Financial capacity and the demand for audit quality

Abstract

Prior research documents that financial capacity could be positively or negatively associated with the demand for audit quality. We re-examine this relation using changes in local real estate prices as exogenous shocks to corporate financial capacity. Using auditor size, auditor industry specialisation, and auditor fees as measures of audit quality, we find robust evidence that an increase (decrease) in financial capacity significantly reduces (increases) the demand for audit quality, and that this relation is more pronounced when firms are more financially constrained, when external monitoring by institutional investors and financial analysts is weaker, and when there is more negative news about real estate price changes. Our study enriches the related literature by describing a more complete and dynamic relationship between audit quality and financial capacity.

Keywords:

• audit quality
• financial capacity
• real estate
• financial constraint

JEL classification:

• G12
• M41
• M42

Acknowledgement

We thank workshop participants at the Central University of Finance and Economics, Jinan University and Singapore Management University. Lim and Yue acknowledge financial support from the School of Accountancy Research Center (SOAR) at Singapore Management University. Pingui Rao acknowledges financial support from the National Natural Science Foundation of China (grant number 71872071 and 71728006) and the Fundamental Research Funds of the Central Universities in China (grant number 19JNLH08).

Data availability

Data used in this study are available from public sources identified in the study.

Disclosure statement

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

Notes

1 This is labelled as the ‘collateral channel’ of real estate prices in Chaney et al. (2012).

2 For example, corporate governance may simultaneously affect both financial capacity and audit quality.

3 Two recent studies, Fung et al. (2015) and Gunn et al. (2017) use brokerage house closures/mergers as a natural experiment to identify exogenous increases in information asymmetry. Their results show that firms’ demand for audit quality increases after they experience exogenous reductions in analyst coverage.

4 For example, to check the validity of this approximation, Chaney et al. (2012) manually collect information on the ownership status of a firm’s headquarters from the 10K forms filed with the SEC for the year 1997. They find that only 2% of the firms (out of 1,578 firms) that report owning no real estate assets in Compustat, report owning their headquarters in their 10K forms. On the other hand, of the 1,815 firms that report owning some real estate assets in Compustat, 44% actually report owning their headquarters in their 10K forms. This under-representation of real assets in Compustat implies that the assumption that real estate assets are located in a firm’s headquarters is conservative and tends to underestimate the impact of a firm’s collateral on its investment.

5 Using a Propensity Score Matching (PSM) research design to control for self-selection, Lawrence et al. (2011) do not find a significant difference in quality between Big N and non-Big N auditors. However, DeFond et al. (2016) report that Lawrence et al.’s results are sensitive to the research design choices inherent in PSM. Using an alternate matching procedure, DeFond et al. (2016) find that Big N auditors have higher audit quality.

6 A large portion (about 35%) of our sample firms do not have RE values. We follow Chaney et al. (2012) closely to calculate real estate values, but as in Chaney et al. (2012), there are many missing data to estimate the real asset values. Chaney et al. (2012) start with an initial sample size of 50,858 firm-year observations, but are left with only 27,543 observations with real estate values in their sample. The loss in sample observations is about 45% in their study.

7 Our proxies of audit quality are measured at one-year after the changes in financial capacity. We assume that firms are able to adjust their demand for audit quality one year after the change in financial capacity. This assumption appears reasonable because the firms could appoint a new auditor or retain the existing auditor at the annual general meeting, which typically occurs at the end of the year. We also repeat the analysis after relaxing this assumption by measuring audit quality at year t+2 instead. Our main results are robust when audit quality is measured over this longer horizon of two years.

8 Other continuous variables, except for State and MSA price index are winsorized at 1%. Our results are similar if we also winsorize State and MSA price index at 1%. This winsorization is needed to prevent extreme values from having an undue influence on the results. Our results are not sensitive to the cut-off used (whether 5% or 1%) in the winsorizing process. We also use Robust Regression to estimate the coefficients of untransformed RE Value following the suggestion in Leone et al. (2019). Our results are generally consistent with this alternative specification.

9 The other comparative statics are computed analogously. When audit quality is proxied by Big N membership, a one standard deviation decrease in RE_Value (State) and RE_Value (MSA) is associated with a 3.19% and a 2.48% increase in BIGN, respectively. When audit quality is proxied by auditor industry specialisation, a one standard deviation decrease in RE_Value (State) and RE_Value (MSA) is associated with a 1.40% and a 1.34% increase in MSHARE, respectively.

10 Although Big N auditors is one of our dependent variables, it may be positively associated with auditor industry specialisation and audit fees. Accordingly, as a robustness check, we include an additional control, BIGN, in models (3) to (6). Our untabulated results indicate that our main inferences remained unchanged after the inclusion of BIGN.

11 RE Value equals zero for 9,959 observations in our sample.

12 There are 7,140 observations with auditor changes (change from BigN to Non-BigN auditor or change from Non-BigN to BigN auditor) in our sample.

13 We obtain audit committee data from the table of Directors and Directors Legacy of the Institutional Shareholder Services (ISS) database. Based on this data, we construct two variables: AC_Size is the proportion of audit committee member to total board members. AC_Indep is the proportion of independent audit committee members to the total number of audit committee members.

14 We obtain CEO shareholdings from Thomson Reuters database. CEO_Holding is the proportion of shares held by the CEO to the total outstanding shares.

15 There are many missing data for audit committee variables because the original database only covers S&P 1500 firms. Due to these missing values, we use the modified zero-order regressions suggested by Greene (2003). This method has fewer assumptions about missing values and substitutes a zero for missing values and adds the three indicator variables DAC_Indep, DAC_Size, and DCEO_Holding, which are coded as one if the corresponding variable is missing, to the regression.

16 We also control for internal control weakness because previous studies report that firms with material internal control weaknesses are generally associated with poor financial reporting quality (Doyle et al. 2007, Ashbaugh-Skaife et al. 2008). The untabulated results remain robust to controlling for internal control weaknesses, except in the model where audit quality is measured by AFEE and when real estate values are measured at MSA. In addition, we remove clients audited by Arthur Andersen because our results could be confounded by the demise of Andersen in 2002, which is likely to have undue influence on earnings quality and audit quality demand (Krishnan and Visvanatham 2008, Cahan and Zhang 2006). The untabulated results indicate that the removal of Andersen clients from the sample does not change our inferences.

17 We obtained similar results when we define a firm to be financially constrained if its WW Index (HP Index) falls in the top quartile of the whole distribution, and unconstrained otherwise.

18 The variable HTECH is omitted in the table due to collinearity.

19 Boone and White (2015) argue that the diverse holdings of quasi-indexers make gathering private information on their portfolio firms more costly, leading to greater demand for firm transparency and enhanced public information production to minimise these costs. On the other hand, dedicated investors have less influence on public information production because they likely rely more on private information. Although we use quasi-indexers to measure institutional ownership in our main tests, we also repeat the analyses with total institutional ownership as well as dedicated institutional ownership as alternative measures. Our results are similar with these alternative measures.

20 Note RE Value is still negative and significant, suggesting that positive shocks also have effects.

Additional information

Funding

This work was supported by National Natural Science Foundation of China: [Grant Number 71728006,71872071].