Interpretation, Dissertation Writing and Defence

Introduction and Context

Completing an accounting and finance dissertation involves more than collecting data and reporting statistical results. The researcher must explain what the evidence means, relate it to theory and previous studies, recognise uncertainty, and communicate conclusions in a form that is useful to academic and professional audiences. Interpretation is therefore the bridge between analysis and knowledge. Dissertation writing is the process through which that knowledge is organised, justified and documented. The defence is the formal opportunity to demonstrate that the research question, method, evidence and conclusions form a coherent and defensible whole.

This chapter focuses on the final intellectual and professional stages of a dissertation. It addresses interpretation of quantitative and qualitative findings, comparison with prior evidence, implications for managers and policymakers, limitations, reproducibility, academic writing, visual presentation and oral defence. The discussion is particularly relevant to accounting and finance, where evidence may influence reporting practice, investment decisions, internal control, regulation, taxation and corporate governance.

Learning Objectives

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

1. Distinguish description, interpretation, inference and evaluation.
2. Interpret accounting and finance findings in relation to theory and prior research.
3. Explain statistical and qualitative results without overstating the evidence.
4. Derive practical, managerial, policy and theoretical implications.
5. Identify limitations and explain their consequences for validity and generalisability.
6. Structure and write a coherent dissertation using appropriate academic conventions.
7. Prepare reproducible research materials and ethically responsible documentation.
8. Design effective tables, figures, presentations and oral defence responses.

1. From Findings to Interpretation

1.1 Description is not interpretation

A finding is an observed result; an interpretation explains its meaning. For example, a regression may show that firms with stronger internal controls report fewer material misstatements. The descriptive statement reports the result. Interpretation asks why the relationship may exist, whether it supports the proposed theory, how large and reliable it is, and whether alternative explanations remain plausible.

A useful distinction is:

• Description: what the data show.
• Interpretation: what the result may mean.
• Inference: what may reasonably be concluded beyond the observed sample.
• Evaluation: how the finding compares with theory, prior evidence and practical objectives.

A dissertation should move through these stages in order. Conclusions should not be based on an interpretation that has not been anchored in the reported evidence.

1.2 Statistical interpretation

For a linear regression, a common model is:

Here, y_i is the outcome for observation i, x_i is the explanatory variable, β₀ is the intercept, β₁ is the expected change in y associated with a one-unit change in x, and ε_i is the unexplained component. The interpretation of β₁ requires attention to measurement units, model specification and assumptions. In a multiple regression, the coefficient is interpreted as the association with y holding the included control variables constant.

A coefficient is not automatically a causal effect. Causal language requires a credible identification strategy and assumptions such as appropriate temporal ordering, absence of important confounding, and reliable measurement. If the design is observational and cross-sectional, terms such as “is associated with” or “is consistent with” are usually more defensible than “causes”.

Statistical significance and practical significance are different. A small effect may be statistically significant in a very large sample but economically unimportant. Conversely, a practically important estimate may not be statistically distinguishable from zero in a small sample. A responsible interpretation considers the estimate, its uncertainty, effect size, context and decision consequences.

For an estimate b with standard error SE(b), an approximate 95% confidence interval under suitable large-sample or distributional assumptions is:

The interval expresses a procedure for estimating a range of plausible population values; it should not be described as a 95% probability that a fixed parameter lies within one particular interval. The multiplier 1.96 applies to the standard normal approximation. For small samples, a relevant t critical value is generally more appropriate.

1.3 Interpreting non-significant results

A non-significant result does not prove that no relationship exists. It may reflect a genuinely small effect, insufficient statistical power, noisy measurement, an unsuitable model, restricted variation or an inappropriate operationalisation of the construct. The researcher should report the estimate and confidence interval rather than simply stating that the hypothesis was “rejected”. Where relevant, equivalence testing or power analysis may help distinguish evidence of no meaningful effect from inconclusive evidence.

