Economics in 30 Minutes — Part 4 Companion Page: Thinking Tools and Evidence

Companion to "Economics in 30 Minutes," Part 4: Thinking Tools and Evidence.

This page has the definitions, the arithmetic the video only summarizes, the sources behind what it says, the questions from the video with room to go further, and prompts you can give an AI tutor. Work through the questions before you read the hints at the bottom.

1. Key terms

Opportunity cost. The value of the best alternative you give up when you choose something. Only the best alternative counts, because you could only have done one of them. For college, the biggest item is usually the job you'd otherwise have held, not the tuition. And the job is worth more than its paycheck: four years of work is also four years of experience, the skills, contacts and track record that make you more valuable later. The paycheck is the part of the forgone job that is easy to count; the experience is the part that is easy to forget, and for some trades it is the larger part. (Set against it: college builds skills too, and for some careers it is the only route to them. The comparison is between two ways of becoming more valuable, not between college and nothing.)

Accounting cost versus economic cost. Accounting cost is what shows up on bills: tuition, fees, books. Economic cost adds the opportunity cost. Expenses you'd have paid anyway (ordinary rent and food) belong to neither; only the extra expense caused by the choice counts.

Absolute advantage. Being able to produce more of something in a day than someone else. Maya has the absolute advantage in both fish and coconuts.

Comparative advantage. Having the lower opportunity cost of producing something. A coconut costs Maya two fish and costs Leo one, so Leo has the comparative advantage in coconuts; a fish costs Maya half a coconut and costs Leo a whole one, so Maya has it in fish. Everyone has a comparative advantage in something, including the person who is worse at everything, because what matters is the ratio, not the level. This is why trade can benefit both sides and why specialization raises total output without anyone working longer.

Why people trade. Comparative advantage is the supply-side reason: two people with different opportunity costs can both gain by specializing. It is not the only reason. People with identical abilities still trade when they want different things: two neighbors with the same garden swap tomatoes for cucumbers because one likes tomatoes less. Differences in tastes, in what people already have, and in when they want it all create trades; the island story holds tastes fixed on purpose so the production side can be seen on its own. The video says comparative advantage is "one of the main reasons" anyone trades, and this is the other main one.

"Better" on the island. When the video says specializing leaves the island better off, it means production: the same number of coconuts and more fish. What Maya and Leo want or need is a separate question, taken up when they trade and in the entry above on tastes. The reason production alone settles it here is that the island ends up with no less of anything and more of something, so nobody has to be made worse off for someone to gain; economists call a change like that a Pareto improvement, and it is the only sense in which "more" is unambiguously better. If specializing had meant more fish and fewer coconuts, you would need to know how much each of them values fish against coconuts before you could call it an improvement.

Terms of trade. The price at which the two sides exchange. Both gain only if it lies strictly between their two opportunity costs (between one and two fish per coconut on the island). Where it lands in that range decides how the gain is split; the video's two-coconuts-for-three-fish is one example in which the gain happens to be equal.

Positive and normative. A positive statement is about what is or what would happen: "a tenth of the price would leave land scarce and allocate it by queue or connection." A normative statement is about what should happen: "it should go to whoever needs it most." The first can be wrong and checked against evidence; the second depends on a value you have to state. Most arguments about economics mix the two, and most of the heat comes from the mixing. Question 2 of "Your turn" asks for one of each and asks you to keep them apart; the longer course returns to the distinction every time a market outcome is compared with another.

Sunk cost. Money or time already spent that can't be recovered. It is the same whether you continue or stop, so it has no place in the decision; only future costs and benefits do. The sunk cost fallacy is letting it steer you anyway ("we paid, so we're staying").

Correlation. Two things moving together: when one is high, the other tends to be high (or low). A correlation is a fact about the data. It says nothing by itself about which causes which, or whether a third thing drives both.

Causation. One thing changing because another changed. The question every method in the video tries to answer: what would happen to similar people, or similar markets, if this one thing were different?

Confounder (common cause). A third thing that moves both of the things you're looking at. Summer heat raises ice cream sales and puts more people in the water.

Reverse causation. The effect runs the other way from the story. Big fires draw more fire trucks; the trucks don't make the fires bigger.

