Economic Modelling 53 (2016) 156–165 Contents lists available at ScienceDirect Economic Modelling journal homepage: www.elsevier.com/locate/ecmod The impact of earnings management on the performance of ASEAN banks Yueh-Cheng Wu a, Irene Wei Kiong Ting b, Wen-Min Lu c,⁎, Mohammad Nourani d, Qian Long Kweh e a Department of Cultural Vocation Development, National Taipei University of Technology, 1, Sec. 3, Zhongxiao E. Rd., Taipei 10608, Taiwan Department of Finance and Economics, College of Business Management and Accounting, Universiti Tenaga Nasional, Sultan Haji Ahmad Shah Campus, 26700 Muadzam Shah, Pahang, Malaysia c Department of Financial Management, National Defense University, No. 70, Sec. 2, Zhongyang North Rd., Beitou, Taipei 112, Taiwan d Faculty of Economics and Administration, University of Malaya, Jalan Universiti, 50603 Kuala Lumpur, Malaysia e Department of Accounting, College of Business Management and Accounting, Universiti Tenaga Nasional, Sultan Haji Ahmad Shah Campus, 26700 Muadzam Shah, Pahang, Malaysia b a r t i c l e i n f o Article history: Accepted 20 November 2015 Available online xxxx Keywords: Performance evaluation Data envelopment analysis Earnings management Liberalization ASEAN a b s t r a c t Southeast Asian ﬁnancial liberalization policies have enthused both performance evaluation (a pro) and earnings management (a con). Using a sample of ASEAN commercial banks for the period 2007–2014, this study decomposes their banking performance into managerial and proﬁtability efﬁciencies. An efﬁciency analysis reveals that Singaporean banks obtained the highest overall and proﬁtability efﬁciencies, while Bruneian banks had the lowest rates of banking performance. In the stage of managerial efﬁciency, the most inefﬁcient banks are those of the Philippines, whereas the greatest level is related to Malaysian banks. A frontier projection analysis suggests that Singaporean banks and Malaysian banks are generally more efﬁcient in managing their expenditures and longterm assets in generating income in the long run. With respect to the con, a regression analysis indicates that loan loss provisions are negatively related to banking performance. Overall, it is advisable that policy makers with oversight function should promote performance evaluation from a multidimensional perspective, and keep an eye on estimates of loan loss provisions at banks over years because increases/decreases in loan loss provisions mean decreases/increases in net income or return on assets. © 2015 Elsevier B.V. All rights reserved. 1. Introduction According to a report by the ADB (2011), Asian emerging economies hold approximately half of the world’s total exchange reserves, and the whole region is also the major exporter and importer of capitals. This phenomenon attests the underdeveloped intra-regional ﬁnancial system to efﬁciently channel surplus funds. Consequently, effective regional coalition and cooperation are needful in dealing with this immaturity. Among all, the Association of Southeast Asian Nations (ASEAN)1 has succeeded to create a strong integration among its members of Southeast Asian nations. Since the 1970s, many emerging countries in Asia have experienced substantial rectiﬁcations in their ﬁnancial liberalization policies. Speciﬁcally, developing countries in Southeast Asia have experienced various forms of ﬁnancial liberalizations in promoting efﬁcient allocation of resources to achieve greater economic growth during ⁎ Corresponding author. E-mail addresses: [email protected] (Y.-C. Wu), [email protected] (I. Wei Kiong Ting), [email protected] (W.-M. Lu), [email protected] (M. Nourani), [email protected] (Q.L. Kweh). 1 The ASEAN was established on August 1967 by the signing agreement of ﬁve countries, namely, Indonesia, Malaysia, the Philippines, Singapore, and Thailand. Since then, membership has expanded to include Brunei, Cambodia, Laos, Myanmar, and Vietnam, constituting the ten Member States of ASEAN. http://dx.doi.org/10.1016/j.econmod.2015.11.023 0264-9993/© 2015 Elsevier B.V. All rights reserved. the last few decades. Subsequently, a number of institutional reforms occurred following Asian ﬁnancial crisis and global ﬁnancial crisis erupted in 1997 and 2007–2008, respectively, to strengthen the regulatory and supervisory frameworks. As a result of ﬁnancial liberalization policies, the international capital mobility has increased between partner countries through different international agreements such as ASEAN (Lim, 2005). However, it might be a blessing in disguise or a curse. On the one hand, substantial capital ﬂows into the host countries. In addition, the increased competition as the consequence of liberalization policies stimulates ﬁrms to put more cautions on their activities such as cost management, risk monitoring, and resource allocation (Gardener et al., 2011). That is, opening up the economy to international investors leads to higher efﬁciency of ﬁrms by means of intensifying the competition within a domestic market. According to the seminal works of McKinnon (1973) and Shaw (1973), ﬁnancial liberalization yields higher economic growth through increasing interest rate level, which enhances the competition among the market players, while the allocation of resources are well realized. Therefore, ﬁnancial liberalization is likely to increase higher savings, which eventually nurtures economic growth by ameliorating the investment quantity and quality, i.e. efﬁcient allocation of resources (Reinhart and Tokatlidis, 2003). This is the reason why evaluating efﬁciency becomes very important in recent Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 years where many of the emerging countries have undergone full liberalization policies or are under the process of being liberalized, i.e. partial liberalization. On the other hand, liberalization enhances competition, which dampens ﬁrms’ proﬁtability. Therefore, underperforming ﬁrms will be expelled from the marketplace because lower proﬁtability increases the risk of bankruptcy (Baik et al., 2011; Becchetti and Sierra, 2003; Bolt et al., 2012). Put differently, this will create an incentive for managers to sham their corporate performance in order to attract investors. In fact, past studies prove that pressurized ﬁrms with high chance of bankruptcy are more inclined towards engaging “earnings management2 (EM)” practices (Beneish et al., 2012; García Lara et al., 2009). Therefore, the dramatic shift in liberalization policies in emerging countries has predisposed market players to manipulate the information about ﬁnancial reporting. It has been repeatedly reported that the intensiﬁed competition, as the result of liberalization and government deregulation, brings new opportunities for economic prosperity. However, the ﬂip side of the coin tells a different story. Internal managers are more prone towards hyperbolizing the ﬁrm’s performance to market participants within a competitive environment. Therefore, ﬁrms may report higher earnings compared to other rivals to attract investors. In a comprehensive review of the EM literature, Healy and Wahlen (1999) conclude that “the evidence is consistent with ﬁrms managing earnings to window-dress ﬁnancial statements prior to public securities’ offerings, to increase corporate managers’ compensation and job security, to avoid violating lending contracts, or to reduce regulatory costs or to increase regulatory beneﬁts” (p. 368). In their review, Healy and Wahlen (1999) emphasize that past studies have focused on whether EM exists and why, yet the empirical issue about the effect of earnings management practices on efﬁciency has not been well explored (p. 380). In this vein, the banking sector has been the centerpiece in the eyes of policy makers due to its invaluable importance to the economic development; hence, the sound and well-functioning banking sector is a potent engine of economic growth. Given the above discussion, this study aims to estimate the performance of ASEAN banking institutions and address the relationship between EM practices and banking performance. To measure banking performance, frontier efﬁciency analysis, which has received considerable attention from researchers, is more appropriate than a single-dimensional ratio analysis. Data envelopment analysis (DEA), introduced by the inﬂuential work of Charnes et al. (1978), has been coined as the most frequent frontier efﬁciency approach used by researchers, particularly among banking studies (Liu et al., 2013). The value of DEA lies in its ability to transform various performance aspects into a single efﬁciency score (Yeh, 1996) through evaluating the relative performance of a decision-making unit (DMU) compared to its peers or competitors operating within the same group (Liu et al., 2013). However, the traditional DEA models are not sufﬁcient to measure the banks’ complex production process because these models assume the system as a single black box that converts inputs to outputs (Chiu and Chen, 2009; Dong et al., 2014; Moradi-Motlagh and Babacan, 2015; Wang et al., 2014). Accordingly, the detailed sources of inefﬁciency could not be identiﬁed when applying the traditional DEA models. Therefore, we adopt dynamic network DEA (DN-DEA) to deal with inefﬁciencies of interacting divisions that are embedded inside the banks’ production process. The purpose of this study is twofold. First, we apply the newly developed DN-DEA model called dynamic network slacks-based measure (DNSBM), formulated by Tone and Tsutsui (2014), which deals with 2 Earnings management is deﬁned as the use of managerial judgment to manipulate the ﬁnancial reporting with the purpose of either inﬂuencing contractual outcomes which depend on the ﬁnancial reports or misleading the stakeholders (and investors) about the company’s performance (Healy and Wahlen, 1999). 