The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1030-9616.htm

Big data
Big data analytics and auditor analytics and
judgment: an experimental study auditor
judgment
Ranto Partomuan Sihombing
Department of Accounting, Faculty of Economics and Business,
Airlangga University, Surabaya, Indonesia and Department of Accounting, 201
Faculty of Economics and Business, Soegijapranata Catholic University,
Received 2 August 2022
Semarang, Indonesia, and Revised 23 November 2022
5 February 2023
I Made Narsa and Iman Harymawan Accepted 7 February 2023
Department of Accounting, Airlangga University, Surabaya, Indonesia

Abstract
Purpose – Auditors’ skills and knowledge of data analytics and big data can influence their judgment at the
audit planning stage. At this stage, the auditor will determine the level of audit risk and estimate how long the
audit will take. This study aims to test whether big data and data analytics affect auditors’ judgment by
adopting the cognitive fit theory.
Design/methodology/approach – This was an experimental study involving 109 accounting students
as participants. The 2
2 factorial design between subjects in a laboratory setting was applied to test the
hypothesis.
Findings – First, this study supports the proposed hypothesis that participants who are provided with
visual analytics information will rate audit risk lower than text analytics. Second, participants who receive
information on unstructured data types will assess audit risk (audit hours) higher (longer) than those
receiving structured data types. In addition, those who receive information from visual analytics results have
a higher level of reliance than those receiving text analytics.
Practical implications – This research has implications for external and internal auditors to improve
their skills and knowledge of data analytics and big data to make better judgments, especially when the
auditor is planning the audit.
Originality/value – Previous studies have examined the effect of data analytics (predictive vs anomaly)
and big data (financial vs non-financial) on auditor judgment, whereas this study examined data analytics
(visual vs text analytics) and big data (structured and unstructured), which were not tested in previous
studies.
Keywords Big data, Data analytics, Audit risk, Audit hours
Paper type Research paper

1. Introduction
Today, big data has been used in almost al aspects of the decision-making and business
strategy of large companies. For example, a company in the USA may process one billion
data elements every day to understand its competitive environment (Griffin and Wright,
2015). They also say that much data processing can lead to bias because of human cognitive
limitations in processing information. Thus, big data will change the audit process to be

Accounting Research Journal

Vol. 36 No. 2/3, 2023
The researchers would like to thank Dr Monika Palupi Murniati, a Senior Lecturer at Accounting pp. 201-216
Department, Soegijapranata Catholic University, who has given input in designing the experimental © Emerald Publishing Limited
1030-9616
scenario. DOI 10.1108/ARJ-08-2022-0187
ARJ carried out (Griffin and Wright, 2015; Warren et al., 2015; Appelbaum et al., 2017; Tang and
36,2/3 Karim, 2019; Salijeni et al., 2021).
According to Vasarhelyi et al. (2015), the emergence of new forms of automatically
generated data with increasingly large sizes drives technology adoption for the
assurance process. The use of data analytics can assist auditors in reducing cognitive
errors caused by the large and varied number of data (Ahmad, 2019). Previous studies
202 examined the effect of data analytics on auditor judgment and found that data analytics
can improve auditor judgment (Brown-Liburd et al., 2015; Kend and Nguyen, 2020;
David and Abeysekera, 2021).
We suspect that other types of data analytics can affect auditor judgment. A recent
study conducted by Koreff (2021) found that the data analytics model (predictive vs
anomaly) and data type (financial vs non-financial) did not affect auditor judgment.
Based on Koreff’s (2021) suggestion, this study aims to examine other data analytics
models (visual vs text) and big data types (structured vs unstructured) on auditor
judgment.
The development of both visual and text analytics software that is increasingly
sophisticated today can assist auditors in identifying big data, both structured and
unstructured data types, in supporting auditor judgment (Aldhizer, 2017; Sun, 2019). In
particular, Sun (2019) explains two benefits that auditors get when using audit data
analytics. First, they gain deeper insight by identifying information from structured and
unstructured data. Second, they get better support in making judgments because structured
and unstructured data can be used as additional audit evidence in large and diverse
amounts.
We adopt cognitive fit theory to explain the logic of the hypothesis formulation.
According to Vessey (1991), this theory explains the relationship between technology
suitability and the task to perform. He added that individuals would be faster and more
precise in making decisions when the technological aids used can match the decisions to be
made.
Our study findings contribute to the literature and practice. First, participants who
receive information from visual analytics will assess lower audit risk than text
analytics. Second, participants who receive unstructured data type information will
assess audit risk (audit hours) higher (longer) than participants who receive structure
data type information. Moreover, participants rely more information derived from
visual analytics than that from text analytics. The practical contribution is aimed at
external auditors so that they would be more confident in using big data and data
analytics in audit panning.
Section 2 of this paper discusses literature review, Section 3 discusses theory and
hypotheses development, Section 4 discusses the research methods and experimental
assignment. Section 5 discusses the test results. At the end, Section 6 presents the discussion
and conclusions.