Missing denominator. A count presented without the exposure it should be divided by. "Most crashes happen near home" is a count; crashes per mile near and far is the comparison.

False positive and false negative. In the video's plain words: seeing a cause that isn't there, and refusing a cause that is. Statisticians call these Type I and Type II errors. Formally, a Type I error is rejecting a true null hypothesis (concluding there is an effect when there is none), and a Type II error is failing to reject a false one: failing to detect an effect that exists. Failing to reject a hypothesis is not the same as concluding it is true. The cartoons in the video are analogies for these mistakes, not tests of a stated null hypothesis.

Randomized experiment. Assigning the treatment by chance (a coin, dice, a random-number table) and comparing the groups. Because chance did the assigning, the groups are alike in everything else on average, including things nobody measured. Not perfectly alike in every sample, which is why the size of the experiment matters: enough cases let you tell the effect apart from luck.

Selection bias, and Heckman's correction. The reason the "compare everyone offered with everyone not offered" rule matters. People who choose to take up a program differ from those who don't, in motivation, need, or options, and those differences affect the outcome on their own. Compare participants with non-participants and you measure the program plus the difference in the people. James Heckman's 1979 paper showed how to correct for this when you can model who selects in (the "Heckman correction" or two-step estimator, for which he shared the 2000 Nobel), and more importantly showed how large the bias can be when you don't. Random assignment is the cleaner answer, because it removes the choice; when there is no random assignment, selection is the first thing a careful reader looks for.

Intention to treat. The cleanest comparison is by assignment: everyone offered the program versus everyone not offered it, whether or not they took it up. Comparing only the people who completed the program with the rest quietly puts choice back into the groups; the video's rule is to start by comparing everyone offered with everyone not offered.

Natural experiment. A situation where something outside anyone's control, or a rule that applied to some people and not others, created a comparison nobody designed. Snow's brewery, with its own well, is one. Economists look for these constantly, because occasionally a policy is assigned by chance or arrives unexpectedly. The best-known: the Vietnam draft lottery, which assigned draft eligibility by birth date, so men with low and high lottery numbers were alike in everything but their chance of serving; Joshua Angrist used it to estimate what military service did to later earnings (about a 15 percent earnings penalty for white veterans in the early 1980s). The Mariel boatlift of 1980 dropped 125,000 Cubans into Miami's labor market in a few months, and David Card compared Miami's wages and unemployment with similar cities. The 1992 New Jersey minimum-wage increase, with Pennsylvania next door unchanged, let Card and Krueger compare fast-food employment on the two sides of the line. In each case the question is the same as in Snow's: what competing explanation does the comparison rule out, and what does it leave standing?

The Lucas critique (plain version). A relationship estimated under one set of rules may not survive a change in the rules, because the people in the data respond to the change. The Phillips curve is the standard example.

2. The island, in full

Daily rates (the video's numbers, illustrative):

Alone, four days, two days of each. Maya: 8 fish and 4 coconuts. Leo: 2 fish and 2 coconuts. Island total: 10 fish and 6 coconuts.

Specialized. Leo picks coconuts all four days: 4 coconuts. Maya picks for one day and fishes for three: 2 coconuts and 12 fish. Island total: 12 fish and 6 coconuts. Two more fish, same coconuts, same hours.

Trade. Any price strictly between 1 and 2 fish per coconut leaves both better off. At the video's example price of 1½ fish per coconut, Leo sells 2 coconuts for 3 fish: Leo ends with 3 fish and 2 coconuts (one fish better than alone); Maya ends with 9 fish and 4 coconuts (one fish better than alone). At exactly 1 or exactly 2 fish per coconut, one of them gains nothing. Outside the range, one of them refuses.

Try it: at 1¼ fish per coconut, who gets more of the gain? At 1¾?

3. The sources behind the video

College enrollment in downturns. Lisa Barrow and Jonathan Davis, "The upside of down: Postsecondary enrollment in the Great Recession," Federal Reserve Bank of Chicago, Economic Perspectives 36 (4), 2012. Enrollment rose during the 2007–09 recession, consistent with a lower opportunity cost of attending when jobs are scarce; the authors also note other factors (financial aid, demographics). The video uses this as the mechanism, not as an automatic law: a weaker job market can lower the earnings students give up, making college more attractive even if tuition stays the same.

