A.Friedman ICEF-2022 LECTURE NOTES for INTERMEDIATE MICROECONOMICS Contents 1. CONSUMER’S THEORY ............................................................................................. 4 1.1 Budget constraint.......................................................................................................... 4 1.2 Preferences ................................................................................................................... 7 1.3 Utility ........................................................................................................................... 8 1.4 Consumer’s choice ..................................................................................................... 10 1.5 Comparative statics of Marshallian demand .............................................................. 12 1.6 Expenditure minimization problem and duality in consumption ............................... 15 1.7 Slutsky decomposition and Slutsky equation ............................................................. 17 1.8 Alternative approach to consumer’s theory: revealed preferences ............................ 19 1.9 Slutsky substitution effect .......................................................................................... 20 1.10 Measuring changes in consumer’s welfare .............................................................. 22 1.11 Price indices (optional)............................................................................................. 28 1.12 Sample exercise with solution .................................................................................. 30 2. CHOICE UNDER IN-KIND INCOME ....................................................................... 31 2.1 In-kind income ........................................................................................................... 31 2.2 Consumption-leisure model (Individual labour supply) ............................................ 33 2.3 Consumption choices over time: intertemporal choice .............................................. 35 2.4 Production and consumption over time (optional) ..................................................... 39 2.5 Applications of NPV rule: exhaustible resources (optional) ...................................... 41 2.6 Sample exercise with solution .................................................................................... 43 3. CHOICE UNDER UNCERTAINTY ........................................................................... 45 3.1 Gambles and contingent commodities ....................................................................... 45 3.2 Expected utility .......................................................................................................... 50 3.3 Willingness to pay to avoid risk ................................................................................. 52 Application 1. Obtaining additional information ............................................................. 53 Application 2. Demand for Insurance .............................................................................. 54 3.4 The Arrow–Pratt measure of risk aversion ................................................................ 57 3.5 Reducing risk via diversification................................................................................ 58 3.6 Sample exercise with solution .................................................................................... 59 4. GAME THEORY ......................................................................................................... 62 4.1 Simultaneous-move games ......................................................................................... 62 4.2 Sequential-move or extensive-form games ................................................................ 66 4.3 Repeated games .......................................................................................................... 70 1 A.Friedman ICEF-2022 4.4 Sample exercise with solution .................................................................................... 73 5. THE FIRM ................................................................................................................... 75 5.1 Modeling the firm’s technological opportunities ....................................................... 75 5.2 Profit maximization and Cost minimization .............................................................. 77 5.3 Cost minimization ...................................................................................................... 77 5.4 Profit maximization in case of perfect competition ................................................... 85 5.5 Sample exercise with solution .................................................................................... 86 6. PERFECT COMPETITION ......................................................................................... 89 6.1 Perfect competition .................................................................................................... 89 6.2 Equilibrium and efficiency ......................................................................................... 93 Application 1: per unit tax analysis .................................................................................. 96 Application 2: price ceiling .............................................................................................. 98 Application 3: price support program .............................................................................. 98 6.3 Sample exercise with solution .................................................................................... 99 7. GENERAL EQUILIBRIUM AND WELFARE ECONOMICS ................................ 101 7.1 General equilibrium in exchange economy .............................................................. 101 7.2 Pareto optimum in exchange economy .................................................................... 105 7.3 Welfare theorems for exchange economy ................................................................ 108 7.4 Production economy ................................................................................................. 110 7.5 Pareto efficiency in economy with production ........................................................ 114 7.6 Sample exercise with solution .................................................................................. 118 8. MONOPOLY ............................................................................................................. 120 8.1 Pure monopoly ......................................................................................................... 120 8.2 Sources of monopoly and regulatory responses: ...................................................... 122 8.3 Monopolistic price discrimination ........................................................................... 126 8.4 Sample problem with solution.................................................................................. 136 9. OLIGOPOLY ............................................................................................................. 