Research Philosophy and Design

Introduction

An accounting and finance dissertation is not simply a report of information collected from annual reports, databases, interviews, or questionnaires. It is a reasoned investigation in which the researcher makes assumptions about reality, knowledge, evidence, and the appropriate way to answer a research question. These assumptions are expressed through a research philosophy and translated into a research design.

Research philosophy concerns how a researcher understands reality and knowledge. Research design concerns the structure of the investigation: what will be studied, from whom or from what sources data will be obtained, when data will be collected, and how evidence will be analysed. A defensible dissertation aligns its research question, philosophical position, methodology, methods, evidence, and conclusions.

For example, a study asking whether audit committee independence is associated with financial reporting quality may appropriately use archival data and statistical analysis. A study asking how small-business owners understand the usefulness of management accounting information may require interviews and interpretive analysis. Neither design is inherently superior; suitability depends on the research problem.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain the meaning and importance of research philosophy.
2. Compare positivist, interpretivist, critical, and pragmatic perspectives.
3. Distinguish methodology, method, research design, and research strategy.
4. Select exploratory, descriptive, explanatory, experimental, case-study, or mixed-method designs.
5. Evaluate validity, reliability, credibility, ethics, and feasibility.
6. Justify a coherent design for an accounting or finance dissertation.

1. The Research Logic: From Question to Evidence

A research project normally proceeds through a chain of decisions:

Research problem → research question → philosophical assumptions → approach and design → sampling and data collection → analysis → conclusions.

A research question should determine the design rather than the researcher selecting a fashionable method first. The following distinctions are important.

• Ontology: assumptions about what exists and what reality is like. A positivist may treat financial performance as measurable independently of the researcher; an interpretivist may emphasise that performance is also understood through organisational meanings and social practices.
• Epistemology: assumptions about how knowledge can be obtained and justified.
• Methodology: the overall logic and rationale of the inquiry.
• Methods: specific techniques such as regression, interviews, document analysis, surveys, or observation.
• Research design: the planned structure connecting the question to evidence and conclusions.
• Paradigm: a broad set of assumptions about reality, knowledge, values, and inquiry.

A useful distinction is between deduction and induction. Deductive research begins with theory or hypotheses and tests them against evidence. Inductive research begins with observations and develops concepts or explanations from patterns in the data. Many projects use abduction, moving iteratively between evidence and theory to find the most plausible explanation.

2. Major Research Philosophies

2.1 Positivism

Positivism holds that systematic observation and measurement can produce objective knowledge about a reality that exists independently of the researcher. It is commonly associated with hypothesis testing, quantitative data, standardised instruments, statistical analysis, and the search for regularities or relationships.

In accounting and finance, positivist studies may examine whether leverage predicts default risk, whether audit quality is associated with reporting quality, or whether a tax incentive changes investment behaviour. Variables are operationalised so that concepts can be measured consistently.

Typical features include:

• clearly specified hypotheses;
• large or representative samples where feasible;
• structured questionnaires or archival databases;
• statistical estimation and tests of association or difference;
• attention to reliability, validity, replicability, and generalisability.

Positivism is useful when the purpose is to estimate prevalence, compare groups, test relationships, or evaluate effects. However, measurement may simplify complex organisational realities. Statistical association does not by itself establish causation, especially where omitted variables, reverse causality, or selection bias are possible.

2.2 Interpretivism

Interpretivism argues that social reality is constructed and understood through meanings, interactions, language, and context. Accounting numbers are not only technical outputs; they may also reflect professional judgement, organisational power, institutional expectations, and competing interpretations.

Interpretivist research commonly uses semi-structured interviews, participant observation, focus groups, organisational documents, and qualitative case studies. The researcher seeks to understand how participants interpret events rather than imposing categories entirely in advance.

For example, an interpretive study might ask how finance managers experience the introduction of artificial-intelligence forecasting tools, or how employees understand performance targets. Quality is assessed through credibility, transparency, reflexivity, rich contextual description, and a convincing connection between data and interpretation.

