Research methods in
accounting
Introduction
Kinney paragraphs
What are you trying to find out, conceptually? (observed in practice)
What concepts or theories underlie your idea?
Why is an answer important and to whom?
Who will care about your answer and why should they care?
How will you find the answer, operationally? (how to measure?)
What research method and data will you use to find the answer?
Libby boxes
Libby boxes = predictive validity framework
Very useful for archival, survey and experimental research
Includes 5 links which determine a study’s internal and external validity
Internal validity = do we measure what we want to measure? Does the conceptual level matches the
operational level?
External validity = does it makes sense in the real world? Is it relevant?
LINKS:
1. Start from a good and relevant research question (rely on theory to build hypotheses)
2. Operationalization of independent variable (measurement is important)
3. Operationalization of dependent variable (measurement is important)
4. Statistical test of causation (in a lab or field experiment)/association (in survey or archival studies)
5. Includes relevant control variables
1
,Structure of a research paper
Structure:
1. Introduction
2. Literature review and hypotheses development
3. Methodology
4. Results
5. Discussion and conclusion
Research methods
Methods:
1. Archival studies
2. Field studies
3. Field experiments
4. Surveys
5. Laboratory studies
6. Laboratory experiments
Archival studies
Large datasets => powerful tests of association
BUT no causality + hard to devise measures that precisely capture the constructs specified in the theory
they seek to test
Field studies
Rich contextual insights
BUT small samples so no generalizability + attribution
Field experiments
TEST CAUSALITY through manipulation! + rich contextual insights
BUT small samples so small power and what about generalizability?
2
, Surveys
Measurement of specific constructs => generalizability
BUT no causality + self-reported measures (inaccurate or dishonest responses)
Laboratory studies
Controlled setting + do not intervene
BUT no causality + limits contextualization
Laboratory experiments
TEST CAUSALITY through manipulation + controlled setting
BUT no generalizability + contextualization
Archival accounting research
Single-equation regression model = the dependent variable is expressed as a linear function of 1 or more
explanatory variables
If there are causal relationships in such a model, the dependent and explanatory variables flow in
1 direction only!! (from explanatory variables to the dependent variable; x y)
Simple/two-variable regression analysis= study the dependence of 1 variable on 1 explanatory variable
Multiple regression analysis = study the dependence of 1 variable on multiple explanatory variables
Chapter 1: The nature of regression analysis
The modern interpretation of regression
Regression analysis = the study of the dependence of 1 variable (dependent variable) on 1 or more other
variables (explanatory/independent variables)
Goal: to estimate or predict the (population) mean or average value of the dependent variable in terms
of the known or fixed values of the independent variable
Statistical vs. Deterministic relationships
Stochastic/random variables = variables with probability distributions (Ci = C0 + C1 *Xi + ui)
Deterministic/functional variables = variables with no error term (Ci = C0 + C1 *Xi)
Regression vs. Causation
We test ASSOCIATION!!
Regression does NOT necessarily imply causation!
We need theory to draw causal conclusions (a statistical relationship in itself cannot logically
imply causation)
How can we draw causal conclusions based on archival data?
1. Identify a quasi-experimental setting (compare before and after) to assume randomization
3
accounting
Introduction
Kinney paragraphs
What are you trying to find out, conceptually? (observed in practice)
What concepts or theories underlie your idea?
Why is an answer important and to whom?
Who will care about your answer and why should they care?
How will you find the answer, operationally? (how to measure?)
What research method and data will you use to find the answer?
Libby boxes
Libby boxes = predictive validity framework
Very useful for archival, survey and experimental research
Includes 5 links which determine a study’s internal and external validity
Internal validity = do we measure what we want to measure? Does the conceptual level matches the
operational level?
External validity = does it makes sense in the real world? Is it relevant?
LINKS:
1. Start from a good and relevant research question (rely on theory to build hypotheses)
2. Operationalization of independent variable (measurement is important)
3. Operationalization of dependent variable (measurement is important)
4. Statistical test of causation (in a lab or field experiment)/association (in survey or archival studies)
5. Includes relevant control variables
1
,Structure of a research paper
Structure:
1. Introduction
2. Literature review and hypotheses development
3. Methodology
4. Results
5. Discussion and conclusion
Research methods
Methods:
1. Archival studies
2. Field studies
3. Field experiments
4. Surveys
5. Laboratory studies
6. Laboratory experiments
Archival studies
Large datasets => powerful tests of association
BUT no causality + hard to devise measures that precisely capture the constructs specified in the theory
they seek to test
Field studies
Rich contextual insights
BUT small samples so no generalizability + attribution
Field experiments
TEST CAUSALITY through manipulation! + rich contextual insights
BUT small samples so small power and what about generalizability?
2
, Surveys
Measurement of specific constructs => generalizability
BUT no causality + self-reported measures (inaccurate or dishonest responses)
Laboratory studies
Controlled setting + do not intervene
BUT no causality + limits contextualization
Laboratory experiments
TEST CAUSALITY through manipulation + controlled setting
BUT no generalizability + contextualization
Archival accounting research
Single-equation regression model = the dependent variable is expressed as a linear function of 1 or more
explanatory variables
If there are causal relationships in such a model, the dependent and explanatory variables flow in
1 direction only!! (from explanatory variables to the dependent variable; x y)
Simple/two-variable regression analysis= study the dependence of 1 variable on 1 explanatory variable
Multiple regression analysis = study the dependence of 1 variable on multiple explanatory variables
Chapter 1: The nature of regression analysis
The modern interpretation of regression
Regression analysis = the study of the dependence of 1 variable (dependent variable) on 1 or more other
variables (explanatory/independent variables)
Goal: to estimate or predict the (population) mean or average value of the dependent variable in terms
of the known or fixed values of the independent variable
Statistical vs. Deterministic relationships
Stochastic/random variables = variables with probability distributions (Ci = C0 + C1 *Xi + ui)
Deterministic/functional variables = variables with no error term (Ci = C0 + C1 *Xi)
Regression vs. Causation
We test ASSOCIATION!!
Regression does NOT necessarily imply causation!
We need theory to draw causal conclusions (a statistical relationship in itself cannot logically
imply causation)
How can we draw causal conclusions based on archival data?
1. Identify a quasi-experimental setting (compare before and after) to assume randomization
3
