Unlike and connected
Part 1 drew the economy one slice at a time, and the slices are right. This part asks why they are no longer enough, and why the textbook has to be rebuilt rather than extended. The classical figures are not wrong. The questions now put to them are about an economy of agents who are unlike one another and tied to particular others, and the figures answer those questions with exactly the parts they hold still.
The argument has nine steps. Section 1 shows how economic news is usually explained: one curve shifts. Section 2 shows what the textbook knew while it was being built, and set aside for its readers’ sake. Sections 3 and 4 put two of those things back, agents who are unlike one another and agents tied to particular others. These are the two assumptions Part 1 never relaxes head-on (Part 3, section 4). Section 5 explains why the mathematics grew harder once they were put back, and section 6 why the evidence changed. Section 7 rereads the old arguments about what economics is as arguments about what to hold still. Section 8 is about the students, teachers and economists who reached the same diagnosis before this site, and what they built. Section 9 lists the tasks that understanding an economy now takes, and Part 3 builds its archive on them.
Fold the page with the control in the corner. Each figure raises the dimension its flat drawing presses into the page: the price of oil under a line of inflation, the assumption under each date, time under a path of prices, each method under a stack. The page lies on the wheel Part 1 built, and the wheel changes as this part goes on. Its agents become unlike one another, then tied to particular others, and the methods at its hub light one by one. What changes stays changed, and at the end the wheel of Part 1 and the wheel of Part 2 stand one after the other.
1One curve shifts
1How the textbook is used outside the classroom: one curve moves, and everything else is held.
In the year to July 2022 consumer prices in Korea rose 6.3 percent, the fastest rise since 1998. Many explanations written at the time named one shift. Demand had recovered after the pandemic, so the demand curve moved out. Or energy had become dear, so the supply curve moved in. Either answer is Figure 1.1 with one curve moved. It is how the classical textbook is most often used outside the classroom, and it is the use the textbook trains for: hold everything else, move one curve, read off the new price.
The habit has a name. James Kwak (2017) called it economism: the lessons of a first course, chiefly supply and demand, applied to questions the course was not built to answer, and presented as if economists agreed on them. His chief example is the minimum wage, where one pair of curves is taken to settle a question on which empirical studies divide. Economists in policy made the same point from the other side. In his foreword to the Post-Crash Economics Society’s report (2014), Andy Haldane of the Bank of England argued for more history, institutions, money and banking, and a wider range of methods in teaching.
What the one-curve answer holds still was moving at the same time, and by more than prices. Brent crude cost 18 dollars a barrel in April 2020 and 123 in June 2022. The won fell from 1,097 to the dollar in January 2021 to 1,427 in October 2022, so each imported barrel cost more again. The Bank of Korea held its base rate at 0.5 percent until July 2021 and raised it to 3.5 percent by January 2023. Firms that buy from one another passed costs along their supply chains, and households that differ in what they earn and own met the same prices differently.
Held fixedin Figure 1.1, everything but one market.
Folded, the inflation line rises by what a barrel of oil cost in won, the dollar price times the exchange rate, which more than doubled between January 2021 and June 2022. The flat line is the path’s shadow. Then the wheel beside the page shows the one-curve answer as a state: the goods market moves, the households and firms behind its curves are held, and everything else is left out. It shows the economy of 2022, in which oil from abroad, the exchange rate, the central bank and the ties between firms all moved together. Last, it returns to the one-curve state, with the values it held written on the cells it held.
A shifted curve answers a question about one market with everything else held. The questions now put to it are about the economy the holding leaves out.
2What the textbook set aside
2Most of what the textbook leaves out was known while it was being written.
The textbook of Part 1 took its shape between 1890 and the late 1950s. Its parts came from Marshall’s Principles (1890), Hicks’s IS–LM (1937), Samuelson’s Foundations (1947) and his textbook Economics (1948), Arrow and Debreu’s general equilibrium (1954) and Solow’s growth model (1956). Over the same years, and earlier, economists and others found most of what this part is about.
