Optimization of Accounting information System in Public Sector for Sustainable Risk Management Under Big Data Analytics. Does forensic Accountants’ Skill Generate Differences?

The RBV theory founded by Barney (1991) was employed to illuminate the reason to which numerous entities have been operated efficiently and effectively as well as how an entity could operate more fruitfully. Under the circumstances of BD, organizational capital resources placed its emphasis on insights from BDA to facilitate the shift in organizational processes and structure. While the physical capital resources covered with software reinforced the entity to perform the accumulation and analysis of BD, the human capital resources reflected on the competence and expertise of data scientists utilized to conduct analysis and management of information pertaining to the organizational operations (Gunasekaran, et al., 2017).

BDA has been broadly acknowledged as an advanced type of technologies and architectures which have been shaped up to economically extract value from very great volumes of a series of data by facilitating high velocity procurement, breakthrough, and/or analysis (Mikalef, et al., 2018). Numerous academic notes have advocated that BDA consisted of three elements such as BDA management capability, BDA technology capability, and BDA talent capability (Akter, et al., 2016).

It is well knowledge that AIS is a software program that organizations use to manage and oversee their economic and financial activities, as well as to enter and process financial data (Al-Hattami, et al., 2021). In this context, “AIS” meant an information system that gathered data, organized it, ran analyses on it, stored credentials for future use, and then communicated the results to both internal and external stakeholders so that they could make informed decisions (Andiana, 2022).

The term “sustainability risk” was used to refer to risks that were associated with issues pertaining to social justice or the environment and a number of different strategies that might be utilized to exert influence over entities (Anderson, 2005). This study defined SRM as a risk management that prioritizes the effects of both internal and external stakeholder value creation, in line with the work of Schulte and Knuts (2022).

The field of forensic accounting has long been thought of as a haven for those seeking specialized knowledge in areas such as law, accounting, auditing, and evaluation (Renzhou, 2011). Building on the perspectives of Crain, et al. (2015), it is suggested that these accountants broaden their skill sets to include AIS, digital forensics, auditing processes, and cognitive abilities.

The study’s purpose is accomplished through the use of a research design that employs a hypothetico-deductive technique. This approach involves reviewing the relevant literature, formulating hypotheses, constructing a theoretical framework, and deriving logical conclusions based on the research findings (Sekaran, 2006). The choice of research technique is influenced by the research paradigm as it has a substantial effect on the whole approach (Collis and Hussey, 2013). Therefore, a deductive procedure employing a survey questionnaire and adopting a positivist perspective is used. To perform the statistical analysis and make generalizations about the entire population, researchers typically use survey methods to collect data from a sample (Collis and Hussey, 2013). A quantitative technique utilizing a structured questionnaire is employed to validate the relationships in the proposed model (Selem, et al., 2022).

The survey method has been felicitous for capturing data in a context in which the variables examined were related with organizations as well as professional practices (Hair, et al., 2017). Simultaneously, the selection of the survey technique to data procurement has been deemed to be appropriate with investigations in the management and organization research literature (Awan and Sroufe, 2019; Chundakkadan and Sasidharan, 2019) as it has been a common technique for contacting an enormous sample size of a specific population at a low cost (Heeringa, et al., 2017). Remarkably, there has been lack of readily available secondary database on digital transformation within PSOs, especially in developing nations like Vietnam’s. Building on the proposals of Hennessy and Patterson (2011), the research instruments were the most important things to be focused when using survey technique. Consequently, the development of research instrument is commenced.

The following procedures are intended to address the measurement instrument. Initially, a clear explanation of the constructs is provided. Subsequently, the existing literature and assessment tools used in previous research are examined to assess these similar components. Once the measuring constructs are finalized, the survey questionnaire is created with closed-ended questions that offered numerous options using a five-point Likert scale. This tool is highly valuable and efficient for quantitative research (Grove and Gray, 2018). The reason for using closed questionnaires is their ease of completion for participants and suitability for analysis (Fiaz, et al., 2018). Additionally, it has received widespread recommendations from researchers in this field, further attesting to its quality and reliability (Revilla, et al., 2014). Thus, the five-point Likert scale to evaluate the degree of the respondents’ approval with the statements (1 = “vigorously disagree” and 5 = “vigorously agree”) was applied for all measurement items in the questionnaire.

