Dissertation Tips

This is a page of miscellaneous dissertation tips based on questions PA 803 students have asked in the past. Students are also invited to send in their own tips for this page.

General Considerations

Have a story line. In the introduction and also in later chapters, make a list of the headings and subheadings in the chapter. You should be able to use these headings, more or less, in sentences in the introduction to the chapter, weaving them together so that they may a story about your problem statement. This lets the introduction be a roadmap to the rest of the chapter. Each section should end with a summary paragraph or summary subsection that keeps the story line in front of the reader.
Stay on message. Avoid controversial claims that are not central to your dissertation. Avoid unnecessary assertions with which others may disagree. Avoid or be extremely careful with phrases such as "It is accepted that" or "Obviously."

The Introductory (Problem Statement) Chapter

Sections. Example section headings might be Introduction; Background; Research Statement (with dependent variable and major independent variables introduced); Theoretical Importance; Policy Importance; Summary (with road map to rest of dissertation). Consult your committee chairperson about this outline.
No abstracts. In dissertations, the chapters have introductions but not abstracts. A statement of the research problem should appear early in the introduction (ex., by paragraph three, and certainly by the second page). Combine your abstract and introduction.
Frame a testable thesis. You must have a research problem, not just a research topic. A dissertation must be analytic, not descriptive. You cannot just describe what processes are. One may have a theory/model that is plausible and well-documented in the literature, but still to be testable there must be a possibility of its being disconfirmed. In your write-up, it is better if your counterpose models, not just try to support one very general one. You must be able to answer two questions your committee will have: (1) What constitutes evidence your model "works"?, and (2) What would be evidence of it not working? Keep in mind that in a dissertation, you need more than a topic. You need a "research problem". A problem is something that is (1) theoretically important or important in policy terms; (2) is something about which people do or could disagree (it is not a matter of describing facts), and (3) can be investigated empirically (usually though not necessarily quantitatively). Ideally, there are at least two theories which suggest alternative explanations of something. That something must have variance across units of analysis. Each theory implies more specific hypotheses. Whether the hypotheses are true or not helps the researcher decide which theory is more in conformity with observations. Key terms in the list of hypotheses form the list of variables needed in data collection.
- It is almost always easier to show effects on preferences, rankings, and other subjective opinions than it is to show effects on outcomes. At a minimum, be very clear which it is you are doing. If you wish to look at outcomes it is all the more important to be comparing models, not "proving" one model.
- If you do look at outcomes, it will be easier if outcomes are defined as immediate outcomes (ex., passage of environmental legislation) rather than as long-term outcomes (effects on pollution levels).
Have a single statement. In the process of introducing, stating, and summarizing your research problem in the introductory chapter, it is quite possible inadvertently to have multiple conflicting statements. Be sure you have a single research statement and be clear as accomplishing the listed specific research tasks will be the standard by which the dissertation is evaluated. If you intended to have two research problems, reconsider. Having two research purposes/statements will not necessarily lead to a better dissertation nor to completing the doctoral degree faster or less problematically. Unless your committee is insisting on two research problems, conventional wisdom would say to stick to one.
Be clearly empirical or clearly normative. Analysis may be normative or empirical. It is very important to be clear about which you are doing. For focus, it is probably best to do one or the other. For instance, if a draft refers to the normative concept of "the public interest" at various times, one must be clear if one wishes to conduct a normative study or not. For instance, does the researcher wish to define effectiveness in terms of "the public interest"? If civic engagement is examined in relation to agency effectiveness, for instance, is the researcher describing civic engagement as an empirical correlate of effectiveness, or recommending civic engagement as a norm needed to avoid planning agency capture?
Limit discussion of theory and literature. Normally the introduction simply identifies the major schools of thought and states which school is most guiding the author, giving general reasons why. Put another way, the introductory chapter haa highlights from the theoretical and research literature, but avoids a detailed literature review. Since your degree is in public administration, while you are definitely not limited to "PA theories," if a notable one is clearly salient, it should be discussed.