1.4 Interpreting qualitative evidence

Qualitative interpretation involves identifying patterns, meanings, mechanisms and contrasts in interviews, documents, observations or case materials. Themes should be supported by evidence, but quotations should not substitute for analysis. A strong interpretation explains how a theme answers the research question, how it relates to the conceptual framework, and whether contradictory cases alter the explanation.

For example, interviews may reveal that finance staff regard a new enterprise resource planning system as technically effective but organisationally disruptive. The interpretation should examine the relationship between technology, professional routines, incentives and power—not merely list participants’ comments.

2. Connecting Findings to Theory and Prior Evidence

2.1 Theory as an interpretive framework

Theory provides concepts and proposed relationships that help explain findings. In accounting and finance, relevant perspectives may include:

• Agency theory, which examines conflicts between principals and agents and the role of contracts, monitoring and disclosure.
• Stakeholder theory, which views the organisation as accountable to groups affected by its activities.
• Positive accounting theory, which considers how managers’ incentives and political or contractual pressures may influence accounting choices.
• Institutional theory, which explains how rules, norms and professional expectations shape organisational behaviour.
• Signalling theory, which considers how disclosure may communicate private information or quality to external users.
• Resource-based perspectives, which examine whether valuable, rare and difficult-to-imitate resources contribute to performance.

Theory should not be used as decoration. The dissertation should specify which theoretical mechanism is being examined and identify whether the evidence supports, qualifies or challenges it.

2.2 Comparing a result with previous studies

Comparison with prior evidence should be analytical rather than a catalogue of citations. For each important result, consider:

1. Is the direction of the relationship similar or different?
2. Is the magnitude comparable?
3. Are the constructs and measures equivalent?
4. Do the samples, countries, periods or industries differ?
5. Are the research designs and estimation methods comparable?
6. Could institutional, regulatory or economic conditions explain the difference?

An apparently inconsistent result may be theoretically valuable. For instance, a governance mechanism may improve reporting quality in a highly regulated market but have little effect where enforcement is weak. The difference may suggest a moderating institutional condition rather than a failed study.

2.3 Theoretical contribution

At bachelor level, a contribution need not develop a new grand theory. It may provide a careful application of an established theory to a new setting, compare competing explanations, clarify a measurement issue, or show that a relationship is conditional on firm size, ownership or regulation. Claims should be proportionate to the design and evidence.

Useful contribution statements answer three questions:

• What was known before?
• What does this study add?
• Why does the addition matter?

3. From Findings to Implications

3.1 Managerial implications

Managerial implications translate evidence into decisions without turning correlation into a prescription. A study of working-capital management might suggest that finance managers should monitor collection periods and inventory practices, but the recommendation must consider liquidity risk, service levels, industry conditions and implementation costs.

A useful implication identifies the decision-maker, action, expected benefit, conditions and possible trade-off. For example: “For medium-sized manufacturing firms, introducing monthly exception reports for overdue receivables may improve cash-flow visibility, provided that credit policies and customer relationships are reviewed alongside the reporting system.”

3.2 Policy and regulatory implications

Policy implications should distinguish what the evidence directly supports from what requires further research. If a study finds inconsistent climate-risk disclosures, possible responses include guidance, assurance requirements, standardised metrics or enforcement review. However, policy design must also consider compliance costs, comparability, proportionality, unintended incentives and the risk of boilerplate reporting.

3.3 Implications for investors and other users

Investors may use findings to assess information quality, risk or valuation relevance. Auditors may consider implications for risk assessment and evidence gathering. Boards may use results to review oversight and control structures. The dissertation should identify the relevant user and explain how the evidence could affect a decision, rather than claiming that all stakeholders will benefit equally.

3.4 Ethical implications

Accounting and finance recommendations may affect employees, creditors, customers, taxpayers and communities. Ethical analysis should consider fairness, privacy, transparency, conflicts of interest and distributional consequences. A financially efficient policy may still be problematic if it shifts risk to less powerful stakeholders or relies on intrusive data collection.