157 multiple divisions connected by links of network structure within each period vertically and also combines the network structure by means of carry-over activities between two succeeding periods horizontally. The second important objective of the study is to investigate the inﬂuence of controversial EM practices on the divisional efﬁciency scores of ASEAN banking institutions. To measure EM practices, we follow Adams et al. (2009) to use the ratio of loan loss provisions to loans. As a robustness check, we also include the ratio of loan loss reserves to loans as another measure of EM practices. To the best of the authors’ knowledge, this study makes an early attempt by examining the effect of EM on the efﬁciency of ASEAN banking institutions. To articulate the contributions of our study, we would like to highlight two points. First, efﬁciency has become a contemporary major issue to ﬁnance sectors due to the increment of competition, globalization, technological innovation and increased deregulation (DangThanh, 2012). According to Quiggin (2011), the global ﬁnancial crisis in 2007–2008 had a major impact on the ﬁnancial systems of many countries. Therefore, it is crucial for banking sectors to evaluate their efﬁciency level in order to compare competitiveness and further enhance their productivity. As a result, this paper provides a direct economic contribution by estimating the managerial and proﬁtability efﬁciencies of ASEAN commercial banks in order to speed up countries’ ﬁnancial development and economic growth. Banks with well-functioning and efﬁcient ﬁnancial systems are less likely to be suffering ﬁnancially during ﬁnancial crises (Moradi-Motlagh and Babacan, 2015). In contrast, it calls for countries with banks’ low efﬁciency scores to increase banks’ ﬁnancial autonomy in order to face economy challenges. We hope to bridge the gap and shed some light on the literature, speciﬁcally in the ASEAN context. Second, banking institutions are not exempted from EM practices, rather they are also more susceptible to this upshot compared to nonﬁnancial organizations (Grougiou et al., 2014) due to their wideranging ﬁnancial products and complicated operation which lead to information opacity (Levine, 2004) and asymmetry (Mülbert, 2009). Moreover, as highlighted by Greenawalt and Sinkey (1988), the higher chance of earnings manipulation in banking institutions might be attributed to the subjective judgment that managers have to make in regard to expected reserves for losses. Speciﬁcally, during the period of high proﬁt, banks’ managers incline to smooth the earnings by recording more loan losses and vice versa (Cornett et al., 2009). While the theoretical and empirical literature supports the ampliﬁcation in both ﬁrm efﬁciency and EM practices followed by liberalization, the question remains that how EM could be observed from testing its relation to efﬁciency of banks. Hence, this study makes another contribution by examining the effects of earnings management on bank efﬁciency, which is lacking in the literature. The remainder of this paper unfolds as follows: Section 2 reviews the extant literature on banking efﬁciency studies. Section 3 describes the data collection and the methodology used. Section 4 presents the empirical ﬁndings and the discussion, and Section 5 concludes the paper. 2. Literature review 2.1. Efﬁciency studies in the ASEAN banking sector In this section, we try to map out the studies addressing the efﬁciency analysis of banking institutions in the ASEAN alliance. We note that the application of efﬁciency evaluation in ASEAN economies is very scant at the moment. Hence, we also focus our attention to East Asian studies of banking efﬁciency, where necessary, in order to provide a beﬁtting background of the subject matter. Laeven (1999) estimates the technical efﬁciency of East Asian banks for the period from 1992 to 1996. The input variables include interest expenses, labor expenses, other operating expenses and the output variables including loans and securities. The efﬁciency results were surprising since the scores were increasing or constant during the pre-crisis 158 Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 period in which one might expect a declining efﬁciency trend. The author, therefore, tries to shed some light on the risk-taking behavior of commercial banks. The ﬁndings assert that banks with low efﬁciency scores have endured the crisis due to taking fewer loans and thus less risk. The ﬁrst ASEAN banking efﬁciency paper appearing in the literature, authored by Karim (2001), is the analysis of scale and cost efﬁciencies using the stochastic cost frontier approach. The author uses the banking data for four countries during 1989 to 1996, including Indonesia, Malaysia, the Philippines, and Thailand, out of ten members due to data unavailability for the remaining countries. Using the intermediation approach, three inputs (wages and salaries, land, buildings, and equipment and interest on deposits) and ﬁve outputs (commercial and industrial loans, other loans, time deposits, demand deposits, and securities and investments) are selected for the efﬁciency analysis. His ﬁndings on proﬁt and cost efﬁciency suggest that Thai banks are the least inefﬁcient followed by Indonesian and Pilipino banks while Malaysian banks perform the best. The author binds the inefﬁciency of Indonesian and Pilipino banks to their restrictive regulatory systems. His results also support the consolidation policy where the larger banks incline towards higher cost efﬁciency. Consistently, Williams and Nguyen (2005) examine the proﬁt efﬁciency using stochastic frontier approach for the Southeast Asian commercial banks in the period of 1990 to 2003. The authors utilize the unbalanced dataset of 231 commercial banks from ﬁve countries (Indonesia, Korea, Malaysia, the Philippines, and Thailand) to create a common frontier. This is disputatious, however, the efﬁciency results can be controlled for signiﬁcant cross-country differences. Williams and Nguyen apply different country economic indicators to control for cross-border differences. Their key ﬁndings support the policy of bank privatization, which leads to higher proﬁt efﬁciency. Gardener et al. (2011) provide an empirical efﬁciency analysis of ﬁve selected ASEAN banking institutions, including Indonesia, Malaysia, the Philippines, Thailand, and Vietnam for the period of 1998 to 2004. The authors estimate the technical and cost efﬁciencies using DEA with two outputs (net loans and other earning assets) and three inputs (ﬁxed assets, deposits and personnel costs). Their key ﬁndings show that the efﬁciency of banks over the sample period has reduced interpreting the weak restructuring of post-1997 crisis. Malaysian and Vietnamese banks perform better in terms of technical and cost efﬁciency while Indonesia and Thailand possess the least technical and cost-efﬁcient banks in the post-crisis era. Furthermore, countries with higher economic growth rates tend to be more efﬁcient. In summary, the above literature of efﬁciency studies in the East Asian banking sectors put forward the importance of cross-country comparison in which the efﬁciency scores differ between countries and also before and after the crisis event. The restructuring of ﬁnancial institutions takes some time to be effectively implemented, and an immediate analysis might not be appropriate to judge the true inﬂuence of restructuring policies. In addition, due to limitation in data, the above studies fail to consider more countries into analysis in order to provide a more inclusive picture of ASEAN alliance. Therefore, our study aims to ﬁll the gap by addressing a more recent period in efﬁciency analysis as well as a broader group of ASEAN commercial banks. 