- Sizes without a typical size. Vilfredo Pareto (1896) found that the number of people with income above x falls like a power of x. Robert Gibrat (1931) found that firms grow in proportion to their size, which spreads their sizes far apart. George Zipf (1949) found that the sizes of cities fall with their rank like a power. A typical agent does not describe such a population.
- People who reason about one another, and only so far. Keynes (1936) described investors who guess what other investors will guess. John von Neumann and Oskar Morgenstern (1944) and John Nash (1950) built the theory of such games. Herbert Simon (1955) proposed that people choose with limited attention and stop at good enough.
- Prices that need not settle. Herbert Scarf (1960) showed that the classical adjustment of prices need not reach an equilibrium at all.
- People who reach particular others. Stanley Milgram (1967) passed letters through acquaintances to strangers. Mark Granovetter (1973) found that jobs travel through weak ties.
Some of this entered the core: game theory is Stage 4, and George Akerlof’s hidden quality (1970) is Figure 5.3. Some entered only a field course: behavioral economics is 8.9. Networks and the distribution of sizes are not in Part 1 at all. The choice was made for the reader. A student has to be able to solve a textbook figure by hand and see the whole of it on one page. A representative household that trades with anyone in one market can be solved that way, with calculus. A thousand unlike households tied to particular others cannot.
Held fixedhow long each finding took to reach any textbook; the marks give only where it arrives in Part 1.
Folded, each finding rises to the assumption it breaks, and the timeline becomes the table of assumptions that Part 3 draws in its Figure 3.4. Two rows hold only findings that Part 1 never takes in: a representative agent, and no connections. Then the wheel takes each group of findings in turn, and the textbook’s state, in which every agent of a kind is the same size and anyone trades with anyone, puts them out again.
The textbook’s simplicity was a kindness to its readers. It gave them what a student can solve by hand and see on a page, and it was not written in ignorance of what that leaves out. Computers and data have changed what a student can solve.
3Unlike
3Relaxes the representative agent: one firm, one household, one way of reasoning, each standing for all.
The first assumption Part 1 keeps to the end is that one agent can stand for all of its kind. Individuals may differ, but only their total matters. This section puts back three of the findings of section 2, each where Part 1 makes the assumption: firms of every size (Stage 3), households of every income (Stage 6), and players who reason only so far (Stage 4). Folded, the wheel’s agents become unlike one another and stay so for the rest of this part.
Firms, households, players
3.1No representative firm
Relaxes one firm stands for all.
The firms of Stage 3 are identical, so one firm can stand for the industry. Vilfredo Pareto noticed in the 1890s that the number of people with income above x falls roughly like a power of x. The same shape describes city sizes (Zipf’s law) and the sizes of American firms (Axtell 2001). When newcomers prefer to link to those already well connected (Barabási and Albert 1999), the result is a power law too, if a less steep one: a few giants and a long tail of everyone else.
Held fixedin Stage 3, size: every firm alike.
Folded, every firm rises as tall as it is large, and the mean becomes a plane at height 4 that cuts through the skyline. Seventy-two percent of the firms stand below it, and the largest is about nine and a half times the mean. Xavier Gabaix (2011) argued that this matters for macroeconomics: if firm sizes are as unequal as Zipf’s law says, shocks to the largest firms do not average out, and they move total output. On the wheel beside the page, Part 1’s 600 firms are drawn first as the representative firm draws them, all the same size, and then as they are; stopping the largest stops a seventh of the flow through firms.
When sizes follow a power law there is no typical firm. The mean is a plane that runs through no actual firm, and a few giants can move the whole economy.
3.2Unlike households
Relaxes one household stands for all.
