding information risk pricing. Thus, interpretations of our findings are subject to this caveat.

5 See Bushman and Smith (2001), Fields et al. (2001), Healy and Palepu (2001) and Lambert (2001) for further discussion on the issue.

6 Consistent with lower information quality in multiple-segment firms, Gilson et al. (2001) argue that the complex operations of multiple-segment firms make the tasks of analysts more difficult, leading to less accurate forecasts. Krishnaswami and Subramaniam (1999) report that information asymmetry decreases after the breakup of multiple-segment firms and those firms tend to raise more capital after the break-up.

7 Another possible way to reduce the diversification discount is to increase the disclosure level to provide investors with more information. Bens and Monahan (2004) find that, on average, firms that disclose more information have a lower diversification discount. Berger and Hann (2003) and Ettredge et al. (2005) report that the adoption of the Statement of Financial Standards (SFAS) No. 131 (‘Disclosures about segments of an enterprise and related information’) improves the quality of disclosure for multiple-segment firms. SFAS No. 131 requires firms to disclose segment information based on how management internally evaluates the operating performance of different business units. In contrast, previous accounting standards (SFAS No. 14, ‘Financial reporting for segments of a business enterprise’) requested firms to disclose segment information based on the different industries operated by a firm. After the adoption of SFAS No. 131, firms began to disclose a greater number of segments and more disaggregated information (Berger and Hann, 2003).

8 However, Khurana and Raman (2004) fail to find similar evidence in other countries except the United States. They explain that Big 4 auditors in the United States have incentives to provide high-quality service, whereas Big 4 auditors in other countries do not, possibly due to different levels of legal liabilities that auditors face.

9 Notably, we use an ex ante measure of expected returns (i.e. the implied cost of equity capital) instead of realized returns. Ex post realized returns have long been used as a proxy for expected returns because expected returns are not observable. A stream of research, however, warns that realized returns can be a noisy proxy for expected returns, because realized returns are affected by unexpected cash flow and return news (e.g. Pastor et al., 2008).

10 Based on prior studies (e.g. Hail and Leuz, 2006; Dhaliwal et al., 2007), we compute multiple estimates of the implied cost of equity using four models. The average of the four models is used to obtain a single estimate per firm-year observation. Hail and Leuz (2009) explore alternative ways to aggregate the four proxies and find that these techniques have little impact on their results. To save space, we did not elaborate on the construction of each estimate. See the Appendix of Hail and Leuz (2006) for variable construction.

11 Following Francis et al. (2005), we estimate the measure of accruals quality and disaggregate it into innate and discretionary components. We specifically refer to Equations (1) and (8) of Francis et al. (2005) to construct DiscAQ.

12 The major findings are not sensitive to this definition. Results are similar if we define industry at either the two-digit or three-digit SIC level. We also use the total number of segments to measure the degree of diversification (#Seg) for further analyses.

13 For all CoE estimates exceeding 0.50, we set it equal to 0.50 following Dhaliwal et al. (2007). They exhibit high correlations as expected. For example, the lowest Pearson correlation is 0.73 between Gode and Mohanram's (2003) measure and Gebhardt et al.'s (2001) measure.

14 Consistent with Fan and Wong's (2005) finding that firms with higher ownership divergence are more likely to hire Big 4 auditors to mitigate agency problems, the tendency to hire Big 4 increases with the number of segments, which, we expect, is related to the degree of information asymmetry. About 98.1 per cent of firms with six or more segments hire Big 4 auditors, while only 90.9 per cent of single-segment firms do so.

15 Due to data requirements, our sample firms are generally composed of large firms. We argue that this sample selection bias works against finding a significant association between Big 4 and the cost of equity. In the latter part of this paper, we show that the effect of high-quality audits is indeed more salient among small firms than large ones, because small firms have a relatively greater level of information asymmetry than do large firms.

16 We also run regression Equation 2 below by performing one regression with a full sample. The results are not tabulated for simplicity, because they are qualitatively identical (i.e. the coefficient c₃ is significantly negative).