2.2. Data envelopment analysis in the banking sector There are many advantages for using DEA as a performance evaluation technique: (i) it handles multiple input and output variables at the same time; (ii) it has unit invariance meaning that the units of input and output variables do not inﬂuence the analysis; (iii) it does not require the predetermination of the parameters; (iv) it provides the weights for input and output variables using a mathematical method which then suggests the areas for improvement; (v) it renders an objective analysis (Cooper et al., 1999; Lu et al., 2015; Sueyoshi and Sekitani, 2009). The literature on frontier efﬁciency methodology, particularly DEA,3 is fruitful with numerous research works focusing on methodological development, application-centered and both theory and application studies (see Cook and Seiford (2009) and Emrouznejad et al. (2008) for methodological and theoretical developments and see Liu et al. (2013) for a survey of application-embedded studies). Liu et al. (2013), who review high-ranked DEA papers published during 1978 through 2010, indicate that application-embedded papers account for nearly two-thirds of all published papers, and banking studies cover 10.3% of this category (the most popular ﬁeld). Since the invention of novel DEA by Charnes et al. (1978), the groundbreaking work of Sherman and Gold (1985), where the authors examine the operating efﬁciency of bank branches, paved the way for the application of DEA in banking sectors. Sherman and Gold’s argument about the uniqueness of DEA technique embraced by a number of banking researchers (Barth et al., 2013; Berg et al., 1993; Elyasiani and Mehdian, 1990; Parkan, 1987; Pasiouras, 2008; Rangan et al., 1988). Berger and Humphrey (1997), a survey-based study, and Thanassoulis (1999), an informative study, motivate the researchers by providing the potential areas that need to be addressed in the domain of banking efﬁciency and the scope for enhancing the role of DEA in banking, respectively. However, as mentioned before, the banks’ complex production process requires more sophisticated techniques to account for internal structures within the black box. Fortunately, following the pioneering work of Färe and Grosskopf (1996), who were the initiators to investigate the “black box”, many researchers developed new methodologies to overcome the shortcomings. While a rising number of studies pointing to the meaningfulness of decomposing the banks’ performance into sub-divisions (Avkiran, 2009; Lin and Chiu, 2013; Lo and Lu, 2006; Luo, 2003; Seiford and Zhu, 1999; Yang and Liu, 2012), the application of DN-DEA in banking is still in its embryonic stage (Avkiran, 2014a; Fukuyama and Weber, 2013; Kao and Liu, 2014; Wanke et al., 2014). For example, Avkiran (2014a) assesses the dynamic efﬁciency of Chinese commercial banks combined with network structure; Wanke et al. (2014) measure the efﬁciency of Brazilian banks using dynamic SBM; Kao and Liu (2014) propose a relational network model applied to a set of Taiwanese commercial banks; and Fukuyama and Weber (2013) provide an example of dynamic network DEA using a large sample of Japanese banks. Hence, our study contributes to the scarce literature of DN-DEA in banking, and it is the ﬁrst study to apply this technique in a cross-country sample. 3. Research design 3.1. Descriptions of the DEA model and data The efforts to test the hypothesis of this study focus on nine emerging economies: Brunei Darussalam, Cambodia, Indonesia, Lao People’s Democratic Republic, Malaysia, the Philippines, Singapore, Thailand, and Vietnam. These countries share a main trait, in which they are members of the ASEAN. Therefore, we argue that the sample banks in this study are by no means more inﬂuential in countries with larger bank populations. Note that our data screening process leaves us with no banks from Myanmar, which is also one of the ASEAN members. This is not surprising because information is sometimes lacking in emerging economies. All data are extracted from the Bankscope database for the period from 2007 to 2013. We collect ﬁnancial data for as many commercial banks as possible for each country and screen the initial dataset in the following ways. Firstly, we focus only on commercial banks, which have similar products and services, whereby each of the banks is treated as a DMU for the DEA analysis. Secondly, we eliminate banks with no 3 According to a comprehensive survery of frontier efﬁciency analysis in ﬁnancial institutions, mostly banking, by Berger and Humphrey (1997), DEA is the most frequently used approach for efﬁciency evaluation. Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 complete ﬁnancial data for the DEA analysis over the sample period. Finally, we remove banks with missing data for measuring the testing variable, viz. earnings management. As such, we have a balanced panel dataset that is made up of 138 commercial banks in each year, in particular: Brunei Darussalam (1), Cambodia (9), Indonesia (55), Lao People’s Democratic Republic (4), Malaysia (1), the Philippines (20), Singapore (8), Thailand (20), and Vietnam (20). Relying on the Bankscope database for our dataset, we highlight that the ﬁnal dataset for analysis does not reﬂect a market concentration problem; instead, it is about data screening based on data availability and selections of DEA and regression variables. In the selection of DEA variables, it should be noted that we examine the operating processes of banks using intermediation approach in line with prior studies4 (Avkiran, 2014a; Avkiran and Thoraneenitiyan, 2010; Bhattacharyya et al., 1997; Miller and Noulas, 1996; Sturm and Williams, 2004). The intermediation approach stresses on the situation whereby banks transform sources of production into outputs such as earnings assets and ultimately income in the nature of their business. The logical ﬂow is that banks input staff (personnel expenses) who incurs other operating expenses to produce intermediates including loans, deposits, and other earnings assets, which ultimately generate net interest income to the banks. In the network production process, ﬁxed assets such as buildings and ofﬁce equipment are used over long-term periods by staff at work, while liquid assets are considered as another inputted carry-over item that will generate net interest income together with the intermediates. In regards of the DEA analysis, we present Table 1 to summarize the descriptive statistics of variables used. Furthermore, we also perform correlation analysis for the variables. Table 2 shows that all correlation coefﬁcients are signiﬁcantly positive, indicating that the inputs, carryovers, intermediates, and outputs are isotonically related. In other words, the selected variables are appropriate for further analysis using the stipulated dynamic network DEA model. Finally, as the number of banks meets the requirement that the number of DMUs should be larger than double or triple the number of variables used for the DEA analysis, we conclude that the developed DEA model has high construct validity. The traditional DEA models assume a production process as a single ‘black box’ that transforms inputs to outputs. Therefore, every activity has to be categorized as ‘input’ or ‘output’. This would create a problem when there is a complex production process that requires multiple inputs and outputs. Accordingly, the network DEA models overcome the abovementioned shortcoming by considering multiple divisions of production within the black box while evaluating the overall efﬁciency as well. The network structure allows the evaluation of the connectivity between inner linking activities (Kao, 2009; Tone and Tsutsui, 2009), hence, it will enable us to build processes of banks’ inner business activities. This study also considers the linking activities between two succeeding periods, which allow us to take the effect of time on performance measure into account. More speciﬁcally, we incorporate the time effect by means of carry-over activities between the subsequent periods. As such, the idea of dynamic DEA (Tone and Tsutsui, 2010) observes the long-term ﬂuctuated trends of banks through the years. In addition to the above, in traditional DEA models, the relative efﬁciency for each DMU is measured under the assumption of the proportional changes of input and output variables, meaning that the models are radial. In fact, radial models may lack objectivity in terms of reﬂecting the real input/output conditions for each organization. Furthermore, these models assume that inputs and outputs can be adjusted according to their ratios, which cannot be adopted under certain circumstances. As a solution, DNSBM, a model proposed by Tone and Tsutsui (2014), is a non-radial model which takes the possibility of 4 In an investigation of major DEA applications in banking literature in top journals across 2004–2009, Avkiran (2011) reaches the conclusion that “there is no clear agreement among the selection of inputs and outputs beyond the general observance of the intermediation approach to bank behavior” (Avkiran, 2011, p.326). 