The households of Stage 6 are added up into one. The IS curve in Figure 6.2 rests on its consumption function, which says how total spending moves with total income. That is enough as long as every household spends the same share of an extra won, because then it does not matter who receives the won. W. M. Gorman (1953) set out the condition under which households can be added up this way. Korean households do not meet it. In 2025 the bottom fifth of households by income spent 134 percent of their disposable income, drawing on savings and on help from others, while the top fifth spent 56 percent.
Held fixedin Figure 6.2, the household: one, spending the same share of every won.
Folded, each fifth rises by its share of all disposable income, and beside it stands its share of all spending. The top fifth has 42 percent of the income and does 34 percent of the spending. The bottom fifth has 5 percent of the income and does 9 percent of the spending. The survey measures what each fifth spends of what it has, not what it would spend of an extra won. Studies that ask that second question find the same order. Tullio Jappelli and Luigi Pistaferri (2014) asked Italian households how much of an unexpected transfer they would spend, and those with little cash on hand said they would spend much more of it. Greg Kaplan, Giovanni Violante and Justin Weidner (2014) estimate that about a third of American households live hand to mouth. In the models of Kaplan, Benjamin Moll and Violante (2018), where such households are common, a cut in the interest rate raises spending less through each household’s choice between now and later and more through the income it brings.
A household that stands for all spends 69 percent of every won, and no fifth of Korean households does. Who receives a won decides what it does to total spending.
3.3Games real people play
Relaxes every player reasons all the way.
The equilibria of Stage 4 assume that each player knows the others will reason as far as logic goes. Keynes (1936) compared investing to a newspaper beauty contest in which you win by picking the face others will pick. In the experimental version everyone chooses a number from 0 to 100, and the winner is closest to two-thirds of the average. The only Nash equilibrium is 0. In Rosemarie Nagel’s (1995) experiments, first-round guesses averaged 36.7: most players reason one or two steps and stop. When others reason only a little, guessing 0 loses.
Held fixedin Stage 4, how far players reason: all the way.
Folded, each level stands at its own height: level 0 on the page at 50, each step up a third lower, Nash at the top at 0. The guesses trace a staircase down to equilibrium, and the bars show how few players climb it. In the ultimatum game, too, play departs from the selfish equilibrium. Across many studies, summarized by Colin Camerer (2003), offers of 40 to 50 percent are the most common, and offers below about 20 percent are rejected about half the time. On the wheel, the strategy cell lights over players who differ in how far they reason.
Equilibrium is the top of a staircase of reasoning. Most people climb one or two steps, so the best guess depends on how far the others climb, not on how far logic goes.
4Connected
4Relaxes the assumption that anyone trades with anyone.
The second assumption Part 1 keeps to the end is that nobody is connected to anybody in particular. Every buyer reaches every seller through the market, and every bank stands alone. Four findings put connections back: how far apart people are (4.1), what one broken tie does (4.2), how a won of demand travels between firms (4.3), and how a loss travels between banks (4.4). Folded, the wheel gains its networks cell, and its households, firms and banks are tied in turn.
People, ties, firms, banks
4.1Who trades with whom
Relaxes anyone can trade with anyone.
The market of Figure 1.2 has no geography. Every buyer can reach every seller, so the distance between any two people is one step. Real exchange runs through acquaintances, suppliers and neighbors, and the question becomes how far apart people are.
In 1967 Stanley Milgram asked people in Nebraska and Kansas to pass a letter, through people they knew by first name, to a stranger in Massachusetts. The letters that arrived took about six steps. In 1998 Duncan Watts and Steven Strogatz showed why this is unsurprising. Start with a ring in which each person knows only the nearest neighbors: distances are long, and friends of friends know each other. Then rewire a small share of ties at random. Average distance collapses long before local clustering does.
Held fixedin Figure 1.2, who can reach whom: everyone, in one step.
Folded, each person rises by the number of steps it takes to reach them from you. On the plain ring the far side stands fourteen steps high. Eight shortcuts pull the whole ring down to eight. Mark Granovetter (1973) found that people often hear about jobs through acquaintances rather than close friends: the weak ties are the shortcuts.