2. Literature review
The auditors’ main task is to ensure that the procedures and judgments they carry out are
supported by sufficient and valid data (Balios et al., 2020; Deniswara et al., 2020). However,
today’s rapid technological developments have led to the emergence of various forms of very
large and complex amounts of data (big data). According to Gepp et al. (2018), big data refers
to a collection of structured or unstructured data, generally described according to 4Vs:
volume, variety, velocity and veracity. Volume refers to large data sets, and variation
refers to data formats such as quantitative forms, text, images, videos and other formats.
Velocity refers to the increasingly frequent availability of new data at very high speeds. Big data
Veracity refers to the quality and relevance of data that can change dramatically over time. analytics and
The term big data analytics (BDA) refers to an analytical technique used to examine all
auditor
procedures for receiving large amounts of data that helps individual to find errors or fraud
so that it is useful in making a decision (Deniswara et al., 2020). judgment
A systematic review study found that BDA helps auditors make better judgments in
various tasks performed (Appelbaum et al., 2018; Ahmad, 2019; Aboagye-Otchere et al.,
2021). In the planning aspect, the auditor makes judgments in determining the level of risk
203
and the estimated time needed to complete the audit assignment (Mactavish et al., 2018;
Dagiliene_ and Kloviene_ , 2019). Although BDA can assist the auditor in processing data, it is
the auditor who analyzes and evaluates data patterns through a cognitive process (Brown-
Liburd et al., 2015; Griffith et al., 2021). Besides, Griffith et al. (2021) found that auditors with
sufficient cognitive capacity can complete analytical processing well.
Holt and Loraas (2021) examined the factors that influence auditor judgment from the
perspective of cognitive load theory. This theoretical perspective explains that overload
information can cause individuals to experience excessive stress, so that decision-making is
made not based on information processing but on heuristics or ending information
processing earlier to reduce cognitive stress (Chandler and Sweller, 1991). In contrast to Holt
and Loraas (2021), our study looks at it from the perspective of cognitive fit theory. This
theory presents the idea that to eliminate cognitive load, we must create an appropriate
structure. If the appropriate structure in presenting the task can be aligned with how
decision-makers think about the task, then the cognitive load will be reduced allowing
individuals to process information fully (Vessey, 1991).
The focus of our study is to examine whether there is structural suitability between the
types of data analytics (visual analytics vs text analytics) and the types of big data
(structured vs unstructured) with the judgments made by the auditors. We adopted the
concept described by Aldhizer (2017). The author explained that visual analytics would be
more fit in processing data and producing structured-type information. On the other hand,
text analytics will be more suitable for processing data and producing information of an
unstructured type. We present a detailed explanation in the hypothesis development section.

3. Theory and hypotheses development

This section explains the theory used to formulate the hypothesis. As we explained earlier,
the cognitive fit theory is a theory that explains the relationship between the suitability of
the technology used and the decision-making process (Vessey, 1991). Furthermore,
individuals will make better decisions when they get appropriate technology support for the
decisions to be taken.
The auditor will choose to use data types (structured vs unstructured) or data analytics
(visual vs text). Based on the literature, visual analytics will be more suitable for decisions
with structured types. In comparison, text analytics will be more fit for making decisions
with unstructured data types. Koreff (2021) adds that auditors will experience a higher
cognitive fit when concluding familiar data processes.
Data analytics can assist auditors in processing a large number of data and different
results from company transactions digitally so that the audit process can run effectively and
efficiently (Appelbaum et al., 2021; Hamdam et al., 2021; Holt and Loraas, 2021; Rakipi et al.,
2021). According to Elgendy and Elragal (2014), data analytics applies algorithms to analyze
data sets; extract unknown patterns, relationships, and information; and detect important
relationships between data variables.
ARJ Previous studies have found that data analytics in audits affects auditor judgment
36,2/3 (Brown-Liburd et al., 2015; Ahmad, 2019; Hamdam et al., 2021; Holt and Loraas, 2021). In
particular, Brown-Liburd et al. (2015) mentioned that auditors could use data analytics to
extract information from many data to help them focus on high-risk audit areas. In addition,
Ahmad (2019), Hamdam et al. (2021) and Holt and Loraas (2021) found that data analytics
could reduce cognitive errors and biased judgments and help auditors to gain a deeper
204 understanding of transactions in the company.
Two data analytics models tested in this study are visual analytics and text analytics.
Visual analytics is a method of analytical reasoning facilitated by interactive visual displays
(Risch et al., 2008). According to Aldhizer (2017), visual analytics can produce more detailed
analysis for structured data types to more easily reveal hidden relationships between data
elements. Moreover, visual analytics can help reduce the risk of undetected material
misstatement, including the risk of fraud.
The visual analytics model often used to detect fraud is heat maps (Singh and Best,
2016; Leite et al., 2018; Amin et al., 2020). Moreover, Aldhizer (2017) explains that this
model provides a graphical representation of different data elements, such as the
perceived level of fraud risk in the client’s business processes based on the probability of
occurrence and the level of materiality. Different colors represent these representations.
For example, when the possibility of fraud is low and the materiality level is small, it is
marked green. When the possibility of fraud and its materiality is moderate, it is marked
yellow. Finally, when the possibility of fraud is high and the materiality is large, it is
marked in red.
Text analytics is the process of extracting and generating useful information from
unstructured data (Chen et al., 2012). The text analytics model can identify the intention to
commit corruption based on the analysis of relevant keywords to determine the level of
audit risk and determine to what extent the level of forensic audit involvement is (Sun and
Vasarhelyi, 2018; Wang and Xu, 2018).
The text analytics model widely used to detect fraud is link analysis. According to
Aldhizer (2017), link analysis is a method to identify who is talking to whom, about what
and when. This method can also evaluate the relationship between individuals and
organizations. For example, this analysis can pinpoint the sender and recipient of an
email or social network to identify suspects and associates who committed fraud.
Through concept extraction, text analytics complements visual analytics. Text analytics
can also analyze structured data, which is not possible with visual analytics software. The
analysis results from text analytics can then be integrated with visual analytics to provide a
deeper view of the potential fraud schemes (Aldhizer, 2017).
Based on the explanation of the two methods above, we consider that auditors
provide a low audit risk assessment when receiving information from visual analytics
compared with text analytics. In addition, the estimated audit time (budget audit hours)
will be shorter when the auditor is given information on the results of visual analytics
compared with text analytics because visual analytics helps auditors to focus on high-
risk audit areas so that the required audit time is more efficient (Amani and Fadlalla,
2017). The first hypothesis to be tested is as follows:

H1a. Participants who receive visual analytics information will assess audit risk as
lower than text analytics.
H1b. Participants who receive visual information will assess shorter audit hours than
text analytics.
In addition to the data analytics model that has been discussed previously, the suitability of Big data
the type of big data with the analytics model will affect the auditor’s judgment. In the analytics and
context of big data, non-financial data is divided into two, namely, structured and auditor
unstructured data (Eberendu, 2016; Richins et al., 2017). They explain that structured data is
generated through a company’s transaction processing systems, such as point-of-sale
judgment
systems, inventory management systems and customer/supplier relationship management
systems. This type of data is highly organized to be easily stored in the company database. 205
In contrast, unstructured data comes from social media such as Facebook, Twitter,
YouTube and email in various forms such as text, audio and video (Edge et al., 2018).
Most types of big data generated are unstructured data. One of the reasons why it is
important to understand unstructured data is to detect the possibility of financial fraud
(Sharma et al., 2016; Tang and Karim, 2019). In particular, the study results by Sharma et al.
(2016) found that fraud detection from unstructured data is a way to reduce fraud to a
certain level. Furthermore, David and Abeysekera (2021) and Yang et al. (2018) found that
the greater the number and variety of unstructured data types, the more the audit risk and
audit costs would be. Thus, we consider that unstructured data types cause higher audit
risks and costs than structured data types. The second hypothesis tested is as follows:

H2. Participants who receive unstructured data type information will assess audit risk
(audit hours) higher (longer) than structured data types.
Aldhizer (2017) explains that the appropriate data types must support the data analytics
model. According to him, the visual analytics model will fit when analyzing structured data
types. On the other hand, the text analytics type model will fit when analyzing unstructured
data.
Hamdam et al. (2021) explained that visual analytics could assist auditors in finding
meaning and framing relevant information patterns for decision-making. This visual
analytics approach requires integrating data from structured databases (Keim et al., 2008;
Dilla and Raschke, 2015; Singh and Best, 2016; Rodríguez-Quintero et al., 2021). In particular,
Rodríguez-Quintero (2021) found that the magnitude of the data integration process and the
effectiveness of subsequent audits depend on the maturity level of the data traces stored in
the operational system, the quality of the information contained in the business database
and the structural complexity of the process and object of analysis.
Based on statistical data, 80% of business data is unstructured and needs to be extracted
through text analytics to be useful for business people (Anandarajan et al., 2019). In contrast,
the text analytics approach is suitable for unstructured data types (Sun and Vasarhelyi,
2018; Wang and Xu, 2018). Sun and Vasarhelyi (2018) explained that text analytics could
analyze the content of social media posts and news articles that can inform auditors about
potential litigation risks, business risks, internal control risks and fraud risks by
management. The third hypothesis tested is as follows:

H3. Participants who receive text analytics information with unstructured data types
will assess audit risk (audit hours) higher (longer) than visual analytics with
structured data types.

4. Research method
We use the experimental method to test the effect of the independent variable on the
dependent variable. The factorial design applied is 2
2 between subjects. According to
ARJ Geng and Kalargiros (2022), the experimental method is more suitable for testing causal
36,2/3 relationships, especially those related to judgment.

4.1 Participants
The participants of this study were undergraduate students of the accounting study
program at Unika Soegijapranata Semarang who had passed auditing courses and
206 computer accounting practicum. In auditing courses, they have sufficient knowledge about
how an audit plan will be carried out, whereas in the accounting computer practicum course,
they have sufficient knowledge about the benefits of data analytics in analyzing big data.
Furthermore, students acted as auditors at a hypothetical public accounting firm called
“KAP Harmony,” which audited a hypothetical company named “PT Tera Medika” in 2021.
Their task as participants was to make judgments, namely, determining audit risk and
estimated audit hours.
Apart from academic considerations, according to Liyanarachchi (2007), there are no
different results of experimental studies with real participants or students. Furthermore,
Nahartyo (2013) explains that students’ use as participants in an experiment can be done if
the researcher only wants to examine how cognitive processes are involved in making a
decision process. To improve internal validity, students’ use as subjects will give better
results than real participants (Mattila et al., 2021). For example, a study conducted by
Parsons (2007) involved students as participants in examining the impact of voluntary
disclosure of information on the decision-making process.
Based on the results of the questions that passed the manipulation check, the number of
the data that could be processed was 109 (90.83%) out of 120 students who participated in
this experiment. Three people did not pass Stage 1 of the manipulation test, two did not pass
Stage 2 of the manipulation check and two did not pass the Stage 3 manipulation test. The
remaining four people filled out the case material given incompletely. Table 1 presents the
demographic data of the subjects.
Table 1 shows that the subjects who participated in this study mostly were women with
55%, while the number of men was 45%. Subjects aged 20 years were 48.6%, while those
aged 21 years were 51.4%. Subjects who got an A for auditing courses are 33%. The number
of subjects who got an A was 40.4%. The remaining 26.6% got lower than A. The
number of subjects who got a grade point average (GPA) < 3.00 was 17.4%. Meanwhile,
the number of subjects scored between 3.01 and 3.5 was 40.4%. A total of 42.2% were the
subjects who got a score > 3.5.

Variable Range Frequency %

Gender Male 49 45
Female 60 55
Age (years) 20 53 48.6
21 56 51.4
Auditing course grades A 36 33
A 44 40.4
< A 29 26.6
Grade point average (GPA) <3.00 19 17.4
3.01–3.5 44 40.4
Table 1. > 3.5 46 42.2
Subject demographic
data Source: Author’s own creation
The subject’s demographic data in gender, age, grades and cumulative achievement index Big data
(GPA) were further analyzed to determine whether the data were normally distributed. analytics and
Based on the skewness and kurtosis test results, the data is normally distributed because the
skewness value is between 1 and þ1, and the kurtosis value is between 3 and þ3 (Field,
auditor
2013). Analysis of variance tests conducted on the demographic variables of gender, age, judgment
grades and GPA as covariates did not show a significant effect (p-value > 0.05) on the
dependent variable.
207
4.2 Independent variables
The independent variables tested in this study are data analytics (text and visual analytics)
and big data (structured and unstructured). We manipulated by informing participants
about the findings of the big data and data analytics analysis conducted by the “KAP
Harmony” Information Technology (IT) divison.
Text analytics were manipulated by informing how text analytics analyzes big data and
the resulting output. In the text analytics output section, we explain that text analytics can
identify the possibility of fraud from both structured data (sales contract data) and
unstructured data (email conversation data) to help auditors focus more on data that has the
potential for fraud to occur. The same way we do for visual analytics manipulation. The
difference is that manipulation of text analytics is not accompanied by visual output, while
visual analytics is accompanied by visual output in visual heatmaps (Aldhizer, 2017). Text
analytics is given a score of 1, while visual analytics is given a score of 2.
We manipulated structured data by informing concepts and examples of structured data
types. Sales contract data as one of the structured data types owned by “PT Tera Medika”
was described. Meanwhile, manipulation of unstructured data was done by informing the
concept of unstructured data. Email conversation data is unstructured data owned by “PT
Tera Medika” (Eberendu, 2016; Richins et al., 2017). Additionally, structured data is easy to
analyze because of the small amount stored in the company database. In contrast, the
amount of unstructured data is not easy to do because a lot of it is caused by the transaction
activities carried out by “PT Tera Medika” online (Anandarajan et al., 2019). Structured data
was given a score of 1, while unstructured data was given a score of 2.

