Testing The Order And Framing Effect In Budgetary Decision-
Making: Experimental Study
Dhias Kharismadani and I Made Narsa
Department of Accounting, Faculty of Economics and Business, Universitas Airlangga, Surabaya-Indonesia
i-made-n@feb.unair.ac.id
Keywords: Belief-Adjustment Theory, Framing Effect, Order Effect, Recency Effect.
Abstract: One of the important roles of management accountants, nowadays, is to provide input for managers in
decision-making. The presentation order of the input and response modes used, either sequentially or
simultaneously, can lead to order effects (either primacy or recency) that result in biased decision-making.
Order effects in this study were analyzed using the Belief-Adjustment Model (Hogarth and Einhorn, 1992).
Belief adjustment theory is applied as the main framework of this research and framing effects were added
to get an explanation of the interactions and effect. This study aims to empirically examine the moderating
effect of information framing on the belief revision process in the budget preparation. A laboratory
experimental method was used in this study with a 2x2 between-subjects design and participants consisted
of 108 accounts managers from a newspaper’s business group. The hypotheses were tested using Two-Way
ANOVA and found the existence of the recency effect. The finding indicates that in the step-by-step
(sequential) response mode, the occurrence of recency effects can be mitigated by information framing.
1 INTRODUCTION
A perspective shift about the management
accountant’s role in a company is clear in this
modern era. The traditional perspective about the
management accounting system is not enough to
describe the profession of management accountant.
Brewer (2008) proposed a new framework regarding
the management accountant’s role in increasing the
value of controlling interest, which is poured into
four pillars, namely: leadership, strategic
management, operational alignment, and continuous
improvement and learning.
The strategic management accountant’s role is to
facilitate discussion among various functions in the
firm. The report created by the management
accountant is expected to help as a strategic
decision-making tool to align all strategies in the
firm. The management accountant is responsible for
the availability of input in management decision-
making. Information generated by the management
accountant is accumulated in sequence, starting from
varian analysis, through capital budgeting, until
employee performance evaluations.
Sequential patterns of information presented can
affect managers in decision-making due to bounded
rationality (an individual condition in which they
have limitations in processing information
systemically and rationally because of limited
information, time, memory capacity, etc.); hence,
creating biased and non-optimal decisions. Biased
decision-making is caused by the usage of heuristic
strategy, which is a simplification in the decision-
making process. One of the biases that can happen
due to sequential information gathering is order
effect, which means that managers are making
judgments not based on the available information,
being are more affected by the sequence of the
consecutively obtained information. Order effect
consists of primary effect, which means that one
tends to give more weight to the information that
was obtained first, so the last decision made is
affected by this first-obtained information, and
recency effect, which means that one tends to give
more weight to the last-obtained information, so the
last decision made is affected by this last-obtained
information.
This order effect is very likely to happen in
management accounting; one of the examples is in
annual budget prepartion, which requires a few
Kharismadani, D. and Narsa, I.
Testing The Order And Framing Effect In Budgetary Decision-Making: Experimental Study.
In Proceedings of the Journal of Contemporary Accounting and Economics Symposium 2018 on Special Session for Indonesian Study (JCAE 2018) - Contemporary Accounting Studies in
Indonesia, pages 287-293
ISBN: 978-989-758-339-1
Copyright © 2018 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
287
months to process the related information. This
research is referring to the belief-adjustment model
by Hogarth and Einhom (1992) with the assumption
that managers are familiar with budgeting, and using
a few short serial pieces of information for the sake
of simplifiying the experimental cases.
2 LITERATURE REVIEW AND
HYPOTHESES
Tversky and Kahneman (1981) explain the framing
effect phenomena through prospect theory, which
states that framing adopted by managers may affect
the decision taken. Managers process the
information obtained into solutions or decisions for a
problem based on the adopted framing. Framing
effect is an effect on the judgment made because of
the information delivery method. The same
information delivered with a different method will
cause a different perspective. Framing effect can
happen because of an information selection process
that emphasizes only certain parts.