138 9.2 The Stackelberg model ............................................................................................. 140 9.3 Price-setting oligopolists: Bertrand model with homogenous good ........................ 142 9.4 Price leadership or Dominant firm model (optional) ............................................... 143 9.5 Repeated interactions ............................................................................................... 144 9.6 Bertrand model with differentiated goods ................................................................ 147 9.7 Sample exercise with solution .................................................................................. 149 10. FACTOR MARKETS .............................................................................................. 151 10.1 Demand for factors ................................................................................................. 151 10.2 The supply of factors and competitive equilibrium ............................................... 155 10.3 Monopsony and monopoly in factor markets ......................................................... 156 11. ASYMMETRIC INFORMATION .......................................................................... 159 2 A.Friedman ICEF-2022 11.1 Types of asymmetric information problems .......................................................... 159 11.2 Adverse selection and the market for lemons ........................................................ 159 11.3 Adverse selection at insurance market ................................................................... 162 11.4 Private and Government Response to Adverse Selection Problem........................ 165 11.5 Spence model of job market signaling ................................................................... 166 11.6 Screening ................................................................................................................ 168 11.7 Sample exercise with solution ................................................................................ 177 12. EXTERNALITIES AND PUBLIC GOODS ........................................................... 181 12.1 Simple Model of Consumption Externalities ......................................................... 181 12.3 Government Regulation ......................................................................................... 185 12.4 Efficient Provision of Public Good ........................................................................ 186 12.5 Private Provision of Public Good ........................................................................... 187 12.6 The Commons Problem .......................................................................................... 189 12.7 Sample exercise with solution ................................................................................ 191 3 A.Friedman ICEF-2022 1. CONSUMER’S THEORY Economists assume that consumers choose the best bundle of goods they can afford. Thus we have to describe more precisely what we mean by the “best” and what we mean by “can afford”. We start with the concept of affordable bundle. 1.1 Budget constraint Bundle is a vector of commodities. In an economy with N commodities bundle is described by vector x x1 , x 2 ,, x N , where x i stays for the quantity of good i . Each commodity can be consumed only in nonnegative amount, i.e. x i 0 . Key assumption. Consumer is a price taker, i.e. price per unit of a commodity is not affected by the number of units purchased. Denote per unit price of commodity i by pi , then p p1 , p2 ,, p N is a price vector. If consumer purchases bundle x at prices given by p, then he spends p1 x1 p2 x 2 p N x N or simply px . Suppose that consumer has money income M , then he can afford all bundles that cost no more than M . We call the set of all affordable consumption bundles at prices p p1 , p2 ,, p N and income M the budget set of the consumer. Budget set contains bundles x 0 that satisfy the following constraint p1 x1 p2 x 2 p N x N M , which is called budget constraint. In case of two goods we can illustrate budget set graphically using budget line. x2 M p2 Budget line slope p1 p2 Budget set 0 M p1 x1 4 A.Friedman ICEF-2022 Budget line is the set of bundles that cost exactly x 2 ( M p1 x1 ) / p2 , then the slope of budget line is: M: p1 x1 p2 x 2 M . As dx 2 p 1 . dx1 p2 Note: 1) vertical and horizontal intercepts represent bundles in which only one of the commodities is consumed; 2) the slope is equal to negative of the price ratio and reflects the opportunity cost of consuming good 1 (in order to consume more of good 1 you have to give up some consumption of good 2). Change in the price of good 1 results in rotation of budget line. x2 ~ p1 p1 M p2 ~ p1 p2 p1 p2 M ~ p1 0 M p1 x1 Change in income brings a parallel shift of budget line. ~ MM x2 M p2 p1 p2 ~ M p2 0 ~ M p1 M p1 x1 5 A.Friedman ICEF-2022 Non-linear budget constraints Quantity discount x2 100 4 M 100, p2 1 p1 4, x1 15 , and p1 2 for each additional unit 2 0 15 35 x1 In-kind transfer of commodity 1, equal to x1 x2 x1 M / p2 M p1 M x1 p1 x1 The numeriare Budget set is not affected if all prices and income change proportionally. We can divide all p M ~ prices and income by the price of the second good: 1 x1 x 2 or ~ p1 x1 x 2 M . In p2 p2 this case we use second good as a numeriare. 2-good case There are more than 2 commodities, but we can treat good 2 as a composite commodity with price equal 1. Then good 2 represents the amount of all other (than good 1) commodities that agent can purchase by spending $1. Notation: AOG 6 A.Friedman ICEF-2022 1.2 Preferences Preferences are defined over bundles, not goods. Bundle is a vector of commodities: x x1 , x 2 ,, x N Notation: weak preference relationship ‘bundle x is at least as good as bundle y’: x y strict preference relationship: x y x x y but not y indifference relationship x ~ y x y and y x Assumptions Completeness: consumer can compare any two bundles and tell, which one he/she prefers or whether he/she is indifferent between them: for any two bundles x and y we have that x y or y x or both. Transitivity: for any three bundles x, y, z such that x y and y z we have x z The set of all bundles that are indifferent to a given bundle is called indifference curve: IC x x : x ~ x Implication of transitivity: different indifferent curves do not intersect (prove!). Non-satiation: more is better x2 These bundles are x preferred to ~ ~ x ~ x is preferred to these bundles x1 With non-satiation assumption, indifference curves cannot slope upward. We will always assume completeness, transitivity and non-satiation if the opposite is not postulated explicitly. In the second year course we also used the assumption of diminishing MRS but we are not going to use it as a default assumption this year. 7 A.Friedman ICEF-2022 Marginal rate of substitution of good 1 for good 2 MRS12 (i.e. putting of good 2 in place of good 1) is the maximum amount of good 2 a person is willing to give up to obtain one additional unit of good 1 or (in the limit) it is the rate at which commodity 2 must decrease as commodity 1 increases by an infinitesimal