Comparative advantage. David Ricardo, On the Principles of Political Economy and Taxation (1817), chapter 7, the England-and-Portugal example. The island version with two people is the standard classroom form.

The Phillips curve. A. W. Phillips, "The relation between unemployment and the rate of change of money wage rates in the United Kingdom, 1861–1957," Economica 1958. The warning that the relation would shift once people expected the inflation: Milton Friedman, "The role of monetary policy," American Economic Review 1968 (his 1967 presidential address), and Edmund Phelps, Economica 1967. The general point that estimated relationships can change when policy changes: Robert Lucas, "Econometric policy evaluation: a critique," 1976. U.S. inflation and unemployment both rose in the 1970s; oil shocks (1973, 1979) made it worse. The video's one-line summary ("it held, until they leaned on it") compresses a decade of argument, and it says in passing that the oil shocks pushed prices up and output down as well, so the 1970s were not a clean test of the expectations story (see, for example, Donald Kohn's 2008 Federal Reserve speech on the lessons of the 1970s); the handout caption "the people in the model can see you pulling the lever" is the idea.

Smoking. Richard Doll, Richard Peto and others, "Mortality in relation to smoking: 50 years' observations on male British doctors," BMJ 2004; 328: 1519. Lifelong smokers died about 10 years earlier on average than non-smokers. The cartoon's "10 yrs" calendar is that average, not a personal prediction.

John Snow and the Broad Street pump. John Snow, On the Mode of Communication of Cholera, 2nd edition, 1855. The pump handle was removed on September 8, 1854, when cases were already falling. The brewery on Broad Street had its own deep well and the workers drank beer; of more than seventy workers, none was severely ill. The widow in Hampstead, about four miles away, had water from the Broad Street pump brought to her because she preferred its taste; she and a niece who visited died. The dominant theory at the time was miasma, bad air. The map on the slide is schematic.

Supported Work. The National Supported Work Demonstration (1975–79), run by MDRC; eligible applicants at ten sites were assigned at random to a supported-job offer or a control group. Summary and Findings of the National Supported Work Demonstration (MDRC, 1980). Earnings gains for women who had been on welfare long-term (AFDC); no lasting gains for the youth group; mixed for the other groups. Public-use data: ICPSR study 7865.

RAND Health Insurance Experiment (1971–82). Families were assigned to insurance plans with different cost-sharing. People who paid more used less care, including care judged effective, with little measured effect on average health for most people. Joseph Newhouse and the Insurance Experiment Group, Free for All? (Harvard, 1993); RAND research brief RB-9174. A later reanalysis found unequal refusal and dropout across plans and reporting differences, and still concluded that use responds to price while being less sure of the exact size: Aviva Aron-Dine, Liran Einav and Amy Finkelstein, "The RAND Health Insurance Experiment, three decades later," Journal of Economic Perspectives 27 (1), 2013.

Balsakhi tutoring (India). Abhijit Banerjee, Shawn Cole, Esther Duflo and Leigh Linden, "Remedying education: evidence from two randomized experiments in India," Quarterly Journal of Economics 122 (3), 2007. In Vadodara and Mumbai, schools were assigned which grade received a remedial tutor (a balsakhi); the comparison is the same grade across schools. Test scores rose, most for the weakest pupils; the gains faded in the following years. J-PAL evaluation summary.

The free dinner (cartoon U) and the robots. The callback is to Part 1 (Keynes 1930; Musk 2024–25); sources there. A word on the premise, since Part 4 now has the tools to judge it: a ninety-percent fall in the price of ordinary goods is at least imaginable, because the robots can make more of them; a ninety-percent fall in the price of housing is the part that seems unlikely, because most of the price of a house in the places people want to live is the land under it, and the robots can't make more of that. That is the video's own point about what stays scarce, applied to the exercise's premise. Treat the ninety-percent world as a thought experiment for sorting what is scarce from what is merely expensive, not as a forecast.

The bar, the fire trucks, the cinema, the bakery. Illustrations, not reports.