A few shortcuts make a large world small. Distance collapses long before neighbors stop knowing one another, which is why jobs, prices and panics travel faster than any local picture suggests.
4.2Chains, not cores
Relaxes an average describes the network.
Peter Bearman, James Moody and Katherine Stovel (2004) mapped the romantic relationships among the students of one American high school over eighteen months. Models of how infections spread through such networks had assumed a dense core of very active people. Instead, about half of the students who had been in a relationship were linked in one long, sparse structure with short branches, close to a spanning tree. In such a structure there are few redundant paths: breaking one tie can split it, and one new tie can join two distant parts.
Held fixedby an average, how many partners each student has: about two in the chain, two and a half in the core.
Folded, every tie raises a wall as tall as the number of pairs of students who would be cut off from each other if it broke. In the chain the middle ties stand 144 pairs high; in the core no tie stands higher than 23, and those only where one student hangs off the edge. On the wheel, the networks cell lights over households tied to particular firms.
Two networks with similar averages can differ completely in what one broken tie does. The structure, not the count, decides what can travel through a network and what can be cut.
4.3Connected firms
Relaxes a firm buys only labor and capital.
The firm of Figure 3.1 turns labor and capital into output. Real firms also buy from one another: steel and tires for a car, power and freight for nearly everything. Wassily Leontief published the first table of these purchases, for the United States, in 1936, and the Bank of Korea publishes one for Korea every year. Its 2024 table has 33 sectors. Each entry of the table’s Leontief inverse says how much output one won of final demand for one sector draws from another, counting the suppliers’ suppliers.
Held fixedin Figure 3.1, what a firm buys from other firms: nothing.
Folded, each sector rises by the output that the chosen won draws from it, round by round. Without ties only the sector itself would rise, by one won. One won of final demand for transport equipment draws 2.36 won of output in all. The largest part of the first round, 0.24 won, goes to other makers of transport equipment, the parts makers. Basic metals, electrical equipment, chemicals and fabricated metal get about 0.05 won each. Daron Acemoglu, Vasco Carvalho, Asuman Ozdaglar and Alireza Tahbaz-Salehi (2012) showed that when a few sectors supply many others, shocks to those sectors do not average out across the economy, as the giant firms of Figure 2.3 do not. Daron Acemoglu, Ufuk Akcigit and William Kerr (2016) found in American data that changes in demand travel up the chains, to suppliers, and changes in productivity travel down them, to customers.
A firm’s output is other firms’ input. A won of demand for one sector reaches the sectors that supply it, and a shock to a sector that supplies many does not stay where it began.
4.4Contagion between banks
Relaxes a bank stands alone.
The bank of Figure 8.3 fails or survives on its own depositors. In 2008 losses at a few institutions spread through the claims banks held on one another. Daron Acemoglu, Asuman Ozdaglar and Alireza Tahbaz-Salehi (2015) showed that a densely connected system spreads a small loss thin and absorbs it, while a ring passes it from bank to bank and loses several. A loss larger than the system’s combined buffers is carried to every bank, in the dense system and on the ring alike. Against losses that large, the safer system is made of clusters joined by few ties, which keep the loss inside one cluster.
Held fixedin Figure 8.3, who owes whom: nobody; the bank stands alone.
Folded, each bank rises to the round in which it fails, and the ones that survive stay on the page. On the ring a loss of 4 makes a short staircase of three banks, while the dense network loses only the first. Raise the loss above 10 and the ring becomes a spiral ten rounds tall, while the dense network falls all at once, in round 2.
The dense web that spreads a small loss thin carries a large one to everyone, all at once. Whether density protects depends on the size of the shock, not on the network alone.
5Why the mathematics got harder
5Unlike and connected agents need more than calculus: fixed points, dynamics, statistics and computation.
A student who reaches the third year finds that economics has become topology, measure theory, dynamic programming and econometrics. This is often read as a taste for difficulty. It follows from the two assumptions this part relaxes.