4. Limitations, Validity and Generalisability

4.1 Purpose of a limitations section

Limitations are not admissions of failure. They define the boundaries within which the conclusions are credible. A strong limitations section identifies the issue, explains its likely direction or consequence, and states how future research could address it.

Common limitations include:

• Construct validity: the measure may not fully represent the concept.
• Internal validity: alternative explanations may remain.
• External validity: findings may not generalise to other settings.
• Statistical conclusion validity: sampling error, low power or model problems may affect inference.
• Data quality: missing, self-reported, estimated or inconsistent data may introduce bias.
• Temporal limitation: a single period may not capture changing conditions.
• Researcher influence: interpretation may be affected by assumptions or positionality.

4.2 Reliability and validity in mixed designs

Reliability concerns consistency of measurement or procedure; validity concerns whether the research supports the intended interpretation. A reliable measure can still be invalid. For qualitative work, credibility, dependability, confirmability and transferability are useful complementary criteria. Triangulation, audit trails, member reflection, negative-case analysis and transparent coding can strengthen credibility, although no technique eliminates judgement.

4.3 Avoiding overclaiming

Claims should match the design. A survey showing that respondents report higher trust after a disclosure intervention may demonstrate perceived change, but not necessarily behavioural change or long-term economic impact. A case study may provide rich explanation but limited statistical generalisation. The appropriate conclusion may be analytical or theoretical transferability rather than population estimation.

5. Reproducibility and Research Integrity

Reproducibility means that another researcher can understand and, where data and permissions allow, repeat the analytical process. It is especially important in accounting and finance because results may influence investment, assurance and policy.

A reproducible dissertation should document:

• data sources, access dates and inclusion criteria;
• variable definitions, transformations and coding decisions;
• exclusions, missing-data treatment and outlier rules;
• software, packages, versions and operating environment;
• analysis scripts or sufficiently detailed procedures;
• survey instruments, interview protocols and coding frameworks;
• robustness checks and deviations from the research plan.

Reproducibility does not permit disclosure of confidential or personal data. Sensitive material should be anonymised, aggregated, deposited under controlled access, or accompanied by synthetic examples and a clear data-availability statement. Ethical approval, consent restrictions, copyright and contractual data licences must be respected.

A useful file structure separates raw data, processed data, scripts, outputs and documentation. Raw data should be preserved and never overwritten. Version control and a research log help establish what changed and why. AI tools may assist with drafting code, checking prose or suggesting visualisations, but outputs must be verified, confidential information must not be exposed, and the researcher remains responsible for accuracy and authorship.

6. Dissertation Architecture and Academic Writing

A typical dissertation contains the following intellectual sequence:

The exact structure may vary by department and research design, but coherence is essential. The research question should be visible in the introduction, operationalised in the methodology, answered in the findings and revisited in the conclusion.

6.1 Writing the discussion chapter

A reliable paragraph pattern is:

1. State the principal finding.
2. Explain its meaning.
3. Relate it to theory.
4. Compare it with prior evidence.
5. Discuss a plausible alternative explanation.
6. State the implication and boundary of the claim.

The discussion should not repeat every table. It should select the results that answer the research question. Unexpected findings deserve careful attention rather than concealment. Possible explanations should be labelled as explanations or interpretations unless directly tested.

6.2 Academic style

Effective academic writing is precise, economical and evidenced. Prefer concrete verbs and specify the population, period and measure. Avoid unsupported phrases such as “obviously”, “proves” and “everyone agrees”. Distinguish the researcher’s evidence from a cited author’s claim. Use consistent terminology and define abbreviations on first use.

Citation is required for ideas, evidence, data, theories, methods and quotations that are not the researcher’s own. Referencing style must follow institutional requirements. Plagiarism includes unattributed paraphrase, not only copied wording. A similarity report is a screening tool, not a substitute for scholarly judgement.

