The effects of non-audit fees on audit fees during the financial crisis

This study aims to investigate the impact of non-audit fees, auditor tenure, and auditor size on audit fees during the 2008/2009 global financial crisis. The analysis covers the impact on companies in Germany, France, Italy and Spain, as these countries are the four largest in the Eurozone. They were affected at distinct levels by the financial crisis, and this might provide important and contrasting results for analysis. The study starts in 2005, when Eurozone countries were submitted to similar accounting frameworks. This study contributes to the literature by taking into account two indicators of crisis and using several panel models to capture the presence of the financial crisis. The two indicators are the change in gross domestic product (GDP) and a year-based effect. The main indicator considers positive change in GDP, which characterizes the pre-crisis period, and negative change in GDP as characteristic of the in-crisis period. The robust indicator and the various panel models consider the pre-crisis period to be 2005 to 2008 and the in-crisis period, 2009 to 2013.

The current study contributes to the literature by using a GMM (generalized method of moments) approach to examine the impact of non-audit fees, auditor tenure, and auditor size on audit fees in the four largest Eurozone countries, before and during the recent financial crisis. In addition, it contributes by using several panel models (ordinary least squares (OLS), GMM, and fixed and random effects) to examine the impact of non-audit fees, auditor tenure, and auditor size on audit fees in the pre-crisis and in-crisis periods in the four countries collectively. Moreover, its novelty is that it considers the parallel movement of information and the transmission of information between markets, for audit and non-audit fees in the two periods, using a panel co-integration and a panel bivariate generalized autoregressive conditional heteroscedasticity (GARCH) model for the four countries.

The remainder of this paper is organized as follows: Section 2 provides a literature review and develops the hypotheses to be tested; Section 3 provides information on the dataset; Section 4 depicts the methodology; Section 5 analyses the empirical findings of the study; finally, Section 6 concludes and suggests some policy implications.

Salehi, Jafarzadeh, and Nourbakhshhosseiny (2017), applying logistic regression, found that there is no significant association – either before or after 2010 – between pressure on audit fees and audit quality for a sample of 104 firms listed on the Tehran stock exchange. Alharasis and Mustafa (2024) asserted that non-family-owned firms take extra steps to prevent the agency problem, using OLS regression to study 200 family and non-family-owned firms in Jordan from 2005 to 2020. Khaksar, Salehi, and Lari DashtBayaz (2022a), using a multiple regression, found that political connections are negatively associated with the growth of narcissism and overconfidence among auditors on the Tehran stock exchange, in a sample of 768 year-firm observations during the period 2013–2018.

Moradi, Salehi, and Mahdi Balah (2024) provided evidence that the personality traits of Iraqi auditors affect their professional audit judgment; they have conducted regression factor analysis for a sample of 309 participants. Lari Dashtbayaz, Salehi, and Mozan (2023) suggested that Covid-19 affected job satisfaction in Iran without influencing organizational commitment, using 305 questionnaires from a total of 1,500 responses and applying Cochran’s formula.

Wahab, Bagherpour Velashani, and Salehi (2023) found significant and non-significant results, using panel data regressions, for the relationship between audit quality and earnings management for 28 Iraqi banks and 8 Oman banks listed on the stock exchanges of those countries during the period 2005–2021. Rasaei and Shiri (2020), using panel regression, found that the financial crisis contributed to the relationship between management ability and audit fees for 91 firms listed on the Tehran stock exchange, between 2014 and 2019.

Lari Dashtbayaz, Salehi, and Mousavi Shiri (2025) found a negative and significant relationship between illogical expectations and audit report readability when measured with the Fog index, but no association when readability was measured with the Flesch index; they used a sample of 128 firms listed on the Tehran stock exchange between 2014 and 2020. Mottinger (2024) suggested that corruption risk led auditors to increase their audit fees and audit effort across the EU27 countries for a sample of 29,607 firm-year observations spanning the period from 2011 to 2021, using panel regressions.

Khaksar, Salehi, and Lari Dashtbayaz (2022b) found that firms on the Tehran stock exchange that had strong political relations might increase audit market competition and lower audit quality; they studied a sample of 768 year-firm observations during the period 2013–2018, using a multivariate regression model. Van Limpt and Dekeyser (2024) provided evidence of the impact of non-audit services on auditor independence for a sample of public interest entities across the EU from 2013 to 2018 using panel regression.

Tanyi and Cathey (2024) found that auditors responded more to bank risks in the period after the financial crisis than in the period before it, for a sample of banks in the USA from 2003 to 2016, excluding the period 2008 to 2010, using the OLS specification. Chung and Cho (2020) found that audit fees decreased in the crisis period of 2008–2009 from the pre-crisis period of 2006–2007, considering 3,688 Korean public firm-year observations from 2006 to 2009, using the method of regression.

