Transcript
Economics in 30 Minutes, Part 4: Thinking Tools and Evidence
The narration, word for word, as heard in the video. Times are approximate. Plain-text version for pasting into an AI tutor · Back to the companion page
Introduction (0:00)
This is the last part, and it's a toolbox. Four tools. Opportunity cost, which you met in Part 1. Comparative advantage, one of the main reasons anyone trades with anyone. Sunk cost, which takes thirty seconds. And the hard one, how you tell whether one thing caused another, which is where economists spend most of their careers and where most of the arguments you hear on the news go wrong. Then we go back to the robots from Part 1 and see what you'd say differently now. Two pauses in this part, and a few shorter questions along the way.
Opportunity cost: what college costs (0:39)
Start with a question that sounds like it has an obvious answer. What does college cost? Tuition, fees, books, rent. Add it up and you have what an accountant would call the cost, and it's not wrong. It's just not the whole thing. The biggest item is missing, and it's missing because nobody ever sends you a bill for it.
The job you didn't take. Four years as a barista, or an electrician's apprentice, or whatever you'd actually have done. That paycheck is part of what college cost you, because you gave it up to go. Economists call the whole thing, the bills plus the best alternative you passed over, the economic cost. The most expensive thing about college may be the paycheck you didn't get.
Two rules for counting it. Only the best alternative counts, not all of them; you could only have taken one job. And expenses you'd have paid anyway don't count. You'd have eaten and slept somewhere whether or not you enrolled, so only the extra rent is a cost of college, not all of it. Here's a way to see that this is real and not an accountant's trick. When the job market weakens, the paycheck a student would give up gets smaller, so college gets cheaper in the way that matters, even if tuition hasn't moved an inch. Economists have looked for that pattern in enrollment during downturns, and the companion page has the evidence.
So, for you. What job would you have had if you hadn't gone to college, or if you're deciding now, what job do you have? Was it really available, or is it the job you'd like to have had? Whatever the honest answer is, that's the part of the cost that was yours and nobody else's. Take a second with it.
Comparative advantage: Maya and Leo (2:35)
Tool two, and I'll need two people. Maya and Leo live on an island, and each day each of them can fish or pick coconuts, one or the other. Maya catches four fish in a day, or picks two coconuts. Leo catches one fish, or picks one coconut. Maya is better at everything. Leo is worse at everything. Economics is about to save his career.
Here's the move. Stop asking who's better and ask what each of them gives up. A day Maya spends on coconuts gets her two coconuts and costs her four fish, so a coconut costs Maya two fish. A day Leo spends on coconuts gets him one coconut and costs him one fish. A coconut costs Maya two fish. It costs Leo one. Leo makes the cheaper coconut, and that's true even though he's the worse coconut-picker.
Give them four days. If they each work alone and split their time, two days fishing, two days picking, the island ends up with ten fish and six coconuts. Now let Leo do nothing but coconuts, and let Maya pick for one day and fish for three. Six coconuts, same as before. Twelve fish. Two more than before, and nobody worked a minute longer. The extra fish came from nowhere except putting each person where what they give up is smallest.
Leo would like those explained. He didn't catch them. He didn't get better at anything. But without him taking the coconuts, Maya couldn't have spent three days fishing, and the two extra fish wouldn't exist.
Now they trade. Any price between one fish and two fish per coconut works, because that's between what a coconut costs Leo and what it costs Maya; where in that range it lands decides who gets more of the gain. The screen shows one example, two coconuts for three fish, and at that particular price each of them ends up with the coconuts they had before and one extra fish. So they are both better off than before. This is the benefit of specializing not according to what they are best at but where their comparative advantage is. Leo sells the cheaper coconut, and Maya buys it, even though she could have picked it herself faster.
That's the whole idea, and it's a big reason countries trade, and firms specialize, and a dentist hires a plumber instead of fixing her own pipes. Comparative advantage isn't about who is better. It's about who gives up less.