amount to keep the individual on the same indifference curve. Suppose that indifference curve (IC) is described by the function x 2 f x1 . Then MRS 12 is equal to the absolute value of the slope of IC x 2 x1 0 x 1 MRS 12 x lim lim x x1 0 f x1 x1 f x1 f x1 x . x1 x Implication of diminishing MRS: convex to the origin indifference curves. x2 x 2 f x1 Diminishing MRS12 x MRS12 x x1 Examples of preferences Perfect substitutes: goods that can be substituted for each other at a constant rate Perfect complements: goods that have to be consumed in fixed proportions “Bads” (violates the assumption of non-satiation) 1.3 Utility Utility function is a function that assigns a number to every possible consumption bundle such that more preferred bundles get assigned larger numbers than less preferred bundles and vice versa, larger numbers are assigned to more preferred bundles. Denote this function by u . Note: utility function allows ranking bundles by their amount of utility but it does not allow precise comparisons of how various bundles are valued relative to each other. Such function is called ordinal as it only orders the bundles. 8 A.Friedman ICEF-2022 As a result of its ordinal nature utility function is not unique. We can multiply utility by some positive number and get different utility function that represents the same preferences. In fact, we can use any positive monotonic transformation. This can be justified in the following way. Suppose that u is a utility function that represents preferences of some consumer and gu is some arbitrary increasing function. Let us demonstrate that v g u represents the same preferences as u . Consider two arbitrary consumption bundles x and y . As u is a utility function that represents preferences of considered consumer then x y iff ux uy . Since gu is increasing then ux uy implies that gux guy , which means that vx vy . Now, let us go in the opposite direction. Suppose that vx gux guy vy . Is it possible that ux uy ? If ux uy then gux guy as gu is increasing, and we get a contradiction. Thus ux uy iff vx vy , that is v gu represents the same preferences as u if gu is increasing. Construction of utility function by assigning numbers to ICs x2 ue x u ~x e ~ x ux e 0 45 ux u ~x x1 Examples of utility functions Perfect substitutes: ux1 , x 2 x1 x 2 Perfect complements: ux1 , x 2 minx1 , x 2 Commodity 2 is a “Bad”: ux1 , x 2 x1 x 2 Cobb-Douglas utility function: ux1 , x 2 x1 x 2 Quasiliniar utility function: ux1 , x 2 x1 x 2 9 A.Friedman ICEF-2022 How to calculate MRS Marginal utility of commodity i - change in total utility due to an increase in consumption of u this commodity by (an infinitely small) additional unit: MU i . x i This concept allows calculating MRS as a ratio of marginal utilities. We start with an indifference curve ux1 , x 2 u and take a full differential u1 x1 , x 2 dx1 u2 x1 , x 2 dx 2 du 0 . Rearranging we get an expression for MRS: MRS 12 x dx 2 dx1 u ux / x1 MU 1 . ux / x 2 MU 2 Question. Check that MRS for gu is the same as MRS for u if g 0 . Comment on MRS and convexity Diminishing MRS (under nonsatiation) implies convex to the origin indifference curves. Definition: S is a convex set if for any a and b from S a convex combination a 1 b is also in the set S for any 0 1. Preferences are convex if for any level of utility u the set of weakly preferred bundles x : ux u is convex (see the left panel). The right panel illustrates the case of nonconvex preferences. x2 x2 x1 x1 1.4 Consumer’s choice Utility maximization problem m ax ux1 , , x N xi 0 p1 x1 p2 x 2 p N x N M 10 A.Friedman ICEF-2022 Solution of this problem is called x p, M - Marshallian (ordinary) demand and the value of the problem v p, M ux p, M is called indirect utility function. Note: non-satiation implies that income is always exhausted, i.e. budget constraint is satisfied as equality (prove it!). To solve the problem we setup a Lagrangean: ℒ ux1, , x N M p1 x1 p2 x 2 pN x N FOCs for interior solution: ℒ/xi = ux x i pi 0 ux x i pi ℒ/ = M p1 x1 p2 x 2 pN x N 0 MRS ij ux x i pi p i ux x j p j p j Graphical solution for N=2 x2 Interior optimum p1 x1 p2 x 2 M , MRS 12 x p1 / p2 x1 0, x 2 0. Increase in utility x 0 x1 Question. Consider any bundle with x1 0, x 2 0 on budget line such that MRS 12 p1 / p2 . Explain, why this bundle is not optimal. Other interpretation of interior optimum condition: marginal utility per dollar spent has to be the same for all goods: MU 1 x MU 2 x . p1 p2 FOCs are not sufficient to guarantee a maximum: bundles A and B both satisfy the FOC but only bundle B gives the maximum utility while A violates second order condition (SOC). Thus we should check SOC. If non-satiated preferences satisfy diminishing MRS then SOC will hold automatically. 11 A.Friedman ICEF-2022 x2 B A x1 Corner solutions. x2 x x2 MRS 12 x MRS 12 x 0 p1 p2 x1 (a) x1 0, x 2 0 и MRS 12 x 0 p1 p2 p1 p2 x x1 (b) x1 0, x 2 0 и MRS 12 x p1 p2 1.5 Comparative statics of Marshallian demand Comparative statics - comparison of two equilibria. Marshallian (or ordinary) demand functions x1 p1 , p2 , M and x1 p1 , p2 , M we derive from utility maximization problem. Income change Normal good - a good for which an increase in income increases consumption ceteris paribus. Inferior good - a good for which an increase in income decreases consumption ceteris paribus. Neutral good - a good for which an increase in income does not affect consumption ceteris paribus. 12 A.Friedman ICEF-2022 Income elasticity of demand- the percentage change in quantity demanded with respect to a percentage change in income: XM lim M 0 X M X / X M M / M X . For normal goods XM 0 and for inferior goods XM 0 . Income consumption curve x2 x1 0 Own price changes: derivation of individual’s demand curve x2 M p2 Price-consumption curve 0 p3 1 p2 p2 1 p2 p11 p2 p10 p2 x1 p1 p13 p12 p11 p10 Demand curve for good 1 x1 13 A.Friedman ICEF-2022 Ordinary good - a good for which an increase in its price decreases consumption ceteris paribus (individual demand curve is downward sloping) Giffen good - a good for which an increase in its price results in an increase in consumption ceteris paribus (individual demand curve is upward sloping) x2 Giffen good 0 x1 Price elasticity of demand - the percentage change in quantity demanded of good X with respect to a percentage change in its price: Xp lim p 0 x x X p x px X / X px / px X . X X For ordinary good pX 0 and for Giffen good pX 0 . Cross price changes (impact of a change in the price of one good on the quantity demanded of another good) Substitutes (goods that satisfy similar wants): an increase in the price of one good leads to an increase in the quantity demanded of the other good. Complements (goods that tend to be used together): an increase in the price of one good leads to a decrease in the quantity demanded of the other good. Unrelated goods: an increase in the price of one good has no impact on the quantity demanded of other good. Cross price elasticity of demand - the percentage change in quantity demanded of good X with respect to a percentage change in price of good Y : Xp lim p 0 x Y X pY pY X / X pY / pY X . X X X Substitutes pY 0 , complements pY 0 , unrelated goods pY 0 . 14 A.Friedman ICEF-2022 1.6 Expenditure minimization problem and duality in consumption Alternative objective of the consumer: attain a given level of satisfaction in a cheapest way. Expenditure minimization problem: m in p1 x1 p2 x 2 x i 0 u x 1 , x 2 u Solving this problem we get compensated (or Hicksian) demand x p, u . In some textbooks compensated demand is denoted by h p, u . Plugging it into the objective function we obtain the corresponding expenditure function e p, u px p, u . Graphical derivation of compensated demand. Start with identifying the set of bundles that give desired utility, i.e. draw the desired IC Illustrate iso-expenditure lines: p1 x1 p2 x 2 =const Find the point that lies on IC and on the lowest iso-expenditure line: u x1 , x2 u , for interior solution ( x1 0, x2 0 ) we have MRS12 x p1 / p2 Note: compensated demand curve reflects only SE, while ordinary demand curve shows both SE and IE. 15 A.Friedman ICEF-2022 p1 p1 Negative IE Positive IE ~ p ~ p x1 p, u x1 p, M x1 p, u 0 x1 p, M x1 (а) normal good 0 x1 (b) inferior (but not Giffen) good Duality in consumption 1. From utility-maximization (UMP) to expenditure minimization (EMP) Let x solves UMP under p, M and u ux v p, M . Consider EMP under p and u . If x solves UMP under p, M then x solves EMP under p, u and ep, u M . 2. From expenditure minimization (EMP) to utility-maximization (UMP) Now, let us start with expenditure minimization problem. Suppose that x solves EMP under p, u . Let us fix the consumer’s income M e p, u and consider utility maximization under p, M . Then x solves UMP under p, M and ux u . 16 A.Friedman ICEF-2022 Summary of duality results: x p, M x p, v p, M (1) M e p, v p, M (2) x p, u x p, e p, u (3) u v p, e p, u (4) 1.7 Slutsky decomposition and Slutsky equation Price change Substitution effect the effect of a price change on quantity demanded due exclusively to the fact that its relative price has changed Income effect the effect of a price change on quantity demanded due exclusively to the fact that the consumer’s real income has changed Under Hickes approach constant real income means that consumer can attain the same indifference curve, i.e. keeps utility constant. Slutsky decomposition: x i x iSE x iIE x ih x i0 x i x ih 17 A.Friedman ICEF-2022 x i x iSE x iIE . Relative changes: p1 p1 p1 Sign of Hicksian SE: x1SE 0. p1 Derivation of Slutsky equation Consider the change in quantity of good i demanded due to the change of good j price. Differentiate (3) with respect to p j : x i p, u x i p, M x i p, M e p, u . p j p j M p j Expenditure will increase proportionally to the quantity of the good consumed (This result is e p, u x j p, u known as Shephard’s lemma and follows from the envelope theorem): p j and due to (3) x j p, u x j p, e p, u x j p, M , where M e p, u . Plug back and rearrange: x i p, M x i p, u x p, M x j p, M i . p j p j M IE SE Own price version of Slutsky equation: Own SE can never be positive: x i p, M x i p, u x p, M x i p, M i pi p M i IE own SE x i p, u 0 . Prove it! pi Substitution effect is zero in case of kinked IC (for example, in case of Leontieff preferences) Question: illustrate Slutsky decomposition for the case of Leontieff preferences. Sign of IE depends on the nature of the good (normal or inferior). Summary: impact of an increase in the price of a good Type of good SE IE TE=SE+IE Normal Inferior but not Giffen + Giffen + + Question: illustrate Slutsky decomposition for the case of Giffen good 18 A.Friedman ICEF-2022 1.8 Alternative approach to consumer’s theory: revealed preferences Idea: if two bundles x and ~ x are affordable but only x is chosen, then x is revealed ~ preferred to x . Thus, if bundle x x1, x2 is chosen under p1 , p2 , M then it is revealed preferred to any bundle from the budget set x 0 : p1x1 p2 x2 M x2 The weak axiom of revealed preference (WARP) If bundle x is revealed preferred to ~x and the two bundles are not the same, then it cannot happen that ~x is revealed preferred to x . x2 M p2 ~ M ~ p2 0 Violation of WARP ~ x~ p, M x p, M M p1 ~ M ~ p1 x1 ~ x x , then ~ Check WARP: if p1 ~ x1 p2 ~ x 2 M and ~ p1 x1 ~ p2 x 2 M for any p, M and ~ ~ p, M . Application of revealed preferences Replacement of per unit subsidy on good 1 by the lump sum subsidy of equivalent money cost 19 A.Friedman ICEF-2022 x2 M sx1 p2 x p, M sx1 M p2 1 sx p2 0 x M p1 M p1 s x1 Question: prove that the vertical red segment gives the cost of subsidy in terms of good 2. 1.9 Slutsky substitution effect Under Slutsky approach constant real income means that consumer can afford the initial bundle. Thus we pivot the budget line around the original choice. Slutsky decomposition with Slutsky SE: x i x iSE x iIE x icomp x i0 x i x icomp Slutsky substitution effect is derived under the assumption that real income stays constant in a sense that the old optimal bundle is still affordable under new prices: M comp px 0 or in our example M comp M p1 x10 . Observed response= x i x i x i0 Compensated response - the change in quantity demanded resulting from changing the price while simultaneously compensating the individual with income= x iSE x icomp x i0 x i x iSE x iIE . Relative changes: p1 p1 p1 20 A.Friedman ICEF-2022 x2 M comp p2 M p2 x comp x x 0 p1 p2 0 p10 p2 x1 x IE 1 x SE 1 Sign of SE Own substitution effect is always non-positive x1SE 0 (from revealed preferences). p1 Sign of IE x iIE x i x icomp . p1 p1 21 A.Friedman ICEF-2022 Note: M M M comp p1 x10 or p1 IE x iIE M 0 x i x . As . Plug back: 1 p1 M x10 x10 0 , then x iIE x iIE 0 if 0 , i.e. if good i is inferior, p1 M x iIE x iIE 0 if 0 , i.e. if good i is normal p1 M x iIE x iIE 0 if 0 , i.e. if good i is neutral with respect to income p1 M Own price Slutsky decomposition x1 x1SE x IE x10 1 . If p 0 , then Hicksian SE = Slutsky SE. p1 p1 M 1.10 Measuring changes in consumer’s welfare Demand curve as a marginal valuation schedule Consider a quasi-linear utility function ux1 , x 2 vx1 x 2 , where v 0 and v 0 , v0 0 . Let p2 1 . In case of interior solution MRS 12 x v1 p1 or vx1 p1 . 1 1 $ Inverse demand function Ordinary (Marshallian) consumer surplus (CS) – the difference between what the consumer is willing to pay and what he has to pay. CS gross x10 p1 x1 dx1 vx1 dx1 v x10 v0 v x10 as v0 0 . x10 x10 0 0 22 A.Friedman ICEF-2022 $ Inverse demand function CS x10 CS gross x10 p10 x10 v x10 p10 x10 u x10 , M p10 x10 M as x 2 M p1 x1 . Note: the derived equality between CS and utility adjusted for the income is correct for the quasilinear preferences, do not apply it if preferences are different. CS is calculated as the area below Marshallian inverse demand function above the market p x p dx or as an area to the left from Marshallian demand CS x x p dp , where p is the cut-off price at which the quantity price CS x10 x10 0 function 0 1 1 0 1 1 1 p p10 1 1 1 demanded becomes equal to zero. Application of CS Suppose the price of good 1 goes up from p10 to p1 . What is the resulting change in consumer’s welfare? It equals to the change in consumer’s surplus CS CS x1 CS x10 ux1 , M p1 x1 M u x10 , M p10 x10 M ux1 , M p1 x1 u x10 , M p10 x10 . 23 A.Friedman ICEF-2022 $ CS after price increase Initial CS Reduction in CS due to the price increase Problems with the concept of CS In the presence of income effect CS is only an approximate measure of consumer welfare. Reason: demand curve shows the relationship b/w price and quantity demanded holding other things fixed, including money income. But with fixed money income the value that an individual puts on an additional unit of a good may depend on the amount that he has already spent on previous units of the good. As a result, price is not identical to consumer’s marginal valuation of the associated unit of output. Other measures of consumer’s welfare Compensating variation (CV) - the change in money income just necessary to offset the change in utility induced by the price change By definition vp0 , M u0 v p, M CV . From duality: M CV ep, u0 and M e p 0 , u 0 , which implies CV e p, u0 e p0 , u0 . 24 A.Friedman ICEF-2022 CV is measured at the new prices. Equivalent variation (EV) - the change in money income that is equivalent in its effect on the individual’s utility to a change in the price of a commodity. M e p, u , which implies EV e p, u ep , u. By definition: v p, M u v p0 , M EV . From duality: M EV e p0 , u and 0 EV is measured at the initial prices. 25 A.Friedman ICEF-2022 Relationships b/w EV, CV, CS EV and CV can be represented as areas bound by the compensated demand curves. To measure CV we need a schedule that shows how the quantity demanded varies with price, assuming that as price changes, the consumer’s money income is adjusted to keep him at initial level of utility. This is a compensated demand curve that reflects only Hicks SE. Inverse compensated demand function also can be interpreted as marginal valuation: it gives marginal valuation for arbitrary preferences (not only quasi-linear one) as marginal valuation of each dollar is not affected by the price. Let’s prove that EV and CV can be represented as areas bound by the compensated demand curves. We start with writing down Shephard’s lemma for good 1 and initial utility level: e( p, u 0 ) x1 ( p, u 0 ) . p1 By integrating this function with respect to good 1 price from p10 to p1 , we can find p1 x1 ( p1 , p 2 , u )dp1 0 p10 p1 e( p1 , p 2 , u0 ) dp1 e( p1 , p2 , u0 ) e( p10 , p2 , u0 ) CV . 0 p 1 p 1 Thus compensating variation can be represented as the area under corresponding compensated demand curve x1( p1 , p 2 , u0 ) between initial and new prices. Similarly, from Shephard’s lemma with new utility level we get expression for equivalent variation: p1 x ( p ,p 1 p10 1 2 , u)dp1 p1 e( p1 , p 2 , u) dp1 e( p1 , p2 , u) e( p10 , p2 , u) EV p 1 p0 1 Thus equivalent variation is represented by the area under another compensated demand curve that corresponds to the new level of utility x1 ( p1 , p 2 , u) . 26 A.Friedman ICEF-2022 EV CS CV Finally, we can compare the three measures for the change in consumer’s welfare, resulting from the considered price increase: EV CS CV , where CS - is the corresponding change in consumer surplus. We can explain this result intuitively. Consider the relationship between CV and CS . Price increase reduces the purchasing power of income (real income), which in its turn leads to a fall in demand for normal good. Due to the negative income effect Marshallian demand (that includes income effect) lies to the left from corresponding compensated demand curve (that reflects only substitution effect) for any price above p10 . As a result the area bounded by Marshallian demand is smaller than area bounded by compensated demand curve x1( p, u0 ) . Implication: Marshallian CS is only an approximation of the true CS measured as an area below the compensated demand curve. Question. Why Marshallian CS is widely used? If price of normal good goes up then EV< CS <CV. 27 A.Friedman ICEF-2022 If price of inferior good goes up then EV> CS >CV. For the case of neutral good EV= CS =CV as Marshallian and compensated demand curves coincide. 1.11 Price indices (optional) Ideally, the changes in the cost of living would be measured by the change in money income that is necessary for the consumer to achieve the same level of utility in the given year as in the base year (index based on CV) Then, if the consumer’s money income increases more (less) then this measure of the cost of living we can infer that he is better (worse) off. As compensated demand is not observable this ideal measure of cost of living (ICLI) cannot be used. Instead we use some approximations: Laspeyras price index (LPI) and Paashe price index (PPI). LPI - the ratio of the sum of given year prices weighted by the base year quantities to the sum N of base year prices weighted by the base year quantities: LPI t p x 0 i p 0 i t i i 1 N i 1 . 0 i x x2 px 0 / p20 u LPI u0 M LPI p 20 M ICLI p 20 x p, u 0 M 0 / p20 x x p ,u 0 x 0 0 p1 p 20 0 p10 p 20 x1 Using base year quantities to weight the prices of goods in a different year, LPI does not allow for the fact that a consumer tends to substitute away from goods that become relatively expensive. It implies that an individual whose income is indexed in accordance with LPI can 28 A.Friedman ICEF-2022 purchase base year bundle, i.e. he can never be worse off. Moreover he could be better off by substituting away from relatively expensive goods. Thus agent might be better off even if his income increases slightly less then LPI. Conclusion: LPI overstates increases in the true cost of living. Paashe price index (PPI) - the ratio of the sum of given year prices weighted by the given year quantities to the sum of base year prices weighted by the given year quantities: N PPI t p x t i t i p 0 i t i i 1 N i 1 . x PPI understates increases in the true cost of living. PPI gives a minimum estimate of the increase in the TCL since it assumes (erroneously) that N had the consumer received in the base year an amount of income equal to p i 1 0 i x it he would choose given-year bundle. Instead the consumer would tend to buy relatively more of the commodities which in base year were cheaper than in given year. This implies that for an individual whose income is indexed in accordance with PPI, current year bundle was affordable in the base year. Thus he is never better off. Moreover an agent could be worse off even if his income rises a bit more than PPI. x2 px px p, u p2 p2 p0 x p2 p 0 x p 0 , u p2 u0 u p1 p2 x0 x x p 0 , u p10 p2 u PPI 0 x1 What determines the magnitudes of the errors in the LPI and PPI? the extent to which relative prices change, the extent to which the consumer substitutes b/w the commodities when relative prices do change, individual versus representative agents (average bundle) index. 29 A.Friedman ICEF-2022 1.12 Sample exercise with solution Bob uses his monthly income (M) to pay for water services and all other goods (represented by a composite commodity). The price of the water services is p per m3, and the price of composite commodity is 1. Bob’s preferences are represented by differentiable utility function. The local water company cannot cover its cost and considers two options to solve the problem. It could raise the price by 10%. In this case Bob’s utility level is reduced from u 0 to u 1 . Alternatively, the water company may keep per unit price constant but in addition introduce fixed per month charge that results for Bob exactly in the same utility loss. Which scheme brings more revenue to the water company? Which scheme results in greater water conservation? Provide graphical and analytical solution. Solution Graphical solution Let x stays for water consumption and y -for AOG. Revenue of water company is given by the sum of revenue from sales (price multiplied by quantity) and fixed charge. As his income is the same, then water expenditure equals M y . Graphically we compare TR1 p1 x p1 , M M y p1 , M TR2 F p 0 x p 0 , M F M y p 0 , M F . and From the graph we get TR1 TR 2 and x x x x . Thus the second scheme brings more revenue but the first scheme provides greater water conservation. 0 1 0 2 y BC1 TR1 1 u0 u BC2 TR2 BC-initial x1 x2 x0 x Algebraic solution As bundles x 1 , y1 and x 2 , y 2 provide the same utility, then the change in quantity demanded is due to Hicksian SE only. We know that own SE is nonpositive (proof to be provided at class). As relative price goes up when we proceed from x 2 to x 1 and ICs are smooth (due to differentiability of utility function) then x SE x 1 x 2 0 . Thus x 1 x 2 , which means that water conservation is higher under the first scheme. Due to nonsatiation with lower consumption of x we can have the same utility only with increased consumption of y : ux 1 , y1 ux 2 , y 2 and x1 x 2 implies y1 y 2 . Thus TR1 M y1 M y 2 TR2 . 30 A.Friedman ICEF-2022 2. CHOICE UNDER IN-KIND INCOME 2.1 In-kind income Let us suppose that instead of money income consumer is endowed with commodities bundle: x1 , x 2 , i.e. he owns x 1 of the first good and x2 units of the second one. This bundle is called initial endowment and it is always on the budget line as it can be consumed without market trade. If he sells this bundle at market prices, he gets money income M p1 , p2 p1 x1 p2 x 2 . Note: now income depends on prices. x2 x2 p1 x1 p2 Agent sells good 1 Initial endowment x2 Agent buys good 1 p1 p2 x1 0 x1 p x x1 2 2 p1 The effect of price change on budget set in case of in-kind income x2 First good price increase Become affordable x2 p1 x1 p2 x2 Become unaffordable p1 p2 0 x1 p1 p2 p x x1 2 2 p1 x1 Utility maximization problem with in-kind income 31 A.Friedman ICEF-2022 m ax ux1 , x 2 ,, x N xi 0 N N p x p x . i 1 i i i 1 i i If x - solution of UMP under price vector p , then we say that consumer is a net buyer of good i if x i x i , consumer is said to be a net seller of good i if x i x i . As income depends on prices, the income effect is different. On the one hand, the monetary income changes as the same endowment will generate higher monetary income under increased prices. On the other hand, if we fix the monetary income, its purchasing power is reduced. Slutsky equation with in-kind income Differentiate demand for good i with respect to p j : dx i p, px x i p, M x i p, M px . dp j p j M p j Using Slutsky equation for fixed monetary income and rearranging we get dx i p, px x i p, u x p, M x i p, M xj i xj dp j p j M M x i p, u x p, M x j x j i p M j Total IE SE Own-price version: dx i p, px x i p, u x p, M x i x i i . dpi pi M Conclusion: in case of in-kind income the sign of IE depends on the type of the good (normal/inferior) and on the type of the agent (net buyer/net seller). The following table summarizes the signs of income effect and the overall effect of the own price change taking into account that own substitution effect can never be positive. Good type Agent type Normal good Inferior good Net buyer Net seller Net buyer Net seller Income effect + + Total effect +/ +/ 32 A.Friedman ICEF-2022 2.2 Consumption-leisure model (Individual labour supply) One of the applications of the model with in-kind income deals with labour supply. Assume that we have only two commodities: leisure l and aggregate consumption c . Initial endowment is given by T , C . Time endowment is divided between leisure l and labour L : l L T . Denoting the prices by w - wage rate (price of leisure) and p -price of consumption good we get the following budget set: pc wl pC wT and 0 l L , c 0 . Assuming that preferences are represented by utility function uc, l that increases in both c and l , we can find the consumer’s choice from the following utility maximization problem: max uc, l s.t. pc wl pC wT 0 l T, c 0 . Interior solution: MRS lc w / p and pc wl pC wT . Reservation wage: w / p MRS lc (l T , c C ) c Increase in utility C wT p w/ p c C L l 0 l T Analysis of the wage rate increase dl w, c wT dw l comp l l comp l T l L . w M w M SE IE SE: leisure becomes more expensive, thus SE reduces leisure and increases consumption of aggregate commodity 33 A.Friedman ICEF-2022 IE: Agent is a net seller of labour (he can never be a net buyer). Selling labour time at higher price raises his real income. c w / p c0 w0 / p c l IE 0 T l SE l If leisure is normal, then it goes up due to increase in income and IE>0. If leisure is inferior, then it goes down and IE<0. Leisure Normal good Inferior good Substitution effect Income effect + (if SE dominates) + (if IE dominates) Total effect Derivation of individual labour supply p 1 c Labour supply w2 w2 w1 w1 w0 w0 c L 0 0 L1 L2 (а) T l L0 L1 L2 L (b) 34 A.Friedman ICEF-2022 Possibility of backward bending labour supply (if leisure is normal) c w w2 Labour supply w1 IE dominates w2 w1 SE dominates w0 w0 c L0 0 (а) L1 L2 T l L0 L2 L1 L (b) Question. Explain why an increase in the basic wage rate per hour offered to a worker may decrease the number of hours she wishes to work while an overtime premium offered to the same worker may increase the number of hours she wishes to work? 2.3 Consumption choices over time: intertemporal choice Let us assume that there are two periods: current period ( t 0 ) and future period ( t 1 ). Individual gets income of Y 0 in current period and Y1 in future period. This bundle corresponds to his endowment point (the bundle of present and future consumption that can be consumed without market trade). Assume that individual can borrow and lend at the same market interest rate r . If individual consumes less than he earns in the current period, then the difference Y0 c 0 is saved and in the next period the agent gets additional income equal to Y0 c0 1 r . Thus his future consumption equals c1 Y1 Y0 c 0 1 r . If currently the agent wants to consume more than he earns, then he has to borrow c 0 Y0 and in the future he will repay the debt together with interest payments, thus his future consumption equals c1 Y1 c 0 Y0 1 r Y1 Y0 c 0 1 r . It means that irrespective of whether agent borrows or lends, his budget constraint is c1 Y1 Y0 c 0 1 r . If we open the brackets and put consumption in the LHS, then the budget constraint can be rewritten as c 0 1 r c1 Y0 1 r Y1 . 35 A.Friedman ICEF-2022 In the LHS we have the future value of the life-time consumption and in the RHS- the future value of the life-time income. By dividing both sides by 1 r the budget constraint can be stated in terms of present values c0 c1 Y Y0 1 . 1 r 1 r This budget constraint states that present value of lifetime consumption has to be equal to the present value of endowment. Graphically this intertemporal budget constraint can be represented by a straight line that goes through endowment point and has a slope of 1 r . c1 Saves in period 0 Borrows in period 0 Assume that agent derives utility from consumption in both periods uc 0 , c1 . Let more consumption in either period be preferred to less, so that utility increases as we move further from the origin. If we assume diminishing MRS then we get convex indifference curves. The MRS between current and future consumption reveals the intensity of individual’s preferences for consumption in different periods of time. If we write down MRS 01 1 , then is the rate of time preference. A person is said to be impatient if when consumption levels are the same in both periods is positive, meaning that person is willing to forego more than $1 of future consumption to increase current consumption by $1. A person is said to be patient if when consumption levels are the same in both periods is negative, meaning that person is willing to forego less than $1 of future consumption to increase current consumption by $1. Intertemporal utility-maximization problem: m ax uc0 , c1 c 0 0 ,c1 0 c0 c1 Y Y0 1 1 r 1 r 36 A.Friedman ICEF-2022 In case of interior solution the point of tangency of budget constraint with IC indicates the optimal consumption bundle. If c 0 Y0 , then agent is called a net lender (saver), if c 0 Y0 , then agent is called a net borrower. c1 Net lender 1+r c0 Conclusion: if financial market are perfect (agents can lend and borrow at the same interest rate), then consumption decision is determined by the present value of life-time income, not the income in current or future period alone. Comparative statics In this model interest rate plays a role of price. An increase in the interest rate brings two effects: substitution effect and income effect. Due to substitution effect current consumption falls as it becomes relatively more expensive. The sign of income effect depends on whether we deal with net lender or net borrower (as consumption stands for aggregate commodity it is treated as a normal good in each period). For net borrower an increase in the interest rate decreases wealth and results in a fall in current consumption. So for net borrower both effects move in the same direction and current consumption definitely falls. An increase in the interest rate increases the wealth of net lender and under given prices results in an increase in current consumption. So for net lender current consumption falls and saving increases if substitution effect dominates and current consumption rises together with fall in saving when income effect dominates. We can get the same results from the analysis of Slutsky equation. Income effect is proportional to the amount saved (S), that is why it may become dominant if S is large enough: c 0 c 0comp c (Y 0 c 0 ) 0 . r r M SE IE 37 A.Friedman ICEF-2022 Agent type Net borrower Net lender Substitution effect - Income (wealth) effect + Total change in c 0 +/ Change in borrowing/lending Borrowing= Lending= = c0 Y0 c0 0 0 if SE dominates = Y 0 c 0 c 0 0 if IE dominates Conclusions. Individual demand for borrowing is downward sloping, so does the aggregate demand. Individual supply of lending could be backward bending (upward sloping under low saving). Lending-borrowing equilibrium. Note: below we assume that backward bending part of individual supply disappears in process of aggregation. r Supply of lending r 0 Demand for borrowing B L Lending, borrowing More than 2 periods and bonds pricing Consider a bond that pays a fixed coupon amount x each period (starting from the next period) until a maturity date T and at T the face value F is paid. If we denote the discount factor by 1/(1 r ) then the price of the bond i P x 2 x T 1x T F . Let S 1 2 T 1 T then S 2 T T 1 S 1 T 1 which gives S 1 T 1 . 1 38 A.Friedman ICEF-2022 The price of a consol (perpetuity that never matures) we get as a special case, where F=0 and 1 1/(1 r ) x T=. Thus the price is PCONSOL x 2 x T x x x . 1 1 1/(1 r ) r The price of a bond has an inverse relationship with the rate of interest. 2.4 Production and consumption over time (optional) Suppose that investment (productive) opportunities are available but consumer has no access to the financial market. Investment opportunities are described by the PPC. PPC In equilibrium ct Qt and I S . If PPC is given by Q1 F Y0 I , then the consumption and production decision is given by the solution of the problem max uc 0 , c1 s.t . Q1 F I , c 0 Q0 Y 0 I , c1 Y1 Q1 . Suppose that investment (productive) opportunities are available and in addition consumer can borrow and lend at the same market interest rate. Consumption Production 39 A.Friedman ICEF-2022 c1 W0 . The level of 1 r wealth is the intercept of budget line with the horizontal axes. The highest attainable budget line is tangent to the PPC. The corresponding highest level of wealth equals W0 . Then Each budget line is associated with a specific level of wealth: c 0 consumer chooses the best bundle under given level of wealth. Investment financing. Net lender I Y 0 c 0 Q0 c 0 , net borrower I Y 0 c 0 c 0 Q0 own saving lending own saving borrowing On aggregate lending=borrowing, which implies that aggregate saving equals aggregate investment. Separation of consumption and production decisions m ax uc 0 , c1 c1 Y Q1 Q0 1 1 r 1 r Q0 Y 0 I , Q1 F I s.t . c 0 Note that production decision does not depend on consumers preferences. FOC: F I 1 r . Investment are chosen to maximize the consumers wealth, i.e. Y Q1 Y F I Y F I maxW0 max Q0 1 max Y 0 I 1 Y 0 1 max I . 1 r 1 r 1 r 1 r Separation theorem If markets for intertemporal claims are perfect, individuals can separate investment decisions (aimed at maximizing wealth) and consumption decision (dependent on consumer’s time preferences). Present value rule Due to separation theorem production decision can be delegated to managers. Managers that maximize the wealth of the firm will be making the correct investment decisions for all the owners individually regardless of the possibly differing time-preferences of the owners. Net present value of investment project in two-period model: NPV I I F I . 1 r If we have T periods and Rt I is the net income in period t , then NPV I R0 I R1 I R2 I R I T 1 T 1 2 1 r 1 r 1 r 40 A.Friedman ICEF-2022 Discrete case. If the number of investment projects is finite and these projects are mutually exclusive, then the project with the maximum possible present value should be chosen (given that it is positive). If projects are not mutually exclusive then all projects with positive NPVs should be adopted. 