7. Tables, Figures and Visualisation

Tables should answer a question. A descriptive-statistics table should define variables, report an appropriate number of decimals and identify the sample and missing-data rules. Regression tables should state the dependent variable, coefficient type, standard-error treatment, controls, sample size and model statistics. Significance stars must be explained and should never replace confidence intervals or substantive interpretation.

Figures are useful for trends, comparisons, processes and conceptual relationships. Every figure needs a title or caption, labelled axes, units, source and an explanation in the text. Avoid truncated axes that exaggerate differences, unnecessary three-dimensional effects and colour combinations that reduce accessibility. Financial amounts should identify currency and scale, such as “£ million”.

A visualisation is not evidence merely because it is attractive. The design should preserve proportionality and make uncertainty visible where relevant. For example, a line chart showing annual audit fees should identify whether nominal or inflation-adjusted values are used and should not imply continuity if observations are only available in selected years.

8. Preparing and Conducting the Oral Defence

The defence tests ownership, reasoning and scholarly judgement, not memory alone. Preparation should include a short explanation of the research problem, a one-sentence answer to the research question, the rationale for the method, the principal finding, the main limitation and the contribution.

A professional presentation normally follows this sequence:

1. Problem and motivation.
2. Research question and objectives.
3. Conceptual framework or hypotheses.
4. Method and evidence.
5. Principal findings.
6. Interpretation and contribution.
7. Implications and limitations.
8. Conclusion.

Slides should support speech rather than reproduce the dissertation. Use readable text, consistent terminology and a small number of informative visuals. Report uncertainty and avoid presenting every robustness test unless it is central to the argument.

When answering questions, listen fully, pause, answer directly and distinguish evidence from speculation. If a point is uncertain, acknowledge it and explain how it could be investigated. A useful response structure is: answer, evidence, qualification. For example: “The result is consistent with the agency explanation because the association remains after including ownership controls. However, because the design is observational, I cannot exclude reverse causality.”

Difficult questions often concern sample choice, omitted variables, measurement validity, causal language, contradictory literature, ethical decisions and practical relevance. These are not attacks; they are opportunities to demonstrate critical understanding.

9. Applied Case: Interpreting a Dissertation on ESG Disclosure and Cost of Debt

The following is an illustrative case.

Background

A student investigates whether stronger environmental, social and governance (ESG) disclosure is associated with a lower cost of debt among listed manufacturing companies. Agency theory suggests that better disclosure may reduce information asymmetry and monitoring costs. The sample contains 120 firm-year observations from 30 companies over four years.

Problem and approach

The student constructs an ESG disclosure index from annual reports and estimates a regression in which the dependent variable is the interest rate on new borrowings. Firm size, leverage, profitability and year indicators are included as controls. The estimated ESG coefficient is −0.18 percentage points and its 95% confidence interval is [−0.32, −0.04]. The result is statistically distinguishable from zero under the specified model.

Analysis and interpretation

The coefficient indicates that, holding the included controls constant, a one-unit increase in the index is associated with an estimated 0.18 percentage-point reduction in borrowing cost. The interpretation depends on the index scale: if the index ranges from 0 to 1, a 0.10 increase corresponds to an estimated 0.018 percentage-point reduction, or 1.8 basis points, because 1 percentage point equals 100 basis points. The result is consistent with the information-asymmetry mechanism, but it does not establish that disclosure causes cheaper debt. Firms with better management, stronger controls or lower unobserved risk may both disclose more and borrow more cheaply.

Implications

Managers may view decision-useful ESG disclosure as part of creditor communication rather than a purely reputational exercise. Lenders may incorporate disclosure quality into due diligence, while regulators may consider comparability and assurance. These implications should be conditional: disclosure can create costs, and poorly verified information may increase rather than reduce uncertainty.

Limitations and outcome

The index involves judgement, the sample is limited to listed manufacturers, and reverse causality remains possible. The student therefore concludes that the evidence supports an association consistent with the theory, not a universal causal effect. A stronger future design could use a longer panel, independent assurance data, fixed-effects methods and a policy change that creates plausibly exogenous variation.