Batrancea, Batrancea, and Moscviciov (2019) indicated that the authorities needed to choose better practices for tax collection for a sample covering 44 nations from five continents, using multilevel regression analysis. In a later study, Batrancea, Nichita, et al. (2022) found that the authorities in post-communist countries achieved higher tax compliance and lower tax evasion than those in non-post-communist countries, using a sample of 2,786 self-employed taxpayers from eleven post-communist and non-post-communist countries that were doing business in five economic sectors. Inegbedion and Okoye-uzu (2024), using structural equation modelling, found that desk and field tax audits had a positive and significant impact on tax compliance for a sample of 205 owners and managers of SMEs in Nigeria.

Hazaea, Al-Matari, Omer, Farhan, and Zhu (2024) did not find any significant relationship between internal audits and commercial banks’ performance in five Arab countries: Saudi Arabia, Yemen, Sudan, Somalia, and Djibouti. They studied 34 commercial banks and used structural equation modelling. Moscviciov, Batrancea, Batrancea, and Batrancea (2010) made judgments about the past and future financial performance of companies listed on the New York Stock Exchange in the technology, software and computer service sectors, based on the Du Pont System, analyzing the return on investment, liquidity, profitability, activity and financial leverage from 2004 to 2008.

Batrancea (2021a) showed that short-term and long-term financial equilibrium, measured by current ratio, quick ratio, and debt to equity ratio, were significantly influenced by specific financial performance ratios over a decade (2007Q1 to 2020Q3), using panel methodology for 34 major companies trading on the New York Stock Exchange. Batrancea, Nichita, Balci, and Akgüller (2023) found a strong influence of wellbeing-related infrastructure on economic growth across 151 Western Europe regions and 61 Central and Eastern Europe regions, with different indicators of wellbeing, during the period 2001–2020, using a panel methodology.

Batrancea (2021b), for a sample of 28 EU nations, found that the global market and consumption behavior changed significantly from December 2019 to October 2020. Batrancea (2023), using panel models, showed that economic growth for eleven countries in Central and Eastern Europe over the decade from 2011 to 2020 was significantly impacted by the income, rate of young employees and educational attainment level in the two regions.

Batrancea, Balci, Akgüller, and Gaban (2022) found that their independent variables (exports, imports, foreign direct investment inflow and outflow, social contributions and wages) affected economic growth across a sample of 36 European countries during the period 2018Q3-2021Q3, using a panel methodology. Batrancea, Malar, and Batrancea (2022) found that more investments and green policies would be needed to sustain economic growth across a sample of fifty countries for the second half of the century.

Batrancea et al. (2020) found that non-performing loans, bank credit, and inflation had a negative effect on sustainable economic growth in India, Brazil, and Romania in the period during and after the global financial crisis, while CO₂ emissions had a positive effect. Batrancea, Batrancea, and Moscviciov (2009) studied the origins of the global financial crisis in the US, which were the decrease in consumption and the lack of liquidity in the capital market.

Batrancea, Gómez, Nichita, and Dragolea (2023) reported on a brief cross-cultural study of the most important markets around the world in order to analyze the extent of bribery, during the period from 2016 to 2020. All of the 25 countries and territories in the study were considered at low risk of bribery in this period.

Auditors should be careful when they account for audit risk for companies during a financial crisis. They might take into account audit attributes such as auditor tenure, non-audit fees, and auditor size, rather than audit fees. Thus, the hypotheses that are examined in this study are the following:

The data were extracted from the Refinitiv Eikon database. The sample includes 628 German, 628 French, 208 Italian and 112 Spanish companies in non-financial industries, because financial companies apply different accounting measures. The analysis was based on the four largest Eurozone countries, as these have a common accounting framework (IFRS) and a common currency, and the period of the financial crisis was close to the same in them all. The period under investigation goes from 2005 to 2013 ^[1].

The current research on the relationship between audit fees (AF) and non-audit fees (NAF) follows the previous studies of Barkess and Simnett (1994), Ezzamel, Gwilliam, and Holland (1996), Antle, Gordon, Narayanamoorthy, and Zhou (2006), Patel, Prasad, and Sauve (2010), and Chan, Chen, Janakiraman, and Radhakrishnan (2012). The aim of exploring a potential association between auditor tenure (TEN) and audit fees (AF) is based on the previous work of Krishnan and Krishnan (1997), while the relationship between auditor size (BIG4) and audit fees (AF) was explored in the studies of Becker, DeFond, Jiarnbalvo, and Subramanyam (1998), Chen, Lin, and Zhou (2005), and Corbella, Florio, Giotti, and Mastrolia (2015), among others. Finally, the inclusion of leverage (LEV) and return on assets (ROA) as control variables follows the previous study of Campa (2013).

