Auditors' Response to Political Connections and Cronyism in Malaysia

Authors


  • This paper has benefited from the insightful comments of an anonymous reviewer, Jere Francis, Bin Srinidhi, Chung-ki Min, Dan Simunic, Sami Heibatollah, Suresh Radhakrishnan, Phyllis Gul, Judy Tsui, and numerous other individuals and participants at various workshops and conferences. Thanks are also due to Mazlina Mat Zain, Sunny Sun, and Alvis Lo for their assistance in data collection and data analyses. Financial support from the Hong Kong Research Grants Council is gratefully acknowledged.

ABSTRACT

This paper extends the literature on the role of political economy in financial reporting and auditing by testing two hypotheses. The first hypothesis predicts that there will be a greater increase in audit effort and audit fees for Malaysian firms with political connections, as a result of the Asian financial crisis, than for non-politically connected firms because these firms have a higher risk of financial misstatements. The second hypothesis predicts that the audit fees of politically connected firms will decline when capital controls are introduced by the government as a ploy to financially assist politically connected firms to rebound from the crisis, and thus reduces the risk of financial misstatements. The results show that there is a greater increase in audit fees for firms with political connections than for non-politically connected firms as a result of the Asian financial crisis. However, there is a decline in audit fees for politically connected firms after the capital controls are implemented.

1. Introduction

A recent strand of the accounting and finance literature uses cross-country data to investigate how the incentives of corporate executives, investors, regulators, auditors, and other market participants are shaped by the institutional structure of the country in which firms are domiciled (Ball, Kothari, and Robin [2000], Ball, Robin, and Wu [2003]). These institutional structures include a country's legal/judicial system, securities laws, taxation regime, and political economy. The incentives are expected to shape the properties of accounting numbers “through a complex interplay of accounting standards, legal, market, regulatory, and political pressures and reporting discretion exercised by managers” (Bushman and Piotroski[2006, p. 2]). In particular, prior studies suggest that political economy1 is related to various aspects of financial reporting. For example, Bushman, Piotroski, and Smith [2004] employ a cross-sectional country-level analysis to show that firms with higher government share ownership are associated with a lower level of financial transparency. Bushman and Piotroski[2006, p. 4] also show that firms in countries with more state involvement in the economy “speed recognition of good news and slow recognition of bad news in reported earnings relative to firms in countries with less state involvement.”2 However, several concerns have been raised regarding cross-sectional studies. These include limited sample sizes, the likelihood of endogeneity in the availability of variables at the country level, noisy variables, and the severe correlated omitted variables problem (see Miller [2004] for a detailed discussion of these limitations). In conclusion, Miller [2004] suggests that future research should focus on a country or a region of the world that demonstrates a more general issue in international research. Further, such a “more focused approach would free authors from needing variables available across a wide range of countries, allowing variables to be designed that more cleanly capture the construct being measured” (Miller[2004, p. 266]).

In this spirit, this study focuses on Malaysia and considers the impact of political factors and macroeconomic changes on the behavior and responses of auditors who are both important participants in the market and directly connected with the credibility of financial reporting (see Bushman, Piotroski, and Smith [2004] for a discussion of the role of audit quality). More specifically, this study examines the response of auditors, in terms of audit effort and audit fees of politically connected (PCON) firms (as opposed to non–politically connected (non-PCON) firms) to the 1997 Asian financial crisis and the subsequent implementation of capital controls. The financial crisis followed by the capital controls and political factors provides a unique setting to examine the interplay and effects of macroeconomic changes, and institutional factors, on accounting institutions. The interplay of these components in Malaysia is expected to influence the property of accounting numbers, which, in turn, are likely to affect auditors' assessment of audit risk. By focusing on a particular country, this study overcomes some of the criticisms of cross-country studies, including potential difficulties in data analyses and interpretation that arise as a result of differences in legal regimes, judicial efficiencies, and cultural and even perhaps anthropological factors (Miller [2004]).

Using the Malaysian setting, Johnson and Mitton [2003] (hereafter, J&M), in a related study, argue that PCON firms were harder hit during the 1997 crisis than other firms due to their inefficiencies and the inability of the government to bail out these favored firms during the crisis, at least initially. However, the government's subsequent implementation of capital controls in 1998 was primarily intended to benefit PCON firms. Consistent with the arguments, J&M document that stock returns of PCON firms were lower relative to other Malaysian firms in the early phase of the Asian crisis. However, once the capital controls were imposed the returns of these favored firms were higher on average.

The analysis in J&M provides insights into how the market reacted to the financial crisis and subsequent capital controls for PCON firms. However, there are other components in the system, such as auditors, that are also likely to be affected by the dynamics of the changes in the social contract. More specifically, auditors are likely to feel assured that the government would support PCON firms before the financial crisis. However, with the change in the social contract as a result of the financial crisis, the players in the system are expected to respond and adjust relatively quickly to the change. Accordingly, from an audit perspective, the analysis in J&M suggests that auditors would have assessed PCON firms as having a higher audit risk during the financial crisis relative to other firms due to (1) a greater likelihood of business failure and (2) a greater likelihood of misreporting and overstatement of earnings in order to avoid debt default. As a result, auditors are expected to expend relatively more effort, hence leading to higher audit fees, ceteris paribus. This implies that the increase in audit fees between the period before the financial crisis (i.e., firms with year-ends before July 1997) and the financial crisis period should be higher for PCON firms than for non-PCON firms. However, after the capital controls were introduced, there was less likelihood of business failure and less incentive for PCON firms to misreport and overstate earnings. As a result, the audit risk for PCON firms, as assessed by auditors, is likely to have decreased in 1998 following the capital controls, leading to less audit effort and lower audit fees. The empirical results from this study indicate that this is the case. This study provides complementary evidence supporting the J&M analysis based on an observation of changes in audit fees in Malaysia during the Asian financial crisis.

This paper contributes to the corporate governance and accounting literature in the following ways. First, by identifying and examining the role of corporate political connections in audit pricing, this study adds to the recent literature that shows links between the political economy and accounting institutions.3 More specifically, prior cross-country studies have shown that there are linkages between the political economy and financial reporting (see Bushman, Piotroski, and Smith [2004], Bushman and Piotroski [2006]). A natural extension of this line of research is to examine if auditors' behavior is also affected by political factors. Thus, this paper examines auditors' response to firms with political connections (rather than government ownership4) and provides country-specific evidence that overcomes some of the criticisms of cross-country studies. Second, J&M show that PCON firms had comparatively poor stock returns during the financial crisis but that these stock returns improved once capital controls were imposed by the government. This paper provides evidence from an auditing perspective that sheds some light on the real financial conditions of the PCON firms during the crisis. In so doing, this paper demonstrates how auditors, an important component in the system, responded to the financial crisis and the subsequently implemented capital controls. Third, the idea of “cronyism,” and its linkage to financial reporting incentives, has been discussed in the literature (see, for example, Ball, Robin, and Wu [2003]), but there has been little evidence provided of cronyism in the accounting and finance literature. By using “audit” analysis, this study demonstrates how PCON firms suffered more during the financial crisis and recovered just as quickly after capital controls (a ploy to assist “cronies”) were introduced, hence providing some concrete evidence of crony capitalism “at work.” Finally, this study provides new insights into an institutional phenomenon that affects corporate governance unique to the Malaysian corporate sector. Ball, Robin, and Wu [2003] identify political connections in Malaysia as a factor that could contribute to the lower levels of financial reporting quality,5 which is consistent with the findings of this study from the perspective of auditors' assessments of audit risk for firms.

The next section of the paper provides a background for the study and this is followed by the hypotheses development section. The fourth section discusses the design and sample selection procedures while the results are presented and discussed in the fifth section. The final section presents a summary and conclusions of the study.

2. Background

2.1 auditing in malaysia

All big international audit firms have offices in Malaysia and at the time of this study audit more than 80% of the listed companies. Besides audit work, quality assurance procedures are closely monitored by the audit firms' head offices based in the United States (see, for example, Favere-Marchesi [2000] for a discussion of audit quality in Malaysia). Professional auditing standards require auditors to assess both client-related risks and institutional factors when performing auditing procedures aimed at reducing audit risk to an acceptable level. However, to understand the role of PCON firms and cronyism in financial reporting and auditors' risk assessments, it is important to understand a fundamental characteristic of the East Asian economic system, the relationship-based system.