4. Questions from the video, with room to go further

4.1 Opportunity cost (the first short stop)

1. What job would you have had if you hadn't gone to college, or what job do you have now? Was that alternative really available to you?

2. How does that alternative change the cost of college for you personally? Which earnings would you give up, what work could you keep doing while enrolled, and what extra expenses would college cause? And beyond the paycheck: what would four years in that job have taught you, and what would it have been worth later?

3. Which living expenses would you pay either way? Why shouldn't they all count as a cost of college?

4. In a recession, what happens to the opportunity cost of college? What would you expect enrollment to do, and what else might be going on?

4.2 Comparative advantage (the second short stop)

5. Is it counterintuitive that Leo's coconut has the lower opportunity cost even though Maya is better at both jobs? Why?

6. What does each person give up to produce one coconut? Why is that a different question from who picks more coconuts in a day?

7. Specializing produced two extra fish with no extra hours. Where did they come from?

8. If two coconuts trade for three fish, why can both gain? What happens at a price of one fish per coconut? Two fish? Two and a half?

9. The surgeon's lawn: when should she mow it herself?

4.3 Sunk cost

10. Think of something you continued because you'd already spent money or time on it. What part was unrecoverable? At that moment, what were your remaining choices, and was continuing a mistake, or did the future benefit still justify it?

4.4 Correlation and causation

11. Suppose a report says most driving crashes happen near people's homes. Does that show that driving near home is more dangerous per mile? What else would you need to know? Would driving farther from home automatically make you safer?

12. Ice cream sales and shark attacks rise and fall together (the pause). Should we ban ice cream? What causal claim would that policy need? What could make both rise? What information would separate your explanation from an effect of ice cream itself?

13. Give another pair of things that are correlated without one causing the other. Is your explanation a common cause, coincidence, or something else?

14. Larger fires attract more fire trucks. What is wrong with concluding that fewer trucks would mean less damage? Could the trucks also have a real effect in the other direction?

15. A shop uses more staff on its busiest days. Does that show that adding staff creates customers? What comparison would help find out?

16. "Everybody dies, so smoking can't hurt me." What kind of mistake is that? What does it mean for something to be a cause if it doesn't work every time?

17. In the Phillips-curve story, why might people behave differently once a policy becomes predictable? What would a model miss if it kept their behavior fixed?

18. Pick one of the three experiments. What was the question, what was assigned at random, and what was measured?

19. What could go wrong after random assignment? How might refusing the treatment, dropping out, or affecting the other group change the comparison?

20. Snow had no experiment. List his clues and say, for each one, which explanation (pump water or bad air) it fits better.

4.5 Your turn (the closing pause)

21. Revisit your 90-percent-cheaper-world answers from Part 1. Which prediction would you change after Parts 2 and 3? Which line moved, in which markets?

22. What would stay scarce? How would it be allocated, and how should it be? Keep the prediction and the recommendation apart.

23. Your own biggest opportunity cost this year: the best alternative you gave up, rather than what you spent. Was it worth it?

24. (From Part 3, for practice.) Draw the beer market after a heat wave, then after a hop fire. Which line moves in each case?

5. Using an AI tutor

These prompts ask the AI to guide your thinking, not give answers. Paste the set-up prompt first, then the prompt for the question you are working on.

Set-up prompt (paste first)

> I am a student working on exercises about opportunity cost, comparative advantage, sunk cost, and the difference between correlation and causation. Act as a tutor, not an answer key. Do not give me final answers. Ask me one question at a time. Make me state my assumptions and show my arithmetic before you comment. If I ask for the answer, remind me to try first and give me a hint instead. At the end, ask me to summarize what I concluded and what I am still unsure about.

Opportunity cost (questions 1–4)

> Here is the job I think I would have had instead of college, and what I think college cost me in total: [paste]. Ask me whether that job was really available to me, which of my expenses I'd have paid anyway, and whether I've counted the same months twice (as forgone earnings and as something else). Then ask me what happens to that cost in a recession.

Comparative advantage (questions 5–9)

> Here are my numbers for Maya and Leo and my answer to who should pick coconuts: [paste]. Ask me what each person gives up to pick one coconut, and don't accept "Maya is better" as an answer to that question. Then give me a price of fish per coconut and ask me whether both would agree to trade at it, and who gains more. Finally, ask me for a comparative-advantage example from my own life where the "better" person shouldn't do the job.