With one representative agent in one market, an equilibrium is where two curves cross, and calculus finds it. With many markets and many agents there is no picture to look at. Kenneth Arrow and Gérard Debreu (1954) proved that a competitive equilibrium exists with a fixed-point theorem from topology. The simplest theorem of that kind, Brouwer’s, says that a continuous map of a closed, bounded, convex set into itself leaves at least one point where it is. John Nash (1950) had proved in the same way that every finite game has an equilibrium in mixed strategies. Existence was settled, but whether markets reach the equilibrium was not. Herbert Scarf (1960) built an economy of three people and three goods in which the classical adjustment of prices circles the equilibrium forever. The adjustment raises a price when its good is in excess demand. Hugo Sonnenschein, Rolf Mantel and Debreu (1972–74) then showed that adding up unlike households puts almost no shape on total demand. Beyond continuity and a few identities, any shape is possible. The representative agent was the way out. If the total is assumed to behave like one household, the shape returns.
Held fixedin Figure 5.1, time: equilibrium is reached at once.
Folded, time rises off the page. The representative economy’s prices reach the center and climb straight up it. Scarf’s prices climb around the center in a spiral that never closes in. The equilibrium exists in both economies, and only one of them reaches it.
The other tools arrived the same way. Time and uncertainty need dynamic programming (Richard Bellman, 1957) and the measure theory beneath probability. Data need a model of how they were produced (Trygve Haavelmo, 1944). They also need a design that separates cause from coincidence, which Joshua Angrist and Jörn-Steffen Pischke (2010) called the credibility revolution. Unlike agents tied in networks often have no solution in closed form at all, and are computed instead. Agent-based models (Joshua Epstein and Robert Axtell, 1996) and heterogeneous-agent macroeconomics solve them by simulation. The Journal of Economic Literature’s classification lists these tools side by side. The table below says which part of the economy calls for which.
| What the model contains | The mathematics it needs | JEL | First seen |
|---|---|---|---|
| one market, a curve on each side | calculus, comparative statics | C6 | Figure 1.1 |
| a household or a firm that chooses | constrained optimization | C6 | Figure 2.1 |
| every market at once | fixed-point theorems (topology) | C6, D5 | Figure 5.1 |
| players who watch one another | game theory, fixed points | C7 | Figure 4.1 |
| time and uncertainty | dynamic programming, probability | C6, D8 | Figure 2.3 |
| data | probability models, research designs | C1, C9 | Figure 7.2 |
| many unlike agents | distributions, computation | C63, D3 | Figure 2.4 |
| agents tied in networks | graphs, linear algebra, computation | D85, C63 | Figure 2.8 |
| choices that depart from consistency | experiments | C9, D91 | Figure 2.5 |
| prices that need not settle | dynamical systems, computation | C6, C63 | Figure 2.10 |
The mathematics of economics grew harder as the economy it drew became unlike and connected. Where a page cannot solve the model, a proof, an estimate or a computer has to.
6Why the evidence changed
6Most of what economists now publish measures unlike agents with designed comparisons.
The other half of the change is in evidence. Daniel Hamermesh (2013) classified every full-length article in three top general journals, the American Economic Review, the Journal of Political Economy and the Quarterly Journal of Economics, in one year of each decade from 1963 to 2011. In 1963 half of them (50.7 percent) were pure theory, and in 2011 a fifth (19.1 percent). Empirical work on data the authors had collected themselves rose from 8.7 to 34.0 percent, and experiments from none to 8.2 percent. In 2011 nearly three articles in four were empirical, and four in five had more than one author, against one in six in 1963.
Two things drove the change. Records of unlike people became available: tax and benefit records, surveys that follow the same households for years, the accounts of firms. And the design of studies changed. Natural experiments, randomized trials and the other designs that Joshua Angrist and Jörn-Steffen Pischke (2010) called the credibility revolution ask what would have happened otherwise, which a regression line through a scatter cannot say. The statistics of Stage 7 is where this begins. In 2019 Raj Chetty and Gregory Bruich taught an introductory course at Harvard, Using Big Data to Solve Economic and Social Problems, that starts from records and questions and reaches the models afterwards.