159 Table 1 Summary statistics of inputs, intermediates, and outputs. Mean Input Personnel expenses (X2) Other operating expenses (X3) Input (carry-over) Fixed assets (X1) Liquid assets (X4) Intermediate Loans (Z1) Other earning assets (Z2) Deposits (Z3) Output Net interest income (Y1) Standard deviation 75.865 90.096 1st Quartile 140.196 159.157 3rd Quartile 7.000 8.000 89.986 189.728 7,422.737 18,799.629 66.750 82.750 6.000 67.750 398.000 5,060.750 5,516.260 14,150.085 296.550 3,591.500 5,347.262 9,516.799 2,556.250 4,101.750 7,424.737 18,799.629 400.000 5,062.750 277.247 535.101 31.700 227.000 Note: 1. All variables are denoted in USD million. 2. Deﬁnitions of the variables are as follows. Personnel expenses are total staff costs in year t. Other operating expenses are total operating costs other than staff costs in year t. Fixed assets are total tangible property, plant, and equipment in year t − 1. Liquid assets are resources that could be converted into cash quickly in year t − 1. Loans are temporary funds provided to customers at interest in year t. Other earning assets include ﬁnancial investments in stocks and bonds in year t. Deposits are monies kept by customers at banks in year t. Net interest income refers to the excess income generated from loans and other earnings assets over expenses associated with interests on deposits in year t. non-proportional changes of inputs and outputs into account. This model deals with slacks when estimating the divisional and overall efﬁciencies. Considering differences of input and output slacks concurrently, this study uses non-oriented efﬁciency to estimate banks’ performance. The dynamic network process is shown in Fig. 1. 3.2. Modelling the dynamic network SBM Consider the dynamic network processes presented in Fig. 1 that deals with n banks (j = 1,…,n) consisting of k divisions (k = 1, …, K) over T terms (t = 1, …, T). At each term, banks have common mkt inputs (i = 1,b…, mkt ) consisting of k divisions and rkt outputs (i = 1,…, rkt ) consisting of k divisions. Let xkiot (i = 1,…, mkt ) and ykrot (i = 1,…, rkt ) denote the observed input and output values of bank j consisting of k dividenotes the continuity of link ﬂows sions at term t, respectively. z(k,h) jt (carry-overs) between terms t and t + 1. This study symbolizes the category link as Cbad. In order to identify them by term (t), bank (j), divi(p = sions (k) and item (i), this study employs the notion Ck,bad pot ; t = 1, …, T; k = 1, …, K) for denoting bad link values where 1, …, mk,bad t nbad is the number of bad links. These are all observed values up to the denote the observed carry-over input values of DMU j term T. Let mk,bad t consisting of k divisions at term t. Using these expressions for production, this study expresses the target banko (o = 1,…,n). Therefore, this Table 2 Spearman correlation coefﬁcients. X2 Personnel expenses (X2) Other operating expenses (X3) Fixed assets (X1) Liquid assets (X4) Loans (Z1) Other earning assets (Z2) Deposits (Z3) Net interest income (Y1) X3 X1 X4 Z1 Z2 Z3 Y1 1.000 0.959 1.000 0.875 0.944 0.938 0.692 0.881 0.946 0.926 0.738 1.000 0.859 1.000 0.845 0.975 1.000 0.628 0.767 0.707 1.000 0.944 0.946 0.859 1.000 0.975 0.767 1.000 0.951 0.937 0.877 0.941 0.949 0.691 0.941 1.000 Note: 1. All of the coefﬁcients are signiﬁcant at the 1% signiﬁcance level. 2. See Table 1 for the deﬁnitions of the variables. 160 Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 Fig. 1. The conceptual framework of the dynamic network production processes for banks. study deﬁnes the non-oriented efﬁciency by solving the program as follows: − for term T. This implies that the optimal slacks for term t in (3) are all zero. þ Let ρ∗o denote the overall efﬁciency during the term T. Where skit , skrt , " − k ;bad are slack variables denoting, respectively, input excess, outand spt puts shortfall, link excess, and link deviation. This objective function is an extension of the non-oriented SBM model (Tone, 2001) and deals with excesses in both input resources and undesirable (bad) links. The numerator is the average input efﬁciency and the denominator is the inverse of the average output efﬁciency. This study deﬁnes the non-oriented overall efﬁciency as a ratio that ranges between 0 and 1, and attains 1 when all slacks are zero. This objective function value is also units-invariant. " !# − k ;bad Xmk sk− Xmk;bad spt 1 XT XK 1 t k it t w 1− þ i¼1 xk p¼1 t¼1 k¼1 t T mkt þ mk;bad C k;bad iot t pot " !# ρo ¼ Min ð1Þ kþ X X X k 1 1 srt T K rt k w 1 þ r¼1 yk t¼1 k¼1 t T rkt rot s:t: X − n xkiot ¼ xk λk þ skit i ¼ 1; …; mkt ; t ¼ 1; …; T; k ¼ 1; …; K j¼1 i jt jt Xn þ yk λk −skrt r ¼ 1; …; r kt ; t ¼ 1; …; T; k ¼ 1; …; K ykrot ¼ j¼1 r jt jt Xn λkjt ¼ 1ðt ¼ 1; …; T Þ Xn ðk;hÞ k Xnj¼1 ðk;hÞ h z λ jt ¼ z λ jt ; ∀ ðk; hÞðt ¼ 1; …; T Þ j¼1 jt j¼1 jt X − n k;bad k k ;bad C k;bad ¼ C λ þ s p ¼ 1; …; mk;bad ; t ¼ 1; …; T; k ¼ 1; …; K jt pot pt t j¼1 p jt Xn Xn C k;bad λ jt ¼ C k;bad λ jtþ1 p ¼ 1; …; mk;bad ; t ¼ 1; …; T−1; k ¼ 1; …; K t j¼1 i jt j¼1 i jt − − þ λkjt ≥0; skit ≥0; skrt ≥0; skpt ;bad ≥0 ð2Þ If the optimal solution for Eq. (1) satisﬁes ρ∗o = 1, banko is called nonoriented overall efﬁcient or brieﬂy overall efﬁcient. XK wk k¼1 t ρt ¼ " 1− 1 Xmk sk− it t mkt þ mk;bad t " XK 1 k w 1þ k k¼1 t rt T i¼1 xkiot − þ k ;bad Xmk;bad spt Xrk skþ t rt r¼1 yk rot !# t p¼1 !# C k;bad pot ð3Þ In Eq. (3), ρo ¼ T1 ∑t¼1 ρt . If all optimal solutions of satisfy ρt = 1, banko is called non-oriented term efﬁcient or brieﬂy term efﬁcient 1− ρkt ¼ − k ;bad Xmk sk− Xmk;bad spt t it t þ i¼1 xk p¼1 mkt þ mk;bad C k;bad iot t pot " !# þ 1 Xrkt skrt 1þ k r¼1 yk rt rot 1 !# ð4Þ In Eq. (4), ρt = ∑Kk = 1wkt ρkt . If all optimal solutions satisfy ρkt = 1, banko is called non-oriented term efﬁcient or brieﬂy term efﬁcient with the divisions k at the term T. This implies that the optimal slacks with the divisions k at term t in Eq. (4) are all zero. 3.3. Regression model In a DEA-application involving multivariate analysis to ﬁnd the effect of contextual factors on performance measures, there are two main recent streams for a second-stage analysis. First approach suggests using a maximum likelihood estimation of a truncated regression, proposed by Simar and Wilson (2007). Second approach advocates the use of maximum likelihood estimation of ordinary least squares (OLS) or Tobit regression, proposed by Banker and Natarajan (2008). McDonald (2009) asserts that DEA analysis followed by second stage-analysis involving OLS could yield a valid estimation of the contextual factors, and it is less restrictive as compared to the truncated model proposed by Simar and Wilson (2007). Through OLS, Banker and Natarajan (2008) show that consistent estimators of the regression coefﬁcients could be obtained despite the fact that efﬁciency scores range between zero and one. As mentioned by Liu et al. (2016), the status of development in a secondstage analysis for DEA studies has left practitioners with some confusion about the true use of methodology. Although the abovementioned evidence suggests the use of OLS, many research works have also used truncated regression with a bootstrapping approach for robustness checks. Thus, we perform both approaches in our regression model. In this study, the regression results are adjusted for year-speciﬁc and country-speciﬁc effects. Speciﬁcally, we employ panel data estimation procedures, which adjust for the time-series and cross-sectional effects. Note that the Breusch-Godfrey serial correlation Lagrange multiplier (LM) test suggests that panel data regression is a better estimation technique as compared to pooled regression in this study, while the Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 Hausman test indicates that ﬁxed-effects model (FEM), instead of random-effects model (REM), should be applied. Therefore, we employ the ﬁxed-effects panel data regression model to examine the relationship between earnings management and overall efﬁciency. In this study, the following regression models are run to test the hypotheses: X OEit ¼ α 0 þ α 1 LLPLit þ α 2 LLRLit þ α 3 LNASSETSit þ α 4 GROW it þ α 5 LIABit þ þ Yri X þ Countryi þ εit ð5Þ where: The non-oriented DNSBM overall efﬁciency score based on variable returns to scale in year t. Loan loss provisions scaled by loans in year t. LLPLit Loan loss reserves scaled by loans in year t. LLRPit LNASSETSit The natural logarithm of total assets in year t. GROWit The growth rate of net income in year t. Total liabilities scaled by total assets in year t. LIABit Year-speciﬁc effect dummy variables. ∑Yri ∑Countryi Country-speciﬁc effect dummy variables. OEit Based on Adams et al. (2009), we argue that earnings management in banks commonly happens through loan loss provisions and loan loss reserves due to the nature of their discretionary choices. The intuition is that increases in loan loss reserves and/or loan loss provisions would mean decreases in net income and deﬁnitely the ratio of earnings to assets. Therefore, the book value of equity is also reduced because loan loss provisions and loan loss reserves ﬂow through the ﬁnancial statements. LNASSETSit is used to control for the ﬁrm size of the sample banks, while GROWit is a measure to control for the growth opportunities of banks. LIABit could serve as a proxy of individual bank risktaking. If bank managers engage in earning management, we would expect to have signiﬁcantly negative coefﬁcients on α1 and α2 because the good-looking earnings today have to be paid off in the future, in which case, it would be shown in this study through the dynamic performance measure that is measured over a long-term period. 4. Empirical ﬁndings and discussion 4.1. Dynamic network performance analysis Table 3 shows the results of the dynamic network DEA model for banking institutions in the ASEAN region. Speciﬁcally, the table presents both yearly and averages of overall, managerial, and proﬁtability efﬁciency scores. While the overall efﬁciency of ASEAN member countries seems to drop during 2008 to 2012 and then increases in 2013, the Singaporean banking sector tends to swim against the tide. Also, Singaporean banks appear to be more efﬁcient in terms of overall efﬁciency followed by Cambodia and Malaysia, with average scores of 0.622, 0.511, and 0.421, respectively. The overall efﬁciency of the total sample shows a monotonic decrease over the period with 62.8 per cent room for improvement on average. In addition, Table 3 also provides the breakdown of overall efﬁciency into managerial and proﬁtability efﬁciencies. In the ﬁrst division, i.e. managerial efﬁciency, we can see ﬂuctuating trends in countries’ banking performance; however, the total sample average is again at declining trend. Although the managerial efﬁciency of ASEAN banking institutions suggests the poor performance of banking sectors in the region, it happens to be more efﬁcient when compared to proﬁtability efﬁciency. The results indicate the slightly better performance of managerial division, particularly in the last three years, pointing that most countries have enhanced their capabilities to manage the human resources but have failed to create salient proﬁt-making capacities using their managerial abilities. Nonetheless, there exists large room for improvement in both managerial and proﬁtability efﬁciencies. For 161 Table 3 Overall, managerial, and proﬁtability efﬁciencies of banks in ASEAN countries for the period of 2008–2013. Country Overall efﬁciency Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam Total sample Managerial efﬁciency Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam Total sample Proﬁtability efﬁciency Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam Total sample 2008 2009 2010 2011 2012 2013 Average 0.295 0.693 0.484 0.587 0.620 0.319 0.585 0.357 0.387 0.450 0.210 0.559 0.466 0.407 0.519 0.289 0.630 0.311 0.372 0.417 0.215 0.553 0.451 0.465 0.375 0.284 0.609 0.322 0.369 0.410 0.201 0.511 0.409 0.342 0.337 0.249 0.603 0.318 0.312 0.373 0.132 0.353 0.324 0.227 0.259 0.217 0.690 0.296 0.242 0.311 0.160 0.398 0.311 0.282 0.417 0.240 0.614 0.288 0.245 0.310 0.202 0.511 0.408 0.385 0.421 0.266 0.622 0.315 0.321 0.378 0.239 0.751 0.425 0.654 0.881 0.229 0.694 0.287 0.482 0.430 0.234 0.638 0.448 0.561 0.833 0.246 0.724 0.296 0.452 0.430 0.239 0.639 0.426 0.473 0.648 0.236 0.742 0.306 0.416 0.414 0.267 0.696 0.429 0.509 0.602 0.233 0.756 0.305 0.359 0.411 0.273 0.726 0.398 0.527 0.765 0.212 0.739 0.266 0.315 0.387 0.254 0.639 0.328 0.502 0.772 0.191 0.694 0.263 0.318 0.346 0.251 0.681 0.409 0.538 0.750 0.225 0.725 0.287 0.390 0.403 0.350 0.642 0.549 0.524 0.390 0.408 0.529 0.426 0.310 0.478 0.186 0.479 0.486 0.293 0.308 0.331 0.580 0.326 0.312 0.411 0.190 0.473 0.476 0.459 0.190 0.331 0.572 0.341 0.331 0.415 0.141 0.330 0.392 0.189 0.133 0.263 0.572 0.334 0.268 0.344 0.030 0.161 0.312 0.087 0.120 0.232 0.680 0.327 0.211 0.289 0.067 0.202 0.314 0.134 0.087 0.288 0.596 0.316 0.187 0.293 0.161 0.381 0.421 0.281 0.205 0.309 0.588 0.345 0.270 0.372 instance, bank institutions in ASEAN region can improve their managerial and proﬁtability efﬁciencies by 59.7 per cent and 62.8 per cent, respectively, in order to be fully efﬁcient. On average, the statistical ﬁndings5 indicate the superior efﬁciency of Singaporean and Malaysian banks; that is to say, the highest overall and proﬁtability efﬁciencies are attributed to Singapore being at 62.2 per cent and 58.8 per cent, respectively, and 75 per cent as the largest managerial efﬁciency score is related to Malaysia. Yet Bruneian and the Philippine banking sectors are ranked as the most inefﬁcient members of ASEAN coalition in which the former scored the least in overall and proﬁtability efﬁciency and the latter scored the least in managerial efﬁciency. The unreported results show that only one bank is found to be nonoriented efﬁcient. This high level of inefﬁciency encourages us to report the frontier projections. The potential improvements for inefﬁcient DMUs are determined based on the banks on the efﬁcient frontier, i.e. benchmark units (Avkiran, 2014b). Table 4 provides the average excess and shortage of each variable for all member countries. A positive percentage implies the shortage of resources (inputs) and a negative percentage implies the excess of resources (outputs). The ﬁndings in Table 4 suggest that the ASEAN members on average have to cut their personnel expenses and other operating expenses by 59.4 per cent and 59 per cent, respectively. The carry-overs are approximately the same as the primary inputs for banks in which these two input quantities have to be reduced by 60.7 per cent (ﬁxed assets) and 59.9 per cent (liquid assets). The three intermediates act as dual-role variables in production process; meaning that they are outputs for the ﬁrst division and inputs for the second division. Consequently, the 5 This study applied Kruskal–Wallis test, a non-parametric statistical analysis, to examine whether differences exist among efﬁciency performance of countries in the region. For brevity purpose, we did not report the table. As the signiﬁcance level of 1%, we prove that there is a signiﬁcant difference among ASEAN countries in terms of efﬁciency scores. 