2.2 relationship-based system

Rajan and Zingales [1998] identify the economic systems of several East Asian countries as a relationship-based system (as opposed to an arm's length system). This system, which is characterized by cronyism and low levels of transparency, works well in jurisdictions with weak corporate governance mechanisms and where contracts are poorly enforced. In such situations, power relationships substitute for contracts because they can achieve better outcomes than a primitive contract system (Walker and Reid [2002]). However, the relationship-based system suppresses the price system and in so doing misallocates large external capital inflows that come from predominantly arm's length lenders. As pointed out by Jackson[1999, p. 6], “the abundance of inexpensive capital, in combination with local banks based on personal relationships rather than real business plans, resulted in widespread misallocation of capital into speculative and noncompetitive sectors and enterprises.” Since foreign investors are aware of this misallocation of capital and other abuses, they minimize risk by keeping their claims short term in order to exit at the first sign of trouble. The Rajan and Zingales [1998] thesis suggests that the contract between the relationship-based system and the arm's length system of foreign investors creates a “fragile hybrid” that works well in normal times but is prone to shocks, such as when the Asian financial crisis occurred (Walker and Reid [2002]). In Malaysia, anecdotal evidence and widespread press reports suggest that PCON firms exhibit more of the basic characteristics of a relationship-based system such as cronyism and lower levels of transparency than non-PCON firms (Gomez[1990, p. 11], Jomo and Gomez [2000]). Based on the Rajan and Zingales [1998] thesis, PCON firms are expected to face more financial hardships when there is a “shock” such as a financial crisis.

2.3 political connections in malaysia

While there are broad similarities6 in the accounting and regulatory environments7 in Malaysia, the United States, and the United Kingdom, there is, however, an important institutional difference. This relates to the fact that the Malaysian corporate sector is characterized by the existence of politically favored corporations. Most writers agree that a major factor that contributes to the Malaysian government's involvement in the corporate sector is the government's new economic policy (NEP) that commenced in late 1969 following divisive racial riots. The NEP was initiated by the government to increase the participation of indigenous Malays in the economy as a means of addressing the then existing situation where the economy was predominantly run by the Chinese to the exclusion of the indigenous Malays (Bumiputeras).8 Apart from encouraging Bumiputera ownership of firms with easy bank credit facilities, the government created Bumiputera trust agencies under the direction of the United Malay National Organization (UMNO)9 to acquire equity stakes on behalf of the Malays (Mehmet [1988]). The government also selectively picked certain firms to receive favorable investment resources, such as the Heavy Industries Corporation of Malaysia (HICOM) in 1980. HICOM later invested in the steel, cement, and auto industries. This active involvement by the government saw the emergence of many companies by the 1990s that were owned and managed directly or indirectly by UMNO10 (Cheong [1997], Enright, Scott, and Dodwell [1997], Gomez [2005]).

The second form of political favoritism consisted of informal ties among firms run by Malay, Chinese, and Indian businessmen, along with the Prime Minister, Dr. Mahathir, the previous Deputy Minister, Anwar Ibrahim, and a prominent businessman (and later Finance Minister), Daim Zainuddin. These relationships arose as a result of the fact that the Chinese business community actively solicited and developed ties with politically influential individuals in order to advance their business interests (Bowie[1991, p. 103–104]). The Chinese connections that developed were inevitable given Malaysian business being historically dominated by Chinese business. Several Chinese political parties such as the Malaysian Chinese Association and the Gerakan made up a part of the ruling coalition, the Barisan Nasional. By the mid-1990s, the corporate sector was dominated by politically linked companies and businesses.

There are two broad views on the effects that state ownership and political connections have on financial markets. The first, based on the theories of North [1990] and Olson [1993], argues that governments acquire control of firms to provide employment, subsidies, and other benefits to supporters in return for votes, political contributions, and bribes (e.g., see La Porta, Lopez-de-Silanes, and Shleifer [2002], Bushman, Piotroski, and Smith [2004]). The second, a more generous view, is that government ownership of firms is designed to deal with market imperfections (see Gerschenkron [1962], Shleifer [1998]). In Malaysia, the PCON firms are not necessarily owned by the state but are identified as “favored” firms by the ruling government. Bushman, Piotroski, and Smith [2004] identify a number of ways political involvement in the economy can affect financial transparency, of which the following may be relevant in the Malaysian context. First, firms with political connections may suppress firm-specific information to hide expropriation activities by politicians and their cronies. Second, politicians exploit their control over regulatory policies to favor cronies in return for bribes, nepotism, and political support.

3. Hypotheses

The general idea that relationship-based systems are more prone to shock and that corporations controlled by politicians are less transparent is not new and has been identified and discussed previously in the literature (Walker and Reid [2002], Shleifer and Vishny [1994]). As a result of the financial crisis, including the devaluation of the Malaysian dollar, the government faced severe financial hardship and was unable to financially assist PCON firms. Such a setback probably affected the various components of the system, such as the market and auditors. While J&M consider how the market reacted to PCON firms during this period, this paper considers how auditors responded to the crisis and capital controls.

The linkage between changes in the system, such as a financial crisis, and how auditors respond to these changes can be demonstrated via firms' financial reporting practices. Poor inherent firm performance of PCON firms during the crisis due to operating inefficiencies (the result of cronyism), and coupled with the inability of the government to provide protection/support during the crisis, is posited to result in an increased risk of the financial statements being materially misstated, ceteris paribus. Auditors are expected in this situation to use more audit effort in order to collect sufficient evidence to render an appropriate opinion. As a consequence, higher audit fees will be charged as a result of the perceived higher risk by the auditor.11Appendix A provides a brief discussion of the phases of the audit process and auditors' assessment of risk. Results in J&M's study are consistent with the view that these politically favored companies were associated with higher audit risk during the Asian financial crisis. This reasoning suggests the following hypothesis:

  • H1 :  There will be a greater increase in audit fees for PCON firms than for non-PCON firms as a result of the financial crisis.

Anecdotal evidence and numerous press reports suggest that capital controls were introduced by the Malaysian government in September 1998 so that some “strong politicians” could support PCON firms (see, for example, J&M, Appell [1999], Kotler and Kartajaya [2000]). The notion that capital controls are designed to protect politically favored firms is also consistent with the characterization of a “relationship-based system” by Rajan and Zingales [1998]. In fact, they suggest that capital controls facilitate the implementation of policies in a “relationship-based” system whereby politicians and banks channel funds to favored firms (often referred to as cronyism), and that this is easier to implement when a country is relatively isolated from international capital flows. J&M find evidence that PCON firms received subsidies12 and benefited more as a result of capital controls.

Since capital controls and various subsidies assisted politically favored firms financially, it is likely that managers of these firms had less incentive to misstate financial statements after the controls were in place. If this is true, then PCON firms would be perceived by auditors to be associated with relatively less audit risk than during the period (after July 1997 but before September 1998) when the full force of the financial crisis hit the country. This reasoning leads to the second hypothesis:

  • H2 :  Audit fees for PCON companies will decline after the imposition of capital controls.

4. Research Method

4.1 data collection

The initial sample consists of 1,119 firm-year observations for 1996, 1997, and 1998 from the Worldscope database. After screening for financial firms and firms with missing information, and deleting outliers (top 1% of the sample with very high audit fees), complete data are available for 812 firm-year observations of listed nonfinancial firms on the Kuala Lumpur Stock Exchange. Data that are not available from Worldscope are collected from the annual reports of firms domiciled in Malaysia and listed on the Kuala Lumpur Stock Exchange. To provide a stronger test of the theory, the sample of firms most likely to be affected by the financial crisis and the capital controls are selected for testing (see Table 1). As auditors typically conduct tests of transactions in phase III of the audit process, which is two to three months before the year-end (see Arens, Elder, and Beasley[2003, p. 357] and Appendix A), firm observations with years ending at least three months after the crisis and after the capital controls were introduced are selected for the financial crisis sample and the capital controls sample, respectively. The sample for the period before the financial crisis is selected for firms with year ending before June 1997.