$urn:x-wiley:08105391:media:acfi12011:acfi12011-math-0002$ (2)

17 A list of references using similar measures includes the following: volatility (Ang et al., 2006), volume (Datar et al., 1998; Lee and Swaminathan, 2000) and forecast dispersion of analysts (Diether et al., 2002). More recently, Jiang et al. (2005) use firm age, return volatility and average daily turnover as proxies for information asymmetry. In this study, we add firm size, analyst following and forecast dispersion.

18 It is also worth noting that we focus our analysis on the partitioned samples when comparing the differences between samples. In so doing, we control for managerial decision for firm segmentation, which naturally makes our analyses less subject to the endogeneity issue between the number of segments and auditor choice. Similarly, Lamont and Polk (2001) tackle this potential selection issue to examine the diversification discount.

19 The null hypothesis of the underidentification test states that instruments are not relevant to the endogenous variable. The null hypothesis of the overidentification test states that instruments are not correlated with error terms. Thus, suitable instruments should reject the null hypothesis of the underidentification test and not reject the null of the overidentification test.

20 We impose a caliper distance of 1 per cent to achieve better similarity among matched pairs. Our results still hold with those obtained by single nearest-neighbor matching.

21 Based on these empirical findings, the US Government Accountability Office (2003) concludes that mandatory auditor rotation that limits the maximum length of auditor tenure is not necessary. However, the Public Company Accounting Oversight Board (PCAOB, 2011) is currently re-examining the issue. Thus, the debate on the role of auditor tenure is not yet settled. We thank an anonymous referee for pointing out this issue.

22 The tenor of the results does not change when we use the quintile rank of auditor tenure or the natural logarithm of auditor tenure instead of the indicator variable.

23 The results are qualitatively similar even when we replicate our analyses after excluding former clients of Arthur Andersen after 2001.

24 Due to the fact that we classify only one auditor who reports the highest audit fees as a specialist in each industry in each year, the number of observations with a value of one for our Specialist variable equals the number of industry year pairs. The trend of the results does not change if we count two auditors in terms of audit fee revenue as industry specialists. However, we cannot rule out the possibility that the specialist measure merely captures client characteristics such as client size. We thank an anonymous referee for pointing this out.

25 We also use alternative measures of industry specialization, such as an auditor industry share based on client sales or assets (Balsam et al., 2003; Knechel et al., 2007). Although we find directionally consistent results, the difference in the Specialist coefficients between multiple- and single-segment firms is not statistically significant.

