A good EPQ question is narrow enough to answer in 5,000 words with data you can actually reach. AQA awards 20% of marks to AO2 — 'research, critically select, organise and use information' — so the question you choose determines a fifth of your grade before you write anything.
Most EPQ ideas lists are topics, not questions. "Climate change." "Artificial intelligence in medicine." Those are fields, and a student who takes one into February discovers there is no way to finish.
So these thirty are phrased as questions, each narrow enough to answer, with the source you would answer it from.
What makes an EPQ question answerable?
You can say in one sentence what evidence would change your mind.
That is the whole test. "Is social media harmful to teenagers?" fails it — no obtainable evidence settles it, and 5,000 words will produce a summary of other people's arguments. "Does self-reported daily screen time predict self-reported sleep duration among US 12th graders in the 2023 Monitoring the Future survey, controlling for sex and region?" passes, because the dataset exists and the analysis is within reach.
This matters for marks, not just for finishing. AQA allocates 20% of the available marks to AO2, defined as "research, critically select, organise and use information." The operative phrase is critically select — you are being assessed on choosing between sources that disagree. A question answerable from one textbook cannot earn those marks however well it is written.
The practical rule: your question should force you to adjudicate a disagreement.
Thirty questions, by field
Each of these can be answered with material a school student can reach.
Economics and public policy
- Did the introduction of the Sugar Levy in 2018 change the sugar content of UK soft drinks, or the price, or both? (ONS, manufacturer reformulation data)
- Do areas with more betting shops show higher rates of personal insolvency in England and Wales? (Gambling Commission, Insolvency Service)
- Has the rise of remote work changed rental prices differently in commuter towns and city centres? (ONS rental indices)
- Does minimum wage variation across US states correlate with teenage employment rates? (IPUMS, Bureau of Labor Statistics)
- Do countries with higher female labour participation show lower fertility rates, and does the relationship hold after income is controlled for? (World Bank Open Data)
Public health and epidemiology
- Did US counties with earlier mask mandates show different excess-mortality curves in 2020–21? (CDC WONDER)
- Is the urban–rural gap in life expectancy in England widening or narrowing since 2010? (ONS)
- Do areas with more fast-food outlets per capita show higher childhood obesity rates, and is the effect distinguishable from deprivation? (NHS Digital, local authority data)
- Has the introduction of HPV vaccination changed cervical cancer incidence in the UK cohorts old enough to show it? (Cancer Research UK, ONS)
- Does air quality measured at school locations correlate with recorded asthma admissions? (DEFRA air quality data, NHS)
Psychology and education
- Does reported screen time predict reported sleep duration among US adolescents, controlling for grade and sex? (Monitoring the Future)
- Has the attainment gap between free-school-meal and other pupils changed since 2019? (DfE statistics)
- Do students in selective school systems report different academic self-concept from those in comprehensive systems? (PISA)
- Is there a measurable relationship between class size and attainment in the PISA data once national income is controlled for? (PISA)
- Do later school start times correlate with attendance in districts that have adopted them? (US district-level data)
Political science
- Did the 2024 US presidential vote shift more among voters without degrees in counties with higher manufacturing employment loss? (ANES, county election returns)
- Has turnout in UK local elections diverged from general election turnout since 2010? (Electoral Commission)
- Do parliamentary constituencies with younger populations show different Brexit-era voting patterns after controlling for education? (ONS, House of Commons Library)
- Is there a relationship between a country's press freedom score and its vaccination uptake? (RSF index, WHO)
- Do coalition governments pass fewer bills per session than majority governments in comparable parliamentary systems? (national legislative records)
Computer science and data
- Can a simple classifier distinguish AI-generated from human-written student essays, and where does it fail? (your own corpus)
- How much does the readability of Wikipedia medical articles vary by condition prevalence? (Wikipedia API, readability metrics)
- Do open-source projects with codes of conduct show different contributor retention? (GitHub API)
- How accurate are consumer weather forecasts at 1, 3 and 7 days for one location over six months? (your own logged data)
- Does sentiment in local news coverage of a policy predict subsequent polling movement? (news archives, polling data)
Biology, chemistry and environment
- Has the flowering time of one common species shifted in your region over the recorded period? (Nature's Calendar, phenology records)
- Do urban and rural bird survey counts show different species diversity trends? (BTO Breeding Bird Survey)
- How does water hardness vary across supply zones, and does it correlate with published kettle-scale complaints or appliance data? (water company quality reports)
- Has recorded microplastic concentration in published river surveys changed with upstream land use? (Environment Agency data, published surveys)
- Does the caffeine content stated on drink labels match published independent measurements? (literature review plus label survey)
How to turn a topic into a question in one afternoon
Four steps, in order.