162 Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 Table 4 Frontier projections for banks in ASEAN countries (%). Input Carry-over Intermediate Output Country X2 X3 X1 X4 Z1 Z2 Z3 Y1 Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam Total sample −75.2 −25.5 −60.5 −39.9 −39.6 −75.0 −31.6 −69.8 −60.9 −59.4 −73.7 −29.8 −56.6 −44.1 −12.4 −79.8 −32.7 −71.8 −60.3 −59.0 −75.8 −40.3 −60.2 −54.8 −22.9 −77.8 −18.2 −72.2 −61.6 −60.7 −83.4 −56.8 −55.6 −65.7 −68.5 −68.1 −30.5 −63.9 −69.5 −59.9 −60.9 −27.7 −37.1 −26.6 −30.7 −49.4 −9.6 −59.2 −48.2 −41.3 11.2 6.6 10.3 7.2 5.1 −25.7 −15.5 −29.3 −11.1 −5.6 −72.0 −23.2 −37.3 −25.5 −41.0 −61.0 −30.4 −54.7 −54.5 −44.4 8.8 21.3 19.3 32.0 69.1 7.1 101.3 6.9 18.5 21.1 Note: Personnel expenses (X2); other operating expenses (X3); ﬁxed assets (X1); liquid assets (X4); loans (Z1); other earning assets (Z2); deposits (Z3); net interest income (Y1). Negative: Excess of resources. Positive: Shortage of resources. suggestions on potential improvements are mixed for these variables. For instance, other earning assets have to be increased for Brunei, Cambodia, Indonesia, Laos, and Malaysia while it has to be decreased for the Philippines, Singapore, Thailand, and Vietnam. In summary, ASEAN banks could be efﬁcient if they can increase their net interest income on average, as the only output, by 21.1 per cent while performing the required changes in inputs, carry-overs, and intermediate variables. In order to get more insights into the sources of inefﬁciencies, Table 5 provides the frontier projections for banking sectors in each Table 5 Frontier projections by year (%). Panel A: Input and Output Country Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam X2 X3 Y1 2008 2009 2010 2011 2012 2013 2008 2009 2010 2011 2012 2013 2008 2009 2010 2011 2012 2013 −76.7 −23.2 −58.8 −37.2 −35.8 −74.0 −35.6 −67.5 −52.7 −77.4 −31.6 −55.8 −45.3 −50.0 −73.4 −31.3 −66.4 −56.2 −81.2 −37.7 −60.6 −47.5 −41.8 −75.3 −27.6 −67.5 −60.8 −75.7 −24.2 −61.5 −40.1 −58.1 −75.8 −29.6 −70.4 −67.0 −69.7 −15.8 −60.5 −37.1 −27.1 −75.6 −28.1 −73.0 −68.5 −70.6 −20.3 −66.1 −32.0 −24.8 −76.0 −37.4 −73.7 −60.7 −72.0 −22.4 −57.8 −34.8 0.0 −79.9 −42.9 −76.1 −49.9 −76.2 −40.2 −56.2 −39.2 0.0 −80.0 −33.4 −75.8 −54.3 −76.7 −30.7 −55.8 −56.0 −42.1 −80.6 −31.9 −72.0 −57.7 −72.5 −30.1 −50.6 −49.1 −17.9 −76.9 −27.2 −65.5 −64.2 −70.4 −17.1 −53.1 −37.0 0.0 −78.9 −28.5 −70.9 −66.0 −74.2 −38.5 −65.9 −48.2 −14.5 −82.5 −32.5 −70.7 −69.8 0.0 5.1 50.3 2.5 13.7 6.1 46.6 1.8 15.6 0.0 0.4 3.9 26.8 48.5 0.0 37.3 0.0 12.1 0.0 3.8 2.7 6.7 47.9 0.1 118.6 6.7 17.5 12.2 2.7 2.1 8.2 30.5 7.1 167.4 5.4 3.9 40.3 92.8 43.7 100.2 266.3 29.1 114.9 17.0 48.6 0.0 23.1 12.9 47.5 7.4 0.0 123.0 10.4 13.1 2008 2009 2010 2011 2012 2013 2008 2009 2010 2011 2012 2013 −79.7 −29.1 −56.0 −31.9 0.0 −77.2 −13.4 −70.3 −52.9 −76.2 −36.8 −53.6 −47.3 0.0 −72.6 −18.1 −69.0 −53.8 −70.4 −40.0 −55.7 −54.7 −21.7 −73.3 −18.0 −68.6 −56.7 −71.7 −36.9 −59.1 −58.1 −43.3 −77.4 −16.2 −72.5 −61.1 −77.8 −49.4 −67.0 −67.8 −43.4 −82.1 −21.8 −76.2 −71.0 −79.2 −49.5 −69.7 −69.2 −29.2 −84.0 −21.9 −76.7 −74.2 −65.0 −32.3 −42.8 −46.4 −55.6 −58.3 −35.3 −56.7 −65.4 −81.4 −51.7 −49.9 −57.7 −54.3 −66.9 −37.4 −67.4 −65.0 −81.0 −51.6 −51.5 −51.0 −71.9 −66.8 −32.7 −63.9 −62.2 −84.2 −65.8 −60.2 −79.5 −82.6 −72.3 −26.9 −65.7 −72.7 −95.7 −66.7 −62.9 −80.6 −55.9 −73.1 −25.5 −63.1 −72.6 −93.4 −73.0 −66.6 −79.3 −90.7 −71.2 −25.3 −66.5 −79.1 2008 2009 2010 2011 2012 2013 2008 2009 2010 2011 2012 2013 2008 2009 2010 2011 2012 2013 −39.7 39.5 −20.3 19.2 −33.1 −38.0 −15.5 −55.5 −15.2 −47.8 −22.1 5.2 51.0 −18.3 −38.8 −24.5 −54.5 −50.1 −54.7 −13.8 −35.8 −36.8 −59.0 −41.0 −6.5 −46.2 −50.2 −38.7 −29.9 −45.0 −35.9 −65.5 −44.8 −8.3 −62.9 −44.6 −92.4 −56.0 −60.7 −66.8 33.7 −64.2 −9.5 −67.9 −58.3 −92.3 −83.7 −65.9 −90.4 −41.9 −69.3 6.5 −68.1 −70.5 −1.8 0.0 −2.9 −1.9 0.9 −27.6 −16.1 −30.2 −8.5 −1.1 0.1 −0.2 −2.1 2.1 −30.3 −19.0 −32.7 −9.5 2.1 0.5 7.4 −1.6 6.9 −30.9 −21.3 −27.9 −17.0 7.6 2.7 26.7 −1.0 4.9 −22.4 −16.5 −28.5 −18.1 28.9 17.7 17.9 20.2 5.7 −20.4 −10.8 −28.8 −9.3 31.4 18.4 12.9 29.5 10.2 −22.6 −9.4 −27.8 −4.4 −64.7 40.6 −28.7 −29.2 −27.8 −53.0 −41.4 −54.0 −50.3 −62.6 −8.0 −5.9 55.7 −33.7 −51.9 −34.4 −51.9 −40.1 −58.7 −29.3 −34.2 −7.9 −56.6 −54.8 −29.5 −28.3 −55.2 −72.6 −39.3 −49.1 −48.5 −75.6 −63.9 −28.0 −61.6 −59.9 −93.2 −54.1 −58.1 −66.9 −2.0 −70.7 −24.3 −66.4 −66.5 −80.1 −48.9 −47.5 −56.0 −50.4 −71.4 −25.1 −66.2 −55.2 Panel B: Carry−over Country Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam X1 X4 Panel C: Intermediate Country Brunei Cambodia Indonesia Laos Malaysia Philippines Singapore Thailand Vietnam Z1 Z2 Z3 Note: Personnel expenses (X2); other operating expenses (X3); ﬁxed assets (X1); liquid assets (X4); loans (Z1); Other earning assets (Z2); Deposits (Z3); Net interest income (Y1). Negative, excess of resources; positive, shortage of resources. Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 country segregated by years. The table illustrates three panels including input and output, carry-over and intermediate variables for the period of 2008 to 2013. Singapore, a leading country in terms of overall efﬁciency score (Table 3), appears to be more efﬁciently using personnel expenses along with Cambodia where the country stands second in her overall efﬁciency score. The efﬁcient usage of an input factor in performance analysis is one of the main factors to attain higher efﬁciency. Malaysia has also shown tolerable usage of personnel expenses with her overall efﬁciency score being ranked among the top three countries. Indeed, Malaysia performed greater than other countries in consuming other operating expenses as an input factor. Moreover, the carry-over input (ﬁxed assets) consumed by top three countries in terms of overall efﬁciency scores seems to be efﬁciently managed as compared to other countries. For example, Singaporean banks have to reduce their ﬁxed assets quantities by 21.9 per cent in 2013 while the Philippines has to cut down this carry-over item by 84 per cent in the same year. Less efﬁcient countries, namely, Brunei and the Philippines are suffering from over-utilization of the two input factors as well as the carry-over item in the ﬁrst stage, dragging down both the overall efﬁciency and managerial efﬁciency scores to be the least among other countries in the region. Recall that the results of Table 3 ranked Malaysia, Singapore and Cambodia as the leading countries, respectively, and Brunei and the Philippines as being the low-performing nations in terms of managerial efﬁciency. Apparently, the excess of the input factors and the carry-over item are in line with the managerial efﬁciency rankings of the countries. The intermediate variables are the outputs of the ﬁrst stage and the inputs of the second stage. Hence, the variations in terms of excess and shortage of the factors are obtained to achieve the optimum solutions for both managerial and proﬁtability efﬁciency stages. In the majority of cases, banking sectors have to reduce their loans and deposits quantities to attain higher efﬁciency; however, the results for other earnings assets are mixed where, for example, Malaysian banking sector has to target on increasing, and Singaporean banking sector has to focus on reducing this intermediate factor. The substantial excess in the usage of the carry-over item in the second stage, liquid assets, discloses one of the main causes for inefﬁciency of Malaysian banking sector in proﬁtability efﬁciency obtained in Table 3. Singaporean banks, however, appeared to utilize the liquid assets more efﬁciently as compared to other banks in the region and the trend shows improving throughout the years. Finally, the results indicate the inefﬁciencies associated with less efﬁcient banking sectors in the sample, for example, Brunei and the Philippines, are not induced by net interest income, the ﬁnal output. However, the Singaporean banking sector, among the leading sectors in both managerial and proﬁtability stages, need to substantially increase her output quantity in order to be even more efﬁcient. 4.2. The relationship between earnings management and efﬁciency Banking institutions, which contribute much to economic developments, are susceptible to EM practices (Grougiou et al., 2014) due to complicated operation that usually cause information opacity (Levine, 2004) and information asymmetry (Mülbert, 2009). Therefore, as mentioned earlier, this study examines the impacts of earnings management on efﬁciency through regression analysis. Before we run any regression analysis, we test the potential problem caused by multicollinearity. The diagnostic test of variance inﬂation factors (VIF), which is not reported, suggests that multicollinearity problems do not exist in this study, whereby the centered VIF values are all less than 1.5 (Kennedy, 1998). We also perform the diagnostic test of potential heteroskedasticity for the regression residuals, and we ﬁnd evidence of heteroskedasticity. Therefore, the p-values in Table 6 are corrected using White crosssection standard errors. The F-statistics indicate that Eq. (1) is statistically signiﬁcant. The results in Table 6 indicate that loan loss provisions (LLPL) are signiﬁcantly and negatively related to the dynamic overall efﬁciency of 163 Table 6 Regression results. FEM Truncated regression Variable Coefﬁcient Standard P-value Coefﬁcient Standard P-value error error Intercept LLPL LLRP LNASSETS GROW LIAB Year dummies Country dummies Adjusted R2 F-statistic Log-likelihood 1.402*** −0.200*** 0.243 −0.144*** 0.011*** 0.058 Yes Yes 0.131 0.031 0.152 0.014 0.002 0.056 0.000 0.000 0.110 0.000 0.000 0.308 0.804*** −0.682* 0.493*** −0.028*** 0.011*** −0.284*** 0.063 0.378 0.100 0.006 0.003 0.057 Yes Yes 0.000 0.071 0.000 0.000 0.000 0.000 0.838 31.229*** 193.026 Note: *, **, and *** denote the statistical signiﬁcance at the 10%, 5%, and 1% level, respectively. banks in ASEAN countries, suggesting that earnings management today will be paid off in dynamic performance in the long term, consistent with our prediction. However, the positive coefﬁcient on loan loss reserves (LLRP) does not reach the conventional signiﬁcance level. For another sensitivity analysis, we estimate Eq. (1) using truncated regression, following Lu et al. (2014). The truncated regression results remain almost qualitatively the same as those of the FEM. In summary, we ﬁnd that earnings management engaged by bank managers could be observed from loan loss provisions rather than loan loss reserves from the perspective of dynamic performance. 