References

Aggarwal, R. K., and A. A. Samwick, 2003, Why do managers diversify their firms? Agency reconsidered, Journal of Finance 58, 71–118.
10.1111/1540-6261.00519
Web of Science® Google Scholar
Ang, A., R. Hodrick, Y. Xing, and X. Zhang, 2006, The cross-section of volatility and expected returns, Journal of Finance 61, 259–299.
10.1111/j.1540-6261.2006.00836.x
Web of Science® Google Scholar
Ahmed, A., S. Rasmussen, and S. Tse, 2008, Audit Quality, Alternative Monitoring Mechanisms, and Cost of Capital: An Empirical Analysis. Available at: http://ssrn.com/?abstract_id=1124082 (accessed August 2008).
Google Scholar
Armstrong, C., A. Jagolinzer, and D. Larcker, 2010, Chief executive officer equity incentives and accounting irregularities, Journal of Accounting Research 48, 225–271.
10.1111/j.1475-679X.2009.00361.x
Web of Science® Google Scholar
Balsam, S., J. Krishnan, and J. Yang, 2003, Auditor industry specialization and earnings quality, Auditing: A Journal of Practice and Theory 22, 71–97.
10.2308/aud.2003.22.2.71
Web of Science® Google Scholar
Behn, B., J.-H. Choi, and T. Kang, 2008, Audit quality and properties of analysts' earnings forecasts, The Accounting Review 83, 327–349.
10.2308/accr.2008.83.2.327
Web of Science® Google Scholar
Bens, D., and S. Monahan, 2004, Disclosure quality and the excess value of diversification, Journal of Accounting Research 42, 691–730.
10.1111/j.1475-679X.2004.00154.x
Web of Science® Google Scholar
Berger, P., and R. Hann, 2003, The impact of SAFS No. 131 on information and monitoring, Journal of Accounting Research 41, 163–223.
10.1111/1475-679X.00100
Web of Science® Google Scholar
Berger, P., and E. Ofek, 1995, Diversifications effect on firm value, Journal of Financial Economics 37, 39–65.
10.1016/0304-405X(94)00798-6
PubMed Web of Science® Google Scholar
Bhattacharya, U., H. DaoukK, and M. Welker, 2003, The world pricing of earnings opacity, The Accounting Review 78, 641–678.
10.2308/accr.2003.78.3.641
Web of Science® Google Scholar
Brown, S., and S. Hillegeist, 2007, How disclosure quality affects the level of information asymmetry, Review of Accounting Studies 12, 443–477.
10.1007/s11142-007-9032-5
Web of Science® Google Scholar
Bushman, R. M., and A. J. Smith, 2001, Financial accounting information and corporate governance, Journal of Accounting and Economics 32, 237–333.
10.1016/S0165-4101(01)00027-1
Web of Science® Google Scholar
Chaney, P., D. J. Jeter, and L. Shivakumar, 2004, Self-selection of auditors and audit pricing in private firms, The Accounting Review 79, 51–72.
10.2308/accr.2004.79.1.51
Web of Science® Google Scholar
Chen, H., J. Z. Chen, G. J. Lobo, and Y. Wang, 2011, Effects of audit quality on earnings management and cost of equity capital: evidence from China, Contemporary Accounting Research 28, 892–925.
10.1111/j.1911-3846.2011.01088.x
Web of Science® Google Scholar
Choi, J.–H., and T. J. Wong, 2007, Auditors governance functions and legal environments: An international investigation, Contemporary Accounting Research 24, 13–46.
10.1506/X478-1075-4PW5-1501
Web of Science® Google Scholar
Claus, J., and J. Thomas, 2001, Equity premia as low as three percent? Empirical evidence from analysts' earnings forecasts for domestic and international stock markets, Journal of Finance 56, 1629–1665.
10.1111/0022-1082.00384
Web of Science® Google Scholar
Core, J., 2010, Discussion of chief executive officer equity incentives and accounting irregularities, Journal of Accounting Research 48, 271–287.
10.1111/j.1475-679X.2010.00367.x
Web of Science® Google Scholar
Datar, V., N. Naik, and R. Radcliffe, 1998, Liquidity and asset returns: an alternative test, Journal of Financial Markets 1, 203–220.
10.1016/S1386-4181(97)00004-9
Google Scholar
Demirkan, S., S. Radhakrishnan, and O. Urcan, 2012, Discretionary accruals quality, cost of capital, and diversification, Journal of Accounting, Auditing and Finance 27, 496–526.
10.1177/0148558X11409162
Google Scholar
Denis, D. J., D. K. Denis, and A. Sarin, 1997, Agency problems, equity ownership, and corporate diversification, Journal of Finance 52, 135–160.
10.1111/j.1540-6261.1997.tb03811.x
