Economics interview questions:
5 types, fully worked

What the Cambridge Tripos actually demands

The Cambridge Economics Tripos Part I has five compulsory components: British Economic History, Political and Social Aspects of Economics, Mathematics and Statistics, Microeconomics, and Macroeconomics. The breadth is deliberate. Cambridge is looking for an all-rounder: someone with strong quantitative ability who can also construct a clear argument in prose and engage seriously with economic history and institutional questions.

This breadth is precisely what the interview tests. An interviewer might start with a supply-and-demand diagram, move to a philosophical question about economic methodology, then hand you a probability puzzle. The question categories below correspond directly to what you will face.

The six criteria interviewers are assessing

Understanding the evaluation criteria shapes how you should prepare. Cambridge economics interviewers are consistently looking for the following:

Criterion six is the most underrated. Cambridge uses a supervision system where one or two students work through problems with a supervisor each week. The interview is a preview of that relationship. They want to know: can I teach this person? Will they update their view when I show them something new?

This is why interviewers deliberately challenge your answers, sometimes even when they are correct. They are not correcting you; they are watching how you respond to intellectual pressure.

Type 1: Open-ended economic reasoning

The question

This is a deliberately broad question. There is no single right answer. What the interviewer is testing is whether you can give the question structure, acknowledge complexity, and reason through a position, not whether you can recall a fact.

What a weak answer looks like

A weak answer either dives immediately into one position without structure ("Yes, because economic models are supposed to tell us what will happen") or hedges on everything without taking a view ("It depends on the model and the situation"). Both responses show the same underlying problem: the candidate has not yet learned to decompose a broad question.

Worked solution: how to structure your answer

Step 1: Define terms. What do we mean by "good"? What counts as "prediction"? A model is a tool; whether it is "good" depends on the purpose it serves. Be explicit about this.

Step 2: Give structure. There are cases where predictive ability is the core virtue of a model, and cases where it is not. Organising your answer around this distinction is immediately more impressive than an unstructured response.

Case for prediction as a requirement: Policy analysis is the clearest case. If a central bank wants to know whether raising interest rates by 50 basis points will reduce inflation by 2 percentage points, the model's value lies entirely in its forecast. The same is true of fiscal multiplier estimates used by the Treasury. In these settings, a model that cannot generate predictions, even probabilistic ones, is not doing its job.

Case against: Consider stock prices. The Efficient Market Hypothesis implies that if a model reliably predicted future stock returns, arbitrageurs would trade against those predictions until they disappeared. Systematic predictability cannot persist in a competitive market. So a model of stock price dynamics that generates no predictive power is not a failure. It may be telling us something important about the market structure.

More broadly, consider the Lucas critique: when a government uses a model to set policy, rational agents change their behaviour in response to the policy. This can invalidate the model's predictions. The act of predicting can change the outcome being predicted. This is not a defect of the model. It reflects something real about how people respond to expectations.

There is also Goodhart's Law, best known in the anthropologist Marilyn Strathern's phrasing: "When a measure becomes a target, it ceases to be a good measure." Applied to modelling: using a model's prediction as a policy guide can break the relationship the model was built on.

The simplification argument: Milton Friedman argued in his 1953 essay that the realism of a model's assumptions is irrelevant, only its predictive accuracy matters. On this view, a model is "good" if and only if it predicts well. But there are competing views: some models are valued for their insight, clarity, or the questions they allow us to ask, even if their predictions are imprecise.

Conclusion you can defend: A good economic model should be able to generate testable predictions where the purpose of the model requires it. Policy models, demand forecasts, and macroeconomic projections should all be judged partly on predictive accuracy. But models used for conceptual clarity, teaching basic incentive structures, or identifying welfare properties need not generate tight empirical predictions to be valuable.

Type 2: Technical economics knowledge

The question

Worked solution

First-degree (perfect) price discrimination occurs when a firm charges each consumer exactly their maximum willingness to pay, the reservation price. No consumer surplus survives: the firm extracts the entire surplus from every transaction.

The efficiency result is striking: perfect price discrimination produces the same output as perfect competition, because the firm is now willing to sell to every consumer whose willingness to pay exceeds marginal cost. There is no deadweight loss. All the surplus has simply been redistributed from consumers to the firm.