4.3 Dependent variables

The dependent variables in this study are audit risk and estimated audit time. The audit risk
variable measures how high the auditor determines the audit risk. The higher the score, the
higher the audit risk. This variable was measured using a Likert scale of 1–7. The audit time
estimation variable measures how many time changes were made from the 30 h of the initial
budget estimate time set. Each addition or subtraction of 1.5 h increased or decreased the
estimated time by 5%. The measurement scale used is from 100% to 100%. Both
measurements were adopted from a previous study conducted by Koreff (2021). If
participants decide to increase the estimated audit time, they choose a number that has a
positive value. On the other hand, if the participants decided to reduce the estimated audit
time, they chose a negative value number.

4.4 Experimental procedures

Before conducting experiments on actual participants, researchers first conducted a pilot
test. This aims to determine whether each stage of the case scenario that will be given can be
understood well and whether the instructions for filling out have been delivered
correctly and to determine the time allocation for implementation (Schindler, 2019;
ARJ Handoko et al., 2021). In addition, pilot tests are useful for finding out whether there are
36,2/3 errors in the experimental design (Walker, 2014).
The experimental protocol for all participants was carried out in a classroom (laboratory
experiment). Several regular classes have been held because the Covid-19 case in Indonesia
in April 2022 has been sloping. The experiment was divided into four different sessions.
Each session consists of 30 subjects. The goal is to keep the distance to implement the health
208 protocol. According to Nahartyo (2013), we have a high level of control over confounding
variables in laboratory experiments. We can obtain high confidence that there is a causal
relationship between the independent and dependent variables. He also said that laboratory
experiments are suitable for testing theories that explain the causal relationship.
All participants were assigned randomly to one of four cells through an arranged seating
mechanism and the distribution of case materials to be worked on. The goal is that the
participants in each cell can be varied so as not to cause bias. Although the participants in
this study were students, their participation in this experiment was voluntary. They register
first through the Google Form to participate in this experiment. To increase the seriousness
of participants in this experiment, participants can be given incentives that will not directly
affect the test results (Nahartyo, 2013). In this study, researchers gave e-money as an
appreciation for involvement in this experiment.
We adopted the experimental scenario developed by Koreff (2021). Of course, we have
adjusted the experimental design according to the needs of this study. Participants must
follow five stages. Each stage is followed by participants simultaneously according to the
code given by the researchers. The goal is to reduce the threat of maturation because the
subject sees that his partner has already finished working on the case material experiment,
so there is a desire to finish it immediately without being accompanied by accuracy in
reading every information contained in the case material.
After the case material experiment was distributed to participants, the researchers
presented the entire series of assignments that would be carried out. The goal is that
participants are careful in reading all the information presented. Before starting Stage 1,
participants read the introduction, which contains a brief description of the implementation
of the experiment. In Stage 1, participants receive brief information about “PT Tera Medika”
(auditee), the previous year’s audit results, the level of inherent risk, control risk and fraud
risk assessment for the current year. In Stage 2, participants received information about the
capability of the IT division of “KAP Harmony” in conducting big data analysis using data
analytics. Giving the same information in Stages 1 and 2 is a control mechanism used by the
researchers because this study did not use a control group. Each participant gets the same
information that can influence participants in making audit judgments.
In Stage 3, each group received different information according to the manipulations
carried out by the researchers. The participants finished reading the information in Stages 1, 2
and 3. At the end of each stage, the researchers performed a manipulation check. This aims to
motivate the subject to be more thorough in reading important information used to carry out
the task at a later stage. If the participant does not pass the manipulation test, it will not be
included in the hypothesis testing. In Stage 4, each group will make a judgment. In the final
stage, all participants will fill in demographic data.

5. Results
5.1 Manipulation checks
To find out the participants’ understanding of the manipulation of research variables in the
experiment, questions are asked at the end of Stages 1–3 with a different number for each
stage. At the end of Stage 1, there are two manipulation check questions. At the end of
Stage 2, there is one manipulation check question, while at the end of Stage 3, there are two Big data
manipulation check questions. At the end of Stage 1, the first question is to ensure that the analytics and
participant is aware of his or her role as the audit team leader. The second question is about
the level of inherent risk, control risk and fraud risk determined by the auditor for the “PT
auditor
Tera Medika” audit. Questions at the end of Stage 2 are to ensure that participants know judgment
that the IT division of “KAP Harmony” has adequate capabilities in carrying out data
analytics even though they do not have an accounting education background.
At the end of the third stage, there were two different questions for each group according 209
to the treatment received by the participants. The first question ensures that participants
understand “PT Tera Medika’s” structured or unstructured data types. The second question
is to ensure that participants understand the concept of text analytics or visual analytics
carried out by the IT division of “KAP Harmony.” All types of questions are of multiple-
choice.

