A political science thesis in the Philippines needs a research question stated as a testable relationship, one or more null and alternative hypotheses derived from it, and clearly named independent and dependent variables — most commonly a governance instrument, program or demographic trait as the independent variable and a measured attitude, compliance level or participation rate as the dependent variable. The panel checks three things at the hypotheses stage: that each hypothesis follows from the statement of the problem, that the variables are actually measurable with the respondents or records the student can reach, and that a named statistical test exists for each one.
This answer walks through how a political science hypothesis differs from a general research question, the variables that recur in AB Political Science and Public Administration theses in the Philippines, four worked examples across governance, elections, youth participation and public administration, and the test each hypothesis implies. If you have not yet chosen a topic to build hypotheses from, start with the twenty political science thesis topics scoped to one local government unit.
What is the difference between a research question and a hypothesis in political science?
A research question asks; a hypothesis states a predicted answer that a statistical test can accept or reject. “What is the relationship between SK budget utilisation and youth program completion?” is a research question. “There is no significant relationship between SK budget utilisation rate and youth program completion rate” is the null hypothesis a chi-square or correlation test can address directly. Philippine research manuals generally ask for both: the research question in the statement of the problem, and the corresponding null hypothesis (Ho) — sometimes with its alternative (Ha) written out — in a numbered list right after it, one Ho per sub-problem.
Not every political science thesis needs a hypothesis. Purely qualitative, document-based or exploratory designs — a case study of one Bangsamoro transition committee, a content analysis of ordinances — usually state objectives instead, because there is no numeric relationship to test, the same logic that governs which research design a social-science thesis should use when its respondents are officials rather than a random sample. Hypotheses belong to quantitative and mixed designs with a survey, a coded dataset or an experiment component.
What variables show up most often in a Philippine political science thesis?
The independent variable is almost always something the researcher did not create: a law, a local program, a demographic category, or an exposure the respondent already has. The dependent variable is what the study measures as an outcome — usually from a validated or researcher-made scale, an official record, or a self-report.
| Sub-field | Typical independent variable | Typical dependent variable |
|---|---|---|
| Local governance | SGLG compliance status; NTA/devolution transition stage; barangay demographic profile | Perceived responsiveness; service satisfaction; budget utilisation rate |
| Elections and representation | Age group; education level; exposure to campaign material; incumbency status | Voting intention; candidate preference; turnout |
| Youth and civic participation | Social media use intensity; SK membership; voter education exposure | Political efficacy; civic participation level; turnout intention |
| Public administration | Citizen’s charter compliance level; frontline staffing ratio; digitalisation stage | Client satisfaction score; processing time; complaint volume |
Where the design compares groups — officials versus residents, urban versus rural barangays, before versus after a policy — the grouping variable itself becomes a second independent variable, and the hypothesis tests whether the outcome differs across groups rather than only whether two variables correlate. Each variable pairing eventually needs a stated rationale in the significance of the study, in the structure described in what the significance of the study should contain.
How do you write a null and alternative hypothesis for a governance study?
Start from the sub-problem, not the topic. If sub-problem 2 asks “Is there a significant relationship between the respondents’ length of residency and their trust in the barangay council?”, the null hypothesis mirrors it exactly in the negative: “There is no significant relationship between the respondents’ length of residency and their trust in the barangay council.” The alternative hypothesis is the mirror positive, and in a two-tailed test it simply asserts that a relationship exists without predicting its direction: “There is a significant relationship between the respondents’ length of residency and their trust in the barangay council.”
Three checks catch most panel objections before the defense. First, the variables named in the hypothesis must match the variables named in the conceptual framework and the questionnaire — a hypothesis about “trust” needs a trust scale in the instrument, not a proxy question the panel has to infer from. Second, a hypothesis comparing groups (“satisfaction differs between urban and rural respondents”) needs roughly balanced group sizes in the sample, or the test loses power. Third, state one hypothesis per testable relationship rather than one long compound hypothesis — “length of residency and education are related to trust” is two hypotheses, and a panel will ask which test addresses which half.
What are worked hypotheses across political science sub-fields?
The four examples below use invented, clearly labelled sample data to show the shape of a complete hypothesis-to-test chain. Replace the numbers with your own respondents’ figures; the structure is what a panel is checking.
Local governance: SGLG compliance and household satisfaction
Research question: Is there a significant difference in household satisfaction with basic services between barangays with full Seal of Good Local Governance (SGLG) compliance and barangays without it, within one municipality?
Ho: There is no significant difference in mean household satisfaction scores between SGLG-compliant and non-compliant barangays.
Sample data (invented): compliant barangays, mean satisfaction 4.1 of 5 (n = 60); non-compliant barangays, mean satisfaction 3.4 of 5 (n = 60). An independent-samples t-test on this two-group, interval-scale comparison would either reject or fail to reject Ho at the alpha the manual specifies, typically 0.05.
Elections: campaign exposure and vote choice
Research question: Is there a significant relationship between a first-time voter’s exposure to social media campaign material and their stated candidate preference in a local election?
Ho: There is no significant relationship between social media campaign exposure level and stated candidate preference.
Sample data (invented): exposure recorded on a 3-level ordinal scale (low, moderate, high); candidate preference recorded as a nominal category among three candidates. Because both variables are categorical, a chi-square test of independence — not correlation — is the appropriate test for this pairing.