Table 1. 
Distribution of Firms by Year-Ends
Panel A: Pre–financial crisis sample
 With Corporate Political Connections (N= 36)Without Corporate Political Connections (N= 185)
FrequencyCumulative %FrequencyCumulative %
199608  1 0.01
19961217 47.229752.97
199701458.331360.00
199702  361.62
199703366.672776.22
199704169.44 679.46
199705  280.54
19970611 100.00 36100.00 
Panel B: Financial crisis sample
 With Corporate Political Connections (N= 41)Without Corporate Political Connections (N= 234)
FrequencyCumulative %FrequencyCumulative %
1997122048.78120 51.28
199801 458.541758.54
199802  359.82
199803 463.303474.35
199804 273.18 877.77
199805  379.05
19980611100.00 4297.00
199807  297.85
199808  5100.00 
Panel C: Capital controls sample
 With Corporate Political Connections (N= 38)Without Corporate Political Connections (N= 206)
FrequencyCumulative %FrequencyCumulative %
19981219 50.0012862.14
199901460.53 1267.97
199902   369.43
199903471.06 2782.54
199904173.69  384.00
199905   284.97
199906894.74 31100.00 
199907 
1999082100.00  

Since the financial crisis started in July 1997, the financial crisis sample is selected from firms with year-ends from December 1997 to August 1998, providing a total sample of 275 firm observations (including PCON firms N= 41). Since capital controls were introduced in September 1998, the capital controls sample includes all firms with year-ends from December 1998 to August 1999, providing a total sample of 244 firm observations (including PCON firms N= 38). Firms with year-ends before July 1997 are selected for the sample to represent firms before the financial crisis period, providing a total sample of 221 firm observations (including PCON firms N= 36). The final sample for the three periods consists of 740 firm year observations.

4.2 audit fee model

While the majority of studies use the natural logarithm of audit fees as the dependent variable in the audit fee model (see Francis [1984], Gul [1999], Ferguson, Francis, and Stokes [2003]), Simunic [1980] uses audit fees deflated by total assets as the dependent variable. Since most audit fee models use the logarithmic transformation of audit fees as the dependent variable, the tests in this study are conducted using logarithmic transformation as the dependent variable (see, for example, Francis [1984], Ferguson, Francis, and Stokes [2003]). In sensitivity tests reported later the dependent variable is also measured in terms of audit fees deflated by total assets.

Prior research has identified a variety of factors to explain the cross-sectional variation in audit fees (see, for example, Francis [1984], Simunic [1980]),13 including client-specific risks (see, for example, Davis, Ricchuite, and Trompeter [1993], Johnstone and Bedard [2001], Mock and Turner [2002]). Apart from client-specific risks, auditors are also expected to consider macroeconomic and institutional factors in their assessments of audit risks (see, for example, Hayes et al.[2005, p. 200–208]). These institutional factors are likely to affect managers' financial reporting incentives which, in turn, are likely to affect auditors' assessments of audit risk.

The institutional variable examined in this study is PCON versus non-PCON firms. A dummy variable is used to indicate whether or not a firm is politically connected. J&M identify PCON firms on the basis of a search of the entire text by Gomez and Jomo [1997], who provide a detailed analysis of Malaysian corporations and their political connections prior to the Asian financial crisis.14 PCON firms are those that are identified by Gomez and Jomo [1997] as having close relationships with their officers or major shareholders and primarily Mahathir, Daim, and Anwar. The political connection variable is set as PCON= 1 if the firm is one of the firms identified. Appendix B provides a list of PCON firms used in the study.15

In the audit fee model adopted in this paper, several standard control variables are selected, including additional variables that may be unique to the Malaysian setting. The effects of size are controlled for with the natural log of total assets. The debt ratio and the ratio of current assets to total assets are for controlling audit risk. Profitability and liquidity are controlled for by including return-on-assets and the working capital (liquidity) ratio, respectively. Audit complexity is controlled by the number of directly owned subsidiaries and the number of subsidiaries that are located in countries other than Malaysia. Loss sharing between the auditor and the client is controlled for with audit qualifications (i.e., clean versus qualified opinion). Controls for audit quality (big audit firms versus small audit firms)16 and a dummy for firms with losses in the last year are also included. The fiscal year-end is controlled for with a dummy variable set as 1 for those firms with a 31 December fiscal year-end and zero otherwise. Most firms in the sample have a December year-end. To control for the relative financial distress faced by both PCON firms and non-PCON firms, the Altman Z-score17 (Altman [1993]) is computed for each firm and is included as a control variable in all the regressions. Nor and Chin [2002] test the validity of the Z-score model for Malaysian firms and conclude that the model is appropriate for Malaysian firms.

4.3 descriptive statistics

Table 2 reports the distribution of the sample firms in different industries. It shows that the PCON sample comes mainly from three industries and constitutes a significantly smaller number of firms in each of the categories compared with the non-PCON group.

Table 2. 
Number of Observations by Industry
IndustryWith Corporate Political ConnectionsWithout Corporate Political Connections
Pre-financial Crisis SampleFinancial Crisis SampleCapital Controls SamplePre-financial Crisis SampleFinancial Crisis SampleCapital Controls Sample
Consumer products898 26 38 36
Industrial products10 10 9 60 80 72
Properties011 10 11  9
Construction343 22 24 18
Hotels111  3  3  3
Trading/services11 13 13  43 51 47
Plantations222 17 23 18
Mining111  4  4  3
Total36 41 38 185234206

Descriptive statistics are reported in Table 3. Panels A, B, and C contain summary statistics for the pre–financial crisis (PFC), financial crisis (FC), and capital controls (CAPCTRL) sample firms with and without corporate political connections, respectively. Since the descriptive statistics are for both continuous and dichotomous variables, both t-tests and chi-square tests are used to test for differences where appropriate. The descriptive statistics show that for samples in all the three periods, PCON firms are larger in terms of total assets and have a larger number of subsidiaries. The PCON firms also have a lower liquidity ratio (LIQ), a higher proportion of losses, and, as expected, a higher proportion of Bumiputera-controlled firms than non-PCON firms, but the differences are statistically significant only for the capital controls sample. In general, PCON firms have higher average audit fees than non-PCON firms in the three periods. However, the descriptives also show that the average audit fees for PCON firms increased from M$283,732 to M$567,919 during the FC period and declined to M$314,380 during the CAPCTRL period while the average audit fees for the non-PCON firms are similar for the three periods (i.e., M$172,513 for the PFC sample, M$211,449 for the FC sample, and M$188,967 for the CAPCTRL sample).

Table 3. 
Descriptive Statistics
Panel A: Pre–financial crisis sample
VariableWith Corporate Political Connections (N= 36)Without Corporate Political Connections (N= 185)t-test 
MeanStd. Dev.MedianMeanStd. Dev.Median
LAF5.2110.9905.1244.7260.9134.654−2.88*
AF (‘000)283.732292.837168.000172.513178.424105.000−2.20*
SIZE13.8981.26813.72812.9791.18112.888−4.22*
TA (‘000)2286496.3903078860.450916032.500971239.3702588619.130395546.000−2.70*
DA0.2330.5560.1220.1090.1220.071−1.33 
CURRAT0.4660.2160.4690.4380.2130.423−0.72 
LIQ1.5111.0441.2481.6821.3731.3090.71 
BIGa0.7780.4221.0000.8050.3971.000@0.14 
YE0.5000.5070.0000.4540.4990.000@0.26 
LSUBb2.6660.7572.7712.1330.9652.079−3.13*
SUB18.50013.57216.00013.26515.1908.000−1.92 
LFOREIGNb−3.5126.1510.000−5.7306.184−11.513−1.97 
FOREIGN3.0566.3331.0001.7574.8500.000−1.17 
ROA−0.0330.3830.0290.0360.0580.0341.08 
LOSSDUM0.1390.3510.0000.0650.2470.000@2.33 
DUMZ0.1940.4010.0000.2640.4330.000@0.79 
OPINa0.0830.2800.0000.0320.1780.000@2.00 
BUMDUMc0.3530.4850.0000.2720.4460.000@0.89 
Panel B: Financial crisis sample
VariableWith Corporate Political Connections (N= 41)Without Corporate Political Connections (N= 234)t-test 
MeanStd. Dev.MedianMeanStd. Dev.Median
LAF5.6071.1865.2094.8020.9514.740−4.12*
AF (‘000)567.919835.790183.000211.449367.632114.500−2.69*
SIZE14.1171.42313.89213.1251.35513.097−4.29*
TA (‘000)3237247.6804474384.4201079047.0001597952.7706371292.400487474.500−2.02*
DA0.1310.4310.0011.07414.4650.0000.99 
CURRAT0.4390.2030.3820.4330.2130.414−0.15 
LIQ1.3891.2061.0901.6201.6381.2251.07 
BIG0.7800.4191.0000.8080.3951.000@0.16 
YE0.4880.5060.0000.4440.4980.000@0.26 
LSUBa2.5720.9312.7731.6352.9342.197−3.89*
SUB18.36614.37516.00014.18820.7559.000−1.59 
LFOREIGNa−5.3516.424−11.513−5.8616.209−11.513−0.48 
FOREIGN2.1953.6480.0002.3158.0140.0000.15 
ROA−0.0820.4590.000−0.0440.3610.0180.51 
LOSSDUM0.1220.3310.0000.1110.3150.000@0.04 
DUMZ0.2680.4490.0000.2480.4330.000@0.08 
OPIN0.1950.4010.0000.1710.3770.000@0.14 
BUMDUMb0.3680.4890.0000.2540.4360.000@2.11 
Panel C: Capital controls sample
VariableWith Corporate Political Connections (N= 38)Without Corporate Political Connections (N= 206)t-test 
MeanStd. Dev.MedianMeanStd. Dev.Median
  1. *p < 0.05; @chi-square tests.