One more, closer to home. The surgeon is the best lawn-mower on the street, and the joke on the sign says she should never mow it. The careful version: if that hour could otherwise be spent operating, an hour of hers costs a knee replacement and an hour of the neighbor's kid costs a video game. If the alternative is a Sunday she'd rather spend in the garden, mow away. Being better at a job doesn't make it your best use of time; what you'd otherwise do with the hour does.
Here's the question to sit with. Leo is worse at both jobs, and his coconut is still cheaper. If that feels wrong, say why, and then say what each of them gives up to pick one coconut. The feeling usually comes from counting how much each person can do. The answer comes from counting what each person gives up.
Sunk cost (6:23)
Tool three takes thirty seconds. The movie is bad. The ticket is gone. You may leave. The money was spent whether you stay or go, so it has no business in the decision; the only question left is what to do with the rest of the evening. Economists call money you can't get back a sunk cost, and the mistake of letting it steer you the sunk cost fallacy. Everyone makes it. Think of the last time you finished something because you'd already paid for it.
Did this cause that? (6:53)
Now the hard one. Back in Part 1, I drew a face and changed one thing at a time, and you could see what the mouth did. The world doesn't work like that. Several things change at once, and you have to work out which one did what. There are two ways to get it wrong. You can see a cause that isn't there, a false positive. Or you can refuse to see a cause that is there, a false negative. Statisticians call these Type I and Type II errors; I'll use the plain names.
Before either of those, there's an older mistake, which is having no comparison at all. A hundred percent of people who confuse correlation and causation eventually die. True. So does the control group. A number on its own tells you nothing until you ask, compared with what?
Suppose a report says most accidents happen within a few miles of home. Should you drive farther away to be safe? Think about where most of your driving happens. Most of your miles are near home, so you'd expect most of your crashes there too, whether or not those miles are any more dangerous. To compare the risks you need crashes per mile near home and crashes per mile farther away. The number of crashes alone doesn't tell you which is safer. A number without its denominator is a story, not a finding.
Now a false positive. Here's a real correlation. The biggest fires have the most fire trucks at them. The man on the lawn has a plan. Cut the trucks, get smaller fires. The correlation is real; the cause runs the other way. Big fires draw trucks. Whenever two things move together, ask two questions. What else moves both of them? And which way does it run?
Your turn. Ice cream sales and shark attacks rise and fall together, month by month. Should we ban ice cream? If not, what could make both of them go up at once? And then name another pair of things that move together without one causing the other. Pause here.
Economics has another problem of its own: people respond when the policy changes. The famous case has a name. In the nineteen-sixties, inflation and unemployment moved against each other, year after year, and the pattern was real. Governments decided to steer by it: accept a little more inflation, buy a little less unemployment. It held, until they leaned on it. People saw the policy coming, expected the inflation, and raised their prices and wages ahead of it. By the nineteen-seventies the country had high inflation and high unemployment at the same time; the oil shocks pushed prices up and output down as well, so it wasn't a clean experiment, but the pattern policy had leaned on was gone. The people in the model can see you pulling the lever. A pattern found under one set of rules may not survive a change in the rules.
The other way to be wrong. Everybody dies, so smoking can't hurt me. That's a false negative, refusing a cause because the effect isn't total. Nobody needs cigarettes to die. What smoking does is move the appointment up, by about ten years on average in the long study of British doctors. A cause doesn't have to work every time to be a cause. It only has to change the odds, or the timing.
So how do you settle it, when you can? You go back to the happy face. Change one thing, and count. Same bar, same beer, many comparable nights, and each night a coin decides whether the price is five dollars or seven. Then count the pints. Because the coin chose, the cheap nights and the dear nights are alike in everything else, on average, and the difference you count estimates what the price did. Enough nights, and you can tell that effect apart from luck. That's a randomized experiment. Medicine does it with pills. Economists sometimes get to do it too.
Three real ones. In the nineteen-seventies, a program called Supported Work asked whether a year in a supervised, paid job would help people earn more afterward. Eligible applicants were assigned by chance to a job offer or to a comparison group. Earnings rose for women who had been on welfare a long time. For the young people in the program, nothing lasting.
