Welcome to the Course

The Experimental Method

Experimental method is a technique for collecting data in which a researcher deliberately manipulates one aspect of a situation — one variable at a time — and observes the effect on an outcome of interest. It is used across psychology, economics, sociology, and political science, and is the core tool of behavioural economics.

Types of Experiments

Why Run an Experiment? Two Advantages

1. It measures things unavailable in observational data. How much a person privately values a coffee mug, for instance, is not sitting in any existing dataset — it has to be elicited.
2. It cleanly establishes causality. Because only one variable (the treatment) is changed at a time, any difference in the outcome between the treatment and control groups can be attributed to that one change rather than to a confound.

The canonical illustration is the mug experiment: the treatment is "mug given," the control is "no mug given," and the outcome of interest is the participant's valuation of the mug.

(The full results behind this experiment appear later, under the endowment-effect study.)

Worked Example: A/B Testing

A/B testing applies the same logic in industry: A is the incumbent or "champion" system, B is the "challenger." Any measurable outcome — user satisfaction, ad recall, revenue — can be compared across the two.

MSN Hotmail Experiment

An MSN employee tested whether users would stay longer on the MSN homepage if the Hotmail link opened a new tab rather than the same window.

• Champion (A): Hotmail opens in the same window.
• Challenger (B): Hotmail opens in a new tab.
• Result (~900,000 users): the challenger produced an 8.9% increase in clicks on the MSN homepage.
• Scaled up to 2.7 million US users, the effect held.

This is why most websites now open outbound links in a new tab by default: keeping the original page open in the background raises the odds the user returns to it, increasing clicks (and potential ad revenue).

Moderna Vaccine RCT

A randomized controlled trial (RCT) is the same design applied to medicine. Moderna randomly assigned 15,000 participants to a placebo/control condition and 15,000 to a vaccine/treatment condition.

$\text{Vaccine efficacy} = 1 - \frac{\text{Cases}_{\text{treatment}}}{\text{Cases}_{\text{control}}} = 1 - \frac{5}{90} \approx 94.5\%$

Because random assignment makes the two groups identical on average except for the vaccine itself, the drop in infections is attributable to the vaccine.

Exam tip: the logic is identical across the mug study, A/B tests, and RCTs — one manipulated variable, a comparable control group, and a measured outcome. Naming the treatment, control, and outcome is the first step in analysing any experiment question.

Key takeaways

• Experimental method changes one variable at a time and measures its effect on an outcome, isolating causality.
• Three types: lab, artifactual field (lab-in-the-field), and field experiments — they differ by setting and by whether participants know they're being studied.
• Two advantages: measuring otherwise-unobservable quantities, and clean causal identification.
• A/B testing (business) and RCTs (medicine) are the same design: champion vs. challenger, control vs. treatment.
• The MSN and Moderna examples show the design scaling from an initial test to large-sample, real-world confirmation.

Origins of Behavioural Economics

Behavioural economics grew out of a collaboration between psychologists and economists who challenged the classical assumption that people always decide optimally.

Key Figures

Three Illustrations of Bounded Rationality

The Smoker's Puzzle

Consider a smoker whose utility from smoking is $k$ when they have both a cigarette and a lighter, and $0$ otherwise.

A rational "econ" should be equally indifferent between B, C, and D — all yield zero utility. Ask an actual smoker, though, and scenario C (holding a lighter with no cigarette to use it on) is reported as the most frustrating of the three, not equally indifferent. The explanation is developed in later modules — for now it stands as a case where the classical prediction visibly breaks down.

The Goldilocks Mug Choice (Dilip Soman)

From The Last Mile by Dilip Soman: participants were offered three sizes of coffee mug and asked to choose one.

The size people called "about right" changed completely between rounds, yet the justification was identical. People believe they are optimizing like econs, but the "right" choice tracks the position within whatever set is offered, not a fixed, independent preference.

The Endowment Effect (Kahneman, Knetsch & Thaler)

A Cornell University mug was distributed to roughly half the seats in a classroom, at random.

Because the mug was randomly assigned, classical price theory predicts WTA ≈ WTP. The gap observed here is the founding demonstration of the endowment effect: merely owning an item raises the value a person places on it. This is the same treatment/control/outcome design introduced earlier as the canonical "mug experiment."

Key takeaways

• Kahneman and Tversky founded behavioural economics; Kahneman won the Nobel Prize in Economics in 2001 (Tversky had already passed away).
• Thaler (Nobel 2017) popularized the "econs vs. humans" framing (Nudge, Misbehaving); Shiller contributed behavioural finance; Fehr contributed social-preference theory.
• The smoker's puzzle shows "rational" indifference (B = C = D = 0 utility) breaking down behaviourally.
• Soman's mug study shows people rationalize a "just right" choice that is really just the middle option of whatever set they're shown.
• The Kahneman–Knetsch–Thaler mug study is the founding demonstration of the endowment effect: WTA (~$7) far exceeded WTP (~$3) for an identical, randomly-assigned good.

Rationality: Econs vs. Humans

Rationality, in the economic sense, is the classical assumption that a decision-maker maximizes a narrowly-defined material payoff, is wilful, has perfect foresight, and can flawlessly evaluate complex alternatives. Richard Thaler calls this idealized agent an econ (homo economicus).

The Four Assumptions Behind "Econ"

Rationality in everyday English (per the Oxford English Dictionary) means "the ability to think sensibly or logically." In economics it is a technical term: an agent is rational only if it satisfies the four assumptions above — not merely if it is reasonable.

Why Assume This?

These are simplifying assumptions adopted to make formal analysis tractable, not empirical claims about real behaviour. The classical savings decision, for example, models an econ as solving a single optimization problem that packs in all four assumptions at once:

$\max \; \text{Utility} \quad \text{subject to a budget constraint}$

From Econ to Human

Few real people satisfy these assumptions: few are perfectly forward-looking, few are purely selfish (people care about friends and family), and few are as wilful as assumed — most people can set a 5 a.m. gym alarm and still fail to get up. This gap gives rise to what Thaler calls humans: economic agents with empirically-grounded, psychologically realistic assumptions.

Key takeaways

• Classical "rationality" = self-interested + wilful + perfect foresight + unlimited computational ability = an "econ."
• These are simplifying assumptions adopted for analytical tractability, not empirical claims.
• The economic definition of "rational" is technical and narrower than the everyday sense of "sensible."
• Real people ("humans") systematically depart from all four assumptions — they care about others, fail to follow through on plans, and have bounded foresight and cognition.
• Behavioural economics studies "humans" with empirically-founded assumptions, trading analytical simplicity for descriptive accuracy.