Initially, the questionnaire was first designed in English and redesigned in the Vietnamese language through a back-translation procedure to overcome the discrepancies of translation (Brislin, 1970). As the questionnaire in this study was established in divergent contexts – both culturally and environmentally – a pretest with 6 experts was implemented to minimize unanticipated complexity (Alreck and Settle, 1995). Grounded on the feedback of experts, the constructs were modified to remove ambiguous measuring items. Prior to gathering the questionnaire responses, a pilot study was done with multiple participants to ascertain any potential challenges in completing the set of questions. As such, 30 respondents with similar idiosyncrasies to the survey population were demanded to partake in the smallscale pilot test. The Cronbach’s alpha is utilized to assess the reliability and consistency of the research instruments (Hair, et al., 2012). The results indicate that the Cronbach’s alpha coefficient for each item exceeds 0.70. Cronbach’s alpha scores ranging between 0.50 and 0.70 are considered appropriate for ensuring the internal consistency of a reliable scale (Hinton, et al., 2014). The results of the pilot testing demonstrated that the items meet the criteria for assessing their reliability, and all of the items are retained for further analysis. Consequently, the techniques were deemed valid and reliable for collecting empirical data. Remarkably, Ismail, et al. (2018) suggest that the sample and data collected during the pilot study should not be utilized in the main investigation.

The items measured for BDA which comprised BDA Management Capability, BDA Technology Capability, and BDA Talent Capability were inherited from the contributions of Akter, et al. (2016) and Kim, et al. (2012). The AIS which included four components, namely, data input system organization, data processing system organization, data storage system organization, and financial statement system organization were stemmed from the findings of Uyar, et al. (2017) and Romney, et al. (2013). The criteria which were employed to evaluate the SRM consisted of Sustainability Assessment, Sustainability Risk Identification, Sustainability Risk Analysis and Evaluation, and Sustainability Risk Treatment and Communication were emanated from the outcomes of Schultea and Knuts (2022). The measurement scales utilized to assess the FAS were sprung from the work of Digabriele (2008).

In order to accurately reflect the research population and collect necessary primary data, data are obtained through the use of two sample units. Organizations serve as the primary sampling unit, while employees within such organizations are considered the secondary sampling unit. Most of the PSOs in Vietnam have started considering the deployment of digital technologies. The choice of PSOs in the southern regions of Vietnam for this study is based on the idea that empirical research in this area can provide useful references for PSOs across Asia, not just in Vietnam (Huy and Phuc, 2020). Alternatively, as the accountants would play a crucial role in achieving a favorable outcome as a result of the widespread adoption of digital technologies (Zybery and Rova, 2014), the participants in this study are accountants from various PSOs in the southern region of Vietnam. The present study centers on accountants in PSOs who possess the ability to provide perceptive analysis and extensive expertise on relevant subjects from the standpoint of their respective organizations. The selection criteria for participants include the sufficient expertise and knowledge in digital technologies and risk management, as well as a minimum of 5 years of working experience in their respective organizations, which aims to confirm their active involvement in AIS design. Furthermore, participants are requested to answer questions regarding their familiarity with BDA application, AIS, and risk management implementation in a digitalized business environment to ensure that they had sufficient knowledge to complete the questionnaire. This process ensures that the dataset does not include any participants who are unaware of these concerns. The primary data in the current study are captured through a paper- and-pencil questionnaire. To do so, the questionnaires are distributed in person to participants by the researchers. The list of PSOs is acquired with the authorization of the Department of Finance in each province of Southern Vietnam. All the data studied in this research are generated by a solitary participant in each PSO. Prior to requesting employees to participate in the study, the researchers sought approval from the senior management of the PSOs to collect their contact information. Following the subjects’ informed agreement, the researchers personally distributed the surveys to them. By doing this, researchers would be able to inform participants on the correct procedure for completing questionnaires, reduce the potential bias caused by the common technique, and provide information regarding the confidentiality and anonymity of the study’s results. Participants were guaranteed confidentiality and anonymity, and they had the freedom to withdraw from the study at any point, without any restrictions or explanations required.

The sample of this research is formulated on the basis of convenience and snowball sampling. Convenience sampling is a form of nonprobability sampling in which individuals of the target population meet criteria, namely, easy accessibility and availability (Taufique and Vaithianathan, 2018). In the meanwhile, snowball sampling is an acknowledged and viable approach of recruiting research respondents who are difficult to accessible or known to the researchers (Marcus, et al., 2017; Naderifar, et al., 2017; Reagan, et al., 2019; Wohl, et al., 2017). Based on the standpoints of numerous scholars, the sample-to-variable ratio preferred from 5:1 (minimal) to 10:1 (optimal) and the sample-to-item ratio is proper to determine the sample size (Hair, et al., 2010). On the suggestion of Comrey and Lee (1992), a sample size below 50 is deemed extremely weak, ranging from 51 to 100 is-comparatively weak, varying from 101 to 200 is relatively adequate, differing from 201 to 300 is good, and fluctuating from 301 to 500 is exceedingly good, whereas a sample scale of larger than 500 is recognized to be excellent. As proposed by Hair, et al. (2022), the responses that suffer from more than 15% missing items would be excluded from the data set. Therefore, all the responses that suffer from 15% of missing data would be eliminated. Thus, a total of 800 questionnaires are distributed to accountants of PSOs in the southern region of Vietnam from the end of November 2021 to the middle of May 2022. There are 683 valid questionnaires captured with the response rate of 85.38 percent.