Learning points

The case demonstrates that interpretation requires measurement awareness, theory, comparison with alternatives, careful units, proportionate recommendations and explicit limitations. The most persuasive conclusion is not the strongest-sounding one; it is the one most faithfully supported by the research design.

10. Best-Practice Checklist

Before submission and defence, the researcher should verify that:

• the title, question, objectives and conclusion are aligned;
• every objective is answered with evidence;
• tables and figures agree with the text and appendices;
• units, denominators, dates and currencies are clear;
• causal language matches the design;
• limitations explain consequences rather than merely list weaknesses;
• references are complete and consistently formatted;
• ethical approval, consent and data restrictions are documented;
• data and code documentation is sufficient for reproducibility;
• recommendations identify users, actions, conditions and trade-offs;
• the oral presentation can be delivered within the allocated time.

Chapter Summary

Interpretation converts findings into justified meaning. It requires separation of description, inference and evaluation; careful attention to coefficients, uncertainty, measurement and causal assumptions; and meaningful comparison with theory and prior evidence. Accounting and finance dissertations should derive proportionate implications for managers, investors, regulators and other stakeholders while considering ethical and distributional consequences.

Limitations establish the boundaries of credible knowledge and should explain their effect on validity and generalisability. Reproducibility depends on transparent data, coding, analytical and documentation practices, subject to privacy and ethical constraints. Effective dissertation writing creates alignment among the question, literature, method, findings, discussion and conclusion. Tables, figures and oral presentations should clarify the argument rather than decorate it. In a defence, confidence comes from understanding the choices, evidence and limitations of the research.

Key Terms

Interpretation: The process of explaining the meaning and significance of research findings.

Inference: A conclusion drawn from evidence, often extending from a sample to a wider population.

Causal claim: A claim that a change in one factor produces a change in another.

Statistical significance: Evidence that an observed result is unlikely under a specified null model and significance threshold.

Practical significance: The real-world importance or magnitude of a result.

Confidence interval: An interval produced by a statistical procedure to express uncertainty about a population parameter.

Effect size: A quantitative measure of the magnitude of a relationship or difference.

Construct validity: The extent to which a measure represents the concept it is intended to measure.

Internal validity: The credibility of an explanation within the study setting.

External validity: The extent to which findings apply beyond the studied setting.

Generalisability: The justified extension of findings to other people, organisations, settings or periods.

Triangulation: The use of multiple sources, methods or perspectives to examine a phenomenon.

Theoretical contribution: The way research extends, applies, qualifies or challenges existing explanation.

Managerial implication: A decision-relevant consequence for organisational practice.

Policy implication: A consequence relevant to regulation, public administration or standard-setting.

Reproducibility: The ability of another researcher to understand and repeat an analytical process.

Research integrity: Conduct that protects honesty, transparency, accountability and respect in research.

Positionality: Reflection on how the researcher’s perspective may influence the research.

Oral defence: A formal examination in which the researcher explains and defends the dissertation.

Review Questions

5 Short Answer Questions

1. Distinguish description from interpretation in dissertation writing.
2. Why is statistical significance different from practical significance?
3. State two reasons why a non-significant result may occur.
4. What information should a reproducible research record contain?
5. Why must causal language match the research design?

5 Long Answer Questions

1. Explain how a researcher should connect empirical findings to theory and prior evidence.
2. Discuss the main types of dissertation limitation and their implications for validity.
3. Explain how managerial and policy implications can be derived without overstating evidence.
4. Evaluate the role of tables, figures and visualisation in accounting and finance dissertations.
5. Describe how a student should prepare for and respond to questions in an oral defence.

3 Analytical / Application Questions

1. A cross-sectional study finds that firms with independent audit committees have higher reporting quality. Explain what can and cannot be concluded, identify two alternative explanations, and propose a stronger research design.
2. A regression coefficient is −0.25 with a standard error of 0.10. Using the approximate 95% interval formula b ± 1.96 × SE(b), calculate the interval and explain its meaning, assuming the approximation is appropriate.
3. Design a discussion section for a dissertation whose findings support one hypothesis, contradict another and provide inconclusive evidence for a third. Explain how theory, prior research, implications and limitations should be integrated.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes interpretation from description?