Held fixedin Figure 7.1, the data: given, and the same for everyone.
Folded, the stack comes apart. Each method stands in its own row, each column as tall as that method’s share, so that each can be followed across the decades. On the wheel beside the page the measurement cell lights over households and firms that are unlike one another.
In 2011 nearly three articles in four were empirical. What economists now learn to do most is measure unlike agents with designs that a single scatter does not teach.
7The old arguments, again
7The arguments about what economics is were arguments about what to hold still.
The questions of this part are not new. Economists have argued about them since the subject began. The arguments are worth reading again with the wheel in view, because each is about what a model should move and what it may hold still.
- 1883. In the Methodenstreit, Carl Menger held that economics deduces general laws from individual choice; Gustav Schmoller, that it studies the history and institutions of particular economies.
- 1890, 1932. Alfred Marshall called economics “a study of mankind in the ordinary business of life”. Lionel Robbins defined it as the study of behavior as a relationship between ends and scarce means.
- 1936. John Maynard Keynes argued that employment is set by total spending, which can stay too low; his analysis takes the money wage as given, and he argued that cutting it need not restore employment.
- 1945. Friedrich Hayek argued that the knowledge an economy runs on is dispersed among people who differ, that prices carry it, and that no planner can gather it.
- 1953. Milton Friedman argued that a theory’s assumptions need not be realistic if its predictions are right.
- 1976. Robert Lucas argued that the rules describing behavior change when policy changes, so models must start from choices.
- 1992. Alan Kirman asked whom the representative individual represents: unlike people can add up to a total that none of them would choose.
- 2008. The benchmark model behind much central-bank analysis, Smets and Wouters (2007), had no banks, and models that added credit did not tie banks to one another.
Held fixedwho won each argument; the plans record what each was about.
Read this way, the arguments do not divide economists into those who use models and those who do not. Each side proposes a different state of the same wheel. Marshall’s subject is the whole of it, and Robbins’s is one tool at its hub. Keynes holds the money wage, and with it the labor market that the classics let clear. Friedman accepts the classical state and asks only whether it predicts. Lucas asks what the state is built from, and Kirman whether the one agent at its center can stand for the many. The crisis of 2008 showed what a state without banks cannot see. Paul Krugman (2009) made that charge in public. Folded, the wheel takes each position in turn.
The argument about what economics is has always been an argument about what to hold still. An archive that states, for every model, what it holds can carry that argument instead of hiding it.
8Those who asked before
8Students, teachers and economists reached the same diagnosis. One need is still unmet.
The diagnosis of this part was reached before, from several directions, mostly after 2008.
- Students. At the University of Manchester the Post-Crash Economics Society published Economics, Education and Unlearning (2014), asking for history, institutions, money and banking, and more than one school of thought. In May 2014, 42 student groups from 19 countries signed an open letter for pluralism in economics (ISIPE); within weeks the network had 65 groups in 30 countries.
- Teachers. CORE, begun in 2013 by economists from five countries under Wendy Carlin, wrote The Economy, which starts each unit from a problem, inequality first, with data and strategic interaction from the first week. It is used in hundreds of universities.
- Macroeconomists. The Rebuilding Macroeconomic Theory Project (David Vines and Samuel Wills, 2018) asked for four changes to the benchmark model: financial frictions, limits on rational expectations, heterogeneous agents and better microfoundations.
- Builders of pieces. Opportunity Insights’ course starts from big data. QuantEcon makes models run as code, including heterogeneous agents and Thomas Sargent and John Stachurski’s Economic Networks. The Santa Fe Institute’s Complexity Explorer teaches agent-based models, networks and power laws, and Exploring Economics places the schools side by side.