Interpretivism can reveal processes and meanings that numerical data miss. Its limitations include the time required for fieldwork, the difficulty of replication in an identical form, and limited statistical generalisation. Nevertheless, a carefully bounded case may provide analytical generalisation by developing concepts relevant to other settings.

2.3 Critical Perspectives

Critical research views knowledge and organisational reality as connected to power, inequality, ideology, and historical conditions. It asks not only what happens, but also who benefits, who bears costs, whose voice is absent, and how existing arrangements might be changed.

A critical accounting project might examine how performance measurement affects low-paid workers, how sustainability reporting represents corporate responsibility, or how lending systems disadvantage particular communities. It may analyse reports, regulations, interviews, media, and institutional practices.

Critical research generally recognises that research is not value-free. The researcher should disclose assumptions and consider the consequences of the inquiry. Its contributions may include exposing hidden interests, challenging apparently neutral classifications, and supporting more equitable practice.

Critical approaches must still provide rigorous evidence. Advocacy does not replace analysis, and criticism should not be based on unsupported claims. Researchers must avoid speaking for communities without consent or adequate representation.

2.4 Pragmatism

Pragmatism begins with the research problem and asks which forms of evidence best answer it. It does not require absolute commitment to either objective measurement or subjective interpretation. A pragmatic researcher may combine survey results, financial data, interviews, and case evidence when this produces a more useful answer.

For instance, a study of budgeting software may first survey adoption rates, then interview managers to explain why implementation succeeded in some organisations but failed in others. Pragmatism is therefore particularly compatible with mixed methods, applied research, evaluation, and problem-solving.

Its danger is methodological opportunism: simply collecting whatever data are convenient. A pragmatic design still requires explicit reasoning about assumptions, sequencing, integration, quality, ethics, and limitations.

Comparison of Philosophies

These perspectives are not merely labels. They influence the research question, the role of theory, the relationship with participants, the form of evidence, and the language used in conclusions.

3. Selecting a Research Design

3.1 Exploratory Design

Exploratory research is appropriate when a topic is poorly understood, concepts are unclear, or existing theory is insufficient. It may use interviews, focus groups, open-ended surveys, document analysis, or an initial case study.

An accounting student might explore how small firms decide whether to use cloud accounting. The objective is not to estimate the proportion of all firms that use a particular system, but to identify motivations, barriers, and emerging themes that can support later research.

Exploration should not be described as proof. Its outputs are concepts, questions, propositions, or preliminary explanations.

3.2 Descriptive Design

Descriptive research answers questions such as: What is happening? How common is it? What are the characteristics of a population or phenomenon? A survey of sustainability-reporting practices or a profile of financing sources among start-up firms may be descriptive.

Descriptive statistics can include frequencies, percentages, means, medians, and dispersion measures. A mean is calculated as:

where xᵢ is an observed value and n is the number of observations. The mean is meaningful when observations are measured on a scale for which averaging is appropriate; extreme values can make the median a more representative summary.

Descriptive designs do not automatically explain causes. Reporting that firms with larger audit committees have higher disclosure scores does not establish that committee size caused better disclosure.

3.3 Explanatory Design

Explanatory research investigates why or how variables are related. It commonly uses theory-derived hypotheses and quantitative or qualitative evidence. In a quantitative study, a simple linear model may be expressed as:

Here yᵢ is the outcome for observation i, xᵢ is a predictor, β₀ is the intercept, β₁ is the expected change in y for a one-unit change in x, and εᵢ represents unexplained influences. The model assumes, among other things, that the specified relationship is appropriate and that the error structure supports the intended inference. A regression coefficient indicates association unless the design and assumptions justify a causal interpretation.

Qualitative explanatory research may trace decisions over time, compare contrasting cases, or identify mechanisms through which a policy affects behaviour.

3.4 Experimental Design

An experiment deliberately introduces an intervention and compares outcomes with a control or comparison condition. Random assignment, where feasible, helps balance observed and unobserved characteristics between groups. A basic treatment effect can be represented as:

where ȳₜ is the average outcome in the treatment group and ȳ𝚌 is the average outcome in the comparison group. This difference is a credible estimate of the intervention’s average effect only under appropriate design conditions, including valid assignment and no major contamination or differential attrition.