4.3. Discussions The main impetus of this study is ﬁnancial liberalizations and institutional reforms in the ASEAN countries, which would help in promoting efﬁcient allocation of resources in their banking sectors. This could assist the sectors in achieving a greater economic growth through the strengthened regulatory and supervisory frameworks. The borderless business world has created the situation whereby ASEAN banks need to be able not only to survive locally, but also to compete internationally. In this regard, performance measurement of banks has since become an important subject of study. Speciﬁcally, the relative efﬁciency of banks in ASEAN countries is an important question to be answered. By performing efﬁciency analysis on banks in ASEAN countries through DEA, we are able to reveal their relative competitiveness from a multidimensional perspective. A clear implication of this assessment on applied economics studies is that we scrutinize banking performance in the aspects of managerial efﬁciency and proﬁtability efﬁciency over longterm periods. Before next initiatives to improve economic growth are introduced, economists who are decision makers should ﬁrst examine how well the human capital in the countries performs in terms of managerial efﬁciency. They should next look at the individual but relative performance of banks in the region before ultimately discussing economic growth. In other words, they should work on improving the fundamental issue of managerial and proﬁtability efﬁciencies of ASEAN banks. The current paper ﬁnds that banks in certain countries should focus on ﬁrst improving their managerial efﬁciencies. The frontier projections provide some ideas on how economists may look at the allocation of resources in their banking sectors, which would ultimately result in economic growth. It is also important to highlight the possibility of window-dressed performance by the banks in ASEAN countries. As discussed earlier, managers in the banking sector nowadays might have greater tendency to dress up their corporate performance due to the intensiﬁed competition in the region. Economists who intend to promote economic prosperity in their countries may overlook this matter, namely, earnings management. A manipulated ﬁnancial report 164 Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 might, in the end, bring disastrous ﬁnancial catastrophe to the company, and if the company is large enough, the world economy will be affected. The most relevant ﬁnding of this study is the signiﬁcantly negative impacts of earnings management practices on the banking performance of ASEAN banks. Taken together, economists may consider using the DN-DEA model to assess banking performance because it allows researchers to incorporate multiple variables in the performance measurement. Moreover, the ‘black-box’ of banking performance is revealed, while considering dynamism in banking performance over long-term periods. 5. Conclusions This study aims to address two key questions in the literature of bank efﬁciency. Firstly, we provide a unique example of the application of the DN-DEA model in banking that is still in its embryonic stage. More speciﬁcally, we decompose the efﬁciency of ASEAN banking institutions into managerial efﬁciency and proﬁtability efﬁciency using the newly developed DN-DEA model, viz. DNSBM. Secondly, we provide an empirical answer to the important question that what the effect of any EM practices is on the efﬁciency of ﬁrms. For the ﬁrst objective, the ﬁndings reveal that there exists large room for improvement in the efﬁciency domain of ASEAN banks. This will accentuate the need for a better policy formulation in the region in boosting the banking sectors upwards. Indeed, the role of ASEAN alliance could be very inﬂuential to achieve this goal. Likewise, the monotonic decrease in the overall and divisional efﬁciencies of the total sample is alarming. Among all, the overall performance of Singaporean banking sector outstrips that of any other banking sectors in ASEAN region. In detail, Singapore is ranked ﬁrst in proﬁtability division and second in managerial division. Meanwhile, the poor performance of Malaysian banking sector on proﬁtability efﬁciency drags down its overall efﬁciency despite its spectacular managerial efﬁciency scores. In order to determine the reasons behind the inefﬁciencies of ASEAN banks, we investigate the potential areas of improvement relevant to input and output variables. The results suggest the equal attention should be given to carry-overs as well as to inputs where the ASEAN banks have to reduce the input and carry-over variables by approximately 60 per cent. For the second objective, we ﬁnd a signiﬁcant negative relationship between loan loss provisions and dynamic performance of ASEAN banks. However, the loan loss reserves could not satisfy our proposed hypothesis in determining the EM practices of banking institutions. Therefore, we argue that loan loss provisions is an appropriate proxy for banks’ EM practices, which signiﬁcantly dampen the dynamic efﬁciency of ASEAN banks. All in all, more research needs to be done to conﬁrm the generalizability of our results. Indeed, the new approach developed in this study on the relationship between EM practices and performance demands a global appeal. It can be applied to those banking sectors where the sources of inefﬁciencies are unknown. We leave it to future studies to examine further earnings management in banks through other proxies when more data are available in the Bankscope database to the public. Future studies should also include more bank-level and country-level control variables. Dummies on country-level controls may not be adequate to account for country-speciﬁc heterogeneity. Another important area for future research would be to compare various methodologies developed in the domain of DN-DEA among different banking institutions like Islamic banks. Particularly, rather than assuming a common frontier in cross-country samples, researchers may develop a meta-frontier DN-DEA to estimate the performance of ASEAN banks or even world banks. Acknowledgement We sincerely thank two anonymous reviewers for their constructive comments on an earlier version of this study. References Adams, B., Carow, K.A., Perry, T., 2009. Earnings management and initial public offerings: the case of the depository industry. J. Bank. Financ. 33, 2363–2372. ADB, 2011. Financial sector operational plan. ADB Administration and Governance. Asian Development Bank. Avkiran, N.K., 2009. Opening the black box of efﬁciency analysis: an illustration with UAE banks. Omega 37, 930–941. Avkiran, N.K., 2011. Association of DEA super-efﬁciency estimates with ﬁnancial ratios: investigating the case for Chinese banks. Omega 39, 323–334. Avkiran, N.K., 2014a. An illustration of dynamic network DEA in commercial banking including robustness tests. Omega 55, 141–150. Avkiran, N.K., 2014b. Probing organizational efﬁciency. Int. J. Organ. Anal. 22, 149–160. Avkiran, N.K., Thoraneenitiyan, N., 2010. Purging data before productivity analysis. J. Bus. Res. 63, 294–302. Baik, Y.-S., Kwak, B., Lee, J., 2011. Deregulation and earnings management: the case of the U.S. airline industry. J. Account. Public Policy 30, 589–606. Banker, R.D., Natarajan, R., 2008. Evaluating contextual variables affecting productivity using data envelopment analysis. Oper. Res. 56, 48–58. Barth, J.R., Lin, C., Ma, Y., Seade, J., Song, F.M., 2013. Do bank regulation, supervision and monitoring enhance or impede bank efﬁciency? J. Bank. Financ. 37, 2879–2892. Becchetti, L., Sierra, J., 2003. Bankruptcy risk and productive efﬁciency in manufacturing ﬁrms. J. Bank. Financ. 