Web of Science® Google Scholar
Dhaliwal, D., L. Krull, and O. Z. Li, 2007, Did the 2003 Tax Act reduce the cost of equity capital?, Journal of Accounting and Economics 43, 121–150.
10.1016/j.jacceco.2006.07.001
Web of Science® Google Scholar
Diether, K., C. Malloy, and A. Scherbina, 2002, Differences of opinion and the cross–section of stock returns, Journal of Finance 57, 2113–2141.
10.1111/0022-1082.00490
Web of Science® Google Scholar
Easley, D., and M. O'Hara, 2004, Information and the cost of capital, Journal of Finance 59, 1552–1583.
10.1111/j.1540-6261.2004.00672.x
Web of Science® Google Scholar
Easton, P., 2004, PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital, The Accounting Review 79, 73–95.
10.2308/accr.2004.79.1.73
Web of Science® Google Scholar
Easton, P., and G. Sommers, 2007, Effects of analysts' optimism on estimates of the expected rate of return implied by earnings forecasts, Journal of Accounting Research 45, 983–1015.
10.1111/j.1475-679X.2007.00257.x
Web of Science® Google Scholar
Ettredge, M., S. Kwon, D. Smith, and P. Zarowin, 2005, The impact of SFAS No. 131 business segment data on the market's ability to anticipate future earnings, The Accounting Review 80, 773–804.
10.2308/accr.2005.80.3.773
Web of Science® Google Scholar
Fama, E. F., and K. R. French, 1996, Multifactor explanations of asset pricing anomalies, Journal of Finance 51, 55–84.
10.1111/j.1540-6261.1996.tb05202.x
Web of Science® Google Scholar
Fama, E. F., and K. R. French, 1997, Industry costs of equity, Journal of Financial Economics 43, 153–193.
10.1016/S0304-405X(96)00896-3
Web of Science® Google Scholar
Fan, J. P. H., and T. J. Wong, 2005, Do external auditors perform a corporate governance role in emerging markets? Evidence from East Asia, Journal of Accounting Research 43, 1–38.
10.1111/j.1475-679x.2004.00162.x
Web of Science® Google Scholar
Fields, T., T. Lys, and V. Vincent, 2001, Empirical research on accounting choice, Journal of Accounting and Economics 31, 255–307.
10.1016/S0165-4101(01)00028-3
Web of Science® Google Scholar
Francis, J., L. E. Maydew, and H. C. Sparks, 1999, The role of Big 6 auditors in the credible reporting of accruals, Auditing: A Journal of Practice and Theory 18, 17–34.
10.2308/aud.1999.18.2.17
Web of Science® Google Scholar
Francis, J., R. LaFond, P. Olsson, and K. Schipper, 2004, Costs of equity and earnings attributes, The Accounting Review 79, 967–1010.
10.2308/accr.2004.79.4.967
Web of Science® Google Scholar
Francis, J., R. LaFond, P. Olsson, and K. Schipper, 2005, The market pricing of accruals quality, Journal of Accounting and Economics 39, 295–327.
10.1016/j.jacceco.2004.06.003
Web of Science® Google Scholar
Gebhardt, W., C. Lee, and B. Swaminathan, 2001, Toward an implied cost of capital, Journal of Accounting Research 39, 135–176.
10.1111/1475-679X.00007
Web of Science® Google Scholar
Geiger, M., and K. Raghunandan, 2002, Auditor tenure and audit reporting failures. Auditing: A Journal of Practice and Theory 21, 67–78.
10.2308/aud.2002.21.1.67
Web of Science® Google Scholar
Ghosh, A., and D. Moon, 2005, Auditor tenure and perceptions of audit quality, The Accounting Review 80, 585–612.
10.2308/accr.2005.80.2.585
Web of Science® Google Scholar
Gilson, S., P. Healy, C. Noe, and K. Palepu, 2001, Analyst specialization and conglomerate stock breakups, Journal of Accounting Research 39, 565–582.
10.1111/1475-679X.00028
Web of Science® Google Scholar
Gode, D., and P. Mohanram, 2003, Inferring the cost of capital using the Ohlson-Juettner model, Review of Accounting Studies 8, 399–431.
10.1023/A:1027378728141
Web of Science® Google Scholar
Gomes, J., and D. Livdan, 2004, Optimal diversification: Reconciling theory and evidence, Journal of Finance 59, 507–535.
10.1111/j.1540-6261.2004.00641.x
Web of Science® Google Scholar
Graham, J. R., M. L. Lemmon, and J. G. Wolf, 2002, Does corporate diversification destroy value?, Journal of Finance 57, 695–720.
10.1111/1540-6261.00439
Web of Science® Google Scholar
Government Accountability Office (GAO), 2003. Public accounting firms: Required study on the potential effects of mandatory audit firm rotation. Available at: http://www.gao.gov/new.items/d04216.pdf (accessed November 2003).