The practical problem is information: the firm must know every individual's reservation price. This is largely infeasible. In practice, first-degree approximations emerge through negotiation (car dealerships), personalised algorithmic pricing (some airline and hotel platforms), or when consumer history provides rich data about individual valuations.

Third-degree price discrimination involves charging different prices to different groups of consumers, segmented by an observable characteristic: student status, age, geography, time of purchase.

The firm splits its customers into groups it can tell apart, treats each group as its own small market, and sets a higher price for the group that is less sensitive to price and a lower price for the group that is more sensitive. The intuition is simple: if a group keeps buying even when the price rises, you can charge them more; if a group walks away quickly, you keep their price low so you don't lose them.

Same firm, two groups: the one less willing to walk away (P₁) is charged more than the price-sensitive one (P₂).

What "price elasticity" means here, in plain terms: elasticity is just how strongly demand reacts when the price changes. Demand is inelastic when people barely change what they buy as the price rises (a commuter who has to travel at peak time), and elastic when demand drops sharply (a leisure traveller who can change plans, or simply not go). The firm leans on this difference, charging the inelastic group more, and it only works if it can keep the groups apart and stop the cheaper group reselling to the dearer one.

Examples are everywhere: student discounts on software and streaming services (students have lower incomes and more price-sensitive demand); peak vs off-peak rail fares (business travellers have inelastic demand, leisure travellers have elastic demand); pharmaceutical companies charging different prices in different countries.

Key differences

Type 3: Applied economics and game theory

The question

You are shown a payoff matrix known as the Battle of the Sexes:

Worked solution: finding the Nash equilibria

A Nash equilibrium is just an outcome where neither player wishes they had chosen differently, given what the other did. Nobody can do better by changing their own mind alone, so the outcome is stable. The quick test: look at each cell and ask, for each player in turn, "given the other's choice, would I want to switch?" If the answer is no for both, that cell is a Nash equilibrium.

1. (O, O): A gets 2; switching to F would give 0, so A stays. B gets 1; switching to F would give 0, so B stays. Neither wants to move, (O, O) is a Nash equilibrium.
2. (O, F): A gets 0 but could get 1 by switching to F, so A moves. Not an equilibrium.
3. (F, O): B gets 0 but could get 2 by switching to F, so B moves. Not an equilibrium.
4. (F, F): A gets 1; switching to O would give 0, so A stays. B gets 2; switching to O would give 1, so B stays. (F, F) is a Nash equilibrium.

So there are two Nash equilibria: both go to the opera, or both go to the football. In each one the players have successfully coordinated, they end up together, but they disagree about which outcome they would prefer: A would rather it be (O, O); B would rather it be (F, F).

What it illustrates

This game (the "Battle of the Sexes") is the classic coordination problem. Both players would rather be together than apart, so both equilibria beat the mismatched outcomes, but they pull in different directions about which one to land on. Equilibrium analysis tells you which outcomes are stable; it does not tell you which one will actually happen. That gap is the interesting bit, and it is where strong candidates add insight: real coordination usually relies on something outside the payoff matrix, a convention, prior communication, a habit, or who commits first.

Real-world analogies an interviewer might welcome: competing technology standards (VHS vs Betamax), countries trying to agree a shared treaty or standard, or buyers and sellers converging on the same marketplace.

Follow-up questions you might get

Type 4: Mathematics and probability

The question

Why this question is asked

This is a Bayes' theorem question. It tests whether you can update beliefs correctly in the light of evidence, which is the foundation of rational inference and, in a broader sense, much of modern economics and statistics. Getting this right signals strong quantitative instincts.

The naive wrong answer is 1/2 (because there are two coins and you've "randomly" selected one). The error is ignoring the information the Heads outcome provides.