2.1. Order Effect
Research regarding the order effect that is analyzed
using the belief-adjustment model (Hogarth &
Einhorn, 1992) is related to the differences of
information type, including information complexity,
information length, and the response method to
various information. This research uses a simple
information type that involves only one item for
every part of information that needs to be processed
by the subject and is a familiar job for the subject.
Short information (2-12 items) consists of four items
– as mentioned in Ashton and Ashton’s (1988)
experiment. The response modes are either Step-by-
Step (SbS), in which the subject is asked to revise
their belief level sequentially every time information
is obtained, or End-of-Sequence (EoS), in which the
subject is asked to revise their initial belief level
simultaneously at the time the last information is
obtained.
The Belief-Adjustment Model that was
developed by Hogarth and Einhorn (1992) assumes
that one will finish the belief revision assignment
generally using a gradual process of anchoring and
adjustment in which the first opinion (anchor) is
adjusted to a few information pieces that are
obtained next. This research uses an experimental
method in which the subject is asked to revise their
belief about causal hypothesis in a management
accounting context that is classified as an evaluation
assignment, so the encoding of new information
obtained can be either positive or negative.
2.2. Framing Effect
Framing refers to the possibility that a hypothesis
can be evaluated by various alternatives (Tversky &
Kahneman, 1981). Psychology research that tests the
framing effect shows that a small difference in word
arrangement or assignment presentation can change
one’s tendency, so it also affects the alternatives of
the decision taken (Guiral & Esteo, 2006).
Decision-making can be affected by how the
information is framed, either negative frame or
positive frame. Belief level will decrease if positive
information is obtained first and negative
information next. The recency effect can be
moderated using Information Framing (Fischhoff,
1983). The management accountant acts as an
information provider to support management
decision-making in which the information can be
positively or negatively framed. It is possible that
non-optimal decision-making can be avoided by
framing the information relevant to the decision that
is under consideration. Specifically, it can be said
that positive framing can be used to reduce the
unnecessary influence of negative information, and
negative framing can be used to neutralize the
unnecessary influence of positive information
(Rutledge, 1995).
2.3. Belief-Adjustment Model and Order
Effect
Hogarth and Einhorn (1992) imply that the manager
evaluation may be influenced by the order of
information acquisition, which may result in bias;
hence, decision-making becomes less than optimal.
This may be related to how the management
information system is designed to generate
information and how the information is used by the
management accountant (Dillard et al., 1991). The
result of the study conducted by Hogarth and
Einhorn (1992) will be re-examined in this research
(especially in the management accountant context)
to see whether the recency effect is proven to happen
in a simple assignment with short and diverse
information in SbS response mode, and also whether
the primacy effect is proven to happen in a simple
assignment with short and diverse information in
EoS response mode. What is meant by diverse
information is when the information obtained has
different/opposite coding/type, such as positive
JCAE Symposium 2018 Journal of Contemporary Accounting and Economics Symposium 2018 on Special Session for Indonesian Study
288
information, and then followed by negative
information, or vice versa.
Based on the Belief-Adjustment Prediction
Model (Hogarth & Einhorn, 1992) and the result of
previous researches related to short, simple and
diverse information presentation that resulted in the
recency effect in SbS response mode, two
hypotheses can be formulated as follows:
H1: A subject that receives information with the
order of ++-- (positive-negative) will improve
their belief to raise the budget higher than the
subject that receives information with the order
of --++ (negative-positive).
2.4. Framing
The thinking logic for the second hypothesis refers
to Rutledge (1995), who states that if the order of
information received is “positive-negative”, then the
primacy effect should be decreasing when the
information is framed negatively. When the order of
information received is “negative-positive”, the
primacy effect should be decreasing when the
information is framed positively.