Google Scholar
Hail, L., and C. Leuz, 2006, International differences in the cost of equity capital: Do legal institutions and securities regulation matter?, Journal of Accounting Research 44, 485–531.
10.1111/j.1475-679X.2006.00209.x
Web of Science® Google Scholar
Hail, L., and C. Leuz, 2009, Cost of capital effects and changes in growth expectations around U.S. cross–listings, Journal of Financial Economics 93, 428–454.
10.1016/j.jfineco.2008.09.006
Web of Science® Google Scholar
Healy, P., and K. Palepu, 2001, Information asymmetry, corporate disclosure and the capital markets: A review of the empirical disclosure literature, Journal of Accounting and Economics 31, 405–440.
10.1016/S0165-4101(01)00018-0
Web of Science® Google Scholar
Heckman, J., 1979, Sample selection bias as a specification error, Econometrica 47, 153–161.
10.2307/1912352
Web of Science® Google Scholar
Hughes, J., J. Liu, and J. Liu, 2007, Information asymmetry, diversification, and cost of capital, The Accounting Review 82, 705–729.
10.2308/accr.2007.82.3.705
Web of Science® Google Scholar
Jiang, G., C. M. C. Lee, and Y. Zhang, 2005, Information uncertainty and expected returns, Review of Accounting Studies 10, 185–221.
10.1007/s11142-005-1528-2
Web of Science® Google Scholar
Johnson, E., I. Khurana, and J. Reynolds, 2002, Audit firm tenure and the quality of financial reports, Contemporary Accounting Research 19, 637–660.
10.1506/LLTH-JXQV-8CEW-8MXD
Web of Science® Google Scholar
Khurana, I., and K. Raman, 2004, Litigation risk and the financial reporting credibility of Big 4 versus non-Big 4 audits: Evidence from Anglo–American Countries, The Accounting Review 79, 473–495.
10.2308/accr.2004.79.2.473
Web of Science® Google Scholar
Kim, J. B., D. A. Simunic, M. T. Stein, and C. H. Yi, 2011, Voluntary audit and the cost of debt capital for privately held firms: Korean evidence, Contemporary Accounting Research 28, 585–615.
10.1111/j.1911-3846.2010.01054.x
Web of Science® Google Scholar
Kleibergen, F., and R. Paap, 2006, Generalized reduced rank tests using the singular value decomposition, Journal of Econometrics 133, 97–126.
10.1016/j.jeconom.2005.02.011
Web of Science® Google Scholar
Knechel, W. R., V. Naiker, and G. Pacheco, 2007, Does auditor industry specialization matter? Evidence from market reaction to auditor switches, Auditing: A Journal of Practice and Theory 26, 19–45.
10.2308/aud.2007.26.1.19
Web of Science® Google Scholar
Krishnan, G., 2003a, Audit quality and the pricing of discretionary accruals, Auditing: A Journal of Practice and Theory 22, 109–126.
10.2308/aud.2003.22.1.109
Web of Science® Google Scholar
Krishnan, G., 2003b, Does Big 6 auditor industry expertise constrain earnings management?, Accounting Horizon 17(Suppl), 1–16.
10.2308/acch.2003.17.s-1.1
Google Scholar
Krishnaswami, S., and V. Subramaniam, 1999, Information asymmetry, valuation and the corporate spin–off decision, Journal of Financial Economics 53, 73–112.
10.1016/S0304-405X(99)00017-3
Web of Science® Google Scholar
Lambert, R., 2001, Contracting theory and accounting, Journal of Accounting and Economics 32, 3–87.
10.1016/S0165-4101(01)00037-4
Web of Science® Google Scholar
Lambert, R., C. Leuz, and R. Verrecchia, 2007, Accounting information, disclosure, and the cost of capital, Journal of Accounting Research 45, 385–420.
10.1111/j.1475-679X.2007.00238.x
Web of Science® Google Scholar
Lamont, O., 1997, Cash flow and investment: Evidence from internal capital markets, Journal of Finance 52, 83–109.
10.1111/j.1540-6261.1997.tb03809.x
Web of Science® Google Scholar
Lamont, O., and C. Polk, 2001, The diversification discount: Cash flows versus returns, Journal of Finance 56, 1693–1721.
10.1111/0022-1082.00386
Web of Science® Google Scholar
Lang, L., and R. Stulz, 1994, Tobin's q, corporate diversification and firm performance, Journal of Political Economy 102, 1248–1280.
10.1086/261970
Web of Science® Google Scholar
Larcker, D., and T. Rusticus, 2010, On the use of instrumental variables in accounting research, Journal of Accounting and Economics 49, 186–205.