  2. aSince data on auditor type and audit opinion for 36 firms and 18 firms, respectively, are not available, the information for these variables is assumed to be the same as for the year 1997.

  3. bObservations having a zero for LSUB or for LFOREIGN are re-coded to a small positive value (0.00001) to enable a logarithmic transformation.

  4. cThis variable is only available for 34 PCON firms and 158 non-PCON companies in the pre–financial crisis sample.

  5. *p < 0.05; @chi-square tests.

  6. aObservations having a zero for LSUB or for LFOREIGN are re-coded to a small positive value (0.00001) to enable a logarithmic transformation.

  7. bThis variable is only available for 38 PCON firms and 197 non-PCON companies in the financial crisis sample.

  8. *p < 0.05; @chi-square tests.

  9. aObservations having a zero for LSUB or for LFOREIGN are re-coded to a small positive value (0.00001) to enable a logarithmic transformation.

  10. bThis variable is only available for 36 PCON firms and 177 non-PCON companies in the capital controls sample.

  11. LAF= natural logarithm of audit fees

  12. AF= audit fee in thousand dollars

  13. SIZE= natural logarithm of total assets

  14. TA= total assets in thousand dollars

  15. DA= book value of long-term debt to total assets

  16. CURRAT= current assets to total assets

  17. LIQ= current assets to current liabilities

  18. BIG= indicator variable, 1 for big audit firms

  19. YE= indicator variable, 1 for fiscal year ending 31 December

  20. LSUB= natural logarithm of the number of subsidiaries

  21. SUB= number of subsidiaries

  22. LFOREIG= natural logarithm of the number of foreign subsidiaries

  23. FOREIGN= number of the foreign subsidiaries

  24. ROA= net income to total assets

  25. LOSSDU= indicator variable, 1 for loss in the last year

  26. DUMZ= indicator variable, 1 for high Altman Z-score (>2.073)

  27. OPIN= indicator variable, 1 for modified audit opinion

  28. BUMDUM= indicator variable, 1 for Bumiputera-controlled company in which 50% or more of the equity is held by Bumiputera shareholders and institutions

LAF5.3570.9235.2924.7950.9014.687−3.52*
AF (‘000)314.380298.640199.000188.967231.586108.500−2.46*
SIZE13.8701.34213.41413.0781.29513.065−3.45*
TA (‘000)2269506.7603011387.370669962.5001127920.6402401144.010472037.500−2.58*
DA1.6388.1570.0010.1000.7330.000−1.16 
CURRAT0.4200.2070.3970.4150.2030.4000.15 
LIQ1.1270.8250.9751.6571.8641.2152.84*
BIG0.8420.3701.0000.8060.3971.000@0.28 
YE0.4470.5040.0000.3790.4860.000@0.64 
LSUBa2.5210.8552.7401.5533.0802.197−3.79*
SUB16.39510.86216.00013.01016.0669.000−1.62 
LFOREIGNa−5.2086.415−5.756−6.0126.174−11.513−0.73 
FOREIGN2.0533.3530.5002.1017.7000.0000.06 
ROA−0.2120.601−0.040−0.0330.1960.0081.95 
LOSSDUM0.5000.5060.0000.2520.4350.000@9.53*
DUMZ0.1320.3430.0000.2380.4270.000@2.10 
OPIN0.3680.4890.0000.2430.4300.000@2.62 
BUMDUMb0.3890.4940.0000.2490.4330.000@2.97*

4.4 model specification

Based on prior audit fee studies (Francis [1984], Gul [1999], Ferguson, Francis, and Stokes [2003]), the following ordinary least squares regression model is run to test H1:

image(1)
where:
LAF

natural logarithm of audit fees

Control Variables
SIZE

natural logarithm of total assets

DA

book value of long-term debt to total assets

CURRAT

current assets to total assets

LIQ

current assets to current liabilities

BIG

1 for big audit firms, 0 otherwise

YE

1 for fiscal year ending 31/12, 0 otherwise

LSUB

natural logarithm of the number of subsidiaries18

LFOREIGN

natural logarithm of the number of foreign subsidiaries

ROA

net income to total assets

LOSSDUM

1 for firms with loss in the last year, 0 otherwise

DUMZ

1 for firms with high Altman Z score (>2.073), 0 otherwise.

OPIN

1 for modified audit opinion, 0 otherwise

Experimental Variables
PCON

1 for firms with identifiable political connection with high-ranking political figures, 0 otherwise

H2 is tested with a regression model that pools the data for the three periods and examines the interaction terms between PCON and the period dummy variables (i.e., PFC= 1, and 0 otherwise; and CAPCTRL= 1, and 0 otherwise). The interaction terms test whether there is a significant increase in audit fees for the FC period and a significant decline in audit fees after the imposition of capital controls.

5. Results and Discussion

Table 4 reports the correlation matrix for the three periods. The regression results for the FC sample after controlling for industry differences are shown in model 1 of Table 5. The coefficient for PCON is positive and significant (0.251, p < 0.01, 1-tailed), thus providing support for H1. The parameter of 0.251 represents an average audit fee premium of 28.5%. The premium is obtained by calculating the effect of the percentage shift on the natural log of audit fees and is defined as ez− 1, where z is the parameter value for PCON firms (see Ferguson, Francis, and Stokes [2003], Simon and Francis [1988]).19 The signs for the control variables are all in the right direction except for the coefficients for DA, YE, LOSSDUM, and OPIN, which are all insignificant except for YE. Significant results in the predicted direction are obtained for the coefficients for SIZE, CURRAT, LIQ, BIG, LSUB, LFOREIGN, ROA, and DUMZ. The adjusted R2 for the model is 62%.

Table 4. 
Spearman Rank Correlation Matrix
VariableLAFSIZEDACURRATLIQLIQBIGYELSUBLFOREIGNROALOSSDUMDUMZOPINPCONBUMDUM
  1. *p < 0.05.

  2. LAF= natural logarithm of audit fees

  3. SIZE= natural logarithm of total assets

  4. DA= book value of long-term debt to total assets

  5. CURRAT= current assets to total assets

  6. LIQ= current assets to current liabilities

  7. BIG= indicator variable, 1 for big audit firms

  8. YE= indicator variable, 1 for fiscal year ending 31 December

  9. LSUB= natural logarithm of the number of subsidiaries

  10. LFOREIGN= natural logarithm of the number of foreign subsidiaries

  11. ROA= net income to total assets

  12. LOSSDUM= indicator variable, 1 for loss in the last year

  13. DUMZ= indicator variable, 1 for high Altman Z-score (>2.073)

  14. OPIN= indicator variable, 1 for modified audit opinion

  15. PCON= indicator variable, 1 for politically connected firms

  16. BUMDUM= indicator variable, 1 for Bumiputera-controlled company in which 50% or more of the equity is held by Bumiputera shareholders and institutions