Justification. The introductory chapter should contain a "Why This Research Matters" section, usually early on, which establishes the theoretical and/or policy importance of the researcher.
Clarify dependent variable(s). Although this will come up in more detail in the literature review and methods chapters, there should be a section (at least a paragraph) indicating how your dependent variable is to be measured. Be clear about what is a dependent variable and what is not. or instance, in a study of planning effectiveness in relation to civic engagement, there is a difference between studying if civic engagement empirically improves effectiveness, or studying how effective planning departments implement civic engagement. These are two entirely different questions. One's discussion should not sometimes have civic engagement as an independent causing effectiveness, but then sometimes have it as a dependent caused by agency and other attributes. To take another example, the purpose of a study might assert, "The capacity of an agency's performance appraisal systems to support strategic plans is strongly correlated with effectively implementing strategic plans. " However, this is ambiguous because there is a difference between "capacity" and actual "use" of that capacity. Again, these are two different variables. In general, consider your dependent variable carefully, especially asking if it is really two or more separate variables. There are statistical ways to investigate this, too, such as scaling, factor analysis, and structural equation modeling.
Clarify independent variables. Although this will come up in more detail in the literature review and methods chapters, there should be a section (at least a paragraph) indicating examples of independent variables and control variables you will look at. The behavior measures must be different from the effectiveness measures, to avoid tautology.
Clarify control variables. Establishing bivariate relationships is rarely if ever enough in a doctoral dissertation. The research statement must go beyond stating that you expect a certain independent variable to correlate with a given dependent variable. Instead, you should indicate that you will be undertaking a multivariate analysis. While the discussion of control variables will appear in the literature review and particularly the methods chapter, the introduction should still commit to multivariate analysis and give some illustrations of controls to be investigated.
Clarify direction and magnitude. Sometimes the research problem is cast in terms of wishing "to examine the extent to which" the independent variable(s) affect the dependent variable(s). "The extent to which" does not carry what is normally expected in a dissertation: expectations about the direction and magnitude of effect. These expectations are often rooted in theory and past research, but may be implications of the author's model.
Keep the story line prominent and avoid wandering. For example, a section title, "Why Study Organizational Attributes of Local Planning?" is appropriate for the introductory chapter, which should justify the policy and/or theoretical importance of the research question. However, the content in this section needs to keep the story line prominent. For example, the story line might be:
1. Agency effectiveness is a central focus of PA because the central purpose of government is to provide services, including land use planning.
2. There are various causes of effectiveness, and civic engagement may be an important one.
3. Regulatory policy/structure may be a second important cause of effectiveness.
4. The literature also suggests other significant causes such as…..
5. If we understand the causes, we will be in a better position to improve effectiveness (policy importance). Also, understanding which causes are more important than others throws light on the relative salience of one theoretical perspective on government compared to another (theoretical importance).
Give some examples along the way to healp readers understand the generalizations you are making. This should help make your theories and hypotheses more plausible.
Where to put historical background information? One wants not to make historical background the last section of the introduction, which instead should be a summary of the research statement and of its importance. Where background is placed varies by author and the doctoral student should consult his committee chair. Some options are (1) in an appendix; (2) in the literature review section, up front as a prelude to the literature review; (3) in the introductory chapter, after the introductory section, presented as a separate second section heading ("before presenting the research problem and its significance in greater detail, it is appropriate to highlight some aspects of the historical background"). If (2) or (3) are chosen, the background section should be relatively brief highlights as an extended historical section will obscure the story line of the dissertation and may be perceived as "wandering" by your committee.
Road map. Usually the conclusion of the introduction is a summary plus a road map to the chapters which follow.