10.1016/j.jacceco.2009.11.004
Web of Science® Google Scholar
Lawrence, A., M. Minutti-Meza, and P. Zhang, 2011, Can Big 4 versus non–Big 4 differences in audit-quality proxies be attributed to client characteristics?, The Accounting Review 86, 259–286.
10.2308/accr.00000009
Web of Science® Google Scholar
Lee, C. M. C., and B. Swaminathan, 2000, Price momentum and trading volume, Journal of Finance 55, 2017–2069.
10.1111/0022-1082.00280
Web of Science® Google Scholar
Maksimovic, V., and G. Phillips, 2002, Do conglomerate firms allocate resources inefficiently across industries? Theory and evidence, Journal of Finance 57, 721–767.
10.1111/1540-6261.00440
Web of Science® Google Scholar
Mansi, S., W. Maxwell, and D. Miller, 2004, Does auditor quality and tenure matter to investors? Evidence from the bond market, Journal of Accounting Research 42, 755–793.
10.1111/j.1475-679X.2004.00156.x
Web of Science® Google Scholar
Martin, J., and A. Sayrak, 2003, Corporate diversification and shareholder value: A survey of recent literature, Journal of Corporate Finance 9, 37–57.
10.1016/S0929-1199(01)00053-0
Web of Science® Google Scholar
Meyer, M., P. Milgrom, and J. Roberts, 1992, Organizational prospects, influence costs and ownership changes, Journal of Economics and Management Strategy 1, 9–35.
10.1111/j.1430-9134.1992.00009.x
Web of Science® Google Scholar
Mitton, T., and K. Vorkink, 2010, Why do firms with diversification discounts have higher expected returns?, Journal of Financial and Quantitative Analysis 45, 1367–1390.
10.1017/S0022109010000566
Web of Science® Google Scholar
Morck, R., A. Shleifer, and R. W. Vishny, 1990, Do managerial objective drive bad acquisitions?, Journal of Finance 45, 31–48.
10.1111/j.1540-6261.1990.tb05079.x
Web of Science® Google Scholar
Myers, J., L. Myers, and T. Omer, 2003, Exploring the term of auditor–client relationship and the quality of earnings: A case for mandatory auditor rotation?, The Accounting Review 78, 779–800.
10.2308/accr.2003.78.3.779
Web of Science® Google Scholar
Ozbas, O., and D. Scharfstein, 2010, Evidence on the dark side of internal capital markets, Review of Financial Studies 23, 581–599.
10.1093/rfs/hhp071
Web of Science® Google Scholar
Pastor, L., M. Sinha, and B. Swaminathan, 2008, Estimating the intertemporal risk–return tradeoff using the implied cost of capital, The Journal of Finance 63, 2859–2897.
10.1111/j.1540-6261.2008.01415.x
Web of Science® Google Scholar
Pittman, J., and S. Fortin, 2004, Auditor choice and the cost of debt capital for newly public firms, Journal of Accounting and Economics 37, 113–136.
10.1016/j.jacceco.2003.06.005
Web of Science® Google Scholar
Public Company Accounting Oversight Board (PCAOB), 2011. Concept Release on Auditor Independence and Audit Firm Rotation. PCAOB Release No. 2011-006. Available at: http://pcaobus.org/rules/rulemaking/docket037/release_2011-006.pdf (accessed August 16, 2011).
Google Scholar
Rajan, R., H. Servaes, and L. Zingales, 2000, The cost of diversity: The diversification discount and inefficient investment, Journal of Finance 55, 35–80.
10.1111/0022-1082.00200
Web of Science® Google Scholar
Staiger, D., and J. Stock, 1997, Instrumental variables regression with weak instruments, Econometrica 65, 557–586.
10.2307/2171753
Web of Science® Google Scholar
Stulz, R., 1990, Managerial discretion and optimal financial policies, Journal of Financial Economics 26, 3–27.
10.1016/0304-405X(90)90011-N
Web of Science® Google Scholar
Teoh, S. H., and T. J. Wong, 1993, Perceived auditor quality and the earnings response coefficient, The Accounting Review 68, 346–366.
Web of Science® Google Scholar
Willenborg, M., 1999, Empirical analysis of the economic demand for auditing in the initial public offerings market, Journal of Accounting Research 37, 225–239.
10.2307/2491405
Web of Science® Google Scholar

Citing Literature

Volume54, Issue1

March 2014

Pages 135-163

Association between Big 4 auditor choice and cost of equity capital for multiple-segment firms

Abstract

1. Introduction

2. Theory and hypotheses development

3. Methodology