All the calculations for the hypothesized model were performed with the support of SPSS v27 and AMOS v26 through a two-step analytic approach (Anderson and Gerbing, 1988). More instrumentally, the instrument substantiation was carried out through assessing the construct validity and reliability by means of Cronbach’s alpha, exploratory factor analysis (EFA) with the support of SPSS v27, and confirmatory factor analysis (CFA) with the support of AMOS v26. In order to testify the hypothesized interconnections, SEM analysis was leveraged through Amos v26. The maximum likelihood estimation approach was taken advantage to evaluate the measurement and structural model as well (Moon and Kim, 2001). Furthermore, the multi-group SEM analysis was also made use to scrutinize the moderating impacts of FAS.

Among 683 valid survey questionnaires, 61.20 percent was female and 38.80 percent was male. Of these, the group of accountants working in the public schools accounted for 73.94 percent, which was followed by the group working in local government around 22.40 percent. The group working in preventive health centers and public hospitals, public universities, public colleges, public libraries, and museums made up negligible percent around 0.73 percent, 1.17 percent, 0.44 percent, 0.73 percent, and 0.59 percent, respectively, ranking last among the given groups. The age of groups was allocated with 15.67 percent “Below 30,” 47.00 percent “31–40,” 35.43 percent “41–50,” and 1.90 percent “Above 51.” All of the informants obtained an undergraduate degree. Concerning to the year of experience, 15.67 percent belonged to “Below 10” group, 71.30 percent was “10 to Below 20” group, and 13.03 percent was the “20 to Below 30” group.

Prior to implementing EFA, it was compulsory to perform Bartlett’s Test of Sphericity as well as Kaiser– Meyer–Olkin test to gauge the sufficiency of the samples. The outcomes in Table 1 underlined that the sampling was sufficient and the data were appropriate for factor analysis.

The Cronbach’s α and composite reliability were recommended to be higher than the referenced value 0.7 (Fornell and Larcker, 1981a), and all the indicators in the constructs should have loadings exceeding the threshold of 0.5 (Hair, et al., 2010). Additionally, the convergent validity was judged to be sufficient when the average variance extracted (AVE) equaled or exceeded 0.5 (Fornell and Larcker, 1981a). The statistical corroboration of these criteria in Table 2 ascertained the convergent validity and internal consistency reliability.

The comparison between the square root of the AVE and the correlations among the proposed constructs was conducted to determine discriminant validity. The AVE of each latent variable was greater than the off-diagonal correlation scores (Fornell and Larcker, 1981b). Based on the output in Table 3, the measurement scale illustrated the good discriminant validity.

The chi-square/df value was requested to be below 2 (Byrne, 1989). The comparative fit index (CFI), Tucker–Lewis index (TLI), and goodness-of-fit index (GFI) were requested to obtain the score equal to/beyond 0.9 (Schumacker and Lomax, 2004). The rootmean-square error of approximation (RMSEA) was proposed to be equal to/under 0.05 (Joreskog and Sorbom, 1989). The measurement and structural models in Table 4 were in line with the collected data in a flawless manner.

Upon careful analysis of the study’s findings, it is possible that certain inherent limitations may become apparent despite the researchers’ diligent efforts to ensure the research’s credibility. The limits that have been placed will facilitate the development of innovative ways for the ongoing research in this field. The primary challenge that hinders making definitive statements in the analysis of results pertains to the utilization of cross-sectional data. Alternatively, it might be stated that inferences on the study can solely be made in relation to the shared connections among the variables under consideration, as it solely provides fixed viewpoints on compatibility. Hence, it is recommended that future investigations incorporate contemporary statistical methodologies in conjunction with a thorough longitudinal analysis. One of the drawbacks is the significantly small sample size, often insufficient to yield definitive findings. Therefore, it is advisable to proportionally increase the sample size in relation to the amount of time and effort expended. One notable limitation pertains to the composition of the sample. The current study relies predominantly on samples, notably a limited number of informants from a single institution. On the other hand, the generalizability of the findings in this study is limited due to its exclusive focus on data sets of Vietnamese origins. To ensure the validity and applicability of the results, it is necessary to conduct comparable investigations in other nations. In the context of ground testing the suggested model, a higher sample size would be advantageous for comparative analysis. Furthermore, even with the implementation of various strategies designed to enhance the fairness of the individuals involved, there exists a potential for the emergence of self-serving bias from the respondents as a result of the survey inquiries (Oduro and Haylemariam, 2019; Murray, et al., 2011). Future study should consider utilizing secondary data sources as a means to overcome this constraint. In order to enhance the comprehensiveness of the research, it is imperative to consider other statistical approaches, as indicated by Abubakar, et al. (2019), due to the relatively lower predictive capabilities of correlation, regression, and structural equation modeling. This issue may be easily rectified and is not anticipated to pose any future complications.