Youth participation: SK program exposure and political efficacy
Research question: Is there a significant relationship between a youth respondent’s number of Sangguniang Kabataan programs attended and their internal political efficacy score?
Ho: There is no significant relationship between SK program attendance count and internal political efficacy score.
Sample data (invented): program attendance recorded as a count (0 to 6); efficacy measured on a validated 10-item Likert scale summed to a continuous score. Both variables here are continuous or near-continuous, so Pearson’s r, or Spearman’s rho if the efficacy scores are not normally distributed, fits the pairing.
Public administration: digitalisation stage and processing time
Research question: Does the digitalisation stage of a city’s business permitting office (manual, semi-automated, fully online) significantly affect the average processing time of a business permit application?
Ho: There is no significant difference in mean processing time across the three digitalisation-stage groups.
Sample data (invented): manual, mean 14 days (n = 20 applications); semi-automated, mean 8 days (n = 20); fully online, mean 3 days (n = 20). With three independent groups and one continuous outcome, one-way ANOVA is the test, followed by a post hoc test such as Tukey’s HSD if the omnibus result is significant, to identify which pair of stages actually differs.
Which statistical test fits which hypothesis?
| Variable pairing | Test |
|---|---|
| Two groups, one continuous outcome | Independent-samples t-test |
| Three or more groups, one continuous outcome | One-way ANOVA, with a post hoc test |
| Two categorical variables | Chi-square test of independence |
| Two continuous or ordinal variables, linear relationship | Pearson’s r (continuous, normal) or Spearman’s rho (ordinal or non-normal) |
| One outcome predicted by several variables at once | Multiple regression |
The choice of test is not a late-stage decision — it has to be settled before the questionnaire is finalised, because a Likert item collapsed into “agree/disagree” categories forecloses Pearson’s r and forces a chi-square instead. Deciding the scale of measurement for each variable at the same time as writing the hypothesis avoids rebuilding the instrument in Chapter 3, and the wider logic for choosing among these tests is set out in the answer on what statistical treatment of data a thesis needs.
What common mistakes do panels flag in a political science thesis’s hypotheses?
A hypothesis with a variable that never appears in the questionnaire or the conceptual framework — the panel asks where it was measured. A directional hypothesis (“satisfaction is higher in compliant barangays”) paired with a two-tailed test in the results chapter — the direction claimed in Chapter 1 has to match the tail specified in Chapter 3. A hypothesis about “the Philippines” or “Filipino voters” when the actual sample is 150 respondents in one city — the hypothesis should be worded at the scope the data can support. And a compound hypothesis bundling two relationships under one Ho, which makes the results chapter unclear about which test answered which claim. These are exactly the kind of gaps a panel probes at title and final defense; see the pattern of questions in what questions are asked during a thesis defense in the Philippines.
Turn your topic into numbered hypotheses and variables
Tesify drafts the statement of the problem, the numbered null hypotheses and the variable definitions from the topic you choose, matched to the Chapter 1 structure Philippine political science and public administration programs require, and flags when a hypothesis’s variables do not yet appear in your questionnaire. Over 9,000 students have used it to write more than 15,000 chapters, and every word is still written by you.
Frequently asked questions
How many hypotheses should a political science thesis have?
One null hypothesis per testable sub-problem, usually two to five for an undergraduate thesis. A thesis with one sub-problem about description (“what is the level of X”) needs no hypothesis at all; only sub-problems that compare or relate variables need one.
Do I write the alternative hypothesis (Ha) too?
Some Philippine research manuals ask for both Ho and Ha written out; others ask only for Ho, with Ha implied. Check your school’s manual — the format is a formatting rule, not a methods rule, and panels flag its absence only where the manual requires it.
What if my variables are all categorical?
Then chi-square, not correlation or regression, is almost always the right family of test. Categorical-by-categorical pairings, such as region by preferred candidate, are exactly what chi-square tests of independence were built for.
Can a qualitative political science thesis have hypotheses?
Generally no. A purely qualitative design — key-informant interviews, a single case study, document analysis without coding to numeric categories — states research objectives and, sometimes, propositions, not statistically testable hypotheses.
What is the difference between an independent and a moderating variable?
The independent variable is the presumed cause; a moderating variable changes the strength or direction of that relationship for a subgroup. “Trust in the barangay council” moderated by “length of residency” would mean the relationship between an independent variable and trust looks different for long-time residents than for newcomers — a more advanced design than most undergraduate theses attempt, but common at the graduate level.
Should I use a validated scale or a researcher-made questionnaire?
A validated scale — for political efficacy, for instance — strengthens the thesis because its reliability is already established in the literature; cite the original author and report the reliability statistic from your own data collection. A researcher-made questionnaire is acceptable when no validated instrument fits the local governance context, but expect the panel to ask about its pilot testing and reliability check, a step covered section by section in how to write Chapter 3 research methodology.
What sample size do I need to test these hypotheses?
It depends on the test and the effect size you expect to detect, not on a fixed rule of thumb. Chi-square and ANOVA generally need larger per-group samples than a correlation between two continuous variables; a thesis adviser or statistician can confirm the minimum for your specific design before data collection begins.
Do document-based political science theses need hypotheses?
Only if the document analysis produces numeric, comparable data — counts of ordinances by year, budget allocations by category — that a statistical test can compare across groups or time periods. A purely descriptive document analysis states objectives instead.