| Where to get data | What it holds | Fields it serves |
|---|---|---|
| IPUMS | Census microdata, US and international | Economics, demography, sociology |
| CDC WONDER | US mortality, natality, disease surveillance | Public health, epidemiology |
| Monitoring the Future | Annual US adolescent behaviour survey | Psychology, education |
| World Bank Open Data | Cross-country development indicators | Economics, politics |
| ONS | UK population, prices, health, labour market | Almost any UK question |
Find the dataset first. This inverts the usual advice and works better. Pick something you can reach, spend an hour in its documentation, and look for a variable pairing nobody seems to have checked. Narrow questions come from data, not from brainstorming.
Write the question with its variables named. If your question does not contain the specific things being compared, it is still a topic.
Say out loud what result would surprise you. If no result would, the question is not open.
Check that someone disagrees. Find two sources that take different positions. That disagreement is where your AO2 marks live.
What to avoid
Anything requiring a lab. Not because you lack the skill — because of policy. Most universities bar under-eighteens from BSL-2, radioactive and animal work, and require constant direct supervision and safety paperwork filed weeks ahead. If your question needs a lab you do not have access to, change the question rather than waiting.
Questions answerable in one paragraph. "What is the capital structure of a REIT" is a fact, not a project.
Questions with no obtainable evidence. "Would the Roman Empire have survived if…" is fun and unassessable.
Human-subjects research without thought. Surveying your own year group sounds easy and raises consent and ethics questions your school will have a policy about. Check before you design it.
If you want to publish it afterwards
Know that most EPQs cannot be, as written, and choose accordingly now.
Student research journals overwhelmingly want original data and a paper structure — introduction, methods, results, discussion. An EPQ is an extended essay with a production log and a reflective component. Converting one means narrowing the question, adding a proper methods section and cutting the reflection: real work, not reformatting.
If publication matters to you, choose a question from the list above with a dataset attached rather than a purely argumentative one. Then read the venue's rules before you start. The Journal of Emerging Investigators charges $49 at submission, refuses to expedite for any reason, and wants original findings. Note also that Curieux Academic Journal has closed to new submissions — it still appears on most guides.
Where we come in
The Oxford Centre for Advanced Research pairs students with PhD candidates and early-career researchers at Oxford, Cambridge, other leading UK universities and Ivy League institutions. The Oxford Scholar Programme is £2,000 for ten contact hours over ten to fourteen weeks; the Oxford Publication Fellowship is £3,400 and is built around submission.
The single thing a mentor does that matters most for an EPQ happens in the first two sessions: refusing the question you arrived with and making you narrow it. Students who get that right finish in September. Students who do not are still reading in March.
You can get that from a teacher with a doctorate, from a graduate student willing to read one page, or from us. The order in that sentence is deliberate — the first two are free and you should try them first.
Before you commit to a title
Write your question down, then write underneath it the single sentence describing what evidence would change your answer.
If you cannot write that second sentence, you do not have a question yet. Ten minutes spent there saves a term.
Frequently asked questions
What makes a good EPQ question?
It can be answered, with evidence you can get, in 5,000 words. The test is whether you can name in one sentence what data would change your mind. 'Is social media bad for teenagers' fails that test. 'Does reported screen time predict reported sleep duration in the 2023 Monitoring the Future sample' passes it.
How narrow is too narrow?
Rarely a problem in practice — almost every failing EPQ is too broad, not too narrow. A question so specific that the answer is a single number is thin, but you can widen a narrow question in an afternoon. Narrowing a vast one usually means starting again in February.
Can I do an EPQ on something not related to my A-levels?
Yes, and it is often the better choice. The EPQ rewards the research process rather than subject knowledge, so a question in an adjacent field gives you something distinctive to discuss at interview without competing against your own A-level teaching.
Do I need original data?
No. The EPQ does not require it, and a well-argued analysis of existing evidence scores fully. It does matter if you later want to submit to a research journal, since most student journals want original findings — so if publication is a goal, choose a question with a dataset behind it.
How does the EPQ mark scheme affect the question?
AQA allocates 20% of the marks to AO2, defined as 'research, critically select, organise and use information'. The phrase doing the work is critically select. A question that forces you to choose between conflicting sources earns those marks; one answerable from a single textbook does not.
Can an EPQ be published?
Usually not as written. Journals want original data and a paper structure; an EPQ is an extended essay with a production log. Converting one takes real work — narrowing the question, adding a methods section, cutting the reflective material. It is possible, and it is not a re-format.