Panel A: Pre–financial crisis sample (N= 221)
LAF1.0000.608*0.157*0.078−0.0360.0730.0150.582*0.461*−0.032−0.112−0.260*0.0010.207*0.006
SIZE 1.0000.232*−0.206*−0.034−0.0640.156*0.478*0.427*0.058−0.203*−0.242*−0.0930.268*0.051
DA 1.000−0.273*−0.163*−0.0920.1430.133*0.150*−0.236*0.026−0.440*−0.0140.137*0.104
CURRAT 1.0000.351*0.045−0.0800.062−0.043−0.0370.102−0.0230.0830.046−0.129
LIQ 1.0000.028−0.002−0.050−0.0480.337*−0.149*0.371*−0.039−0.063−0.105
BIG 1.000−0.084−0.062−0.0300.0140.0160.1600.045−0.026−0.017
YE 1.000−0.0080.014−0.014−0.029−0.1220.0390.0340.130
LSUB 1.0000.615*−0.1280.006−0.257*−0.0250.239*−0.073
LFOREIGN 1.000−0.059−0.021−0.130−0.0230.152*−0.140
ROA 1.000−0.429*0.366*−0.112−0.059−0.102
LOSSDUM 1.000−0.0120.1120.1030.013
DUMZ 1.000−0.015−0.060−0.020
OPIN 1.0000.0950.085
PCON 1.0000.068
BUMDUM 1.000
Panel B: Financial crisis sample (N= 275)
LAF1.0000.691*0.180*−0.033−0.126*0.0380.0840.639*0.444*−0.069−0.148*−0.1190.0020.255*0.069
SIZE 1.0000.076−0.265*−0.032−0.0030.0990.465*0.377*0.095−0.280*−0.031−0.0900.250*0.061
DA 1.000−0.141*−0.451*−0.0100.0580.201*0.108−0.425*0.208*−0.371*0.299*0.0720.097
CURRAT 1.0000.370*0.020−0.0630.024−0.0090.0800.0860.239*0.0070.012−0.102
LIQ 1.0000.069−0.038−0.0950.0260.588*−0.259*0.377*−0.290*−0.052−0.024
BIG 1.000−0.030−0.080−0.0930.0690.0030.0540.034−0.024−0.035
YE 1.0000.008−0.038−0.139*−0.046−0.0860.0260.0310.149*
LSUB 1.0000.548*−0.115*−0.077−0.143*0.0470.184*0.005
LFOREIGN 1.0000.006−0.0690.0090.0340.060−0.118
ROA 1.000−0.374*0.487*−0.376−0.055−0.004
LOSSDUM 1.000−0.151*0.290*0.012−0.048
DUMZ 1.000−0.1530.020−0.016
OPIN 1.0000.0230.079
PCON 1.0000.095
BUMDUM 1.000
Panel C: Capital controls sample (N= 244)
LAF1.0000.684*0.191*−0.048−0.1100.0460.0690.589*0.425*−0.0950.043−0.152*0.0430.225*0.086
SIZE 1.0000.126*−0.243*−0.032−0.0020.0140.401*0.332*0.114−0.069−0.0190.0260.225*0.027
DA 1.000−0.132*−0.487*−0.0620.0400.256*0.116−0.528*0.296*−0.416*0.356*0.197*0.068
CURRAT 1.0000.443*0.013−0.0570.0460.0290.102−0.0900.090−0.0820.008−0.075
LIQ 1.0000.0010.001−0.109−0.0520.539*−0.439*0.348*−0.360*−0.123−0.056
BIG 1.000−0.066−0.100−0.119−0.054−0.0140.0050.0250.034−0.015
YE 1.0000.0770.140*−0.0180.173*−0.0610.0020.0510.096
LSUB 1.0000.528*−0.202*0.131*−0.1090.0550.187*0.027
LFOREIGN 1.000−0.0640.058−0.0780.1020.090−0.116
ROA 1.000−0.490*0.457*−0.403*−0.196*−0.007
LOSSDUM 1.000−0.233*0.356*0.198*0.112
DUMZ 1.000−0.183*−0.093−0.084
OPIN 1.0000.1040.046
PCON 1.0000.118
BUMDUM 1.000
Table 5. 
Regression Results
Dependent variable: natural log of audit fees (LAF)
 Predicted SignModel 1 Financial Crisis Sample (N= 275) CoefficientModel 2 Pooled Analysis (N= 740) Coefficient
  1. ***p < 0.01; **p < 0.05; *p < 0.1. The asterisks indicate significance levels in a 1-tailed test except for the intercept term (2-tailed).

  2. The above regression results include controls for eight industries with dummy variables (not reported in the table).

  3. LAF= natural logarithm of audit fees

  4. SIZE= natural logarithm of total assets

  5. DA= book value of long-term debt to total assets

  6. CURRAT= current assets to total assets

  7. LIQ= current assets to current liabilities

  8. BIG= indicator variable, 1 for big audit firms

  9. YE= indicator variable, 1 for fiscal year ending 31 December

  10. LSUB= natural logarithm of the number of subsidiaries

  11. LFOREIGN= natural logarithm of the number of foreign subsidiaries

  12. ROA= net income to total assets

  13. LOSSDUM= indicator variable, 1 for loss in the last year

  14. DUMZ= indicator variable, 1 for high Altman Z-score (>2.073)

  15. OPIN= indicator variable, 1 for modified audit opinion

  16. PFC= indicator variable, 1 for pre–financial crisis sample

  17. CAPCTRL= indicator variable, 1 for capital controls sample

  18. PCON= indicator variable, 1 for politically connected firms

Intercept?−1.751***−1.902***
Control Variables
 SIZE+0.451***0.440***
 DA+−0.001−0.001
 CURRAT+1.046***0.975***
 LIQ−0.052**−0.065***
 BIG+0.181**0.216***
 YE0.129*0.083***
 LSUB+0.041***0.050***
 LFOREIGN+0.031***0.026***
 ROA−0.330***−0.260***
 LOSSDUM+−0.187−0.065
 DUMZ−0.220**−0.165***
 OPIN+−0.021−0.058
 PFC? −0.027
 CAPCTRL? 0.067
Experimental Variables
 PCON+0.251***0.270***
 PFC * PCON −0.343***
 CAPCTRL * PCON −0.243*
 Adj. R2 0.6160.583

Table 5 also reports the results for the cross-sectional pooled regression model (model 2) for the three periods (N= 740). The purpose of the model is to test if audit fees for PCON firms increased in the FC period and then declined in the CAPCTRL period. To test the relationships, two dummy variables are included in the regression: PFC and CAPCTRL, with the financial crisis year being treated as the “referent” or “default” dummy variable. To test if the fees change, two interaction terms are included in the regressions: PFC*PCON and CAPCTRL*PCON. In order to interpret the dummy coefficients, the following terms are of interest:

image

For non-PCON firms, with the FC period as the reference year, α1 and α2 represent the fee differences of the PFC period and the CAPCTRL period, respectively. The respective estimates of α1 and α2 are −0.027 and 0.067, and both of them are insignificant, thus suggesting that, for non-PCON firms, there is no significant difference among the three periods.

For the FC period, α3 represents the fee difference between PCON and non-PCON firms. The estimate is positive and significant (0.270, p < 0.001, 1-tailed), suggesting PCON firms had to pay an average fee premium of 31% more than non-PCON firms in the FC period. Further, the interactions are both negative and significant (PFC*PCON=−0.343, p < 0.001, 1-tailed; CAPCTRL*PCON=−0.243, p < 0.06, 1-tailed), suggesting that, relative to the FC period, audit fees for PCON firms are lower in the PFC and the CAPCTRL periods. More specifically, for PCON firms, the parameter −0.343 suggests that the average audit fee is 29% lower in the PFC period and the parameter −0.243 suggests that the average audit fee is 21.6% lower after capital controls are introduced relative to the FC period. Since the interaction terms in the regression model reveal that the effects of PCON are different among periods, they need to be examined separately.

The sum of α3 and α4 represents the fee difference between PCON and non-PCON firms in the PFC period. The estimate of (α34) is −0.073 (0.270 − 0.343) and is insignificant. Finally, the sum of α3 and α5 represents the fee difference between PCON and non-PCON firms in the CAPCTRL period. The estimate of (α35) is 0.027 (0.270 − 0.243) and is insignificant. The direction and levels of significance for all the control variables are similar to those results in model 1. Thus, these results provide some evidence that audit fees for PCON firms went up during the year of the financial crisis and then declined in the period of capital controls, which is consistent with H2. The relationship is depicted in Figure 1.

Figure 1.—.

Changes in audit fees during the pre–financial crisis, financial crisis, and capital controls periods for PCON versus non-PCON firms.

PFC= pre–financial crisis sample

FC= financial crisis sample

CAPCTRL = capital controls sample

None of the variance inflation factors (VIF, not reported here) for any of the variables in the regressions exceed 10, suggesting that multicollinearity is not a problem. The results remain qualitatively the same when White-corrected t-statistics (White [1980]) are computed.

5.1 further extensions

Several other tests are conducted to extend the basic findings in this paper.