A. Description explains why a result occurred, whereas interpretation lists observations.

B. Description reports what the data show, whereas interpretation explains the meaning and significance of the result.

C. Description is always qualitative, whereas interpretation is always quantitative.

D. Description makes causal claims, whereas interpretation avoids all conclusions.

Correct Answer: B. Description reports what the data show, whereas interpretation explains the meaning and significance of the result.

Question 2

In an observational cross-sectional study, which wording is generally most defensible?

A. The independent variable proves the outcome.

B. The independent variable causes the outcome in every firm.

C. The variables are associated, holding the included controls constant.

D. The outcome is irrelevant because causality is not established.

Correct Answer: C. The variables are associated, holding the included controls constant.

Question 3

What is the main purpose of a confidence interval?

A. To show that the sample mean is always correct.

B. To express uncertainty about a population parameter using a specified statistical procedure.

C. To prove that a relationship is causal.

D. To replace the need to report an effect size.

Correct Answer: B. To express uncertainty about a population parameter using a specified statistical procedure.

Question 4

Which is an example of a theoretical contribution at bachelor level?

A. Claiming to replace all existing accounting theories.

B. Applying an established theory to a new setting and identifying a relevant boundary condition.

C. Removing the literature review because the data are original.

D. Reporting a significant p-value without explaining the mechanism.

Correct Answer: B. Applying an established theory to a new setting and identifying a relevant boundary condition.

Question 5

Which practice most directly improves reproducibility?

A. Overwriting the raw dataset after cleaning.

B. Reporting only the final graph.

C. Documenting variable transformations, exclusions, software and analysis procedures.

D. Keeping all analytical decisions undocumented to avoid bias.

Correct Answer: C. Documenting variable transformations, exclusions, software and analysis procedures.

Question 6

A non-significant coefficient necessarily proves that the true effect is zero. Is this statement:

A. True, because non-significance establishes no relationship.

B. True, provided the sample is large.

C. False, because non-significance may also reflect low power, noisy measurement or model limitations.

D. False, because all non-significant results are caused by data errors.

Correct Answer: C. False, because non-significance may also reflect low power, noisy measurement or model limitations.

Question 7

Which feature is most appropriate for a managerial implication?

A. It identifies an action, decision-maker, conditions and potential trade-offs.

B. It recommends the same action for every organisation regardless of context.

C. It repeats the statistical table without identifying a decision.

D. It uses causal language even when the design is correlational.

Correct Answer: A. It identifies an action, decision-maker, conditions and potential trade-offs.

Question 8

In the illustrative ESG case, a 0.10 increase in the index with a coefficient of −0.18 percentage points implies an estimated change of:

A. −0.018 percentage points, assuming the coefficient is linear and the index ranges from 0 to 1.

B. −1.8 percentage points, because 0.10 equals 10 percentage points.

C. +0.018 percentage points, because the coefficient is negative.

D. −18 percentage points, regardless of the index scale.

Correct Answer: A. −0.018 percentage points, assuming the coefficient is linear and the index ranges from 0 to 1.

Question 9

Which response is most appropriate when a defence examiner asks about an unresolved limitation?

A. Deny that any limitation exists.

B. Change the subject immediately.

C. Acknowledge it, explain its consequence and propose how future research could address it.

D. Claim that limitations never affect conclusions.

Correct Answer: C. Acknowledge it, explain its consequence and propose how future research could address it.

Question 10

What should a dissertation discussion chapter primarily do?

A. Repeat every result without evaluation.

B. Introduce an entirely new research question.

C. Interpret the most relevant findings against theory and prior evidence, while stating implications and boundaries.

D. Present only results that support the hypotheses.

Correct Answer: C. Interpret the most relevant findings against theory and prior evidence, while stating implications and boundaries.