27, 2099–2120. Beneish, M.D., Press, E., Vargus, M.E., 2012. Insider trading and earnings management in distressed ﬁrms. Contemp. Account. Res. 29, 191–220. Berg, S.A., Førsund, F.R., Hjalmarsson, L., Suominen, M., 1993. Banking efﬁciency in the Nordic countries. J. Bank. Financ. 17, 371–388. Berger, A.N., Humphrey, D.B., 1997. Efﬁciency of ﬁnancial institutions: international survey and directions for future research. Eur. J. Oper. Res. 98, 175–212. Bhattacharyya, A., Lovell, C.A.K., Sahay, P., 1997. The impact of liberalization on the productive efﬁciency of Indian commercial banks. Eur. J. Oper. Res. 98, 332–345. Bolt, W., De Haan, L., Hoeberichts, M., Van Oordt, M.R., Swank, J., 2012. Bank proﬁtability during recessions. J. Bank. Financ. 36, 2552–2564. Charnes, A., Cooper, W., Rhodes, E., 1978. Measuring the efﬁciency of decision making units. Eur. J. Oper. Res. 2, 429–444. Chiu, Y.-H., Chen, Y.-C., 2009. The analysis of Taiwanese bank efﬁciency: incorporating both external environment risk and internal risk. Econ. Model. 26, 456–463. Cook, W.D., Seiford, L.M., 2009. Data envelopment analysis (DEA)–thirty years on. Eur. J. Oper. Res. 192, 1–17. Cooper, W., Park, K., Pastor, J., 1999. RAM: a range adjusted measure of inefﬁciency for use with additive models, and relations to other models and measures in DEA. J. Prod. Anal. 11, 5–42. Cornett, M.M., McNutt, J.J., Tehranian, H., 2009. Corporate governance and earnings management at large U.S. bank holding companies. J. Corp. Financ. 15, 412–430. Dang-Thanh, N., 2012. Measuring the performance of the banking system: case of Vietnam (1990–2010). J. Appl. Financ. Bank. 2, 289–312. Dong, Y., Hamilton, R., Tippett, M., 2014. Cost efﬁciency of the Chinese banking sector: a comparison of stochastic frontier analysis and data envelopment analysis. Econ. Model. 36, 298–308. Elyasiani, E., Mehdian, S.M., 1990. A nonparametric approach to measurement of efﬁciency and technological change: the case of large US commercial banks. J. Financ. Serv. Res. 4, 157–168. Emrouznejad, A., Parker, B.R., Tavares, G., 2008. Evaluation of research in efﬁciency and productivity: a survey and analysis of the ﬁrst 30 years of scholarly literature in DEA. Socio Econ. Plan. Sci. 42, 151–157. Färe, R., Grosskopf, S., 1996. Intertemporal production frontiers: with dynamic DEA. Kluwer Academic Publishers, Boston. Fukuyama, H., Weber, W.L., 2013. A dynamic network DEA model with an application to Japanese Shinkin banks. Efﬁciency and Productivity Growth. John Wiley & Sons, Ltd, pp. 193–213. García Lara, J.M., Osma, B.G., Neophytou, E., 2009. Earnings quality in ex‐post failed ﬁrms. Account. Bus. Res. 39, 119–138. Gardener, E., Molyneux, P., Nguyen-Linh, H., 2011. Determinants of efﬁciency in South East Asian banking. Serv. Ind. J. 31, 2693–2719. Greenawalt, M.B., Sinkey Jr., J.F., 1988. Bank loan-loss provisions and the incomesmoothing hypothesis: An empirical analysis, 1976-1984. J. Financ. Serv. Res. 1, 301–318. Grougiou, V., Leventis, S., Dedoulis, E., Owusu-Ansah, S., 2014. Corporate social responsibility and earnings management in U.S. banks. Account. Forum 38, 155–169. Healy, P.M., Wahlen, J.M., 1999. A review of the earnings management literature and its implications for standard setting. Account. Horiz. 13, 365–383. Kao, C., 2009. Efﬁciency decomposition in network data envelopment analysis: a relational model. Eur. J. Oper. Res. 192, 949–962. Kao, C., Liu, S.-T., 2014. Multi-period efﬁciency measurement in data envelopment analysis: the case of Taiwanese commercial banks. Omega 47, 90–98. Karim, M.Z.A., 2001. Comparative bank efﬁciency across select ASEAN countries. ASEAN Econ. Bull. 289–304. Kennedy, P., 1998. A Guide to Econometrics. MIT Press, Cambridge. Laeven, L., 1999. Risk and efﬁciency in East Asian banks. World Bank Policy Research Working Paper No. 2255. World Bank, Washington, DC. Levine, R., 2004. The corporate governance of banks: a concise discussion of concepts and evidence. World Bank Publications. Lim, S., 2005. Foreign capital entry in the domestic banking market of Korea. Korean Polit. Sci. Rev. 39, 189–209. Lin, T.-Y., Chiu, S.-H., 2013. Using independent component analysis and network DEA to improve bank performance evaluation. Econ. Model. 32, 608–616. Y.-C. Wu et al. / Economic Modelling 53 (2016) 156–165 Liu, J.S., Lu, L.Y., Lu, W.-M., Lin, B.J., 2013. A survey of DEA applications. Omega 41, 893–902. Liu, J.S., Lu, L.Y.Y., Lu, W.-M., 2016. Research fronts in data envelopment analysis. Omega 58, 33–45. Lo, S.-F., Lu, W.-M., 2006. Does size matter? Finding the proﬁtability and marketability benchmark of ﬁnancial holding companies. Asia Pac. J. Oper. Res. 23, 229–246. Lu, W.-M., Wang, W.-K., Kweh, Q.L., 2014. Intellectual capital and performance in the Chinese life insurance industry. Omega 42, 65–74. Lu, W.-M., Kweh, Q.L., Nourani, M., Huang, F.-W., 2015. Evaluating the efﬁciency of dualuse technology development programs from the R&D and socio-economic perspectives. Omega, http://dx.doi.org/10.1016/j.omega.2015.08.011. Luo, X., 2003. Evaluating the proﬁtability and marketability efﬁciency of large banks: An application of data envelopment analysis. J. Bus. Res. 56, 627–635. McDonald, J., 2009. Using least squares and tobit in second stage DEA efﬁciency analyses. Eur. J. Oper. Res. 197, 792–798. McKinnon, R.I., 1973. Money and Capital in Economic Development. Brookings Institution, Washington DC. Miller, S.M., Noulas, A.G., 1996. The technical efﬁciency of large bank production. J. Bank. Financ. 20, 495–509. Moradi-Motlagh, A., Babacan, A., 2015. The impact of the global ﬁnancial crisis on the efﬁciency of Australian banks. Econ. Model. 46, 397–406. Mülbert, P.O., 2009. Corporate governance of banks. Eur. Bus. Org. Law Rev. 10, 411–436. Parkan, C., 1987. Measuring the efﬁciency of service operations: an application to bank branches. Eng. Costs Prod. Econ. 12, 237–242. Pasiouras, F., 2008. International evidence on the impact of regulations and supervision on banks’ technical efﬁciency: an application of two-stage data envelopment analysis. Rev. Quant. Finan. Acc. 30, 187–223. Quiggin, J., 2011. What have we learned from the global ﬁnancial crisis? Aust. Econ. Rev. 44, 355–365. Rangan, N., Grabowski, R., Aly, H.Y., Pasurka, C., 1988. The technical efﬁciency of US banks. Econ. Lett. 28, 169–175. Reinhart, C.M., Tokatlidis, I., 2003. Financial liberalisation: the African experience. J. Afr. Econ. 12, ii53–ii88. Seiford, L.M., Zhu, J., 1999. Proﬁtability and marketability of the top 55 US commercial banks. Manag. Sci. 45, 1270–1288. 165 Shaw, E., 1973. Financial deepening in economic development. Oxford University Press, New York. Sherman, H.D., Gold, F., 1985. Bank branch operating efﬁciency: evaluation with data envelopment analysis. J. Bank. Financ. 9, 297–315. Simar, L., Wilson, P.W., 2007. Estimation and inference in two-stage, semi-parametric models of production processes. J. Econ. 136, 31–64. Sturm, J.-E., Williams, B., 2004. Foreign bank entry, deregulation and bank efﬁciency: lessons from the Australian experience. J. Bank. Financ. 28, 1775–1799. Sueyoshi, T., Sekitani, K., 2009. An occurrence of multiple projections in DEA-based measurement of technical efﬁciency: theoretical comparison among DEA models from desirable properties. Eur. J. Oper. Res. 196, 764–794. Thanassoulis, E., 1999. Data envelopment analysis and its use in banking. Interfaces 29, 1–13. Tone, K., 2001. A slacks-based measure of efﬁciency in data envelopment analysis. Eur. J. Oper. Res. 130, 498–509. Tone, K., Tsutsui, M., 2009. Network DEA: a slacks-based measure approach. Eur. J. Oper. Res. 197, 243–252. Tone, K., Tsutsui, M., 2010. Dynamic DEA: a slacks-based measure approach. Omega 38, 145–156. Tone, K., Tsutsui, M., 2014. Dynamic DEA with network structure: a slacks-based measure approach. Omega 42, 124–131. Wang, W.-K., Lu, W.-M., Liu, P.-Y., 2014. A fuzzy multi-objective two-stage DEA model for evaluating the performance of US bank holding companies. Expert Syst. Appl. 41, 4290–4297. Wanke, P., Barros, C.P., Faria, J.R., 2014. Financial distress drivers in Brazilian banks: a dynamic slacks approach. Eur. J. Oper. Res. 240 (1), 258–268. Williams, J., Nguyen, N., 2005. Financial liberalisation, crisis, and restructuring: a comparative study of bank performance and bank governance in South East Asia. J. Bank. Financ. 29, 2119–2154. Yang, C., Liu, H.-M., 2012. Managerial efﬁciency in Taiwan bank branches: a network DEA. Econ. Model. 29, 450–461. Yeh, Q.-J., 1996. The application of data envelopment analysis in conjunction with ﬁnancial ratios for bank performance evaluation. J. Oper. Res. Soc. 980–988.