In accounting education or behavioural finance, an experiment might test whether a particular presentation of risk information changes investment decisions. In firms, true randomisation may be difficult, so quasi-experimental designs use naturally occurring policy changes, matched groups, or before-and-after comparisons. Researchers must be cautious because groups may differ for reasons unrelated to the intervention.

3.5 Case-Study Design

A case study is an intensive investigation of a bounded case, such as one company, project, regulatory change, department, or implementation process. It may use multiple sources: interviews, observation, internal documents, financial records, and public reports.

Case studies are suitable when context is central, boundaries can be defined, and the research asks how or why questions. A single case can be exploratory, descriptive, or explanatory. Multiple-case designs allow systematic comparison and replication of emerging explanations.

A strong case study specifies the case boundary, selection rationale, units of analysis, sources of evidence, protocol, and method of triangulation. It should not be selected merely because access is convenient unless convenience is acknowledged as a limitation.

3.6 Mixed-Method Design

Mixed-method research intentionally integrates quantitative and qualitative evidence in one project. Common structures include:

• Sequential exploratory: qualitative exploration followed by a survey or measurement instrument.
• Sequential explanatory: quantitative results followed by qualitative work to explain surprising or important patterns.
• Convergent: quantitative and qualitative data are collected during a similar period and then compared or integrated.
• Embedded: one method is nested within a larger design, such as interviews embedded in an evaluation.

Mixed methods are justified when one form of evidence cannot adequately answer the question. They are not justified merely because using two methods sounds more rigorous. Integration should occur in the research question, sampling, analysis, interpretation, or presentation of findings.

4. Aligning Design with the Research Question

The following guide is useful, although it is not a substitute for judgement.

A useful dissertation justification explains: (1) the question, (2) the philosophical assumptions, (3) the design, (4) the evidence, (5) the analysis, and (6) why alternatives were less suitable.

5. Quality, Sampling, and Evidence

Quantitative Quality

Construct validity concerns whether a measure represents the intended concept. Internal validity concerns whether an observed effect is attributable to the proposed cause. External validity concerns the extent to which findings apply beyond the study setting. Reliability concerns consistency of measurement.

A sample should be connected to the target population. Probability sampling supports statistical generalisation when the sampling frame and response process are appropriate. Non-probability sampling may be suitable for exploratory or hard-to-reach populations, but claims should be limited.

Qualitative Quality

Qualitative researchers commonly consider credibility, dependability, confirmability, and transferability. Strategies include prolonged engagement, careful transcription, negative-case analysis, participant clarification where appropriate, triangulation, an audit trail, and reflexive discussion of the researcher’s position.

Triangulation means comparing different sources, methods, researchers, or theories. It does not mean that all sources must produce identical results. Disagreement may reveal important contextual differences.

6. Ethics, Governance, and Data Protection

Ethical approval should be obtained before collecting data where institutional procedures require it. Participants should receive understandable information about the study, risks, benefits, voluntary participation, withdrawal rights, and use of data. Consent must not be treated as valid where participation is coerced or materially misrepresented.

Accounting and finance research often involves sensitive information: salaries, tax affairs, fraud allegations, customer records, commercially confidential forecasts, and investment decisions. Researchers should minimise data collection, restrict access, remove direct identifiers where possible, store files securely, and establish retention and destruction procedures. Anonymisation must be realistic; combinations of public facts can re-identify organisations or individuals.

Researchers must not fabricate, manipulate, or selectively suppress results. Conflicts of interest, sponsorship, limitations, non-response, missing data, and analytical choices should be disclosed. AI tools may assist with transcription, coding, or language editing only under approved governance arrangements. Researchers remain responsible for accuracy, confidentiality, copyright, bias, and the integrity of the final work; confidential participant data should not be uploaded to an unapproved system.

7. Illustrative Case: Evaluating a Digital Budgeting System

Background

The following is a fictional illustrative case. Meridian Components, a medium-sized manufacturer, introduces a cloud-based budgeting system intended to improve forecast accuracy and managerial participation.