With an analogous approach, we believed that
the same logic also applies in the recency effect bias
that happens in EoS mode response, when the
framing in the opposite model states that when the
order of information received is in the order of
“positive, negative”, the recency effect should be
decreasing, when the information is framed
positively, so the second hypothesis is formulated as
follows:
H2: A subject that receives information with the
inappropriate frame (information with positive-
negative order framed positively, and vice
versa) will improve their belief level of a
budget raised higher than a subject that
receives information with the appropriate
frame (information with positive-negative
order framed negatively, and vice versa)
3 METHOD AND ANALYSIS
3.1. Experiment Design
This study used indoor experiments in which the
independent variables are manipulated by the
researcher. All participants were randomly assigned
to each group in order to increase the probability of
homogeneity between groups in the experiment
(Nahartyo, 2012). The experiment began with the
pilot test and then continued to the core experiment.
The experimental study used in this study is a full
2x2 between-subject factorial pattern with
information presentation order (two levels: positive-
negative, negative-positive), and information
framing (positive, negative). Between-subject design
will compare the effect of different treatments on
different subjects.
This study uses an experimental 110 accountants
and managers involved in the budgeting process in
several companies in one group that engaged with
the daily national print media. Due to the nature of
the newspaper business, they are very sensitive to
the dollar rate (related to print pricing), basic
electricity tariffs, fuel tariffs, minimum wages of
workers and prevailing tax rates in Indonesia. This is
due to an experimental instrument in the form of a
budgeting case, so that participants are at least
expected to understand the terms and objectives of
the questions in the given case.
Before going into the core experiment, a pilot
test is conducted. The purpose of the pilot test is to
know the subject’s understanding of the given case.
The pilot test was conducted on 20 people on the
2013 Accounting Master’s Program. They are
grouped in fives. The pilot test results stated that the
F is significant so the experimental questionnaire
can be conducted.
The experimental subjects were randomly
divided into four groups, with each group getting a
different treatment.
Table 1: 2 x 2 ANOVA Experiment Design
2 x 2 ANOVA Experiment Design
Information Presentation Order
Information Framing
+ -
++-- Sel 1 Sel 3
--++ Sel 2 Sel 4
Prior to the experiment, subjects were asked to
fill in demographic data, and then they were given
several types of treatment. Overall, researchers gave
four different treatments in which each case gives
different types of manipulation to the participants.
All participants were instructed to remain in the
role of a subject who was estimating an increase in
production budget and answering some questions
related to production cost cases. Subjects are given
Testing The Order And Framing Effect In Budgetary Decision-Making: Experimental Study
289
background information as well as some information
related to the company’s production costs in the
current year. Then, the subject is requested to set an
initial conviction level regarding the possibility of
budget increase as the initial anchor.
Next, additional information is given in stages
according to the group. Each group received
different treatment. Each time they receive
additional information, as mentioned above, the
subject is asked to provide a revision of their
conviction level related to a budget increase within a
100-point scale (0-100). After the experiment is
completed, the subjects are asked to fill in a
manipulation check question and then they were
debriefed.
3.2. Operational Definition and Variable
Measurement
The independent variable in this research is the
sequence of information given, in the form of two
positive information items and two negative
information items, and framing information in the
form of positive framing and negative framing.
The dependent variable in this research is the
conviction level revision from participants that may
cause Order Effect. Order Effect is obtained from
the initial conviction level assessment to the last
conviction level assessment.
3.3. Data Analysis and Hypothesis
Testing Technique
A homogeneity test is performed to test
ANOVA’s assumption that each group (category)
has the same variance. The homogeneity test used in
this research is the Levene test. The Levene test
criteria is whether the significance is < 0.05, so the
group variance is not homogenous, but if the
significance is > 0.05 then the data group variance is
homogenous (Ghozali, 2011).
The analysis technique used in this research is
Two-Way ANOV. Analysis of variance is a method
for testing the relationship between one dependent
variable (metric scale) with one or more independent
variables (nonmetric or categorical scales). Two-
Way ANOVA is used to test the relationship
between one metric dependent variable and two or
more categorical independent variables.