5.1.1. Ethnicity Another important element which is unique in the Malaysian institutional structure is the role of ethnicity in the capital market. J&M also consider whether ethnically favored (Bumiputera) firms were associated with lower stock returns during the crisis since the government, and Mahathir, had publicly stated their support for Bumiputera businesses.20 Further, the government also helped Bumiputera firms after the imposition of capital controls (Perkins and Woo [2000]). J&M find that a firm being ethnically favored is not significant and does not have a large effect on the political connections coefficient. They conclude that “political favoritism, and not simply ethnicity was the more important factor in determining the fortunes of Malaysian firms during this period” (p. 372). To test if ethnicity affects the PCON coefficients in the fee regressions, a new variable “Bumiputera” is included in all the regressions reported in Table 5 and the results are presented in Table 6. The Kuala Lumpur Stock Exchange Annual Companies Handbook [1996–1998] is examined to identify whether firms are ethnically favored. This handbook identifies how much of the ownership of each firm falls into the following categories: Bumiputera, non-Bumiputera, foreign, or government. As only 86% of the sample firms are identified in the handbook, the sample size is reduced to accommodate the ethnicity variable in the regressions.

Table 6. 
Regression Results with Bumiputera Ownership
Dependent variable: natural log of audit fees (LAF)
 Predicted SignModel 1 Financial Crisis Sample (N= 235) CoefficientModel 2 Pooled Analysis (N= 640) Coefficient
  1. ***p < 0.01; **p <0.05; *p < 0.1. The asterisks indicate significance levels in a 1-tailed test except for the intercept term (2-tailed).

  2. The above regression results include controls for eight industries with dummy variables (not reported in the table).

  3. LAF= natural logarithm of audit fees

  4. SIZE= natural logarithm of total assets

  5. DA= book value of long-term debt to total assets

  6. CURRAT= current assets to total assets

  7. LIQ= current assets to current liabilities

  8. BIG= indicator variable, 1 for big audit firms

  9. YE= indicator variable, 1 for fiscal year ending 31 December

  10. LSUB= natural logarithm of the number of subsidiaries

  11. LFOREIGN= natural logarithm of the number of foreign subsidiaries

  12. ROA= net income to total assets

  13. LOSSDUM= indicator variable, 1 for loss in the last year

  14. DUMZ= indicator variable, 1 for high Altman Z-score (>2.073)

  15. OPIN= indicator variable, 1 for modified audit opinion

  16. PFC= indicator variable, 1 for pre–financial crisis sample

  17. CAPCTRL= indicator variable, 1 for capital controls sample

  18. PCON= indicator variable, 1 for politically connected firms

  19. BUMDUM= indicator variable, 1 for Bumiputera-controlled company in which 50% or more of the equity is held by Bumiputera shareholders and institutions

Intercept?−1.866***−1.781***
Control Variables
 SIZE+0.462***0.433***
 DA+0.0140.005
 CURRAT+1.029***0.908***
 LIQ−0.023−0.043**
 BIG+0.152*0.210***
 YE0.0500.017
 LSUB+0.042**0.050***
 LFOREIGN+0.033***0.030***
 ROA−0.301**−0.259***
 LOSSDUM+−0.217*−0.089
 DUMZ−0.303***−0.223***
 OPIN+−0.012−0.048
 PFC? −0.027
 CAPCTRL? 0.097*
Experimental Variables
 PCON+0.260**0.308***
 PFC * PCON −0.369***
 CAPCTRL * PCON −0.281**
 BUMDUM+0.131*0.101**
 Adj. R2 0.6100.573

Table 6 reports the results for all the regressions. The results for all the experimental variables are similar and significant, as those reported in Table 5. The directions and significance levels of the control variables are also similar. Ethnicity is significant and positive (0.131 and 0.101, respectively) for both models, thus suggesting that firms with Bumiputera ownership are associated with higher audit risks and thus higher audit fees. While these results are not consistent with J&M, they are, however, consistent with a prior study by Eichenseher [1995], who finds that Bumiputera firms are associated with higher audit fees than non-Bumiputera firms. A possible reason for this result (and the result in this study) is that Chinese firms (as opposed to Bumiputera-controlled firms) with highly concentrated managerial ownership, family networks, and business alliances are likely to have lower agency costs and therefore lower audit risks and audit fees.

These results suggest that, while the market does not react to ethnicity (based on J&M), auditors associate higher levels of risks to these firms, underlying the notion that, while various components of the political economy move together, they are not “lock step” and that various market participants may react differently to certain institutional features. However, the issue of why ethnicity is recognized by the auditors, and not by market participants (in terms of market returns), as constituting higher risks, remains an interesting issue for future research.

5.1.2. Anwar and Mahathir Factor In September 1998, Anwar, the Deputy Prime Minister, was sacked, and it may be useful to evaluate if the results for the FC sample are affected by Anwar-connected firms (N= 12). To do this, a dummy variable to control for Anwar firms is included in the regressions. The results (not reported here) show that the coefficient for PCON increases to 0.292 (from 0.251 in Table 5) and remains significant (p < 0.02, 1-tailed) while the coefficient for the Anwar variable is negative (−0.109) and insignificant. The directions and significance levels of all the control variables remain the same. The results are similar for the pooled analysis. Since J&M suggest that Mahathir-connected firms benefited more from the capital controls, a dummy variable for Mahathir-connected firms is also included in the regressions together with the “Anwar factor” dummy for all the regressions. The results reported in Table 7 show that for model 1 the coefficient for PCON further increases to 0.348 (from 0.260 in Table 6) and is significant (p < 0.02, 1-tailed), while the coefficients for Anwar and Mahathir are negative and insignificant. Model 2 shows that the results are similar for the pooled analysis. Finally, a regression is also run without the PCON variables and replaced with the Anwar and Mahathir “factor” as dummy variables. The results (not reported) show that the coefficients for both the Anwar variable and the Mahathir variable are positive (0.139 and 0.086, respectively) and insignificant. The results are similar for the pooled analysis. Overall, the results suggest that regardless of Anwar or Mahathir connections, PCON firms are perceived as having higher audit risks than non-PCON firms.

Table 7. 
Regression Results with Controls for Anwar- and Mahathir-Connected Firms
Dependent variable: natural log of audit fees (LAF)
 Predicted SignModel 1 Financial Crisis Sample (N= 235) CoefficientModel 2 Pooled Analysis (N= 640) Coefficient
  1. ***p < 0.01; **p < 0.05; *p < 0.1. The asterisks indicate significance levels in a 1-tailed test except for the intercept term (2-tailed).

  2. The above regression results include controls for eight industries with dummy variables (not reported in the table).

  3. LAF= natural logarithm of audit fees

  4. SIZE= natural logarithm of total assets

  5. DA= book value of long-term debt to total assets

  6. CURRAT= current assets to total assets

  7. LIQ= current assets to current liabilities

  8. BIG= indicator variable, 1 for big audit firms

  9. YE= indicator variable, 1 for fiscal year ending 31 December

  10. LSUB= natural logarithm of the number of subsidiaries

  11. LFOREIGN= natural logarithm of the number of foreign subsidiaries

  12. ROA= net income to total assets

  13. LOSSDUM= indicator variable, 1 for loss in the last year

  14. DUMZ= indicator variable, 1 for high Altman Z-score (>2.073)

  15. OPIN= indicator variable, 1 for modified audit opinion

  16. PFC= indicator variable, 1 for pre–financial crisis sample

  17. CAPCTRL= indicator variable, 1 for capital controls sample

  18. PCON= indicator variable, 1 for politically connected firms

  19. ANWAR= indicator variable, 1 for Anwar-controlled companies

  20. MAHATHIR= indicator variable, 1 for Mahathir-controlled firms

  21. BUMDUM= indicator variable, 1 for Bumiputera-controlled company in which 50% or more of the equity is held by Bumiputera shareholders and institutions

Intercept?−1.919***−1.784***
Control Variables
 SIZE+0.465***0.433***
 DA+0.0140.003
 CURRAT+1.039***0.914***
 LIQ−0.020−0.043**
 BIG+0.1250.211***
 YE0.0610.012
 LSUB+0.044**0.050***
 LFOREIGN+0.033***0.030***
 ROA−0.292**−0.257***
 LOSSDUM+−0.222*−0.081
 DUMZ−0.304***−0.224***
 OPIN+0.002−0.048
 PFC? −0.027
 CAPCTRL? 0.095*
Experimental Variables
 PCON+0.348**0.307***
 PFC * PCON −0.370**
 CAPCTRL * PCON −0.279**
 ANWAR?−0.143−0.057
 MAHATHIR?−0.1820.068
 BUMDUM+0.132*0.100**
 Adj. R2 0.6080.572

5.2 robustness tests

The following tests are conducted to check the robustness of the main results that PCON firms paid higher fees during the financial crisis and lower fees after the capital controls were introduced.