Sunk cost (question 10)

> Here is a time I kept going because I'd already paid: [paste]. Ask me what was unrecoverable at the moment I decided, what my remaining options were, and whether I'm judging the decision by what I knew then or by how it turned out.

Correlation and causation (questions 11–20)

> I'm working on a claim that one thing causes another. Here is the claim and my reasoning: [paste]. Ask me, one at a time: compared with what? What else could move both? Could the cause run the other way? What is the denominator? Then ask me what experiment would settle it if I could run one, what would be assigned at random, and what could still go wrong after the assignment.

The experiments (questions 18–19)

> Here is my summary of one of the three experiments, its question, what was randomized, and what was found: [paste]. Ask me who was compared with whom, by assignment or by take-up, and why that matters. Then ask me whether the result would carry over to different people or to a program ten times the size, and what would make me doubt it.

6. Printable versions

The worksheet, the slide handout as shown in the video, and a white-background version for printing are linked at the top of this page.

7. Hints (try the questions first)

Questions 1–4. Compare feasible alternatives over the same months. Count only earnings actually forgone (if you could work part-time while enrolled, only the difference counts), and don't count the same time twice. Rent and food you'd pay either way are not a cost of college; a more expensive campus apartment is, by the difference. Don't stop at the paycheck: an apprentice electrician who skips college gives up a degree, but the one who goes to college gives up four years toward a license and a reputation. Both are investments in future earning power; count what each one builds, not only what each one pays this year. In a recession the forgone paycheck shrinks, so the economic cost of college falls even with tuition unchanged; enrollment tends to rise, though aid, demographics and the state of the colleges matter too.

Questions 5–9. The intuition comes from comparing how much each person can do; the answer comes from comparing what each gives up. A coconut costs Maya two fish because her day is worth four fish; it costs Leo one. The two extra fish come from moving Maya's time into fishing, where her advantage is largest, and Leo's into coconuts, where his disadvantage is smallest. Both gain at any price strictly between 1 and 2 fish per coconut: at 1, Leo gains nothing; at 2, Maya gains nothing; at 2½, Maya would rather pick her own. The surgeon should mow when the alternative use of that hour is worth less to her than the mowing (a Sunday she enjoys outdoors counts); the cartoon assumes the alternative is operating.

Question 10. The unrecoverable part is sunk whether you continue or stop. Judge the decision by the options and information at the time, not by how it turned out; a decision to continue can be right if the future benefit still exceeded the future cost.

Questions 11–15. Near home: you need crashes per mile near and far; most miles are near home. Ice cream and sharks: a common cause (summer: heat, holidays, more people in the water); banning ice cream would require ice cream itself to put people in the water. Separate the explanations by comparing hot days with and without ice cream, or beaches with and without a stand. Fire trucks: reverse causation plus a confounder (fire size); the trucks also reduce damage, which runs the other way again. Staffing: compare similar days with different staffing, or watch what happens when a store adds staff on an ordinary day.

Questions 16–17. A cause doesn't have to be sufficient or necessary; it has to change the odds or the timing. The Phillips story: once people expect the inflation, they build it into wages and prices, and the trade-off the policy relied on moves. A model that keeps behavior fixed misses that the people in it respond to the policy.

Questions 18–20. Compare by assignment (everyone offered versus everyone not offered), not by take-up. After assignment, watch refusals, dropouts, crossover and spillovers; with schools assigned as groups, the pupils in a school aren't independent cases. Snow's clues: deaths cluster around the pump (fits water, but also fits bad air near the pump); the brewery workers with their own well escaped (fits water, not air); the widow four miles away who drank pump water died (fits water, not air). Each clue rules something out; together they make one story much more likely without a single experiment.

Questions 21–23. Cheaper goods means supply moved far to the right in those markets; prices fall, quantities rise, spending can fall even as you buy more. What stays scarce is land in particular places, time slots, attention, originals, the inputs the robots run on; "would" is a prediction about queues, prices, lotteries and connections, "should" needs you to say what you mean by fair or efficient. Question 23 has no hint; it's yours.