5.2 Descriptive statistics

Table 2 provides an overview of the treatment conditions for each cell. The number of
subjects who passed the manipulation of each cell was more than 25 people. This number is
following the rule of thumb presented by Mattila et al. (2021). The audit risk assessment
range is on a scale of 1–7, while audit hours are in the range of 100% to 100%.
Based on Table 2, in the visual analytics condition, the subjects rated audit risk as
lower with structured data types compared with unstructured data types (mean: 2.357;
sd: 0.869 < mean: 3.111; sd: 1.086). For the same condition, subjects rated audit hours lower
with structured data types compared with unstructured data types (mean: 1.964;
sd: 15.051 < mean: 13.148; sd: 15.326).
In the text analytics condition, the subjects rated audit risk as lower with structured data
types compared with unstructured data types (mean: 3.111; sd: 1.423 < mean: 3.888; sd:
1.527). Under the same conditions, subjects rated audit hours lower with structured data
types compared with unstructured data types (mean: 4.629; up to 20.044 < mean: 12.777; up
to 20.111).
For structured data type conditions, it can be seen that the audit risk assessment of
subjects receiving visual analytics treatment is lower than text analytics (mean: 2.357; sd:
0.869 < mean: 3.111; sd: 1.423). Under the same conditions, subjects who received visual
analytics treatment rated audit hours lower than text analytics (mean: 1.964; up to
15.051 < mean 4.629; up to 20.044).
Meanwhile, in unstructured data types, the subjects assessed audit risk as lower with
visual analytics treatment than text analytics (mean: 3.111; sd: 1.086 < mean: 3.888;

Big data
Data analytics Structured Unstructured

Visual analytics (n = 55) Mean audit risk: 2.357 (sd 0.869) Mean audit risk: 3.111 (sd 1.086)
Mean audit hours: 1.964 (sd 15.051) Mean audit hours: 13.148 (sd 15.326)
n = 28 n = 27
Text analytics (n = 54) Mean audit risk: 3.111 (sd 1.423) Mean audit risk: 3.888 (sd 1.527)
Mean audit hours: 4.629 (sd 20.044) Mean audit hours: 12.777 (sd 20.111)
Table 2.
n = 27 n = 27
Dependent variable
Note: SD = standard deviation descriptive statistics
Source: Author’s own creation (n = 109)
ARJ sd: 1.527). Under the same conditions, subjects who received visual analytics treatment
36,2/3 would rate audit risk higher than text analytics (mean: 13.148; up to 15.326 > mean: 12.777;
up 20.111).

5.3 Hypothesis testing

This study uses multivariate analysis of variance (MANOVA) to analyze the main effect and
210 interaction effect on the dependent variable, namely, audit risk and audit hours. Before
testing the hypothesis, we conducted a homogeneity test of variance. Based on the results of
the test of equality of covariance, it can be said that the sample variance is the same
(p > 0.05). Furthermore, the results of the MANOVA test can be seen in Table 3.

5.4 Tests of H1
The results of hypothesis testing can be seen in Tables 2 and 3. H1a states that participants
who receive visual analytics information will assess audit risk as lower than text analytics.
Table 2 shows that the mean risk audit visual analytics is lower than the mean risk audit
text analytics value for structured and unstructured data types. The results of the
MANOVA test in Table 3 show that F = 10.082; p < 0.05. H1a is supported.
H1b states that participants who receive visual analytics information will assess audit
hours shorter than text analytics. Table 2 shows that the mean value of audit hours of visual
analytics is lower than the mean value of text analytics in conditions of structured data
types. However, the mean value of audit hours of visual analytics is higher than the mean
value of audit hours of text analytics in conditions of unstructured data types. The results of
the MANOVA test in Table 3 show that F = 0.748; p > 0.05. H1b is not supported. This
finding is different from the study conducted by Eilifsen et al. (2020), who found that data
analytics can improve auditing efficiency.

5.5 Tests of H2
H2 states that participants who receive unstructured data type information will assess audit
risk (audit hours) higher (longer) than who receive structured data types. Table 2 shows that
the mean value of audit risk for unstructured data types is higher than the mean value for

Variable Type III SS df Mean square F p-value

Panel A: Audit risk

Big data 16.503 1 16.503 10.195 0.02**
Data analytics 16.319 1 16.319 10.082 0.02**
Big data
data analytics 0.001 1 0.001 0.000 0.983
Error 163.482 101 1.619
Total 1,251.00 109
Panel B: Audit hours
Big data 3,939.149 1 3,939.149 12.315 0.001***
Data analytics 239.345 1 239.345 0.748 0.389
Big data
data analytics 283.417 1 283.417 0.886 0.349
Error 32,307.74 101 319.879
Table 3.
Total 42,950.00 109
Results of the
auditor’s judgment Notes: ***p < 0.01; **p < 0.05
experiment Source: Author’s own creation
audit risk for structured data types in both visual analytics and text analytics conditions. Big data
The results of the MANOVA test in Table 3 show that F = 10.195; p < 0.05. analytics and
Moreover, Table 2 shows that the mean value of audit hours of unstructured data type is
higher than the mean value of audit hours of structured data type in both visual analytics
auditor
and text analytics conditions. The results of the MANOVA test in Table 3 show that judgment
F = 12.315; p < 0.01. H2 is supported.
211
5.6 Test of H3
H3 states that participants who receive text analytics information with unstructured data
types will assess audit risk (audit hours) higher (longer) than visual analytics with
structured data types. Table 3 shows that there is no interaction effect between text
analytics and unstructured data types on audit risk assessment because the value of
F = 0.000; p > 0.05. Besides that, Table 3 shows that there is no interaction effect between
visual analytics and structured data types on audit hours because the value of F = 0.886;
p > 0.05. H3 is not supported.

5.7 Supplemental analysis

To complement the results of hypothesis testing, this study examines the effect of data
analytics on the level of reliance in technology. The instrument was adapted from a previous
study conducted by Koreff (2021). We conducted an analysis of variance (ANOVA) test to
find out whether visual analytics and text analytics have an effect on reliance levels.
Table 4, Panel A, shows that the mean of the level of reliance generated from visual
analytics is higher than that of text analytics (mean: 5.687; up to 0.498 > mean: 5.392; up to
0.799). Furthermore, Table 4, Panel B, shows that data analytics has an effect on the level of
reliance (F = 3.990; p < 0.05).