The following GMM model is used for the investigation of the hypotheses:

All of the above variables are defined in the  Appendix.

Finally, while all the hypotheses are examined with the main model above, they are also investigated using an alternative model of the year-based effect (YEAReffect) as robust indicator of the crisis, instead of the change in the GDP (chGDP) variable. The YEAReffect dummy variable takes the value of 1 for years 2009 to 2013 and zero for years 2005 to 2008 during the financial crisis.

The above equation takes the following form with the robust year-based effect (YEAReffect) variable:

All of the above variables are defined in the  Appendix.

Table 1 refers to the three hypotheses of the main model. The first hypothesis is tested by examining the impact of non-audit fees on audit fees, with the interaction term for the change in GDP and the non-audit fees (chGDP*NAF), which is statistically significant and negative across all four countries under examination during the financial crisis. The values in the four countries (Germany, France, Italy and Spain) are quite close to each other.

The second hypothesis is tested by looking at the influence of the interaction term for GDP change and the auditor tenure on audit fees. The coefficients here are negative across all countries, but statistically significant only in Italy, during the financial crisis.

The third hypothesis is tested by considering the relationship between auditor size (BIG4) and audit fees (AF) during the financial crisis through the interaction term for GDP change and the auditor size (chGDP*BIG4). The coefficient for this relationship has a positive sign in Germany and a negative sign in France, Italy and Spain, and is statistically significant in Spain but not in Germany, France or Italy. The variable of non-audit fees has a positive and statistically significant sign across all countries in the whole period.

The other variables (OCF, LEV, ROA, DIS1, DIS2 and DIS3) are the control variables that have an effect on audit fees. The coefficient for operating cash flows with audit fees is statistically significant and positive for all the countries examined here but Germany. The value of the coefficients is quite similar across all the countries. As for the sign, this shows that the higher the operating cash flows, the higher the audit fees paid to auditors, in all countries except Germany. The coefficient for leverage on audit fees is statistically significant and positive in Germany, France, Italy and Spain, with rather dissimilar values. The estimated sign suggests that the higher the leverage, the higher the audit fees.

The coefficient of the return on assets variable on audit fees is statistically significant and positive in Germany and France, but the values are not very close. The difference in the coefficient value between the two countries is above 0.80 units. The coefficients of the three financial distress variables suggest that their effect on audit fees is different across the four countries under investigation. Specifically, the effect of negative net income (DIS1) on audit fees is statistically insignificant, with different signs in Germany, France, Italy and Spain. The relationship between the coefficient of negative working capital (DIS2) and audit fees and between both negative net income figures and working capital (DIS3) and audit fees shows that the former relationship (DIS2) is statistically significant in Italy, with a negative sign. Also, the latter relationship (with the DIS3 variable) is statistically significant in Germany, with a positive sign.

Table 2 shows the impact of the interaction terms for the year-based effect (the robust indicator of crisis during recession) and non-audit fees (NAF), auditor tenure (TEN), and auditor size (BIG4) on audit fees. The coefficients of the above interaction terms are similar, in general, in terms of both sign and value with the respective coefficients of Table 1. These results provide further support to the findings as they are similar for the two indicators of crisis used here.

The coefficients of the crisis variable with the robust indicator of crisis (YEAReffect) in Table 2 are statistically significant in all countries, as shown on Table 1. Not surprisingly, the sign of all the coefficients is positive for all four countries, as previously verified. The value of the two coefficients now is similar to what they previously were. This may be due to the fact that the crisis period indicator referring to the year-based effect is similar to that for the financial crisis as shown in Table 1.

This section considers the skewness, kurtosis, and Jarque–Bera statistics for the four countries collectively (Germany, France, Italy and Spain) in the crisis and pre-crisis periods for the most important variables: auditor tenure, logarithm of non-audit fees, logarithm of audit fees, and auditor size. Statistical metrics are significant for the variables, indicating the presence of skewness, kurtosis and normality level in the data series.

Most of the Pearson correlations between the variables indicate significant correlations in the crisis and pre-crisis periods. There are a few exceptions when the correlation between the variables is weak.

This section considers the impact of auditor tenure, auditor size, and non-audit fees on audit fees in the crisis and pre-crisis periods for the four countries collectively, using a panel OLS regression. There is a significant impact of non-audit fees (at the 1% level) and auditor size (at the 10% level) on audit fees in the pre-crisis period. The results are not significant for the crisis period. R-squared adjusted is found to be quite high in the two periods, while the F-test shows the importance of the model also in the two periods. In addition, the VIF indicator does not show the presence of multicollinearity for the independent variables. Moreover, the Breusch–Pagan tests for random effects on residuals indicate significance only in the pre-crisis period for the three measures.