Problem

After six months, senior management observes that some departments use the system actively while others continue to rely on spreadsheets. Management wants to know both whether forecasting improved and why implementation differed.

Analysis and Design

A purely positivist design could compare forecast errors before and after implementation. Forecast error might be measured as:

where Aᵢ is the actual result for period i, Fᵢ is the forecast, and |·| denotes absolute value. The measure is in the same units as the underlying financial quantity, such as pounds or units. It treats overestimation and underestimation equally but does not show direction.

However, this comparison could be misleading if market conditions changed or if departments adopting the system were already better managed. The project therefore adopts a pragmatic, sequential explanatory design. First, the researcher compares forecast errors across departments and periods. Second, interviews are conducted with finance staff and departmental managers selected to represent high, medium, and low usage. Third, quantitative and qualitative findings are integrated.

Approach and Expected Interpretation

The quantitative stage may show that active users have lower average absolute forecast error. Interviews may reveal that active users received training and had authority to revise assumptions, whereas low-use departments viewed the system as a compliance tool. The combined interpretation is not simply that the software caused improvement. Rather, implementation conditions—training, managerial autonomy, and data quality—may explain the observed relationship.

Outcome

Management redesigns training, clarifies responsibilities, and introduces data-quality checks. The dissertation reports the improvement cautiously, identifies alternative explanations, explains sampling limitations, and distinguishes measured association from proven causation.

Learning Points

1. A single outcome measure rarely explains an organisational process.
2. Mixed methods are valuable when magnitude and meaning are both important.
3. A design must address rival explanations, not only produce attractive results.
4. Practical recommendations should follow from evidence and acknowledged limitations.

8. Common Errors and Best Practices

Common errors include choosing a method before defining the question, confusing correlation with causation, claiming generalisation from a convenience sample, treating interviews as automatically representative, ignoring contradictory evidence, and describing a study as “mixed methods” without integration.

Best practice is to maintain a design matrix linking each research question to concepts, indicators, data sources, sampling, analysis, and quality criteria. Pilot questionnaires and interview guides. Keep an audit trail of decisions. Pre-specify quantitative analyses where feasible, while remaining transparent about justified changes. Use clear operational definitions, protect participants, and write conclusions at the level supported by the evidence.

Chapter Summary

Research philosophy provides the assumptions underlying a study, while research design provides its practical architecture. Positivism emphasises measurement and testing; interpretivism emphasises meaning and context; critical research examines power and inequality; and pragmatism selects and integrates methods according to the problem. Exploratory designs develop understanding, descriptive designs characterise phenomena, explanatory designs investigate relationships and mechanisms, experimental designs estimate intervention effects, case studies examine bounded situations in depth, and mixed-method designs integrate quantitative and qualitative evidence. A strong accounting or finance dissertation aligns its question, philosophy, design, methods, evidence, analysis, ethics, and claims. Quality depends not on using the most complex method, but on producing a transparent, credible, feasible, and ethically responsible answer.

Key Terms

Ontology: Assumptions about what exists and the nature of reality.

Epistemology: Assumptions about how knowledge can be obtained and justified.

Paradigm: A broad framework of assumptions guiding research.

Methodology: The rationale and logic connecting research questions to methods.

Method: A specific technique for collecting or analysing evidence.

Positivism: A perspective emphasising observable measurement and systematic testing.

Interpretivism: A perspective emphasising socially constructed meaning and context.

Critical research: Inquiry that examines power, inequality, ideology, and possible transformation.

Pragmatism: A problem-centred perspective selecting methods for practical fitness.

Deduction: Moving from theory or hypotheses to empirical testing.

Induction: Developing concepts or explanations from observed evidence.

Abduction: Iteratively moving between theory and evidence to develop the most plausible explanation.

Exploratory design: Design used to investigate insufficiently understood topics.

Descriptive design: Design used to characterise patterns, populations, or practices.

Explanatory design: Design used to investigate relationships, causes, or mechanisms.

Experimental design: Design involving a planned intervention and comparison of outcomes.

Case study: Intensive investigation of a bounded case in context.

Mixed methods: Intentional integration of quantitative and qualitative approaches.

