Auditee Religiosity, External Monitoring, and the Pricing of Audit Services

We measure auditee intrinsic religiosity as the religious adherents’ percentage of the county’s population where auditee is headquartered and we measure auditee extrinsic religiosity using number of religious institutions in the county where auditee is headquartered. As can be seen in Table 1 Panel C, the correlation between auditee intrinsic and extrinsic religiosity is low, with a Pearson (Spearman) correlation of 0.11(0.07).

Jha and Chen (2015) use two measures of norm and two measures of networks to capture social capital. These two measures of norms are (1) voter turnout in presidential elections and (2) the census response rate. The two measures of networks are (1) the number of social and civic associations and (2) the number of nongovernment organization (NGOs) in counties. Social and civic associations include physical and fitness facilities, public golf courses, religious organizations, sports clubs, managers and promoters, political organizations, professional organizations, business associations, and labor organizations in the county.

The social capital index in Jha and Chen (2015) includes both auditee extrinsic religiosity and other elements of social capital. The current study tries to investigate the effects of these elements separately and include the natural logarithm of number of religious institutions in the county as a proxy for extrinsic religiosity. However, social capital index still includes other elements that may not be related to religiosity.

Inherent risk is related to the likelihood that environmental factors, before considering the quality of internal controls, will results in material errors. Control risk is related to the likelihood that the internal controls will not prevent or detect material error and detection risk is the likelihood that the audit procedures will fail to detect a materials error.

Jha and Chen (2015) use two measures of networks are the number of social and civic associations and the number of NGO in counties. Social and civic associations include physical fitness facilities, public golf courses, religious organizations, sports clubs, managers and promoters, political organizations, professional organizations, business associations, and labor organizations in the county, but they exclude NGOs with an international focus. Both of these measures are normalized by the population in the county. They construct a social capital index using principal component analysis and use the first component to measure level of social capital in different counties.

See Sect. 5.3 for details.

Correlation between auditee and auditor religiosity in current study is 0.65 (significant at 1 % level) and the correlation would be even higher when county religiosities are used as proxies for auditor religiosity.

ARDA has changed its name from American Religion Data Archive to Association of Religion Data Archive in 2006.We collect data on county and MSA religiosity from the U.S. Religion Census: Religious Congregations and Membership Study: County and Metro Area Files included in ARDA. This study designed and carried out by the Association of Statisticians of American Religious Bodies (ASARB), compiled data on the number of congregations and adherents for 236 religious groups in each county of the United States. Participants included 217 Christian denominations, associations, or communions (including Latter-day Saints, Messianic Jews, and Unitarian/Universalist groups); counts of Jain, Shinto, Sikh, Tao, and National Spiritualist Association congregations, and counts of congregations and adherents from Bahá'ís, three Buddhist groupings, four Hindu groupings, four Jewish groupings, Muslims, and Zoroastrians. The 236 groups reported a total of 344,894 congregations with 150,686,156 adherents, comprising 48.8 percent of the total U.S. population of 308,745,538 in 2010.

Gallup organization conducted interview of over 610,000 participants and it utilizes responses to three questions as measures of religiosity. However, since a methodology may introduce selection bias and we therefore use data provided by ARDA as our main proxy for religiosity.

Although mean LAF for high auditee religiosity subsample (13.548) is lower than that for low auditee religiosity subsample (13.561), medium auditee religiosity subsample has the lowest LAF (13.539). However, the differences between LAFs for different subsamples are not significant.

We would like to thank the reviewer for suggesting the inclusion of separate extrinsic religiosity variable.

We try to use the audit fee model in Jha and Chen (2015) to investigate the impact of auditee religiosity on audit fees. However, we are unable to obtain two variables—DIST FROM SEC and COST OF LIVING. Although two variables are missing, we believe that the effects on our results are insignificant, as COST OF LIVING is not significant in Jha and Chen (2015) Table 1.

We would like to thank the reviewer for suggesting the methodology.

We address the problem of multi-collinearity between AUDITEE_REIG and AUDITOR_RELIG in Sect. 6.

Coefficients for BIG4 and DLOSS are not significant in Table 1 Panel B of Jha and Chen (2015). However, prior research has suggested that these variables should be positively related to audit fees.

Following Jha and Chen (2015), we construct indicator variables that capture whether the firm is headquartered in the West (AZ, AK, CA, CO, HI, ID, MT, NM, NV, OR, UT, WA, WY), the Northeast (CT, MA, ME, NH, NJ, NY, PA, RI, VT), the South (AL, AR, DC, DE, FL, GA, KY, LA, MD, MS, NC, OK, SC, TN, TX, VA, WV), or the Midwest (IA, IL, IN, KS, MI, MN, MO, ND, NE, OH, SD, WI), based on the U.S. census classifications.

In the case of leverage, we have performed regression (1) on two subsamples split of leverage of year t − 1. Since leverage level of year t and t − 1 are likely to be correlated and we have also reestimated the regressions by excluding leverage at year t. Results are qualitatively similar.

Since there are significant number of missing variables for institutional ownership, the overall sample sizes are significantly smaller for tests including institutional ownership.

We also conduct analyses by applying the regression to top and bottom quartiles, results are qualitatively.

First-stage regression is as follows: \({\text{INSTOWN}} = \beta_{0} + \beta_{1} {\text{LNMVE}} + \beta_{2} {\text{VOL}} + \beta_{3} {\text{TURNOVER}} + \beta_{4} {\text{LOGPRICE}} + \beta_{5} {\text{RETURN}} + \beta_{6} {\text{BASPREAD}} + \beta_{7} {\text{AGE}} + \beta_{8} {\text{TOBINSQ}} + \beta_{9} {\text{TANG}} + \beta_{10} {\text{LEV}} + \beta_{11} {\text{DYIELD}} + \beta_{12} {\text{ROA}} + \beta_{13} {\text{SP}} + \varepsilon\) variable definitions are summarized in Table 1 Panel B.

First-stage regression is as follows: \({\text{LEV}} = \beta_{0} + \beta_{1} {\text{MVBV}} + \beta_{2} {\text{LOGSALES}} + \beta_{3} {\text{ROA}} + \beta_{4} {\text{TANGIBILITY}} + \beta_{5} {\text{TAXRATE}} + \beta_{6} {\text{NOL}} + \varepsilon\) variable definitions are summarized in Table 1 Panel C.

Regression is as follows: \({\text{AUDITOR\_RELIG}} = \mathop {1.855}\limits_{(T = 63.47)} + \mathop {0.65}\limits_{(T = 138.90)} {\text{AUDITEE\_RELIG}} + \varepsilon\).