4 RESULTS AND DISCUSSIONS
Overall, subjects participating in this experiment
were 110 people who were randomly assigned to
four treatment groups (group 1 = 28 people; group 2
= 27 people; group 3 = 28 people; group 4 = 27
people). Of the 110 data subjects who participated,
there were two people who did not pass the
manipulation check, so only 108 subjects’ data were
to be processed and analyzed.
An individual characteristic difference test is
conducted to know the even-ness distribution of
each of the experimental subjects. For that reason, a
randomization test is performed. Table 4.1 shows the
randomization test result.
Table 2: Tests of Between-Subjects Effects
De
p
endent Variable:
gr
ou
p
Source Type III Sum of
S
q
uares
df Mean Square F Sig.
Corrected Model 2,447
a
4 .612 .475 .754
Interce
p
t 171,361 1 171,361 133,156 .000
Gende
r
.317 1 .317 .246 .621
Education Level 2,218 1 2,218 1,724 .192
A
g
e .081 1 .081 .071 .905
Workin
g
Perio
d
.302 1 .302 .234 .629
Erro
132,553 103 1,287
Total 810,000 108
Corrected Total 135,000 107
a. R Squared = .018 (Adjusted R Squared = -.020)
JCAE Symposium 2018 Journal of Contemporary Accounting and Economics Symposium 2018 on Special Session for Indonesian Study
290
Based on Table 4.1, it is known that
equivalence test results show that the F-test score
from the gender of the participant = 0.246 (p =
0.621); the F-test score of the participant’s
educational level = 1,724 (p = 0.192; age of
participant has an F-test score = 0.071 (p = 0.905),
and the working period of participant has an F-test
score of 0.234 (p = 0.629).
Thus, it can be
concluded that there is no significant difference
between treatment groups or, in other words, all
treatment groups are equivalent. Based on this
condition, it is expected that the response of each
group is not contaminated by differences in
characteristics between individuals.
A homogeneity test is performed to test
ANOVA’s assumption that each group (category)
has the same variance. Levene’s test of
homogeneity of variance is one of the tools to test
the homogeneity of groups (categories). Levene
testing’s criteria is if significance > 0.05 then
variance of the data group is homogeneous
(Ghozali, 2011).
Table 3: Test of Homogeneity of Variances
Conviction Level Revision
Levene Statistic df1 df2 Sig.
.465 7 99 .858
This research is using Analysis of Variance
(ANOVA). Two-Way ANOVA is used to test the
relationship between one metric dependent
variable with two or more categorical independent
variables.
Hypothesis 1: A subject that receives
information with the order of ++-- (positive-
negative) will raise their conviction level higher
than a subject that receives information with the
order of --++ (negative-positive). Theoretically,
group 1 and group 3 responses (which are the
subjects that received additional information with
the order of ++-- in doing conviction level revision
regarding the budget raise) will be higher than
group 3 and group 4 responses (the subjects that
receive additional information with --++ order).
Table 4.3 panel A presents the results of one-
way variance analysis and it appears that the F-test
value is 3.822 (p = 0.001). These results reflect
significant differences in response between the
treatments relating to the revision of conviction
level in budget increase. Panel B shows the
average response of participants to conviction level
revision dependent variable. Participant responses
in group 2 have a mean of 44.0741 and group 4
have a mean of 47.7778; this shows that their mean
is lower than the participant responses in group 1,
which have a mean of 55.9259, and group 3, which
have a mean of 54.8148. This indicates that
participants in group 2 and group 4 assess the
conviction level revision regarding the budget
increase lower than participants in group 1 and
group 3. This difference is statistically significant,
below 1 per cent (p = 0.001).