Literature Review Chapter

Have sections. Like all chapters, section headings are highly recommended. A section organization, for example, might be (1) introduction reminding readers of the research purpose and giving a road map to the rest of the chapter; (2) brief historical background in terms or intellectual and/or policy history; (3) first cluster of theories with associated research on variables related to your dependent variable'; (4) additional clusters; (4) variables and hypotheses that emerge from the literature review, summarized in a literature review table as well. Note that while hypotheses and variables are mentioned in the literature review (at least illustrative hypotheses if not a full list), the later methods chapter is where measurement and indicators are discussed.
Have consistent sections. Apart from an introduction up front and a summary at the end, the body of the literature review chapter has two general strategies (assuming one leaves out arbitrary strategies like chronological presentation): (1) topic headings are variables or clusters of variables (ex., "economic factors"), with related theories discussed under these headings; or (2) topic headings are theories or clusters of theories (ex., "market theories"), with related variables discussed under these headings. Pick one or the other; don't mix and match headings.
Make sure you have a theory, not just a framework. The purpose of a theory is to suggest some independent variables are more critical than others. A theory that says "everything is related" may be true and accurate, but it is not helpful for analysis. Complexity theory, institutional theory, ans systems theory are examples which sometimes are just frameworks (vocabularies), not testable theories. . Instead you have to give such "theories" a narrower interpretation, showing how the theory points to the importance of certain independent variables more than others. The organization of the literature review section may take this sequence: (1) here is a theory about my topic; (2) this theory suggests these hypotheses; (3) here is what past researchers have found out about the relationship in each hypothesis.
Be careful about the word "model". Don't use "model" as a synonym for "theory" or "framework". A model may be stated or implied by a theory, but the two are not the same. A model is a set of variables which have explicit relationships one to another, often represented in circle-and-arrow diagrams, where the circles are variables (not topics, organizations, actors, or processes: that is, a model is not a conceptual map) and the arrows represent causal relations. Where a hypothesis usually involves two variables, a model typically includes multiple variables. The multiple arrows in a graphic diagram each reflect a hypothesis. Sometimes the set of hypotheses embedded in a model are all ones implied by a particular theory. Because theories may explain more than one thing, they may involve more than one model.
Avoid trivial theories. A dissertation must show theoretical or policy importance. One does not want theory for theory's sake. Some theories are too trivial to serve as a focus of a dissertation. For instance, one dissertation proposal focused on a theory which led to the proposition, "The more organizations that participate in a negotiation, the greater the likelihood that the outcomes will be acceptable to all the parties involved. Thus compromise entails satisficing behavior instead of satisfying, creating a division of goods that is acceptable to all parties but not optimal for any one party." A theory that negotiation involves compromise is trivial; to be a dissertation problem, there would have to be much more.
Discussing multiple theories. While it is desirable that a dissertation be theory based, it is possible that in the literature review you will wish to discuss multiple related theories (ex., agenda-setting theory, multiple streams theory, decision-making theory). If you do, you may expose yourself to a rambling "look at this theory," "look at that theory" discussion in which your story line is lost. Be careful to explain your purpose to the reader in presenting each theory. Are you contrasting them and selecting the most appropriate? Are you synthesizing elements from each of them? Are you inventorying shortcomings of all as a prelude to presenting your own, different theory?
It is a very good idea to have an alternative theory to contrast with your main theory. If, for instance, you are using diffusion theory, you ought to be able to imagine "not-diffusion theory". There ought to be some alternative theory which points to the importance of variables not the ones expected to be important by diffusion theory. That is, your main theory must be falsifiable. A theory which cannot be falsified is not a social science theory but is instead merely a descriptive vocabulary, no better than other general-purpose vocabularies.
Be clear. The connection of theory to literature to hypotheses should be explicit, not loose. Also, many theories have multiple dimensions and the writer should be clear which portions of the theory are to be used/tested and which are not. For instance, diffusion theory has a dimension focused on explaining the adoption of innovations over time, but it also has a dimension focused on enumerating success factors in adoption of innovation. A dissertation may use the success factors dimension, making clear the author is not attempting to validate a diffusion-over-time model. Likewise, be clear about the dependent variable(s) for your research problem, especially if you discuss literature which has a different dependent variable from your own. Do not lead the reader to think you will be investigating different dependent variables that the one or ones actually to be studied.
Citing findings of correlation. Be clear when citing findings of correlation if the author in your literature review has provided evidence of causal sequence, or merely of correlation.
Self-check. It is not uncommon for your dissertation committee members to use Google Scholar or the like to check to see if you have actually reviewed the literature pertaining to any given hypothesis. While you are not expected to be comprehensive when, as usual, there is too much literature to discuss, you are supposed to cover all the most relevant literature. Anticipate this by using appropriate bibliographic search services for each hypothesis. Somewhere in the literature review chapter, if only in an endnote, you should have a paragraph detailing your search procedures.

