Empirical Research in Accounting
Inhoudsopgave
Lecture 1 – Intro to Empirical Accounting Research..........................................2
Lecture 2 – Introduction to Experiments as Research Method...........................4
Lecture 3 – Motivation, Effort, and Performance & Framing Effects...................9
Lecture 4 – Statistical Analysis of an Experiment (in the Area of Feedback).....13
Lecture 5 – Increasing Dynamics...................................................................15
Lecture 6 – Process Measures & Participants.................................................17
Lecture 7 – Leaving the Lab – Field Experiments & Surveys............................20
Lecture 8 – Transition to Archival Research Method........................................24
Lecture 9 – The Role of Accounting Information.............................................30
Lecture 10 – Accounting Standards and Regulation........................................35
Lecture 11 – Externalities of Accounting Regulation.......................................43
Lecture 12 – Real Effects of Transparency......................................................51
Lecture 13 – CSR Research............................................................................58
,Lecture 1 – Intro to Empirical Accounting
Research
Financial accounting involves the process of recording, summarizing,
and reporting the myriads of transactions resulting from business
operations over a period. These transactions are summarized into financial
statements. Financial statements are disseminated to external stakeholder
(e.g. investors, regulators, customers.)
Management accounting is the production, dissemination, and use of
accounting information for managerial decision making.
Economic vs. behavioral
Economic theory: rational decision-makers, optimizing utilities, self-
interest, high computational capacities, strategic foresights, no mistakes,
no relevant information is missed, irrelevant information is dismissed.
Behavioral theory: bounded rationality, satisficing utilities, social
interests, cognitive limitations, room for short-sightedness, biased,
affective decisions, missing of relevant information, and information
overload and use of irrelevant information.
Economic theory and information economics can provide an important
perspective and benchmark. Tensions (=research opportunities) arise
when economic theory predictions meet behavioral theories.
Management control/reporting system has two roles: a decision-influencing
role (align interests) and a decision facilitating role (information about the
costs of the production).
Kinney’s paragraphs (1986) three questions about research: What is the
problem?, Why is it an important problem?, How will it be solved?. So,
What, Why, and How.
There are four kinds of validities:
internal validity, external validity,
construct validity, and statistical
conclusion validity.
Internal validity is about the research
design itself. It consists of construct
validity (how well are constructs
measured) and statistical conclusion validity (are statistical methods and
inferences appropriate). External validity is about for which settings in
reality can the research design be informative.
Validity vs. reliability
Validity: how well does a measure measure what it is supposed to
measure?
,Reliability: when a measurement is repeated multiple times, how much
do results vary.
Reliability is necessary but is not sufficient for validity.
There is a framework for the validity, which is called predictive validity
framework: Libbby boxes.
Link 1 is the research interest. So, is my theory correct that there is a/no
effect of the treatment on the outcome. How could such an effect be
explained? Hypotheses should be formulated on the concept level (not the
operational level). Theory and explanations can be generalized.
You cannot directly measure concepts: how can an effect be measured
empirically. You can make the concepts measurable, and you can analyze
proxies of concepts. Is there an effect of the Concept A measure on the
Concept B measure. There can be discussion about whether a measure is
a suited proxy, this is about construct validity. Links 2 and 3: does the
proxy constitute an appropriate measure of the concept? Link 4 (statistical
validity): is there a statistical association? Link 5: control needed. This
could be direct (empirical): controls, or indirect (randomization).
, Different methods have different strengths and weaknesses. To try to
overcome this, you can use triangulation: using different methods to
investigate a research question from different (methodological) angles.
This is crucial for scientific discovery: if different perspectives add up,
there is a higher confidence in the findings. Explicit triangulation is rare in
studies: it requires expertise in different methods, and it requires open-
minded reviewers. It is also not only useful in academic environment.
Lecture 2 – Introduction to Experiments as
Research Method
Learning goals
Define and recognize an experiment.
Evaluate strengths and weaknesses of the research method
experiments.
Evaluate strengths and weaknesses of within vs. between-subjects
designs.
Interpret results from an “anova graph".
Explain why correlation can easily be confused with causation
An experiment is a scientific investigation in which independent
variables (IV) (sometimes explanatory variable) are manipulated and
their effects on dependent variables (DV) (sometimes explained
variable, outcome variable, variable of interest) are analyzed. It is a test of
a suggested cause-effect-relation (theory) in which the investigator
can control the independent variable. Inferences (=gevolgtrekkingen) in
an experiment come from the comparison of the control group (C) and
treatment group (T). if there is only one difference C and T (=the
manipulation), any difference in an outcome can be interpreted as being
caused by the manipulation. Accounting experiments usually test theories,
which can then be generalized. A convincing theoretical argumentation is
key: the design of an experiment setting follows theory. Different than in
medicine, we usually focus on directional predictions (-, 0, +) instead of
effect sizes.
False positive vs. false negative errors
False positive (type 1): researcher determines that an effect exists,
although there is none.
False negative (type 2); researcher determines that there is no effect,
although there is one.
In experiments we usually work with limited sample sizes.
- Really small N: anecdotical evidence no statistical analyses.