In the crisis period, the impact of non-audit fees and auditor size on audit fees is significant at the 1% and 5% levels, respectively, when a panel fixed model with individual effects is considered. In the pre-crisis period, the impact of non-audit fees on audit fees remains significant at the 1% level, but the impact of auditor size is not significant. In contrast, there is a significant impact of auditor tenure on audit fees at the 5% level. This indicates the importance of non-audit fees in the two periods under investigation, and the impact of auditor tenure (in the pre-crisis period) and auditor size (in the crisis period). R-squared adjusted is high in the two periods and the F-test indicates the significance of the models under investigation in the two periods.

The random method with individual effects was also employed. The random effects model is appropriate for the crisis period, as the Hausman test indicates, but not for the pre-crisis period. The impact of non-audit fees on audit fees is significant at the 1% level, while auditor tenure and auditor size do not have a significant impact on audit fees in the crisis period. Moreover, there are some additional significant results for the control variables in the crisis period, which shows the importance of this random effects model. The random effects model cannot be used for analysis in the pre-crisis period, based on the Hausman test.

The impact of the specific variables on audit fees is also considered for the two periods with the fixed and random panel models with time effects. The Hausman test shows that the random method with time effects is inappropriate in the crisis period. Both methods indicate a significant impact of non-audit fees and auditor size on audit fees; the fixed method with time effects shows an additional impact of auditor tenure on audit fees. The R-squared adjusted is high, but the F-test is not high enough (at only 1.830), although significant at the 1% level. In the pre-crisis period, the Hausman test shows that the fixed method with time effects is inappropriate, but the results show that there is no considerable difference from the findings with the random method.

Last but not least, the GMM method indicates significant results in the two periods for the impact of non-audit fees on audit fees, and in the pre-crisis period for the impact of auditor tenure on audit fees, when all the independent variables and the constant are included as instruments. In addition, the J-specification has a small value and is found not to be significant.

In general, the whole analysis, considering the panel models, indicates that non-audit fees have a more important impact on audit fees than auditor tenure or auditor size.

This section describes the panel co-integration of the markets for audit and non-audit fees in the two periods considered in the study. The panel co-integration is considered both for each country and also between the countries, with and without time dummies, in the two periods. Two models are considered, FOLS and DOLS methods, providing significant results for co-integration for all the markets that are examined.

In addition, a heteroscedasticity test is performed, considering ARCH effects for up to four lags for the audit and non-audit fees in the two periods. ARCH effects were found to be present in the two series in both periods, and thus a bivariate panel GARCH model was run. The results indicate the transmission of volatility and errors between the four markets for the two variables in the two periods. This means that the information transmitted among the four markets for the two variables was important and meant that the markets interacted as additional news was released in the capital markets of the four countries.

The main purpose of the current study was to examine the relationship between audit fees and non-audit fees, auditor tenure, and auditor size during the most recent recession. The four largest Eurozone countries, Germany, France, Italy and Spain, were examined separately and collectively. In addition, the impact on audit fees of the recession (measured by the change in the GDP and the year-based effect) was incorporated.

For the main indicator of crisis, the negative and positive change in GDP that characterizes the period of recession was taken into account with a dummy variable. The robust indicator of crisis (the year-based effect) split the period of analysis into two periods (pre-crisis and in-crisis) with a dummy variable. For the above two indicators of crisis, a GMM model with interactive effects was used.

A statistically significant result was found for all the countries regarding the relationship between audit and non-audit fees, through the interactive terms, with the main and the robust indicators of crisis (change in GDP and year-based effect). The interactive estimations indicated that the relationship between auditor tenure and audit fees, and the relationship between auditor size and audit fees, with the main indicator of crisis (change in GDP) are, respectively, statistically significant in Italy and in Spain. In addition, the estimations of the above relationships with the robust indicator of crisis (year-based effect) show that there is a similar impact. In addition, applying the same principle, several panel models were run, showing significant effects of the audit characteristics on audit fees in the pre-crisis and in-crisis periods in the four countries collectively. Moreover, a significant parallel movement of information and transmission of information between markets for audit and non-audit fees were found with a panel co-integration and bivariate panel GARCH model in the pre-crisis and in-crisis periods in the four countries.

This study implies that the four markets became less efficient during the period of financial crisis, as indicated by the co-movement of information and the transmission of information among the markets for audit fees and non-audit fees. These findings have important policy implications for accountants, auditors and financial market regulators. The fact that during the recent financial crisis volatility and error information spilled over across markets in terms of audit fees and non-audit fees may mean that there is high risk for investors’ decisions when investing in the capital markets during a crisis period when additional information for audit and non-audit fees is released.

The panel equation takes the following form within the pre-crisis and in-crisis periods for all the countries: