Key Models and Equations
The models and equations of the masterclass, collected unit by unit. Each one is derived or motivated in its unit, with the conditions under which it holds and a worked example with real data.
Economics states its models as equations so that they can be checked against data. A demand curve written as a line can be fitted to a price series; an elasticity written as a ratio can be estimated from two observed points; the quantity equation, once written down, tells you which of its four letters you have to assume constant before you can predict anything. This page collects those models and equations in the order of the units, among them the equilibrium condition, the elasticity formula, the profit-maximizing rule, the Fisher equation, the multiplier, the real exchange rate, the Gini coefficient, and the prospect-theory value function, and it states the handful of results that have no formula, such as comparative advantage and the lemons problem, in words.
Each card gives the equation the way the Key Equation box of its unit does, names its symbols, and says when it holds: which things are held equal, whether competition is assumed perfect, whether the result is an identity or a theory. The later cards are the behavioural models that qualify the earlier ones, and each of them names the model it departs from. The page grows as units are published, so an equation you cannot find yet belongs to a unit that is still to come.
The Economist's Toolkit
Opportunity cost
The opportunity cost of a choice is the value of the best alternative that the choice gives up. On a production possibility frontier it is the slope of the curve: \[ \text{opportunity cost of one more of good } A = -\,\frac{\Delta B}{\Delta A} , \] where \(A\) is the good on the horizontal axis and \(B\) the good on the vertical axis, and \(\Delta A\) and \(\Delta B\) are the changes along the curve. The minus sign makes the cost a positive number, since the frontier slopes downward. The cost rises along the curve when resources differ in their suitability for the two goods.
Growth rate and compound growth
\[ g = \frac{\Delta Y}{Y}, \qquad Y_t = Y_0\,(1+g)^t \] \(g\) is the growth rate per year (0.02 for 2 per cent), \(Y_0\) the starting value, and \(Y_t\) the value after \(t\) years of growth at that constant rate. The second formula assumes that the rate does not change; for a varying rate the factors \((1+g_1)(1+g_2)\cdots\) are multiplied year by year.
The rule of 70
\[ \text{doubling time in years} \approx \frac{70}{g} \qquad (g\ \text{in per cent per year}) \] A quantity growing at 2 per cent a year doubles in about 35 years; at 7 per cent, in about 10 years. The rule follows from \(\ln 2 \approx 0.69\) and holds for the growth rates that economies, prices, and populations show.
Index numbers and real values
\[ I_t = 100 \times \frac{P_t}{P_0}, \qquad \text{real value}_t = \frac{\text{nominal value}_t}{I_t/100} \] \(I_t\) is the index in year \(t\) with base year 0, where it equals 100. Dividing a nominal amount by the index (over 100) expresses it in base-year prices. Nominal and real values are never mixed in one comparison, and a real value always names its base year.
Mean, median, and expected value
\[ \bar{x} = \frac{1}{n}\sum_{i=1}^{n} x_i , \qquad \mathbb{E}[X] = \sum_{k} p_k\, x_k \] \(\bar{x}\) is the mean of \(n\) observed values \(x_i\), their sum divided by \(n\); the median is the middle value of the sorted list and is not pulled up by a few very large values, which is why it is the better summary of incomes. The expected value \(\mathbb{E}[X]\) of an uncertain outcome \(X\) is the average of its possible values \(x_k\) weighted by their probabilities \(p_k\), which sum to one.
Supply, Demand, and Market Equilibrium
Market equilibrium
\[ Q_{\mathrm{d}}(P^{*}) = Q_{\mathrm{s}}(P^{*}) , \qquad\text{and for straight lines}\qquad P^{*} = \frac{a - c}{b + d} , \quad Q^{*} = a - bP^{*} . \] The equilibrium price \(P^{*}\) makes the quantity demanded equal to the quantity supplied, \(Q^{*}\); the formulas hold for \(Q_{\mathrm{d}} = a - bP\) and \(Q_{\mathrm{s}} = c + dP\) in a competitive market.
Consumer surplus and producer surplus
\[ CS = \tfrac{1}{2}\,(P_{\max} - P^{*})\,Q^{*} , \qquad PS = \tfrac{1}{2}\,(P^{*} - P_{\min})\,Q^{*} , \] for straight-line curves, where \(P_{\max} = a/b\) is the price at which demand reaches zero and \(P_{\min} = -c/d\) the price at which supply starts (the formula for \(PS\) needs \(c \le 0\), because for \(c > 0\) the supply line meets the quantity axis instead of the price axis and the area is no longer a triangle). In general, consumer surplus is the area below the demand curve and above the price, producer surplus the area above the supply curve and below the price, and both are measured in money per interval.
The tax wedge, its incidence, revenue, and deadweight loss
\[ P_{\mathrm{b}} - P_{\mathrm{s}} = t , \qquad P_{\mathrm{b}} - P^{*} = \frac{d}{b + d}\,t , \qquad \text{revenue} = t\,Q_t , \qquad DWL = \tfrac{1}{2}\,t\,(Q^{*} - Q_t) . \] A tax of \(t\) per item raises the price buyers pay by the share \(d/(b+d)\) of the tax and lowers the price sellers keep by the share \(b/(b+d)\), for the straight lines \(Q_{\mathrm{d}} = a - bP\) and \(Q_{\mathrm{s}} = c + dP\); the side whose curve is steeper, with the smaller slope coefficient, bears the larger share. The government collects \(t\) on each of the \(Q_t\) items still traded, and the surplus lost on the items no longer traded, the deadweight loss, is the triangle between the two curves over the lost quantity.
Consumer Choice and Elasticity
The budget constraint
\[ P_x\,x + P_y\,y = I , \qquad \text{slope of the budget line } = -\frac{P_x}{P_y} \] A consumer with income \(I\) can buy any bundle \((x, y)\) on or below the line. The slope is the relative price of \(x\), the amount of \(y\) that one more \(x\) costs. The constraint applies to any number of goods, with one term \(P\,q\) for each of them.
The consumer’s rule
\[ \frac{MU_x}{P_x} = \frac{MU_y}{P_y} \] At the best affordable bundle, the last euro (or pound, or peso) spent on each good yields the same marginal utility. The rule is called the equal marginal utility per euro rule. It applies to the goods the consumer actually buys. A good she does not buy at all is one whose very first item would already give less utility per euro than the goods she does buy.
The price elasticity of demand (arc form)
\[ \varepsilon = \frac{\Delta Q / \bar{Q}}{\Delta P / \bar{P}} , \qquad \bar{Q} = \frac{Q_1 + Q_2}{2} , \quad \bar{P} = \frac{P_1 + P_2}{2} \] The percentage change in quantity demanded for a one per cent change in price, computed between two observed points with percentages taken relative to the midpoints. It is negative for almost every good; demand is inelastic when \(|\varepsilon| < 1\) and elastic when \(|\varepsilon| > 1\).
The revenue rule
\[ \frac{\Delta TR}{TR} \approx \frac{\Delta P}{P} + \frac{\Delta Q}{Q} = \frac{\Delta P}{P}\,(1 + \varepsilon) \] A price rise raises total revenue \(TR = P Q\) when demand is inelastic (\(|\varepsilon| < 1\)), lowers it when demand is elastic (\(|\varepsilon| > 1\)), and leaves it unchanged at \(|\varepsilon| = 1\). The approximation holds for small changes; for large ones, compute \(PQ\) before and after.
Tax incidence with elasticities
\[ \text{share of the tax borne by buyers} = \frac{\varepsilon_{\mathrm{s}}}{\varepsilon_{\mathrm{s}} + |\varepsilon_{\mathrm{d}}|} , \qquad \text{share borne by sellers} = \frac{|\varepsilon_{\mathrm{d}}|}{\varepsilon_{\mathrm{s}} + |\varepsilon_{\mathrm{d}}|} \] The buyers’ price rises by this share of a tax \(t\) per item and the sellers’ price falls by the rest. The side that responds less to price, the less elastic side, pays more. The formula holds for a tax that is small compared with the price, in a competitive market, and it does not matter which side hands the money to the government.
Firms, Costs, and Competitive Markets
Profit
\[ \Pi = TR - TC \] Profit \(\Pi\) is total revenue \(TR\) (price times quantity sold, \(P \cdot Q\)) minus total cost \(TC\). In economics \(TC\) includes the opportunity cost of every input, including the ones the owner supplies without being paid.
Marginal cost and average cost
\[ MC = \frac{\Delta TC}{\Delta Q}, \qquad ATC = \frac{TC}{Q}, \qquad AVC = \frac{VC}{Q} \] Marginal cost \(MC\) is the extra cost of one more item (here computed over a step of several items, as the change in total cost divided by the change in output). Average total cost \(ATC\) is the cost per item of everything produced, and average variable cost \(AVC\) leaves the fixed cost out. \(ATC - AVC = FC/Q\) is the fixed cost per item, which falls as output grows.
The profit-maximizing rule
\[ MR = MC, \qquad\text{which for a price taker is}\qquad P = MC \] A firm produces the output at which the revenue from the last item equals its marginal cost. The rule picks the output on the rising part of the marginal cost curve; where \(MC\) is still falling, a further item would cost less than the last and should be produced too.
The shutdown rule and the exit rule
\begin{align*} \text{Short run:}\quad & \text{produce if } P \ge AVC \text{ at the chosen output, shut down if } P < \min AVC ; \\ \text{Long run:}\quad & \text{stay if } P \ge ATC, \text{ exit if } P < \min ATC . \end{align*} In the short run the fixed cost is sunk, which means that it is due whether or not the firm produces, so only the variable cost counts; a firm that covers its variable cost pays part of its fixed cost by producing. In the long run every cost can be avoided by leaving, so the price must cover average total cost.
The long-run competitive outcome
\[ P = MC = \min ATC \] With free entry and exit, the long-run price in a competitive industry equals both the marginal cost and the lowest average total cost of the firms, and economic profit is zero. Each firm produces at the scale that makes its average cost as low as possible.
Market Power and Strategy
Marginal revenue of a straight demand line
\[ P = a - bQ \quad\Longrightarrow\quad MR = a - 2bQ \] The marginal revenue line of a straight demand line starts at the same price \(a\) and falls twice as steeply, so it reaches zero at half the quantity at which demand reaches zero. Beyond that point selling more reduces revenue.
The Lerner index
\[ L = \frac{P - MC}{P} = \frac{1}{|\varepsilon|} \] The Lerner index \(L\) is the mark-up over marginal cost as a share of the price, and for a profit-maximizing firm it equals the inverse of the absolute value of the price elasticity of demand it faces. A price taker faces an infinitely elastic demand and has \(L = 0\). The index is named after Abba Lerner (1903-1982), who proposed it in 1934.
The Cournot equilibrium
\[ q_i = \frac{a - c}{(n+1)\,b} , \qquad P = \frac{a + n\,c}{n + 1} \] The output of each of \(n\) identical firms and the market price when demand is \(P = a - bQ\), marginal cost is the constant \(c\), and each firm chooses its quantity taking the others’ quantities as given. For \(n = 1\) the formulas give the monopoly, for \(n = 2\) the duopoly \(q_i = (a - c)/(3b)\), and as \(n\) grows the price falls towards the competitive price \(c\).
The Herfindahl-Hirschman index
\[ \mathrm{HHI} = \sum_{i=1}^{n} s_i^{2} \] The sum of the squared market shares \(s_i\) of all \(n\) firms in a market, with the shares written in per cent, so that the index runs from near 0 for a market of many tiny firms to 10,000 for a monopoly. A merger of two firms with shares \(s_1\) and \(s_2\) raises the index by \(2 s_1 s_2\), because \((s_1 + s_2)^{2} = s_1^{2} + s_2^{2} + 2 s_1 s_2\).
Market Failure and Public Policy
Social marginal cost
\[ SMC = MC + MEC \] The social marginal cost \(SMC\) of one more item is the producer’s private marginal cost \(MC\) plus the marginal external cost \(MEC\) it imposes on others. The market supply curve is \(MC\), so the market settles where demand meets \(MC\); the efficient quantity is where demand meets \(SMC\). For a positive externality the external term is a benefit and is added to the demand curve instead.
The Pigouvian tax
\[ t = MEC \quad \text{at } Q^{*} \] A tax of \(t\) per item equal to the marginal external cost at the efficient quantity shifts the private marginal cost up to the social marginal cost, so the market produces \(Q^{*}\) and the deadweight loss disappears. The revenue \(t \times Q^{*}\) is a transfer from the market to the treasury and no loss to society. A subsidy equal to the marginal external benefit does the same for a positive externality.
The cost-effectiveness condition
\[ MAC_1 = MAC_2 = \cdots = MAC_n = \text{permit price} \] A given total cut in emissions is achieved at the lowest total cost when the marginal abatement cost is the same for every firm. A market in permits brings this about, because every firm cuts until the cost of one more tonne cut equals the price of a permit, and every firm faces the same price. The same holds for a tax, with the tax in place of the permit price.
The price of a lemon
\[ P = \mathbb{E}[\text{quality}] = \theta\,V_{\mathrm{good}} + (1 - \theta)\,V_{\mathrm{bad}} \] When buyers cannot tell good items from bad ones, the most they will pay is the expected value of an item of unknown quality, which is the value of a good one, \(V_{\mathrm{good}}\), and the value of a bad one, \(V_{\mathrm{bad}}\), weighted by the share \(\theta\) of good items among those actually offered. Because the sellers of good items withdraw when this price falls below what their items are worth to them, \(\theta\) itself depends on the price, and the market can unravel until only the worst items are traded. The result is known as the lemons problem or adverse selection.
Measuring the Economy
The expenditure identity
\[ Y = C + I + G + NX \] GDP (\(Y\)) equals consumption by households (\(C\)), investment (\(I\)), government purchases of goods and services (\(G\)), and net exports (\(NX = X - M\)), which are exports minus imports. The identity holds by definition in every country and every year. The letter \(I\) here is investment, which means spending on new capital goods, new buildings, and additions to inventories, and never the purchase of shares or bonds; the \(I\) that stood for income in Unit 3 does not appear in this unit.
Real GDP and the GDP deflator
\[ \text{real GDP} = \frac{\text{nominal GDP}}{\text{deflator}/100} , \qquad \text{deflator} = \frac{\text{nominal GDP}}{\text{real GDP}} \times 100 \] The deflator is an index number, equal to 100 in the base year, that measures the prices of everything produced in the country. Real GDP is stated in the prices of the base year, written “at 2015 prices” or, for a reason explained below, “in chained 2017 dollars”, and the base year must always be named. Taking the inflation out of a nominal series is called deflating it, and it always means dividing by the price index, never subtracting the inflation rate.
The consumer price index and the inflation rate
\[ \mathrm{CPI}_t = 100 \times \frac{\sum_i q_i P_{i,t}}{\sum_i q_i P_{i,0}} , \qquad \pi_t = \frac{\mathrm{CPI}_t - \mathrm{CPI}_{t-1}}{\mathrm{CPI}_{t-1}} \] The index in year \(t\) is the cost of the base-year basket, with quantities \(q_i\), at the prices of year \(t\), divided by its cost at base-year prices, times 100; with expenditure shares \(w_i\) instead of quantities it is the weighted average of the price relatives, which are the ratios \(P_{i,t}/P_{i,0}\) of each price now to its base-year value, so that \(\mathrm{CPI}_t = 100 \sum_i w_i\,(P_{i,t}/P_{i,0})\). The two forms are the same sum written differently, because the share \(w_i\) is \(q_i P_{i,0}\) divided by the cost of the whole basket, and the second form is how statistical offices compute it. The inflation rate \(\pi_t\) is the percentage change of the index, over a year unless stated otherwise. An index of this fixed-basket form is called a Laspeyres index; strictly, when the weights come from a year before the price base, as in every real CPI, it is a Lowe index, but the textbook name is the one in general use.
The unemployment rate and the participation rate
\[ u = \frac{U}{E + U} , \qquad \text{participation rate} = \frac{E + U}{\text{population of working age}} \] The unemployment rate \(u\) is the number of unemployed \(U\) as a share of the labour force \(E + U\), where \(E\) is the number of employed, and not as a share of the population. The participation rate is the share of the working-age population that is in the labour force at all; the United States counts everyone aged 16 and over who is not in an institution or the armed forces, and the OECD usually counts ages 15 to 64.
Growth and the Wealth of Nations
The rule of 70, recalled
\[ \text{doubling time in years} \approx \frac{70}{g} \qquad (g \text{ in per cent per year}) \] Unit 1 derived this rule from \(\ln 2 \approx 0.693\). At 1 per cent a year an income doubles in 70 years, at 2 per cent in 35 years, and at 7 per cent in 10 years; over a working life of 45 years, growth of 1 per cent raises a country’s income by about one half, since \(1.01^{45} = 1.56\), and growth of 3 per cent almost quadruples it, since \(1.03^{45} = 3.78\).
The Cobb-Douglas production function
\[ Y = A\,K^{\alpha}L^{1-\alpha} , \qquad y = A\,k^{\alpha} \quad\text{with } y = \frac{Y}{L},\ k = \frac{K}{L} \] \(Y\) is real GDP, \(K\) the stock of physical capital, \(L\) the number of workers, and \(A\) total factor productivity, the output obtained from given inputs, which measures technology and the efficiency with which the inputs are used. The exponent \(\alpha\) is the share of capital in national income, about one third. The second form divides both sides by \(L\). Since \(L^{1-\alpha}/L = L^{-\alpha}\) and \(K^{\alpha}L^{-\alpha} = (K/L)^{\alpha}\), output per worker \(y\) depends only on \(A\) and on capital per worker \(k\).
Growth accounting and the Solow residual
\[ g_A = g_Y - \alpha\, g_K - (1 - \alpha)\, g_L \] The growth of total factor productivity, the Solow residual, is the growth of output minus the contribution of capital, which is its growth rate weighted by the capital share \(\alpha \approx 1/3\), minus the contribution of labour, weighted by \(1 - \alpha\). It is a residual in the strict sense, and the economist Moses Abramovitz (1912-2000) called it “a measure of our ignorance”: it contains better technology, but also better management, better institutions, and every error in the measurement of \(K\) and \(L\).
The Solow steady state
\[ s\,y^{*} = (\delta + n)\,k^{*} \] Capital per worker stops changing when investment per worker, the saving rate \(s\) times output per worker \(y\), exactly replaces what depreciation (rate \(\delta\)) and population growth (rate \(n\)) take away. In this steady state output per worker is constant, and without technological progress the economy no longer grows per head. A higher saving rate raises \(k^{*}\) and \(y^{*}\) permanently, but it does not raise the long-run growth rate.
Money, Banks, and Central Banks
Present value
\[ PV = \frac{X}{(1 + i)^{t}} \] The present value \(PV\) of a payment \(X\) due in \(t\) years, when money can be lent or borrowed at the interest rate \(i\) per year. A payment far in the future, or a higher interest rate, makes the present value smaller.
The price of a bond
\[ P = \frac{C}{1 + i} + \frac{C}{(1 + i)^{2}} + \cdots + \frac{C + F}{(1 + i)^{T}} \] The price \(P\) of a bond that pays the coupon \(C\) at the end of each of \(T\) years and the face value \(F\) at maturity, when the market interest rate for loans of this length and risk is \(i\). The rate \(i\) that makes the formula true for an observed price is called the bond’s yield.
The balance sheet identity
\[ \text{assets} = \text{deposits} + \text{other liabilities} + \text{equity} \] Every bank’s books balance by construction, because equity is defined as assets minus liabilities. A bank is solvent as long as its equity is positive, that is, as long as its assets are worth more than its debts. It is liquid as long as it can pay what falls due today in cash.
The quantity equation
\[ M V = P Y \qquad\text{and in growth rates}\qquad g_M + g_V = \pi + g_Y \] The money stock \(M\) times its velocity \(V\) equals the price level \(P\) times real output \(Y\). The equation holds by the definition of \(V\). In growth rates, with \(\pi\) the inflation rate and \(g_X\) the growth rate of \(X\), money growth plus the change in velocity equals inflation plus real growth, because growth rates of quantities that are multiplied together add up, approximately, as Unit 7 showed.
The Fisher equation
\[ i \approx r + \pi \] The nominal interest rate \(i\) is approximately the real interest rate \(r\) plus the inflation rate \(\pi\). Lenders who expect inflation demand a nominal rate that covers it, which is why nominal rates are high in countries with high inflation. The approximation is good when both rates are a few per cent; the exact form is in the derivation.
Business Cycles and Stabilization Policy
Okun’s law
\[ u - u_{\mathrm{n}} \approx -\tfrac{1}{2}\,\frac{Y - Y^{*}}{Y^{*}} \qquad\text{and, in changes,}\qquad \Delta u \approx -0.42\,(g - 3.2\unit{\%}) \] \(u\) is the unemployment rate and \(u_{\mathrm{n}}\) its natural rate, both in per cent; \(Y\) is real GDP and \(Y^{*}\) potential output, so the fraction, expressed in per cent, is the output gap. The second form is the line fitted to US annual data from 1949 to 2025 in Figure 2: \(g\) is the growth rate of real GDP and \(\Delta u\) the change of the unemployment rate in percentage points over the same year.
Aggregate demand
\[ Y = C(Y - T) + I(r) + G + NX \] Planned spending is consumption \(C\), which rises with disposable income \(Y - T\) (income after taxes \(T\)), plus investment \(I\), which falls when the real interest rate \(r\) rises, plus government purchases \(G\), plus net exports \(NX\). The equation says what spending depends on; the aggregate demand curve is this total drawn against the price level.
The spending multiplier
\[ \Delta Y = k\,\Delta G , \qquad k = \frac{1}{1 - MPC} \] In the simplest Keynesian cross, with fixed prices, a fixed interest rate, no taxes that rise with income, and no imports, a rise in government purchases \(\Delta G\) raises GDP by the multiplier \(k\) times \(\Delta G\). Every leak out of the spending stream, such as income taxes, imports, and higher interest rates, makes the multiplier smaller; the derivation gives the version with taxes.
The Taylor rule
\[ i = r^{*} + \pi + 0.5\,(\pi - \pi^{*}) + 0.5\,\tilde{y} \] \(i\) is the policy rate, \(\pi\) the inflation rate of the past year, \(\pi^{*}\) the inflation target, \(r^{*}\) the real interest rate that neither stimulates nor restrains the economy, and \(\tilde{y}\) the output gap in per cent. Taylor set \(r^{*} = 2\) and \(\pi^{*} = 2\) in 1993. The rule describes a reasonable policy and the Fed’s historical behaviour; it is not a law the Fed follows.
The expectations-augmented Phillips curve
\[ \pi = \pi^{e} - \beta\,(u - u_{\mathrm{n}}) + v \] Inflation \(\pi\) equals expected inflation \(\pi^{e}\), minus a positive constant \(\beta\) times the excess of unemployment \(u\) over its natural rate \(u_{\mathrm{n}}\), plus a supply shock \(v\) such as an oil price jump. When expectations are anchored, \(\pi^{e}\) is roughly constant and the curve is flat; when they follow actual inflation, the curve shifts every year.
International Trade and Exchange Rates
Opportunity cost and comparative advantage
\[ \text{opportunity cost of a car in wheat} = \frac{\text{hours per car}}{\text{hours per tonne of wheat}} \] A country has a comparative advantage in cars when this ratio is smaller at home than abroad. The ratio is also the slope of the country’s production possibility frontier of Unit 1, which is the number of tonnes of wheat given up for one more car.
The effects of a tariff in a small country
\[ P_{\mathrm{home}} = P_{\mathrm{w}} + t , \qquad \text{revenue} = t \cdot M , \qquad \text{deadweight loss} = \tfrac{1}{2}\, t\, \Delta Q_{\mathrm{s}} + \tfrac{1}{2}\, t\, \Delta Q_{\mathrm{d}} \] A tariff \(t\) raises the home price above the world price \(P_{\mathrm{w}}\) by the full tariff, the government collects \(t\) on the imports \(M\) that remain, and the two triangles of the diagram are lost to everyone: \(\Delta Q_{\mathrm{s}}\) is the extra domestic output made at a cost above the world price, and \(\Delta Q_{\mathrm{d}}\) the consumption given up by buyers who valued it above the world price. The triangle formula holds for straight-line curves. Buyers lose the sum of all four areas.
The balance of payments identity
\[ CA + KA = FA , \qquad\qquad CA = S - I \] The current account balance \(CA\) plus the capital account balance \(KA\) equals the net lending to the rest of the world recorded in the financial account \(FA\) (in the statisticians’ books the two sides differ by a statistical discrepancy, because they are measured from different sources). A country with a current account deficit is a net borrower, and its deficit equals the excess of its investment \(I\) over its national saving \(S\).
The real exchange rate
\[ \epsilon = \frac{E \, P^{*}}{P} \] \(E\) is the nominal rate in home currency per unit of foreign currency, \(P^{*}\) the foreign price level, and \(P\) the home price level. \(E \, P^{*}\) is the price of one foreign basket of goods converted into home currency, and dividing by \(P\) expresses that price in home baskets, so \(\epsilon\) is the number of home baskets that one foreign basket costs. A rise in \(\epsilon\) is a real depreciation, which makes home goods cheaper abroad and foreign goods more expensive at home. With price indices in place of price levels, \(\epsilon\) is an index too, and only its changes have meaning.
Absolute purchasing power parity
\[ E = \frac{P}{P^{*}} \qquad \text{or, equivalently,} \qquad \epsilon = 1 \] Under PPP the exchange rate equals the ratio of the home and foreign price levels, so that a basket costs the same in both countries and the real exchange rate is 1. PPP holds roughly for traded goods over decades and fails badly for whole economies year by year, because many goods and most services are not traded across borders.
Labour Markets and Inequality
Labour demand of a competitive firm
\[ w = P \cdot MP_L = VMP_L \] A firm that takes the price \(P\) of its product and the wage \(w\) as given hires workers until the value of the marginal product of the last worker equals the wage; because \(MP_L\) falls as more workers are hired, the \(VMP_L\) curve slopes down and is the firm’s demand curve for labour. A firm with market power in its product market replaces \(P\) by the marginal revenue of Unit 5, and the curve is then the marginal revenue product \(MRP_L\).
The Mincer earnings equation
\[ \ln w = a + b \cdot \text{schooling} + c \cdot \text{experience} + d \cdot \text{experience}^{2} \] \(w\) is the hourly wage or annual earnings, schooling and experience are measured in years, and the coefficient \(b\) is the return to a year of schooling: a value of \(b = 0.09\) means that one more year of schooling goes with earnings about 9 per cent higher, because a change of 0.09 in a logarithm is a change of about 9 per cent (Unit 1). The coefficient \(c\) is positive and \(d\) is negative, so earnings rise with experience and flatten out, as skills learned on the job accumulate and then stop growing. A review of 1,120 estimates from 139 countries by George Psacharopoulos and Harry Patrinos, published in 2018 in Education Economics, puts the world average of \(b\) at about 9 per cent a year, stable over decades; the usual textbook range is 8 to 10 per cent.
The raw and the adjusted pay gap
\[ \text{raw gap} = \underbrace{\text{part explained by measured differences}}_{\text{hours, occupation, experience, education}} + \underbrace{\text{adjusted gap}}_{\text{unexplained}} \] The raw gap is the percentage difference between the median or mean earnings of women and men. A regression of log earnings on sex and on measured characteristics splits it into the part that goes with differences in those characteristics and the adjusted gap, the difference that remains between women and men with the same measured characteristics. The adjusted gap is not a measure of discrimination, because the characteristics themselves, such as occupation and hours, may be the result of constraints and expectations, and the raw gap is not a measure of unequal pay for equal work.
The Gini coefficient
\[ G = 2A = 1 - 2B, \qquad G = 1 - \sum_{i=1}^{n} (p_i - p_{i-1})\,(L_i + L_{i-1}) \ \text{ for } n \text{ groups} \] \(A\) is the area between the line of equality and the Lorenz curve, and \(B\) the area under the curve; together they fill the triangle below the diagonal, whose area is \(\tfrac{1}{2}\), so \(A + B = \tfrac{1}{2}\). The coefficient \(G\) is \(A\) as a fraction of that triangle, \(G = A / \tfrac{1}{2} = 2A\). \(G\) runs from 0 (everyone has the same income) to 1 (one household has everything). In the formula for grouped data, \(p_i\) is the cumulative population share at the top of group \(i\) and \(L_i\) the cumulative income share, both as fractions; each term is twice the area of the trapezoid under the straight line joining two known points of the curve (a trapezoid of width \(p_i - p_{i-1}\) with parallel sides \(L_{i-1}\) and \(L_i\) has area \(\tfrac{1}{2}(p_i - p_{i-1})(L_i + L_{i-1})\)), so the sum is \(2B\). The true curve bends smoothly and lies below each of these straight segments, so the formula understates \(G\) by a few hundredths when five groups are used.
Absolute income mobility
\[ A = \Pr\!\left(y_{\mathrm{child}} > y_{\mathrm{parents}}\right) \] The share of children whose income \(y_{\mathrm{child}}\) at a given age exceeds their parents’ income \(y_{\mathrm{parents}}\) at the same age, both in prices of the same year, usually household income at about age 30. \(A\) can be high even when relative mobility is low, as long as growth lifts every rank, and it falls when growth slows or reaches only the top of the distribution.
Heuristics, Biases, and Bounded Rationality
Expected value
\[ \mathbb{E}[X] = \sum_i p_i\, x_i \] The expected value of an uncertain outcome \(X\) is the average of its possible values \(x_i\), each weighted by its probability \(p_i\), where the probabilities sum to one. A bet that pays 100 euros with probability 0.4 and nothing otherwise has an expected value of 40 euros. The standard model assumes that the probabilities a person uses are the true ones.
The conjunction rule
\[ P(A \cap B) \le P(A) \] The probability that two events \(A\) and \(B\) both happen, written \(P(A \cap B)\) and read “\(A\) and \(B\)”, can never exceed the probability of either one alone, because the cases in which both happen are a subset of the cases in which \(A\) happens. Adding detail to a story makes it more representative and less probable at the same time.
The anchoring index
\[ \text{anchoring index} = \frac{\text{estimate after the high anchor} - \text{estimate after the low anchor}}{\text{high anchor} - \text{low anchor}} \] The index is 0 when the anchors have no effect and 1 when people just repeat the anchor. Jacowitz and Kahneman found an average of about 0.5 across the fifteen quantities they asked about.
Bayes’ rule
\[ P(H \mid E) = \frac{P(E \mid H)\, P(H)}{P(E \mid H)\, P(H) + P(E \mid \neg H)\, P(\neg H)}, \qquad \frac{P(H \mid E)}{P(\neg H \mid E)} = \frac{P(E \mid H)}{P(E \mid \neg H)} \cdot \frac{P(H)}{P(\neg H)} \] \(H\) is a hypothesis (the person is ill), \(\neg H\) its opposite, and \(E\) the evidence (a positive result). \(P(H)\) is the prior probability before the evidence and \(P(H \mid E)\), read “the probability of \(H\) given \(E\)”, is the posterior probability after it. The second form uses odds, the probability of \(H\) divided by the probability of its opposite, so that a probability of 1 in 1,000 is odds of 1 to 999. It says that the odds after the evidence equal the odds before it multiplied by the likelihood ratio, the probability of the evidence if \(H\) is true divided by its probability if \(H\) is false. In natural frequencies: the posterior probability is the number of true positives divided by the number of all positives.
The Cognitive Reflection Test
The test of Frederick (2005) consists of three questions, each with an answer that springs to mind and a different answer that is correct.
- A bat and a ball cost 1.10 dollars in total. The bat costs 1.00 dollar more than the ball. How much does the ball cost? The intuitive answer is 10 cents, which would make the bat only 90 cents dearer than the ball; the correct answer is 5 cents, since \(x + (x + 1.00) = 1.10\) gives \(x = 0.05\).
- If it takes 5 machines 5 minutes to make 5 widgets, how long would it take 100 machines to make 100 widgets? The intuitive answer is 100 minutes; the correct answer is 5 minutes, because each machine makes one widget in 5 minutes.
- In a lake there is a patch of lily pads. Every day the patch doubles in size. If it takes 48 days for the patch to cover the entire lake, how long would it take for the patch to cover half of the lake? The intuitive answer is 24 days; the correct answer is 47 days, because the patch doubles on the last day.
The score, 0 to 3, measures the disposition to check an intuitive answer rather than knowledge or arithmetic; none of the three needs more than one line of thought, and the point is whether the line gets written.
Prospect Theory and Loss Aversion
Expected utility and the certainty equivalent
\[ \mathbb{E}[u(W)] = \sum_i p_i\, u(w_i), \qquad u(CE) = \mathbb{E}[u(W)], \qquad \text{risk premium} = \mathbb{E}[W] - CE \] A prospect gives wealth \(w_i\) with probability \(p_i\); its expected utility is the average of the utilities \(u(w_i)\) weighted by the probabilities. The certainty equivalent \(CE\) is the sure amount of wealth that the person values exactly as much as the prospect, and the risk premium is how much expected wealth the person gives up to be rid of the risk. The person is risk averse when \(u\) is concave, which means that the curve bends downward because each further euro adds less utility than the one before; the next paragraph shows why a concave \(u\) makes the risk premium positive.
Rabin’s calibration theorem
If an expected utility maximizer with an increasing and concave utility of wealth turns down a 50:50 bet to lose 100 or gain 110 at every amount of wealth, then she turns down a 50:50 bet to lose 1,000 or gain any sum of money, however large. If she turns down a bet to lose 1,000 or gain 1,050 at every wealth, she turns down a bet to lose 20,000 or gain any sum. The result is a theorem, not an estimate, and its message is that aversion to risks of a few hundred euros cannot come from the diminishing marginal utility of wealth.
The prospect-theory value function
\[ v(x) = x^{\alpha} \ \text{ for } x \ge 0, \qquad v(x) = -\lambda\,(-x)^{\beta} \ \text{ for } x < 0 \] \(x\) is a gain or loss relative to the reference point, \(\alpha\) and \(\beta\) measure the curvature for gains and for losses (a power a little below 1 bends the curve the way the square root of Figure 1 does, only less), and \(\lambda\) is the loss aversion coefficient, the number of times a loss weighs more than a gain of the same size. Tversky and Kahneman (1992) estimated \(\alpha = \beta = 0.88\) and \(\lambda = 2.25\) as the medians over their subjects, which is where the rule of thumb that a loss weighs about twice as much as an equal gain comes from.
Decision weights and the prospect-theory valuation
\[ w(p) = \frac{p^{\gamma}}{\left(p^{\gamma} + (1-p)^{\gamma}\right)^{1/\gamma}}, \qquad\qquad V = \sum_i w(p_i)\, v(x_i) \] A prospect with gains or losses \(x_i\) at probabilities \(p_i\) is valued by the sum of the decision weights times the values, and the person chooses the prospect with the larger \(V\). Tversky and Kahneman (1992) estimated \(\gamma = 0.61\) for gains and, with the same formula and the letter \(\delta\) in place of \(\gamma\), \(\delta = 0.69\) for losses; the letter is the paper’s and has nothing to do with the depreciation rate of Unit 8 or the discount factor of Unit 15. The weights do not add up to one, so \(V\) is not an average. The sum shown is the form of the 1979 theory for prospects with at most one gain and one loss; the 1992 version, called cumulative prospect theory, assigns the weights to cumulative probabilities when several outcomes have the same sign, and the two agree for the simple prospects of this unit.
Present Bias, Self-Control, and Nudges
Discounted utility
\[ U_0 = \sum_{t=0}^{T} \delta^{t}\, u(c_t) = u(c_0) + \delta\, u(c_1) + \delta^{2} u(c_2) + \cdots \] A plan of consumption \(c_0, c_1, \ldots, c_T\) is valued today by the sum of the utilities \(u(c_t)\) of each date, each multiplied by \(\delta^{t}\), where \(\delta\) between 0 and 1 is the discount factor per year or per month. The weights \(1, \delta, \delta^{2}, \ldots\) fall geometrically, which is why the model is called exponential discounting. It is the time version of the expected utility of Unit 14, with dates in place of states of the world.
The beta-delta discount function
\[ D(0) = 1 , \qquad D(t) = \beta\,\delta^{t} \quad \text{for } t = 1, 2, 3, \ldots , \qquad 0 < \beta \le 1 \] A reward at the future date \(t\) is weighted by \(D(t)\). The parameter \(\beta\) is the present bias, the extra discount that every future date suffers relative to the present, and \(\delta\) is the long-run discount factor between any two future dates. With \(\beta = 1\) the function is Samuelson’s exponential discounting. The model is also called quasi-hyperbolic discounting. Psychologists had fitted smooth curves called hyperbolas, which fall steeply at first and gently later, to the choices of animals and people, and the beta-delta function copies that shape with a single step at the start followed by a gentle slope.
The preference reversal condition
\[ \beta\delta\, u(y) \;<\; u(x) \;<\; \delta\, u(y) \] In advance, the person plans to pass over a smaller reward \(x\) that comes one step before a larger reward \(y\), and then takes \(x\) after all when it becomes available, exactly when the ratio \(u(x)/u(y)\) lies between \(\beta\delta\) and \(\delta\). The smaller the present bias \(\beta\), the wider the band of choices that reverse.
The participation effect of a default
\[ \text{effect of the default} = s_{\mathrm{out}} - s_{\mathrm{in}} \quad \text{at equal tenure} \] The effect of automatic enrolment on participation is the share of employees in the plan when they must opt out, \(s_{\mathrm{out}}\), minus the share when they must opt in, \(s_{\mathrm{in}}\), with both shares measured at the same length of service, because participation rises with tenure under either rule. In Madrian and Shea’s firm the effect was \(85.9 - 37.4 = 48.5\) percentage points at 3 to 15 months.
The Save More Tomorrow rule
An employee agrees today that her contribution rate will rise by a fixed number of percentage points at each future pay rise, so that take-home pay never falls, and the increases continue automatically until the rate reaches a cap or she opts out (Thaler and Benartzi 2004). The rule uses present bias to defeat present bias: the sacrifice is placed in the future, where \(\beta\) makes it look small, and inertia then keeps it in place.
Behavioural Finance and Asset Bubbles
The dividend discount model
\[ P_0 = \sum_{t=1}^{\infty} \frac{\mathbb{E}[D_t]}{(1+r)^{t}} \qquad\text{and, when dividends grow at a constant rate } g < r,\qquad P_0 = \frac{D_1}{r - g} \] The price \(P_0\) of a share today is the present value of its expected future dividends \(\mathbb{E}[D_t]\), discounted at the rate \(r\) that investors require for holding shares of this risk. If the dividend expected next year is \(D_1\) and dividends are expected to grow by the fraction \(g\) every year, each term of the sum is the previous one multiplied by \((1+g)/(1+r)\), a fraction below one, so the sum is a geometric series, the same one that gave the spending multiplier in Unit 10, and it collapses to the second form; for \(g \ge r\) the series does not converge, so dividends cannot be expected to outgrow the discount rate forever. Both forms state the fundamental value of a share in the standard model.
Shiller’s variance bound
\[ \mathrm{Var}(P_t) \le \mathrm{Var}(P^{*}_t) \] Var, the variance, measures how widely a series spreads around its average. If the price \(P_t\) is the expectation of the perfect-foresight price \(P^{*}_t\) given the information of the day, the price must vary less over time than the perfect-foresight price does, whatever the information set, as long as investors use it rationally and the discount rate is constant.
The cyclically adjusted price-earnings ratio
\[ CAPE = \frac{P}{\overline{E}_{10}} \] The real price of a share or an index divided by the average of its real earnings over the preceding ten years, all in the prices of the same date. For the US market its mean from 1881 to 2024 is 17.5; a high value means that investors pay a lot for each dollar of normal earnings.
The disposition effect (Odean’s measure)
\[ PGR = \frac{\text{realized gains}}{\text{realized gains} + \text{paper gains}} , \qquad PLR = \frac{\text{realized losses}}{\text{realized losses} + \text{paper losses}} \] The proportion of gains realized, \(PGR\), is the share of all winning positions that were sold on days when the account sold something, and \(PLR\) is the same share for losing positions; without a disposition effect \(PGR \approx PLR\). In Odean’s accounts \(PGR\) was 0.148 and \(PLR\) was 0.098, so a winning share was about 50 per cent more likely to be sold than a losing one.
The equity premium
\[ \text{equity premium} = \overline{r}_{\mathrm{shares}} - \overline{r}_{\mathrm{bonds}} \] The average return of shares over a long interval minus the average return of government bonds or bills over the same interval, both in real terms. In the United States it has been about five to six percentage points a year since 1900, depending on the interval and on whether the average is arithmetic or geometric.