5.2.1. Financial Distress Since it is possible that PCON might be picking up the effects of firms with losses, the regressions (as in Table 5) are rerun after deleting all loss-making firms. The results (not reported here) show that the PCON coefficient is 0.198 and significant (p < 0.04, 1-tailed) for model 1. For the pooled sample (N= 621) the PCON coefficient is 0.256 and significant (p < 0.02, 1-tailed). The PFC * PCON coefficient is −0.380 and significant (p < 0.01, 1-tailed) and the CAPCTRL * PCON coefficient is −0.276 and significant (p < 0.08, 1-tailed). The qualitative nature of the results does not change when the Bumiputera dummy is included in the regressions.

5.2.2. Alternative Model Using Audit Fees Deflated by Total Assets Following Simunic [1980], an alternative model with audit fees deflated by total assets is used as the dependent variable with the same control variables. The results (not reported here) show that PCON is positive (0.095) and significant (p < 0.001, 1-tailed) for the FC period (model 1). The adjusted R2 for the model is 0.305. Similarly, for the pooled cross-sectional model, the coefficient for PCON is positive (0.106) and significant (p < 0.001, 1-tailed) while the PFC * PCON interaction term is negative (−0.161) and significant (p < 0.001, 1-tailed), and the coefficient for CAPCTRL * PCON is also negative (−0.097) and significant (p < 0.03, 1-tailed). The adjusted R2 for the model is 0.231. The levels of significance of these results remain unchanged after conducting the various sensitivity tests.

5.2.3. Earnings Management Since PCON firms experienced significant stock price declines in the FC period, it is possible that they may be associated with more aggressive earnings management. As such, the PCON coefficient may be picking up the effects of aggressive earnings management in the model for the FC sample. An attempt is also made to control for earnings management in the regressions by including the absolute value of discretionary accruals as an additional control variable.21 Discretionary accruals are estimated following Kothari, Leone, and Wasley [2005] by including a lagged ROA in the accruals regression to control for firm performance (see Ashbaugh, LaFond, and Mayhew [2003] for the procedure for calculating the discretionary current accruals (REDCA)). The results (not reported here) for the experimental variables in these models (as in Table 5) remain qualitatively similar for all the regressions. For model 1 (N= 259),22 the PCON coefficient is 0.267 and significant (p < 0.01, 1-tailed). Similarly, for the pooled cross-sectional model (similar to model 2 in Table 5) the PCON coefficient is 0.289 and significant (p < 0.01, 1-tailed). The interaction coefficients are both negative and significant (PFC * PCON=−0.350, p < 0.01, 1-tailed; CAPCTRL * PCON=−0.238, p < 0.07, 1-tailed). The coefficients for REDCA in both models are negative and insignificant. The results (not reported here) remain qualitatively the same when the regression models reported in tables 6 and 7 are run.23

5.3 limitations

In evaluating these results, several limitations should be noted. First, the study relies on J&M and Gomez and Jomo [1997] for identification of the PCON firms. It is possible that other firms are also connected to the government but the linkages are not clear and hence the firms identified as PCON may not be complete.24 Second, it is possible, though unlikely, that the higher fees for PCON firms are because auditors are extracting more rents from PCON firms. Unless there are some other compelling reasons, it is safe to assume that auditors (especially the traditional Big 6, who audit about 80% of the firms in the sample) act ethically. Besides, if the rent argument has merit, there should also be a significant result for the PCON coefficient for the PFC sample and the CAPCTRL sample. Third, it is possible that some of the variables, such as size and proportion of foreign subsidiaries, could affect the relationship between PCON and audit fees. However, when the interactions between PCON and each of these variables are included in the regressions none of the coefficients for the interaction terms are significant. Fourth, while the audit fee model is used to test the various hypotheses, this paper's contribution to the “pure” audit pricing literature is limited since the institutional and legal setting is different from prior audit pricing studies which are based largely on the U.S. and Australian settings. Related to this limitation is the fact that the audit fee model based on prior U.S. and Australian studies may be misspecified in the Malaysian setting. It is, however, worth noting that apart from the institutional setting most of the other variables in the audit fee model are firm specific (e.g., size, debt, number of subsidiaries etc.) and have been shown in prior studies to be theoretically related to audit fees (see Simunic [1980], Francis [1984]), thus mitigating some of the concerns regarding the model misspecification. Finally, this paper adopts a supply-side perspective and does not test or consider the demand-side perspective for audit fees.25

6. Conclusion

This paper adds to the stream of literature that aims to provide a deeper understanding of the dynamic relationship among political economy, macroeconomic changes, and accounting institutions. It examines how auditors, an important component in the financial reporting process, react to firms with political connections under different macroeconomic conditions. The first hypothesis predicts that, during the Asian financial crisis, audit fees of PCON firms increased more than non-PCON firms as a result of the higher likelihood of misstated financial statements. The financial crisis and loss of subsidies for PCON firms resulted in higher assessments of audit risks, higher audit effort, and higher audit fees, ceteris paribus. However, the imposition of capital controls in September 1998, which was a screen to provide financial subsidies and other assistance for favored PCON firms, reduced the potential motivation of managers of PCON firms to misstate financial statements. As a result, auditors reduced the levels of assessed audit risk relative to the FC period for PCON firms. The results provide support for both hypotheses and are consistent with the conclusions in the J&M (p. 380) study that the financial crisis implies that “previously favored firms would lose valuable subsidies, and the imposition of capital controls indicated that these subsidies would be restored for some firms.” Several sensitivity tests also give confidence that the results are robust.

These results are consistent with the general idea that institutional arrangements do matter in macroeconomic dynamics (Olson [1982], Blanchard [2000], J&M). More recently, a stream of cross-country studies shows that political economy does matter in financial reporting. For example, Bushman, Piotroski, and Smith [2004] and Bushman and Piotroski [2006] show that firms in countries with lower state ownership are associated with better financial reporting practices. This study extends this line of work and shows how another participant in the capital market, auditors, reacts to these institutional arrangements and financial statement preparers' incentives during a financial crisis. Further, the results are also consistent with the idea that capital controls are associated with cronyism that benefits PCON firms (Morck, Strangeland, and Yeung [2000], J&M).

Appendices

APPENDIX A

Summary of the Audit Process and Auditors' Risk Assessment (see Arens, Elder, and Beasley[2003, p. 356–359])

The overall objective of an audit is to collect appropriate evidence in order for the auditor to issue an appropriate audit opinion (or inability to form an opinion, i.e., a disclaimer). This is done by following an audit process, which is a methodology for organizing an audit to ensure that the evidence collected is both sufficient and competent. Four phases may be identified in a typical audit process.

  • Phase 1: Plan and design an audit approach. This includes client acceptance, understanding client business, initial assessment of audit risk, inherent risk and internal control risk, and developing an overall audit plan. During this phase the auditor assesses the risk of misstatements in the financial statements, which can be reduced if the client has effective controls and accordingly the auditor will decide to collect less audit evidence resulting in lower levels of audit effort.

  • Phase 2: Perform tests of controls and substantive tests of transactions. This includes performance of tests of controls and substantive tests of transactions (i.e., verification of the monetary amounts of transactions by evaluating the client's recording of transactions). Based on these tests the auditor assesses the likelihood of financial misstatements. To justify the reduction in audit effort as a result of effective internal controls, the auditor must conduct tests of controls. Both the tests for controls and substantive tests are often conducted together. The important outcome of these two procedures is an assessment of the likelihood of misstatements in the financial statements. These could be low, medium, or high (or unknown).

  • Phase 3: Perform analytical procedures and tests of details of balances. The extent of these procedures (and hence audit effort) depends on the auditors' assessment of the likelihood of financial misstatements. The higher the likelihood of misstatements in the financial statements, the more the audit effort is in terms of analytical procedures and tests of details of balances. Analytical procedures such as calculating ratios and making comparisons of the ratios with prior years, with similar firms, occur after the client has finished preparing its financial statements. The outcome of the analytical procedures could also affect the extent of the tests of balances. These procedures could occur two to three months before the year-end.

  • Phase 4: Complete the audit and issue an audit opinion. This includes review for contingent liabilities and subsequent events, accumulation of final evidence, evaluation of results, and issue of audit report.

APPENDIX B

Malaysian Firms That Have an Identifiable Connection with High-Ranking Political Figures

Company NamePrimary Connected Major Shareholder/ DirectorPrimary Political Connection
  1. Source: Johnson and Mitton [2003].

