Potential project ideas
Please do not pick one of these.
They exist so that nobody is stuck staring at a blank page in week 2. Every one is a lead, not a project. Each still has to survive the feasibility filter, and each still needs a decision attached to it before it is worth doing.
The best capstones have always come from somebody’s own curiosity. If one of these sparks a better idea two steps sideways, that is what it is for.
Where each dataset lives, and how to get an account, is on the data sources page. Methods for generating your own ideas are in meeting 1.
Archetypes: 1 evaluation · 2 preference & pricing · 3 prediction & risk · 4 diagnosis · 5 feasibility.
Property and place
| The lead | Data | Type | |
|---|---|---|---|
| 1 | What is a water view worth, and does the premium survive when you compare near-identical houses? | ATTOM transactions + Nature Score (Dewey) | 2 |
| 2 | Is flood or fire risk discounted in house prices yet — and did that change after a local disaster? | US Climate Risk × ATTOM (Dewey) | 1 |
| 3 | When a housing market softens, which price tier moves first? | ATTOM + Dwellsy rents (Dewey) | 4 |
| 4 | Can you beat “last sale price, index-adjusted” at valuing a listing — and what are the big misses? | ATTOM (Dewey) | 3 |
| 5 | Where do people move after a mass layoff in their county? | Consumer Address History × WARN (Dewey) | 4 |
Retail and foot traffic
| The lead | Data | Type | |
|---|---|---|---|
| 6 | Did a competitor’s opening take traffic from nearby stores, or grow the whole area? | Weekly Patterns Plus (Dewey) | 1 |
| 7 | Which store locations underperform what their catchment should deliver? | Daily Store Visits + ACS (Dewey) | 4 |
| 8 | Does falling foot traffic predict a closure, and how far ahead? | Weekly Patterns × Business Entity (Dewey) | 3 |
Labour
| The lead | Data | Type | |
|---|---|---|---|
| 9 | Salary-disclosure laws hit different states at different times. Did employers change what they post? | Job Postings (Dewey) | 1 |
| 10 | Which skills appeared in job ads in the last two years that were absent before? | Job Records (Dewey) | 4 |
| 11 | Where does talent actually go after a layoff — same industry, or out? | WARN × People and Resume (Dewey) | 4 |
| 12 | Do postings that name a salary fill differently from those that do not? | Job Postings (Dewey) | 1 |
Food and consumers
| The lead | Data | Type | |
|---|---|---|---|
| 13 | When a price rises, what do shoppers switch to inside the same category? | Convenience Store Transactions, SKU level (Dewey) | 1 |
| 14 | Did delivery apps grow the restaurant category or cannibalise dine-in? | Delivery card data × Weekly Patterns (Dewey) | 1 |
| 15 | What is a “locally sourced” claim worth, in cents per unit? | Your own choice experiment (Qualtrics + Prolific) | 2 |
| 16 | Do neighbourhoods with only convenience stores show worse health outcomes than the food access story predicts? | Global Places × CDC PLACES × ACS (Dewey) | 4 |
Finance
| The lead | Data | Type | |
|---|---|---|---|
| 17 | Does retail attention move prices, or follow them? | Social Sentiment × Global Equity EOD (Dewey) | 3 |
| 18 | Can card spending nowcast a company’s earnings before it reports? | Daily Consumer Spend by Company (Dewey) | 3 |
| 19 | Did PPP loans change small-business survival? | SBA Loans × Business Entity (Dewey) | 1 |
| 20 | Do ESG scores predict anything once you account for firm size? | ESG Scores × Public Company Financials (Dewey) | 4 |
Health, policy and development
| The lead | Data | Type | |
|---|---|---|---|
| 21 | Two hospitals in one city, same billing code, different negotiated price. Why? | Payer Rates (Dewey) | 4 |
| 22 | Who actually wins green public procurement, and are they different firms? | Global Green Government Contracts (Dewey) | 4 |
| 23 | Does lobbying activity precede favourable rule changes, or follow them? | US Lobbying (Dewey) | 1 |
| 24 | Did port congestion in 2021–22 change where goods were shipped from? | Port Performance (Dewey) | 1 |
| 25 | Is public contracting more concentrated where institutions are weaker? | Unified Government Contract Awards + World Bank governance indicators | 4 |
Technology
| The lead | Data | Type | |
|---|---|---|---|
| 26 | What happens to an app’s daily users when a competitor launches? | Daily App Engagement (Dewey) | 1 |
| 27 | Does a firm’s headcount move before its share price does? | Company Insights × Global Equity EOD (Dewey) | 3 |
| 28 | Does disclosing that content is AI-generated change how people rate it? | Your own online experiment (Qualtrics + Prolific) | 2 |
Collect it yourself
Ideas 15, 28 and anything like them need no dataset at all. You write the instrument and field it.
This is worth taking seriously. The most ambitious project in this programme’s history ran online experiments on brand recognition, then persuaded a soda company to donate thousands of cans for a field experiment in dining halls. That was a nine-month MPS team. It took no econometrics.
Meeting 9 teaches instrument design and how large a sample you need. Permissions and ethics covers the IRB, and you should start that in October.
If none of these fit
Good. Bring your own.
Take it to the four filter questions from meeting 1. Could a team of 3–5 execute it in one spring semester? Can you name the data, who holds it, and how you get it — checked, not assumed? Could the answer come out either way? Would anybody act differently depending on it?
Survives all four? Label its archetype and bring it to class. Fails one? Say which. A question you kill in September is a month you do not lose in March.