6. Discussion and conclusions

This section discusses the results of testing the hypothesis based on the theory used. The
study’s results proved that big data and data analytics help auditors make audit plans,
especially in assessing audit risk and audit hours. This study produces three conclusions.
First, this research has succeeded in proving that data analytics has an effect on auditor
judgment, especially in determining audit risk. These results are supported by the arguments

Source Mean SD N

Panel A: Mean (SD)

Text analytics 5.392 0.799 54
Visual analytics 5.687 0.498 55
Total 5.541 0.678 109

Panel B: Level of reliance

Variable Type III SS df Mean square F p-value
Data analytics 1.722 1 1.722 3.990 0.048**
Table 4.
Error 44.456 103 0.432
Data analytics test
Notes: SD = standard deviation; **p < 0.05 results on level of
Source: Author’s own creation reliance
ARJ presented by Murphy and Tysiac (2015) and Krieger et al. (2021), who found that the use of data
36,2/3 analytics can help auditors focus more on high-risk client data to reduce audit risk.
Second, this study has proven that big data, both structured and unstructured, has an
effect on audit risk and audit hours. This result is different from the previous study
conducted by Koreff (2021). The type of structured data that exists in the company’s
database and in small amounts reduce audit risk and audit hours. Meanwhile, unstructured
212 data types increase audit risk and audit hours. This result is supported by a study
conducted by De Martinis and Houghton (2019), which found that audits that focus on high-
risk data are able to reduce audit risk and audit hours.
Finally, from the results of additional analysis, participants who have received
information from visual analytics results have a higher level of confidence than those
receiving text analytics. This is because participants can understand the information
obtained from visual analytics more easily and faster than that from the text
analytics.
The results of this study support cognitive fit theory, which says that individuals will be
faster and more precise in making decisions when the technological aids used can fit the
decisions made. The results of this study found that when participants were provided with
visual analytics information, they tended to assess audit risk lower than text analytics for
both structured and unstructured data types.
This study also has practical implications for external auditors. They should upgrade
their skills and knowledge about big data and data analytics. In addition, they should
increasingly use big data and data analytics in planning audits.
Although researchers have tried to overcome the limitations of the study, this
experimental study has weaknesses. To maintain the health protocol to keep the
distance, this study was conducted four times in different classrooms because there is
no adequate space to accommodate 120 participants while maintaining a distance.
However, a homogeneity of variance test has been carried out to ensure that the
samples have the same variance.
As a future research opportunity, researchers can test other types of data analytics that
can affect audit risk and audit hours. In addition, to improve the consistency of the findings,
this study can be conducted on accounting students at different universities and can also be
conducted on actual auditors, both external and internal auditors.