The RAND health insurance experiment assigned whole families to different insurance plans, some paying much more out of pocket than others, and followed them for years. People who paid more used less care. Including care that works. Paying more didn't make people good at telling the two apart.
And in India, researchers tested a tutoring program for children falling behind by assigning, school by school, which grade got the tutor. Scores went up, most for the pupils who were furthest behind, and some of the gain faded in later years. None of these answers came from an argument. They came from letting chance pick the groups and then counting.
Did that prove causation? It's a powerful way to get at causation, when the experiment is well designed and carried out, because chance did the assigning, so the people who got the program and the people who didn't were alike in everything else, on average, including the things nobody measured. There's still work to do. Some people refuse the offer, some drop out, some affect the people in the other group, and the researchers have to check all of that. The first rule is to compare everyone who was offered the program with everyone who wasn't, rather than only the people who took part, because the moment you compare volunteers with the rest, choice is back in the groups. And a result for these people in this place may not carry over to other people, or to a program ten times the size.
There's also a human problem. Roll again, I wanted him in the other group. The dice only help if you accept the roll. If you keep rolling until you get the person you want, you're choosing the groups again, and the whole thing is gone. Sometimes the hard part isn't the statistics. It's getting people to agree to the comparison.
And when you can't run the experiment? London, eighteen fifty-four. Cholera. Nobody knew what caused it; the respectable theory was bad air. A doctor named John Snow had no germ theory and no trial. What he had was a pile of clues that one explanation fit and the other didn't. The deaths clustered around one water pump. The brewery workers next to it mostly escaped, and they drank beer, not pump water. A widow four miles away died, and she'd had water from that pump carried to her because she liked the taste. A correlation plus a mechanism plus whatever natural experiments turn up. It falls short of proof, and Snow knew it. What it gives you is evidence, and the question to ask of each clue is what competing explanation it rules out. Real data is often messy. That doesn't mean we can't learn anything, just that we have to be careful about interpretation. We don't want to ignore the danger of smoking or oversell the benefits of inflation.
Every method in this section, the coin, the dice, the pump, is trying to answer one question. What would happen to similar people, or similar markets, if this one thing were different? When you hear a claim that one thing caused another, that's the question to ask about it.
Back to the robots (15:33)
Back to Part 1 and the robots now that we have a few more tools. Will scarcity survive robots? If AI and robots make most things abundant, you now know what that is on the graph: supply moving a very long way to the right, in market after market, and prices falling, maybe collapsing. That part of the forecast might be right; the technology is certainly changing fast. But wherever scarcity remains, and we listed what remains, land in a place, a time, your attention, something still has to decide who gets what. Here the dinner is free. The host wants your Saturday.
Here's what I'd want you to take from four parts. Economics doesn't necessarily tell you something you never knew. You already knew that scarce things cost more, that people buy less when prices rise, that a bad harvest makes bread dear. What it does is make you say exactly what you mean. Which line moved. Compared with what. What you gave up. Who gives up less.
Your turn (16:43)
Your turn, last time. Three questions, and the first one sends you back to Part 1. If you wrote your answers down then, get them out.
One. The ninety-percent-cheaper world from Part 1. Housing, food and ordinary goods at a tenth of the price. Which of your predictions would you change now that you've seen the graph move? Say which line moved, and in which markets.
Two. What is still scarce in that world? Name something specific. How would it be allocated, and how should it be? Those are two different questions. Economists call the first one positive, a prediction, and the second one normative, a judgment, and the second needs you to say what you mean by should.
Three. Your own biggest opportunity cost this year. Count the best alternative you gave up, rather than what you spent. Was it worth it? Nobody else can answer that one.
Next (17:50)
That's the overview. If you want to see the tools used on one real decision, the next video is about a job I took. I was paid five thousand dollars to teach a college class. Who got a good deal? The link is in the description, along with the companion pages and the worksheets for all four parts.