Triangulation: Comparison of evidence from different sources, methods, researchers, or theories.

Reflexivity: Critical consideration of the researcher’s influence and assumptions.

Review Questions

5 Short Answer Questions

1. What is the difference between ontology and epistemology?
2. Identify two typical features of positivist research.
3. Why might an accounting researcher choose an interpretivist design?
4. What is the purpose of triangulation?
5. Distinguish a research method from a research design.

5 Long Answer Questions

1. Compare positivist, interpretivist, critical, and pragmatic perspectives in accounting and finance research.
2. Explain how a researcher should align a research question with a research design.
3. Discuss the strengths and limitations of case-study research.
4. Explain the different forms and purposes of mixed-method design.
5. Evaluate the ethical issues involved in using confidential organisational and financial data.

3 Analytical / Application Questions

1. Design a study investigating whether audit committee characteristics are associated with financial reporting quality. Justify the philosophy, design, data, and quality criteria.
2. A bank wants to understand why customers abandon a digital investment platform. Propose an appropriate exploratory or mixed-method design and explain how evidence would be integrated.
3. A university introduces an AI-supported accounting tutoring system. Develop an evaluation design that distinguishes correlation from causation and addresses privacy, consent, and possible algorithmic bias.

Multiple Choice Questions with Answers

Question 1

Which perspective most directly emphasises socially constructed meanings and organisational context?

A. Positivism

B. Interpretivism

C. Critical realism

D. Pragmatism

Correct Answer: B. Interpretivism

Question 2

A study asking how common sustainability-reporting practices are is primarily what type of design?

A. Descriptive

B. Experimental

C. Exploratory

D. Ethnographic

Correct Answer: A. Descriptive

Question 3

Which statement best describes a research design?

A. A single data-collection technique

B. A statistical significance threshold

C. The planned structure linking questions, evidence, and conclusions

D. A list of references

Correct Answer: C. The planned structure linking questions, evidence, and conclusions

Question 4

In the model yᵢ = β₀ + β₁xᵢ + εᵢ, what does β₁ represent?

A. The unexplained error for observation i

B. The intercept when x equals zero

C. The expected change in y for a one-unit change in x

D. The number of observations

Correct Answer: C. The expected change in y for a one-unit change in x

Question 5

Which design is most directly suited to estimating the effect of a deliberately introduced intervention?

A. Experimental design

B. Descriptive design

C. Exploratory case study

D. Cross-sectional interview study

Correct Answer: A. Experimental design

Question 6

A researcher first surveys managers and then interviews selected managers to explain unexpected survey results. What is this design?

A. Sequential explanatory mixed-method design

B. Sequential exploratory mixed-method design

C. Purely descriptive design

D. Single-case experimental design

Correct Answer: A. Sequential explanatory mixed-method design

Question 7

Which is an appropriate limitation when interpreting an association between software use and forecast accuracy?

A. Association always proves causation

B. Departments may differ in other ways that affect accuracy

C. Qualitative evidence can never be relevant

D. A larger sample eliminates all bias

Correct Answer: B. Departments may differ in other ways that affect accuracy

Question 8

What does analytical generalisation involve in case-study research?

A. Calculating a population percentage from every case

B. Applying statistical significance tests to one participant

C. Relating case findings to concepts or theory beyond the case

D. Assuming every organisation is identical

Correct Answer: C. Relating case findings to concepts or theory beyond the case

Question 9

Which practice best supports qualitative credibility?

A. Deleting contradictory evidence

B. Using only one unrecorded conversation

C. Maintaining an audit trail and comparing multiple sources

D. Avoiding discussion of researcher assumptions

Correct Answer: C. Maintaining an audit trail and comparing multiple sources

Question 10

Which statement about AI in accounting research is most appropriate?

A. AI removes the researcher’s responsibility for accuracy

B. Confidential data may always be uploaded to public AI tools

C. AI may assist approved tasks, but governance, confidentiality, bias, and verification remain the researcher’s responsibility

D. AI-generated analysis automatically establishes validity

Correct Answer: C. AI may assist approved tasks, but governance, confidentiality, bias, and verification remain the researcher’s responsibility