Table 4: ANOVA Test Result for Dependent Variable-Conviction Level Revision
Panel A: Table ANOVA
Squared Amount df Average Squares F Sig
Between Group 9,552.775 8 1194.097 3.822 0.001
Intra Grou
p
30,932.411 99 312.449
Total 40,485.185 107
Panel B: Average Treatment
Group
treatment
Mean Standard
Deviation
Percentage
Conviction
Level
Revision
Grou
p
1 55.9259 19.66239 27
Group2 44.0741 16.23369 27
Group3 54.8148 19.48774 27
Group4 47.7778 15.27525 27
Total 50.6481 18.45799 108
Panel C: Contrast Mean
Contrast Si
g
nificance
Group1 vs Group2 0.039
Group2 vs Group3 0.028
Grou
p
1 vs Grou
p
3 0.036
Grou
p
2 vs Grou
p
4 0.039
Grou
p
1 vs Grou
p
4 0.017
Testing The Order And Framing Effect In Budgetary Decision-Making: Experimental Study
291
Based on Table 4.3, it is known that hypothesis 1
is supported because the test results generate a p-
value of 0.001 ( 0.05). The average subject that
receives information with ++-- order experiences an
increase in conviction level regarding the budget
raise higher than the subjects receiving --++
information order. That result shows that the recency
effect is happening because the last conviction level
is more affected by the last information received.
Therefore, H1 proposed in this research is supported
by data.
Hypothesis 2: A subject that receives information
with the inappropriate frame (information with
positive-negative order is framed positively, and
information with positive-negative order is framed
negatively) will improve their belief level of budget
raised higher than the subject that receives
information with the appropriate frame (information
with positive-negative order is framed negatively,
and information with negative-positive order is
framed positively). Theoretically, group 1 and group
4 responses (the subjects that received information
with an inappropriate frame) in doing belief level
revision regarding the budget raised will be higher
than group 2 and group 3 responses (the subjects that
received information with an appropriate frame).
Based on Table 4.3, it is known that hypothesis 2
is accepted because the mean response of group 1
was 55.9259, significantly higher than group 3,
whose mean response was 54.8148 (p = 0.036). It is
known that group 1 comprises subjects that received
++-- information with positive framing, while group
3 comprises subjects that received ++-- with
negative framing. This means that the subject that
receives information with an inappropriate frame
(positive negative information order is framed
positively) will increase their belief level regarding
budget raised higher than subjects that receive
information with an appropriate frame (positive
negative information order is framed negatively).
The mean response of group 2 was 44.0741,
significantly higher than group 4, whose mean
response was 47.7778 (p = 0.039). It is known that
group 2 comprises subjects that received --++
information order with positive framing, while group
4 comprised subjects that received --++ an
information order with negative framing. This means
that subjects that receive information with an
inappropriate frame (negative positive information is
framed negatively) will increase their belief level
regarding budget raised higher than the subjects
receiving information with an appropriate frame
(negative-positive information order is framed
positively). This shows that information framing can
affect the recency effect that may show up due to
obtaining diverse information. (The subject’s last
belief level is more affected by the last information
obtained.) Hence, the proposed H2 in this research is
supported by data.
5 CONCLUSIONS
Based on the result and analysis, we conclude that
the existence of the recency effect is real. The
attention of the subject when the information is
provided in Step-by-Step mode is more weighed on
the last information than the first information. The
recency effect causes bias in budget decisions.
However, this bias can be mitigated by framing the
information with an inappropriate manner. If the
order of information is ++--, then the framing of
information should be positive (appropriate frame).
If the order of information is --++, then the framing
of information should be negative (inappropriate
frame).
The next research can use group participants, not
only individuals, because in reality the budgeting
process in a firm is a participative process that
combines several related opinions. The next research
may also include variables in the form of anchor
strength to further validate the use of framing in
mitigating the recency effects, both in low or high
anchor condition.
REFERENCES
Alvia, L., & Sulistiawan, D. (2012). The Impact of
Cognitive Style to Recency Effect in Stock
Investment: An Experimental Study. Available at
SSRN 2201544.
Asare, S. K., & Messier Jr, W. F. (1991). A review of
audit research using the belief-adjustment model
Auditing (pp. 75-92): Springer.