Worked solution

1. Define events. Let A = coin with P(H) = 1/4 is selected; let B = coin with P(H) = 3/4 is selected. Initially P(A) = P(B) = 1/2.
2. Find P(Heads on first throw) using the law of total probability.
   $$P(H_1) = P(H_1 \mid A)\,P(A) + P(H_1 \mid B)\,P(B) = \tfrac{1}{4}\cdot\tfrac{1}{2} + \tfrac{3}{4}\cdot\tfrac{1}{2} = \tfrac{1}{2}$$
3. Apply Bayes' theorem to update the probability of each coin.
   $$P(A \mid H_1) = \frac{P(H_1 \mid A)\,P(A)}{P(H_1)} = \frac{1/8}{1/2} = \frac{1}{4}, \qquad P(B \mid H_1) = \frac{3}{4}$$

   Observing Heads is more likely if we picked coin B (the biased coin), so after observing Heads we think it is more likely we have coin B. We have updated our prior (50-50) to a posterior (1/4 for coin A, 3/4 for coin B).

4. Find P(H₂ | H₁) using the updated probabilities.
   $$P(H_2 \mid H_1) = P(H\mid A)\cdot P(A\mid H_1) + P(H\mid B)\cdot P(B\mid H_1)$$$$= \frac{1}{4}\cdot\frac{1}{4} + \frac{3}{4}\cdot\frac{3}{4} = \frac{1}{16} + \frac{9}{16} = \frac{10}{16} = \boldsymbol{\dfrac{5}{8}}$$

The answer is 5/8. This is notably higher than 1/2 because observing Heads makes it more likely we are holding the coin that shows Heads more often, so the next throw is also more likely to be Heads.

Type 5: Logic and problem solving

The question

Worked solution

The key first step is to understand the state space carefully. The investor bets exactly ¥10,000 each round, regardless of how much they hold. If they ever reach ¥0 they cannot continue.

Map all possible paths through three rounds:

Collecting all possible outcomes with their probabilities:

The probability of the investor ending up with at least their starting capital after three rounds is 3/8.

Economic interpretation

This is a stylised version of the gambler's ruin problem. Notice something striking: the expected value of playing is zero in each round (50% chance of winning ¥10,000, 50% chance of losing ¥10,000). Yet the investor has only a 3/8 chance of being made whole. The expected outcome is not a useful guide to the probability of any particular outcome.

The asymmetry comes from the absorbing barrier at ¥0: once bankrupt, the game ends. This is why the distribution of outcomes is skewed even when the single-round gamble is fair. The practical implication, relevant to financial economics, is that investors facing ruin risk should not evaluate strategies on expected value alone. Volatility and the possibility of early exit matter independently of the mean.

Preparing effectively across all five types

Given the breadth of question types, generic "interview practice" is not enough. The most effective preparation is targeted by type:

For open-ended questions: practice structuring broad questions before answering them. Identify two or three dimensions of the question, state your structure out loud before developing any one of them, and make sure you state a conclusion you can defend, not just a list of considerations.

For technical economics: know your A-Level microeconomics thoroughly. Price discrimination, game theory, externalities, information asymmetry, and market structures are all in scope. For each topic, be able to define, give an example, compare and contrast, and discuss welfare implications.

For game theory questions: practise the Nash Equilibrium concept until finding pure NE is automatic. Know how to set up the mixed strategy NE calculation and what it means. Be ready to discuss coordination problems, prisoner's dilemmas, and repeated game logic.

For probability: Bayes' theorem is the most tested topic, so know it well and practise problems involving conditional probability, total probability, and Bayesian updating. Talk through your reasoning, interviewers care more about whether you understand what you are doing than whether you arrive at the answer without hesitation.

For problem solving: draw diagrams, decision trees, and tables. Work through cases systematically. Check your answer makes sense, in the investment example, verifying that probabilities sum to 1 is a simple sanity check that shows methodical thinking.

The most important preparation habit

Think out loud. This feels deeply unnatural at first. Most people are trained to work through a problem privately and present only the answer. In a Cambridge interview, the process is exactly the opposite. They want to hear you think. A candidate who says "let me organise this. I think there are two separate cases to consider..." and then reasons through them is far more impressive than one who produces a correct answer silently and then states it.

The supervisor system means interviewers spend every week listening to students reason through unfamiliar problems. They have a very well-calibrated sense of what genuine understanding sounds like versus a memorised answer. Thinking aloud is the most reliable signal of the former.