Held fixedhow well each project succeeds; a mark records only what it sets out to do.
Every need is met somewhere, and one is met nowhere: a common map on which any model, classical or modern, shows what it moves, what it holds still and what it leaves out. Without it, a student who takes CORE’s units, QuantEcon’s code and a pluralist course cannot say how their models relate, and a writer who reaches for one curve cannot say which state of the economy it assumes. Folded, the wheel lights what each project adds, and at the end shows one model in the form such a map would give it.
The needs were found before this site. What is missing is the map that holds classical and modern models on one plane, and Part 3 proposes it.
Where to start
Start from problems and evidence
- CORE EconThe Economy 2.0, Economy, Society, and Public Policy, Doing EconomicsStarts each unit from a real problem (inequality, climate, innovation) and teaches models as tools for it, with power, institutions and data from the first week. Free online.
- Our World in DataLong-run data on living conditions, with the research behind itThe record behind growth, health, poverty and climate, drawn so that it can be read without a course in statistics.
Make the models run
- QuantEconLectures in Python and Julia, founded by Thomas Sargent and John StachurskiEvery model is code the learner runs: dynamic programming, heterogeneous agents, networks. Computation replaces the closed-form special case.
- EconGraphsInteractive graphs for intermediate microeconomicsEvery textbook diagram becomes something the reader can move, so a shift is seen as a cause rather than memorized as a rule.
- The Evolution of TrustNicky Case’s explorable explanation of the repeated prisoner’s dilemmaGame theory taught by playing it, with populations of strategies that evolve.
Add structure and difference
- Social and Economic NetworksMatthew O. Jackson’s book and courseNetworks as an economic object: how ties form, how information and behavior spread, why position in a network matters.
- Complexity ExplorerThe Santa Fe Institute’s courses on complex systemsAgent-based models, scaling and power laws, and dynamics away from equilibrium.
Add identification
Compare the schools, and keep the baseline
- Exploring EconomicsA pluralist platform from students and researchersPlaces the neoclassical sequence beside other schools so that a reader can compare how each frames a question.
- Marginal Revolution UniversityShort videos by Tyler Cowen and Alex TabarrokPrinciples in pieces small enough to reorder, with applications first.
- MIT OpenCourseWareThe sequence studied in Part 1, openThe classical staircase itself, lecture by lecture: the baseline the projects above reorganize.
9What it takes to understand an economy
9Five tasks, which Part 3 makes the rules of its archive.
Writing about the economy from inside the textbook’s assumptions is useful, and much good economic writing does it. Understanding an economy takes more, and the sections above say what. Five tasks recur.
- Say what is held. Every model holds something still. Name it and the value it is held at, as Part 1 named each figure’s hidden axis (section 1).
- Find the assumption that decides. Most results survive most changes of assumption, and a few assumptions decide the answer. Dani Rodrik (2015) calls these critical assumptions; they are the ones to test (sections 3 and 4).
- Measure with a design. A pattern in data is not a relationship until the design separates cause from coincidence (section 6).
- Compute when the page cannot solve. Unlike agents tied to one another are solved by simulation, and the simulation is part of the argument (section 5).
- Keep history. The structure itself changes: the networks, the distributions, the rules. A model fits a period, and the period should be stated (sections 2 and 7).
Held fixedthe order in which they are done; in practice they repeat.
These are not another field course. They are what every unit of a rebuilt textbook should do. Folded, the wheel performs them in turn: a cell is hatched with the value it is held at, the agents change size and are tied, the measured totals stand at the hub, the computation cell lights, and then history. Then the wheel stands as Part 1 left it and as Part 2 leaves it, one after the other, and last in the form Part 3 gives every model.
Classical economics remains the first state of the wheel, and the one every other is measured against. Understanding an economy now means saying which state one is in, and why.
Part 3 builds the map: an archive in which every model says what it moves, what it holds and what it leaves out, and the classical sequence is one way through among several.