- Small N (lab experiment): statistical analyses possible, but high risk
of type 2 error, noise vs. strength of manipulation (effect size).Big N
(archival research): findings can be statistically significant and
correct (no errors), but economically not very important. Going back
Inhoudsopgave
Lecture 1 – Intro to Empirical Accounting Research..........................................2
Lecture 2 – Introduction to Experiments as Research Method...........................4
Lecture 3 – Motivation, Effort, and Performance & Framing Effects...................9
Lecture 4 – Statistical Analysis of an Experiment (in the Area of Feedback).....13
Lecture 5 – Increasing Dynamics...................................................................15
Lecture 6 – Process Measures & Participants.................................................17
Lecture 7 – Leaving the Lab – Field Experiments & Surveys............................20
Lecture 8 – Transition to Archival Research Method........................................24
Lecture 9 – The Role of Accounting Information.............................................30
Lecture 10 – Accounting Standards and Regulation........................................35
Lecture 11 – Externalities of Accounting Regulation.......................................43
Lecture 12 – Real Effects of Transparency......................................................51
Lecture 13 – CSR Research............................................................................58
,Lecture 1 – Intro to Empirical Accounting
Research
Financial accounting involves the process of recording, summarizing,
and reporting the myriads of transactions resulting from business
operations over a period. These transactions are summarized into financial
statements. Financial statements are disseminated to external stakeholder
(e.g. investors, regulators, customers.)
Management accounting is the production, dissemination, and use of
accounting information for managerial decision making.
Economic vs. behavioral
Economic theory: rational decision-makers, optimizing utilities, self-
interest, high computational capacities, strategic foresights, no mistakes,
no relevant information is missed, irrelevant information is dismissed.
Behavioral theory: bounded rationality, satisficing utilities, social
interests, cognitive limitations, room for short-sightedness, biased,
affective decisions, missing of relevant information, and information
overload and use of irrelevant information.
Economic theory and information economics can provide an important
perspective and benchmark. Tensions (=research opportunities) arise
when economic theory predictions meet behavioral theories.
Management control/reporting system has two roles: a decision-influencing
role (align interests) and a decision facilitating role (information about the
costs of the production).
Kinney’s paragraphs (1986) three questions about research: What is the
problem?, Why is it an important problem?, How will it be solved?. So,
What, Why, and How.
There are four kinds of validities:
internal validity, external validity,
construct validity, and statistical
conclusion validity.
Internal validity is about the research
design itself. It consists of construct
validity (how well are constructs
measured) and statistical conclusion validity (are statistical methods and
inferences appropriate). External validity is about for which settings in
reality can the research design be informative.
Validity vs. reliability
Validity: how well does a measure measure what it is supposed to
measure?
,Reliability: when a measurement is repeated multiple times, how much
do results vary.
Reliability is necessary but is not sufficient for validity.
There is a framework for the validity, which is called predictive validity
framework: Libbby boxes.
Link 1 is the research interest. So, is my theory correct that there is a/no
effect of the treatment on the outcome. How could such an effect be
explained? Hypotheses should be formulated on the concept level (not the
operational level). Theory and explanations can be generalized.
You cannot directly measure concepts: how can an effect be measured
empirically. You can make the concepts measurable, and you can analyze
proxies of concepts. Is there an effect of the Concept A measure on the
Concept B measure. There can be discussion about whether a measure is
a suited proxy, this is about construct validity. Links 2 and 3: does the
proxy constitute an appropriate measure of the concept? Link 4 (statistical
validity): is there a statistical association? Link 5: control needed. This
could be direct (empirical): controls, or indirect (randomization).
, Different methods have different strengths and weaknesses. To try to
overcome this, you can use triangulation: using different methods to
investigate a research question from different (methodological) angles.
This is crucial for scientific discovery: if different perspectives add up,
there is a higher confidence in the findings. Explicit triangulation is rare in
studies: it requires expertise in different methods, and it requires open-
minded reviewers. It is also not only useful in academic environment.
Lecture 2 – Introduction to Experiments as
Research Method
Learning goals
Define and recognize an experiment.
Evaluate strengths and weaknesses of the research method
experiments.
Evaluate strengths and weaknesses of within vs. between-subjects
designs.
Interpret results from an “anova graph".
Explain why correlation can easily be confused with causation
An experiment is a scientific investigation in which independent
variables (IV) (sometimes explanatory variable) are manipulated and
their effects on dependent variables (DV) (sometimes explained
variable, outcome variable, variable of interest) are analyzed. It is a test of
a suggested cause-effect-relation (theory) in which the investigator
can control the independent variable. Inferences (=gevolgtrekkingen) in
an experiment come from the comparison of the control group (C) and
treatment group (T). if there is only one difference C and T (=the
manipulation), any difference in an outcome can be interpreted as being
caused by the manipulation. Accounting experiments usually test theories,
which can then be generalized. A convincing theoretical argumentation is
key: the design of an experiment setting follows theory. Different than in
medicine, we usually focus on directional predictions (-, 0, +) instead of
effect sizes.
False positive vs. false negative errors
False positive (type 1): researcher determines that an effect exists,
although there is none.
False negative (type 2); researcher determines that there is no effect,
although there is one.
In experiments we usually work with limited sample sizes.
- Really small N: anecdotical evidence no statistical analyses.
- Small N (lab experiment): statistical analyses possible, but high risk
of type 2 error, noise vs. strength of manipulation (effect size).Big N
(archival research): findings can be statistically significant and
correct (no errors), but economically not very important. Going back