References
Aboagye-Otchere, F., Agyenim-Boateng, C. and Enusah, A. (2021), “A review of big data research in
accounting”, Intelligent Systems in Accounting, Finance and Management, Vol. 28 No. 4,
pp. 268-283, doi: 10.1002/isaf.1504.
Ahmad, F. (2019), “A systematic review of the role of big data analytics in reducing the influence of
cognitive errors on the audit judgement”, Revista de Contabilidad, Vol. 22 No. 2, pp. 187-202, doi:
10.6018/rcsar.382251.
Aldhizer, G. III (2017), “Visual and text analytics – the next step in forensic auditing and accounting”,
The CPA Journal, Vol. 87, pp. 30-33.
Amani, F.A. and Fadlalla, A.M. (2017), “Data mining applications in accounting: a review of the
literature and organizing framework”, International Journal of Accounting Information Systems,
Vol. 24, pp. 32-58, doi: 10.1016/j.accinf.2016.12.004.
Amin, M.M., Zainal, A., Azmi, N. and Ali, N.A. (2020), “Detecting telecommunication fraud with visual
analytics: a review”, IOP Conference Series: Materials Science and Engineering, Vol. 884 No. 1,
doi: 10.1088/1757-899X/884/1/012059.
Anandarajan, M., Hill, C. and Nolan, T. (2019), Practical Text Analytics, Advances in Analytics and Data Big data
Science, Springer Nature, Cham.
analytics and
Appelbaum, D., Kogan, A. and Vasarhelyi, M.A. (2017), “Big data and analytics in the modern audit
engagement: research needs”, Auditing: A Journal of Practice and Theory, Vol. 36 No. 4, pp. 1-27, auditor
doi: 10.2308/ajpt-51684. judgment
Appelbaum, D., Showalter, D.S., Sun, T. and Vasarhelyi, M.A. (2021), “A framework for auditor data
literacy: a normative position”, Accounting Horizons, Vol. 35 No. 2, pp. 5-25, doi: 10.2308/
HORIZONS-19-127. 213
Appelbaum, D.A., Kogan, A. and Vasarhelyi, M.A. (2018), “Analytical procedures in external auditing:
a comprehensive literature survey and framework for external audit analytics”, Journal of
Accounting Literature, Vol. 40 No. 1, pp. 83-101, doi: 10.1016/j.acclit.2018.01.001.
Balios, D., Kotsilaras, P., Eriotis, N. and Vasiliou, D. (2020), “Big data, data analytics and external
auditing”, Journal of Modern Accounting and Auditing, Vol. 16 No. 5, pp. 211-219, doi: 10.17265/
1548-6583/2020.05.002.
Brown-Liburd, H., Issa, H. and Lombardi, D. (2015), “Behavioral implications of big data’s impact on
audit judgment and decision making and future research directions”, Accounting Horizons,
Vol. 29 No. 2, pp. 451-468, doi: 10.2308/acch-51023.
Chandler, P. and Sweller, J. (1991), “Cognitive load theory and the format of instruction”, Cognition and
Instruction, Vol. 8 No. 4, pp. 293-332, doi: 10.1207/s1532690xci0804_2.
Chen, H., Chiang, R.H.L. and Storey, V.C. (2012), “Business intelligence and analytics: from big data to
big impact”, MIS Quarterly, Vol. 36 No. 4, pp. 1165-1188.
Dagilien_e, L. and Klovien_e, L. (2019), “Motivation to use big data and big data analytics in external
auditing”, Managerial Auditing Journal, Vol. 34 No. 7, pp. 750-782, doi: 10.1108/MAJ-01-2018-1773.
David, R. and Abeysekera, I. (2021), “Auditor judgements after withdrawal of the materiality
accounting standard in Australia”, Journal of Risk and Financial Management, Vol. 14 No. 6,
p. 268, doi: 10.3390/jrfm14060268.
De Martinis, M. and Houghton, K. (2019), “The business risk audit approach and audit production
efficiency”, Abacus, Vol. 55 No. 4, pp. 734-782, doi: 10.1111/abac.12178.
Deniswara, K., Handoko, B.L. and Mulyawan, A.N. (2020), “Big data analytics: literature study on how
big data works towards accountant millennial generation”, International Journal of
Management, Vol. 11 No. 5, pp. 376-389, doi: 10.34218/IJM.11.5.2020.037.
Dilla, W.N. and Raschke, R.L. (2015), “Data visualization for fraud detection: practice implications and a
call for future research”, International Journal of Accounting Information Systems, Vol. 16,
pp. 1-22, doi: 10.1016/j.accinf.2015.01.001.
Eberendu, A.C. (2016), “Unstructured data: an overview of the data of big data”, International Journal of
Computer Trends and Technology, Vol. 38 No. 1, pp. 46-50, doi: 10.14445/22312803/ijctt-v38p109.
Edge, D., Larson, J. and White, C. (2018), ‘“Bringing AI to BI: enabling visual analytics of unstructured
data in a modern business intelligence platform”, Proceedings of Conference on Human Factors
in Computing Systems, 2018-April, pp. 1-9, doi: 10.1145/3170427.3174367.
Eilifsen, A., Kinserdal, F., Messier, W. and McKee, T.E. (2020), “An exploratory study into the use of
audit data analytics on audit engagements”, Accounting Horizons, Vol. 34 No. 4, pp. 75-103, doi:
10.2308/HORIZONS-19-121.
Elgendy, N. and Elragal, A. (2014), “Big data analytics: a literature review paper”, Industrial Conference
on Data Mining, Springer International Publishing, pp. 214-227.
Field, A. (2013), Discovering Statistics Using IBM SPSS Statistics, 4th ed., Sage Publications, Thousand
Oaks, CA.
Geng, X. and Kalargiros, M. (2022), “Why does affect matter in accounting: a review of experimental
studies on the effect of affect”, Journal of Accounting Literature, Vol. 44 No. 1, pp. 1-39, doi:
10.1108/jal-11-2021-0005.
ARJ Gepp, A., Linnenluecke, M.K., O’Neill, T.J. and Smith, T. (2018), “Big data techniques in auditing
research and practice: current trends and future opportunities”, Journal of Accounting Literature,
36,2/3 Vol. 40 No. 1, pp. 102-115, doi: 10.1016/j.acclit.2017.05.003.
Griffin, P.A. and Wright, A.M. (2015), “Commentaries on big data’s importance for accounting and
auditing”, Accounting Horizons, Vol. 29 No. 2, pp. 377-379, doi: 10.2308/acch-51066.
Griffith, E.E., Kadous, K. and Young, D. (2021), “Improving complex audit judgments: a framework and
214 evidence”, Contemporary Accounting Research, Vol. 38 No. 3, pp. 2071-2104, doi: 10.1111/1911-
3846.12658.
Hamdam, A., Jusoh, R., Yahya, Y., Jalil, A.A. and Abidin, N.H.Z. (2021), “Auditor judgment and
decision-making in big data environment: a proposed research framework”, Accounting
Research Journal, Vol. 35 No. 1, pp. 55-70, doi: 10.1108/ARJ-04-2020-0078.
Handoko, J., Narsa, I.M. and Basuki, B. (2021), “Role difference and negativity bias relevance in strategy
review: an experiment”, Cogent Business and Management, Vol. 8 No. 1, doi: 10.1080/
23311975.2021.1938928.
Holt, T.P. and Loraas, T.M. (2021), “A potential unintended consequence of big data: does information
structure lead to suboptimal auditor judgment and decision-making?”, Accounting Horizons,
Vol. 35 No. 3, pp. 161-186, doi: 10.2308/HORIZONS-19-123.
Keim, D., Andrienko, G., Fekete, J.D., Gorg, C., Kohlammer, J. and Melancon, G. (2008), “Visual
analytics: definition, process and challenges to cite this version: visual analytics: definition,
process, and challenges”, Information Visualization-Human-Centered Issues and Perspectives,
Springer, Berlin, pp. 154-175.
Kend, M. and Nguyen, L.A. (2020), “Big data analytics and other emerging technologies: the impact on