Methodology Chapter

Have sections. Typical sections in the methods chapter are (!) introduction, again reminding readings of the research purpose and giving a road map to the chapter; (2) definition and operationalization of the dependent variable ; (3) list of hypotheses about the dependent variable (referring back to the literature review chapter); (4) operationalization of variables embedded in the hypotheses; (5) discussion of research design, data collection, sampling, and sample bias' (6) identification of the primary quantitative or qualitative methods, with discussion of whether methodological assumptions have been met; (7) discussion of potential validity issues; and (8) a summary section with roadmap to the rest of the dissertation chapters.
List hypotheses and operationalize variables. The usual wording is of the form,. "the more the x, the more (less) the y." For instance, "The greater the mass media coverage of a policy issue, the greater the subsequent attention to it in the legislative agenda." The list of hypotheses contains the list of variables which must be operationalized in the methods chapter. Also, the list of hypotheses is the list of relationships, research on each of which should have been discussed in the previous (literature review) chapter. Usually the dependent variable(s) are defined and operationalized before the independent variables.
Have multiple hypotheses. It is customary to have several hypotheses, not just one.
Avoid overlap of independent variables with the dependent variable.. In this case, committee members objected to a dissertation design because indicators of an independent variable correlated highly with indicators of the dependent variable, suggesting definitional overlap of constructs.
Have control variables. It is almost a dissertation requirement to have multivariate analysis, not just bivariate, which means you must test for control variables after establishing the main effect in an hypothesis. If your sample size is too small to support statistical analysis of control variables, then at a minimum, the controls will dictate case study selection (ex., you will have to have a minimum of one large and one small city case if city size is a control).
Have a "Model Specification" section. This section should state if other independent variables (or control variables) pertinent to your dependent variable(s) emerged from the literature, and if so, give a reason why they were not included, speculating on what difference this might make. This should be just prior to the "Validity Issues" section or made a subsection of it.
The data collection instrument. Operationalization of variables is often effected through a survey instrument. The survey instrument in entirety, including introductory forematter, should be placed in an appendix and referenced in the methods chapter. Although redundant, The survey items should be quoted in the operationalization section, not forcing the reader to go back and forth between the methods chapter and the appendix.. If there was a pilot study, one should describe it and how it helped in the evolution of the final instrument.
Use of scales. Using existing scales to operationalize a construct is generally desirable. Presumably the original authors established the reliability and validity of their scales, and this should be referenced, possibly in the validity section of the methods chapter. More rigorously, you may cross-validate the scales of others by using confirmatory factor analysis or other procedures (principal components analysis, or simply Cronbach's alpha) to see if indicators cluster in the anticipated scale groupings, and if the level of scalability is sufficient in each construct group.
Adequate sample size. If you are doing your own data collection, do a power analysis to assure you have adequate sample size for your purposes. SPSS SamplePower is one of the programs which implements power analysis. As a very broad rule of thumb, about 200 cases is typical for dissertations where you are doing your own survey data collection. Worry if sample size is smaller as you may not have enough cases to use control variables effectively.
Time series vs. cross-sectional analyses. Sometimes "process" analyses are contrasted with "variance" approaches to a topic, such as diffusion. Methodologically, this is just another way of comparing time series analyses (processes over time) with cross-sectional analyses (variance of effects). If time series data are available, such data is more appropriate for understanding processes. However, from a methods point of view, the process explanation still has to explain observed variance at the current time. In the end, any analysis must explain variance (ex., why some states adopt a policy and others do not), but time series data have the potential to throw more light on the process which resulted in the observed variance. Opting for a process approach does not take the researcher off the hook for explaining variance in the dependent variable(s).
Triangulation. In general, multimethod dissertations are better than monomethod ones. You may wish to recap the main citations in your literature review grouped by their respective methodologies (not findings), then explain what mix of approaches you will be using, and why. This can be desirable even if you decide on a single method.
Check for unidimensionality. More often than not, the dissertation will focus on some concept (ex., "organizational capacity") which must be measured by multiple indicators (usually survey items). It is incumbent on the researcher to determine if this is really one concept or merely an umbrella term for multiple variables. The methods section must explore the dimensionality of key concepts in order to establish the data space (how many variables are in the model: on called "capacity" or several with names like "finances", "human resources", etc.). .
Relating two scales
Avoiding null findings. While your committee will very likely accept a null findings dissertation as valid, your experience will be enhanced and publication more likely if you have positive findings. The danger of null findings will be lessened if you look at multiple contexts, so there can be inter-group comparisons. Also, you are less likely to have null findings if you measure your dependent variable in more than one way, which will allow you to explore variance more and have more to say.
Qualitative dissertations. Qualitative as well as quantitative dissertations must adhere to rigorous analytic design, with most of the foregoing considerations applying.