Ashton, A. H., & Ashton, R. H. (1988). Sequential belief
revision in auditing. Accounting Review, 623-641.
Bamber, E. M., Ramsay, R. J., & Tubbs, R. M. (1997). An
examination of the descriptive validity of the belief-
adjustment model and alternative attitudes to evidence
in auditing. Accounting, Organizations and Society,
22(3), 249-268.
Biyanto, F. (2001). Hubungan pembingkaian informasi
anggaran, tanggung jawab, dan pengalaman terhadap
pilihan keputusan pada investasi berisiko. Universitas
Gadjah Mada.
Brewer, P. C. (2008). Redefining management accounting:
Promoting the four pillars of our profession. Strategic
finance, 27-35.
JCAE Symposium 2018 Journal of Contemporary Accounting and Economics Symposium 2018 on Special Session for Indonesian Study
292
Dillard, J. F., Kauffman, N. L., & Spires, E. E. (1991).
Evidence order and belief revision in management
accounting decisions. Accounting, Organizations and
Society, 16(7), 619-633.
Fischhoff, B. (1983). Predicting frames. Journal of
Experimental Psychology: Learning, Memory, and
Cognition, 9(1), 103.
Ghozali, I. (2011). Aplikasi Multivariate dengan Program
IBM SPSS 19. Semarang: Badan Penerbit Universitas
Diponegoro.
Guiral, A., & Esteo, F. (2006). Are Spanish auditors
skeptical in going concern evaluations? Managerial
Auditing Journal, 21(6), 598-620.
Gunawan, B., & Yusuf, M. H. (2012). PENGARUH
ORDER EFFECT DAN POLA PENGUNGKAPAN
DALAM PENGAMBILAN KEPUTUSAN
INVESTASI. InFestasi, 8(2), 123-136.
Hogarth, R. M., & Einhorn, H. J. (1992). Order effects in
belief updating: The belief-adjustment model.
Cognitive psychology, 24(1), 1-55.
Krishnamoorthy, G., Mock, T. J., & Washington, M. T.
(1999). A comparative evaluation of belief revision
models in auditing. Auditing: A Journal of Practice &
Theory, 18(2), 105-127.
Nahartyo, E. (2012). Desain dan Implementasi Riset
Eksperimen. Yogyakarta: UPP STIM YKPN.
Nasution, D. (2007). Pengaruh urutan bukti, gaya kognitif,
dan personalitas terhadap proses revisi keyakinan.
Universitas Gadjah Mada.
Pei, B. K., Reed, S. A., & Koch, B. S. (1992). Auditor
belief revisions in a performance auditing setting: An
application of the belief-adjustment model.
Accounting, Organizations and Society, 17(2), 169-
183.
Pinsker, R. (2011). Primacy or recency? A study of order
effects when nonprofessional investors are provided a
long series of disclosures. Behavioral Research in
Accounting, 23(1), 161-183.
Rutledge, R. W. (1995). The ability to moderate recency
effects through framing of management accounting
information. Journal of Managerial Issues, 27-40.
Sugiyono. (2008). Metode Penelitian Pendidikan.
Bandung: Alfabeta.
Sukardi. (2011). Metodologi Penelitian Pendidikan.
Jakarta: Bumi Aksara.
Susanto, B., & Nahartyo, E. (2008). Pengaruh
Tanggungjawab, Motivasi Intrinsik, dan
Pembingkaian Informasi Anggaran Dalam
Pengambilan Keputusan Investasi Dengan Group-
Shifts Sebagai Variabel Pemoderasi. Tesis tidak
dipublkasikan.
Tubbs, R. M., Messier Jr, W. F., & Knechel, W. R. (1990).
Recency effects in the auditor's belief-revision
process. Accounting Review, 452-460.
Twersky, A., & Kahneman, D. (1981). The framing of
decisions and the psychology of choice. Science, 211,
453-458.
Testing The Order And Framing Effect In Budgetary Decision-Making: Experimental Study
293