the Australian audit and assurance profession”, Australian Accounting Review, Vol. 30 No. 4,
pp. 269-282, doi: 10.1111/auar.12305.
Koreff, J. (2021), “Are auditors’ reliance on conclusions from data analytics impacted by different data
analytic inputs?”, Journal of Information Systems, Vol. 36 No. 1, pp. 19-37, doi: 10.2308/isys-19-
051.
Krieger, F., Drews, P. and Velte, P. (2021), “Explaining the (non-) adoption of advanced data analytics in
auditing: a process theory”, International Journal of Accounting Information Systems, Vol. 41,
p. 100511, doi: 10.1016/j.accinf.2021.100511.
Leite, R.A., Gschwandtner, T., Miksch, S., Gstrein, E. and Kuntner, J. (2018), “Visual analytics for event
detection: focusing on fraud”, Visual Informatics, Vol. 2 No. 4, pp. 198-212, doi: 10.1016/j.
visinf.2018.11.001.
Liyanarachchi, G.A. (2007), “Feasibility of using student subjects in accounting experiments:
a review”, Pacific Accounting Review, Vol. 19 No. 1, pp. 47-67, doi: 10.1108/
01140580710754647.
Mactavish, C., McCracken, S. and Schmidt, R.N. (2018), “External auditors’ judgment and decision
making: an audit process task analysis”, Accounting Perspectives, Vol. 17 No. 3, pp. 387-426, doi:
10.1111/1911-3838.12182.
Mattila, A.S., Luo, A., Xue, X. and Ye, T. (2021), “How to avoid common mistakes in experimental
research?”, International Journal of Contemporary Hospitality Management, Vol. 33 No. 1,
pp. 367-374, doi: 10.1108/IJCHM-07-2020-0696.
Murphy, M.L. and Tysiac, K. (2015), “Data analytics helps auditors gain deep insight”, Journal of
Accountancy, Vol. 219 No. 4, p. 10. 52-54,56,58, available at: http://sfx.scholarsportal.info/ottawa?
url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&genre=article&sid=ProQ:
ProQ%253Aabiglobal&atitle=DataþAnalyticsþHelpsþAuditorsþGainþDeepþInsight&
title=JournalþofþAccountancy&issn=00218448&date=2015-04-01&
Nahartyo, E. (2013), Desain Dan Implementasi Riset Eksperimen, 2nd ed., UPP STIM YKPN,
Yogyakarta.
Parsons, L.M. (2007), “The impact of financial information and voluntary disclosures on contributions Big data
to not-for-profit organizations”, Behavioral Research in Accounting, Vol. 19 No. 1, pp. 179-196,
doi: 10.2308/bria.2007.19.1.179.
analytics and
Rakipi, R., De Santis, F. and D’Onza, G. (2021), “Correlates of the internal audit function’s use of data
auditor
analytics in the big data era: global evidence”, Journal of International Accounting, Auditing and judgment
Taxation, Vol. 42, p. 100357, doi: 10.1016/j.intaccaudtax.2020.100357.
Richins, G., Stapleton, A., Stratopoulos, T.C. and Wong, C. (2017), “Big data analytics: opportunity or
threat for the accounting profession?”, Journal of Information Systems, Vol. 31 No. 3, pp. 63-79, 215
doi: 10.2308/isys-51805.
Risch, J., Kao, A., Poteet, S.R. and Wu, Y.J.J. (2008), “Text visualization for visual text analytics”,
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence
and Lecture Notes in Bioinformatics), 4404 LNCS, pp. 154-171, doi: 10.1007/978-3-540-71080-
6_11.
Rodríguez-Quintero, J.F., Sanchez-Diaz, A., Iriarte-Navarro, L., Mate, A., Marco-Such, M. and Trujillo, J.
(2021), “Fraud audit based on visual analysis: a process mining approach”, Applied Sciences
(Switzerland), Vol. 11 No. 11, doi: 10.3390/app11114751.
Salijeni, G., Samsonova-Taddei, A. and Turley, S. (2021), “Understanding how big data technologies
reconfigure the nature and organization of financial statement audits: a sociomaterial analysis”,
European Accounting Review, Vol. 30 No. 3, pp. 531-555, doi: 10.1080/09638180.2021.1882320.
Schindler, P.S. (2019), Business Research Methods, 13th Ed., Mc Graw Hill.
Sharma, V., Pandey, B. and Kumar, V. (2016), “Importance of big data in financial fraud detection”,
International Journal of Automation and Logistics, Vol. 2 No. 4, p. 332, doi: 10.1504/
ijal.2016.080339.
Singh, K. and Best, P. (2016), “Interactive visual analysis of anomalous accounts payable transactions
in SAP enterprise systems”, Managerial Auditing Journal, Vol. 31 No. 1, pp. 35-63, doi: 10.1108/
MAJ-10-2014-1117.
Sun, T. and Vasarhelyi, M.A. (2018), “Embracing textual data analytics in auditing with deep learning”,
The International Journal of Digital Accounting Research, Vol. 18, pp. 49-67, doi: 10.4192/1577-
8517-v18_3.
Sun, T.S. (2019), “Applying deep learning to audit procedures: an illustrative framework”, Accounting
Horizons, Vol. 33 No. 3, pp. 89-109, doi: 10.2308/acch-52455.
Tang, J. and Karim, K.E. (2019), “Financial fraud detection and big data analytics – implications on
auditors’ use of fraud brainstorming session”, Managerial Auditing Journal, Vol. 34 No. 3,
pp. 324-337, doi: 10.1108/MAJ-01-2018-1767.
Vasarhelyi, M.A., Kogan, A. and Tuttle, B.M. (2015), “Big data in accounting: an overview”, Accounting
Horizons, Vol. 29 No. 2, pp. 381-396, doi: 10.2308/acch-51071.
Vessey, I. (1991), “Cognitive fit: a theory-based analysis of the graphs versus tables literature”, Decision
Sciences, Vol. 22 No. 2, pp. 219-240.
Walker, L.S. (2014), “Developing your experiment”, in Webster, M., Jr. and Sell, J. (Eds), Laboratory
Experiments in the Social Sciences, 2nd ed., Elsevier, London, p. 143.
Wang, Y. and Xu, W. (2018), “Leveraging deep learning with LDA-based text analytics to detect
automobile insurance fraud”, Decision Support Systems, Vol. 105, pp. 87-95, doi: 10.1016/j.
dss.2017.11.001.
Warren, J.D., Moffitt, K.C. and Byrnes, P. (2015), “How big data will change accounting”, Accounting
Horizons, Vol. 29 No. 2, pp. 397-407, doi: 10.2308/acch-51069.
Yang, R., Yu, Y., Liu, M. and Wu, K. (2018), “Corporate risk disclosure and audit fee: a text mining
approach”, European Accounting Review, Vol. 27 No. 3, pp. 583-594, doi: 10.1080/
09638180.2017.1329660.
ARJ About the authors
36,2/3 Ranto Partomuan Sihombing is a student of Doctoral Program in Accounting at Airlangga
University, Surabaya. He is also a Senior Lecturer at the Department of Accounting, Faculty of
Economics and Business, Soegijapranata Catholic University in Semarang. His areas of research
interest are auditing, sustainability and behavioral accounting.
I. Made Narsa is a Professor at the Department of Accounting, Faculty of Economics and Business,
Airlangga University, Indonesia. His research interest areas are management accounting, auditing and
216 experimental research. I Made Narsa is the corresponding author and can be contacted at: i-made-n@feb.
unair.ac.id
Iman Harymawan is an Assistant Professor at the Department of Accounting, Faculty of
Economics and Business, Airlangga University, Indonesia. His research interests are in the areas of
corporate governance issues, the accounting impact of political and military connection in business
and financial reporting quality.

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com