General Style

Professional tone. Be formal. Avoid phrases such as referring to "mountains of evidence" or "I'll just throw out this concept", for example. Take the time to study journal articles in your literature review from a stylistic as well as substantive viewpoint. What your high school English teacher tried to get across does apply here: avoiding run-on sentences, one topic per paragraph, avoiding page-length paragraphs, aving proper antecedants for pronouns, proofreading, etc., etc.
No plagiarism. In general, if you have text in quotes, you probably need a cited reference. However, direct paraphrases need citations also. It is better to err on the side of over-citation than to later be accused of failing to cite your sources or even of plagiarism.
In-text references. When you refer to authors, always have the (year) following the name, not just the name (unless cited in the same paragraph already).
Abbreviations. For abbreviations of agencies or other things, always spell out in full for the first usage, follow with the abbreviation in parentheses, then you can use the abbreviation freely.
Justification. Left justification, not full justification, is usually preferred by publishers.
Use section headings. In general, each chapter should be divided into sections with section headings. Each section should start with a roadmap paragraph and end with a summary paragraph. Try to avoid extremely short sections, considering if they belong in some other longer section.
Number your sections. Numbering your sections (ex., 2.1) is good and will help you refer to these sections elswhere in your dissertation. Be consistent.in your numbering (chapters are the first digit, major sections the second digit, subsections the third digit, etc.).
Keep verbs parallel. Pick a tense and stick with it, not flipping from present to past tense, for instance (ex., not "technology has not been the panacea many organizations expected or need".). Avoid the tendency to use present tense with recent authors and past tense with earlier authors. By the time your work appears in print, they will all be "earlier".
Avoid footnotes. Most publishers now prefer endnotes to footnotes. Check your committee chair's preference.
Avoid neologisms. Generally, it is a bad idea to use neologisms. Thus, "federal system" is simpler and more professional than "federalistic system".
"Government" vs. "governance". "Government" refers to institutions, as in "government offices" or "local government". "Governance" refers to the process of governing, as in "governance strategies" or "philosophies of governance." Thus "levels of governance" would be correct usage if one were referring to some sort of hierarchical typology of approaches to governing, but "levels of government" would be correct if one were referring to federal versus state versus local institutions of government.
"Affects" vs. "effects". If A affects B, the A has some influence on B. If A effects B, then A brings about the existence of B. Thus if "regulatory structure affects agency effectiveness," then structure influences effectiveness. If "regulatory structure effects agency effectiveness," then structure is a sufficient cause to bring effectiveness about. For this example, probably the author would mean "affects".

Quantitative Data Sources

NC Data. UNC's Data-Net has published NC poll and election data at least since 1992, at http://southnow.org/research-and-data. Click on Data-Net.

Qualitative Content Tips

Newspaper content analysis: NCSU gets the NewsSource database, which covers the News & Observer and others. It is listed at the library's database page. You can enter search terms related to your policy for finite time periods, and it will count the number of hits, which you might use as indicators of press attention over time.
Legislation by state. The National Conference of State Legislatures keeps track of legislation at http://www.ncsl.org/. Enter "legislation" in the search box to get a list of policy topics. Some of the listings will link to a page where all states are listed. Click on North Carolina or some other state to get a list of all legislation on that topic. NCSL also has a legislative librarians section and one could find an expert in NC or other state who might help further.