ADVANCE SYNERGY BHDAhmad Sebi Abu BakarDaim, Anwar
ANTAH HOLDINGS BHDNegeri Sembilan RoyaltyMahathir
AOKAM PERDANA BHDSamsudin Abu HassanDaim
BERJAYA GROUP BHDVincent Tan Chee YiounDaim
CONSTRUCTION AND SUPPLIES HOUSEJoseph Ambrose Lee, Abdul Mulok Awang DamitDaim
CSM CORPORATION BHDBasir Ismail, Samsudin Abu HassanDaim
CYCLE & CARRIAGE BINTANG BHDBasir IsmailDaim
FABER GROUP BHDUMNOUMNO
GADEK (MALAYSIA) BHDYahya Ahmad, Nasaruddin JalilAnwar, Mahathir
GOLDEN PLUS HOLDINGS BHDIshak Ismail, Mohamed Sarit Haji YusohAnwar
GRANITE INDUSTRIES BHDSamsudin Abu HassanDaim
HICOM HOLDINGS BHDYahya AhmadAnwar, Mahathir
HO HUP CONSTRUCTION COMPANY BHDHalim SaadDaim
HONG LEONG INDUSTRIES BHDQuek Leng ChanAnwar
HUME INDUSTRIES (MALAYSIA) BHDQuek Leng ChanAnwar
JT INTERNATIONAL BHDWan Azmi Wan HazmahDaim
KAMUNTING CORPORATION BHDT. K. LimDaim
KFC HOLDINGS (MALAYSIA) BHDIshak IsmailAnwar
KINTA KELLAS PLCHalim SaadDaim
KRETAM HOLDINGS BHDUMNO Youth, Wan Azmi Wan HamzahDaim
KUMPULAN FIMA BHDBasir IsmailDaim
LAND & GENERAL BHDWan Azmi Wan HamzahDaim
MAGNUM CORPORATION BHDT. K. LimDaim
MALAYSIAN RESOURCES CORPORATIONUMNO, Wan Azmi Wan HamzahDaim, Anwar
METROPLEX BHDDick ChanUnspecified
NANYANG PRESS (MALAYA) BHDQuek Leng ChanAnwar
O.Y.L. INDUSTRIES BHDQuek Leng ChanAnwar
PRIME UTILITIES BHDAhmad Sebi Abu BakerDaim, Anwar
RENONG BHDHalim SaadDaim
SAPURA TELECOMMUNICATIONS BHDShamsuddin bin Abdul KadirMahathir
SETRON (MALAYSIA) BHDPenang Bumiputera Foundation, Kamaruddin JaafarAmwar
SISTEM TELEVISION MALAYSIA BHDUMNO CompaniesUMNO
STAR PUBLICATIONS (MALAYSIA) BHDVincent Tan Chee YiounDaim
TANJONG PLCT. Ananda KrishnanMahathir
TIME ENGINEERING BHDHalim SaadDaim
TONGKAH HOLDINGS BHDMokhzani MahathirMahathir
UNIPHONE TELECOMMUNICATIONS BHDShamsuddin bin Abdul KadirMahathir
UNITED ENGINEERS (MALAYSIA) BHDHalim SaadDaim
UNITED PLANATIONS BHDBasir IsmailDaim
WEMBLEY INDUSTRIES HOLDINGS BHDIshak IsmailAnwar
WIJAYA BARU GLOBAL BHDTing Pek Khiing, Robert TanDaim, Mahathir
YTL CEMENT BHDYeoh Tiong LayMahathir
YTL CORPORATION BHDYeoh Tiong LayMahathir

Footnotes

  • 1

    In general terms, political economy represents a range of institutional arrangements that capture important relations between the government and the economy (Bushman, Piotroski, and Smith [2004]).

  • 2

    This practice may be described as “aggressive” financial reporting as opposed to “conservative” financial reporting.

  • 3

    Several prior studies have examined the effects of institutional factors on accounting property numbers across different countries (e.g., Ali and Huang [2002], Hung [2001], Leuz, Nanda, and Wysocki [2003]).

  • 4

    In some cases, these politically connected firms are also state owned.

  • 5

    “Financial reporting quality” is an elusive concept particularly in view of the multiplicity of users and methods of operationalizing the concept (Ball, Robin, and Wu [2003]).

  • 6

    While Malaysia, the United States and the United Kingdom are common law jurisdictions, the development of company law in Malaysia follows more the pattern of the United Kingdom given the fact that Malaysia was a former British colony. Similarly, Malaysian accounting and auditing standards closely follow the standards in the United Kingdom and other commonwealth countries such as Australia and New Zealand. Malaysia is also represented on the International Accounting Standards Committee. For more information on international accounting comparisons, see Deloitte Touche Tohmatsu [1999]. Most of the audits are conducted by the Big 6 and since British Common Law applies, auditors face litigation risk in the event of audit failures. However, the level of litigation in Malaysia is similar to Australia or the United Kingdom and not as high as that in the United States.

  • 7

    See Ball, Robin, and Wu [2003], Appendix B for a summary of Malaysian accounting standards.

  • 8

    Rajan and Zingales[2003, p. 237–238] suggest that governments are likely to use large corporations to help redress any social inequities in the country and, in this case, the wide gap between the wealth of the Chinese and the Bumiputeras.

  • 9

    The leading component party in the ruling coalition, the Barisan Nasional (National Front), is the UMNO. Two other principal (but lesser) constituent members of the Barisan Nasional are the Malaysian Chinese Association (MCA) and the Malaysian Indian Congress (MIC). While all the component parties are involved in the corporate sector, UMNO is by far the most aggressive corporate player. Through its holding companies, UMNO acquired equity in several companies and embarked on an aggressive “conglomeratization” program facilitated by “extensive patronage practiced by the ruling elite” (Gomez[1994, p. 8]).

  • 10

    The majority of the Board of Directors (including the CEO) of these companies are Bumiputeras or indigenous Malays, and UMNO or its nominees holding more than 20% of the shares. The control of companies is achieved through a system of trusteeship, holding companies, interlocking stock ownership, interlocking directorships, and mergers and acquisitions (see Gomez [1990] for discussion).

  • 11

    Auditors of PCON firms are expected to raise their assessments of inherent risk in both phases I and III of the audit process, and expend more effort and hence charge higher fees, ceteris paribus. Moreover, in terms of phase III (i.e., after compliance and transactions tests), it is also likely that during the financial crisis, auditors classify PCON firms as being associated with a “high likelihood of misstatements.”

  • 12

    See J&M for more details on the different types of subsidies.

  • 13

    Non-audit fees are not included as a control variable since they are not disclosed.

  • 14

    See table A1 in J&M's study for a list of the 65 nonfinancial firms identified as having political connections. The political figures that the firms are identified with are also shown.

  • 15

    The last two companies are identified as Mahathir connected based on a report in Fletcher and Poh [1997], which suggests that Yeoh Sock Ping, the son of Yeoh Tiong Lay, had good connections with Mahathir. Yeoh Sock Ping is the managing director of YTL Corporation, which owns 53.2% of YTL Cement.

  • 16

    Two large firms, BDO Binder and Kassim Chan, had larger market shares than Deloitte Touché Tomatsu and, as such, are included as high-quality auditors based on the “quasi rents” argument (see DeAngelo [1981]). Excluding these two firms in the “BIG” category resulted in weaker coefficients for the “BIG” variables in both regression models but the coefficients of the experimental variables are qualitatively the same.

  • 17

    Altman Bankruptcy Score: The model incorporates five weighted financial ratios taken from Altman [1993].1. Working capital/total assets (X1)2. Retained earnings/total assets (X2)3. Earnings before taxes + interest/total assets (X3)4. Market value of equity/total liabilities (X4)5. Net sales/total assets (X5)The above ratios are used to compute the Altman Z-score using the following model:

    image

    Higher scores are associated with less financial distress.

  • 18

    When the square root instead of the log of the number of subsidiaries is used, the significant results for the experimental variables of all the tests still hold.

  • 19

    The premiums are only reported for the results in Table 5.

  • 20

    See Gomez and Jomo [1997] for a description of Bumiputera ownership.

  • 21

    The tests use a dummy variable for discretionary accruals (1 = absolute discretionary accruals above the median values).

  • 22

    The sample is reduced due to missing values for calculating discretionary accruals.

  • 23

    The coefficients and significance levels for the interaction coefficients increase marginally when the Bumiputera variable is included in the regression.

  • 24

    Such a misclassification is likely to work against finding significant results.

  • 25

    See Bedard and Johntsone [2004] for a discussion and support for the supply-side argument.

Ancillary