2.5 Applications of NPV rule: exhaustible resources (optional) Consider exhaustible resources industry (minerals or fossil fuels). The stock of the resource Q is constant. The owner should decide how much of the stock to extract and sell in each period. Consider two-period model. Assume that demand function is stable over time and P q - diminishing in q . Let extraction costs per unit of resource be constant and equal to c in any period. Competitive industry. In case of competitive industry any firm is a price taker. Let pt stay for current price and pt 1 - for future price. If pt 1 c pt c , then it is profitable to postpone extraction and sell in future period 1 r If pt 1 c pt c , then it is profitable to sell now 1 r If pt 1 c pt c , then it is profitable to sell in both periods 1 r Thus in competitive equilibrium extraction takes place in both periods if and only if pt 1 c pt c r , i.e. price minus extraction costs (marginal pt 1 c pt c or pt c 1 r profit) rises at the rate of interest. The result can be explained intuitively. If marginal profit increases less than market interest rate, then it is profitable to extract and sell the resource today and put money at bank deposit. If the opposite is true, it is optimal to postpone extraction. In any case the profit maximizing company has an incentive to change its production decision, which implies that currently there is excess supply (if extraction today increases) or excess demand (if extraction is postponed), which is not compatible with equilibrium. So, in equilibrium marginal profit should grow at the rate equal to market interest rate. The result can be derived formally from the profit maximization of representative firm: m ax q0 ,q1 0 p0 q0 TC (q0 ) p1q1 TC (q1 ) /(1 r ) s.t . q0 q1 Q FOC for interior solution implies: 41 A.Friedman ICEF-2022 p0 TC (q0 ) p1 TC (q1 )/(1 r ) or p1 c p0 c r . p0 c This rule is known as Hotelling rule. As extraction costs are constant, this implies that price will increase over time while extraction falls (due to declining demand). Graphical solution for linear demand functions. Monopolistic industry As monopolist is a price maker, marginal profit equals to MR c , thus he will extract in both MRt 1 c periods if MRt c , which implies that marginal revenue less marginal 1 r MR1 c MR0 c r extraction costs increases at market interest rate: . MR0 c Analytical derivation of the Hotelling rule for the monopolistic industry. m ax TR q0 TC (q0 ) TR q1 TC (q1 ) /(1 r ) q0 ,q1 0 s.t . q0 q1 Q FOC for interior solution implies: TRq0 TC (q0 ) TRq1 TC (q1 )/(1 r ) or MRq1 c MRq0 c r . MRq0 c Graphical solution and comparison with competitive case (linear demand functions and zero marginal extraction costs). 42 A.Friedman ICEF-2022 Graphical analysis suggests that monopolist would be more conservative: he would extract less today and more tomorrow. As a result the prices for a monopolized industry would be initially higher and become lower at later dates. price monopoly Competitive industry time extraction monopoly Competitive industry time 2.6 Sample exercise with solution Explain, why an increase in the basic wage rate per hour offered to a worker may decrease the number of hours she wishes to work while an overtime premium offered to the same worker may increase the number of hours she wishes to work? Solution 43 A.Friedman ICEF-2022 Under overtime premium the initial bundle is just affordable and we observe a pivot of BL around the previous choice. It means that Slutsky IE=0 and SE works in opposite direction to the price change: with increased overtime payment leisure is more expensive, thus person will reduce his consumption of leisure (moves from A to B). Thus, he will work more. Under an increase in the basic wage rate the budget line becomes steeper and individual’s real income increases as he can generate higher income supplying the same amount of labour. Thus in addition to the SE that reduces leisure we observe income effect that increases demand for leisure (in this case leisure must be a normal good) and it might happen that IE dominates the SE so that total demand for leisure increases and, as a result, labour supply decreases as it is demonstrated below. c w / p B C c0 w0 / p A IE c Lpremium SE 0 leisure T LWage _ increase l0 L0 44 A.Friedman ICEF-2022 3. CHOICE UNDER UNCERTAINTY 3.1 Gambles and contingent commodities A state of the world is the outcome of uncertain situation. Contingent commodity is the amount of consumption, the level of which depends on the state of the world occurring. Flipping coin game Suppose for each dollar you bet in a flipping coin game, you win (and get your bet back) if a heads comes up and lose your bet when a tail comes up. States of the world: state 1- tails comes up and state 2- heads comes up. Contingent commodities: consumption if tails comes up (denote by c1 ) and consumption if heads comes up ( c 2 ). Endowment point - consumption bundle of contingent commodities that is available when you make no trades with the market. In case of flipping coin game initially person has income of w in either state of the world. Budget constraint for contingent commodities shows how much of each contingent commodity you can have in each state of the world. Let us denote the bet by z and assume that bet can never be negative, then we have the following system that describes the budget constraint: c1 w z c 2 w z 0 z w If we solve this system with respect to z , then we get the budget constraint of the form: c 2 w w c1 or c1 c 2 w1 where 0 c1 w . w1 endowment 1 c1 45 A.Friedman ICEF-2022 Note that after the state of the world is determined, the person will consume only one contingent commodity that corresponds to the state of the world that takes place. Budget constraint does not extend to the right from endowment point as it was assumed that agent is not allowed to select the other side of the original bet. That is, we do not allow individual to make a bet in which he wins $1 if tails comes up and lose $ if heads comes up. If person is allowed to take both sides of the gamble, then his budget constraint will be a straight line that goes through initial endowment with a slope of . endowment 1 Budget constraint when both sides of the bet can be taken Generalisation Suppose that the terms of the gamble are such that consumption changes by x 1 in the first state of the world (if tail appears) and by x 2 in the second state of the world (if heads appears). Then we have the following system that describes the budget constraint: c1 w x1 z c 2 w x 2 z If we solve this system with respect to z , then we get the budget constraint of the form: c2 w Then the slope of the budget line equals x2 w c1 . x1 dc 2 x 2 . dc1 x1 Fair odds line Each state of the world s can occur with some probability p s , where 0 ps 1 and S p s 1 s 1. 46 A.Friedman ICEF-2022 Expected value of the gamble is the weighed sum of outcomes, where weights are equal to S probabilities: EV x p s x s . s 1 EV of the gain in original flipping coin game is $1 p1 p2 $1 p1 1 p1 . If coin is symmetric then p1 0.5 and EV gamble = 0.5 1 . A gamble with zero expected monetary gain is called a fair gamble. If expected monetary gain is different from zero, then such a gamble is said to be unfair. Gamble with positive expected gain is said to be favourable; gamble with negative expected gain is called unfaivourable. The fair odds line is a budget constraint reflecting the opportunities presented by an actuarially fair gamble (odds - the ratio of the probabilities of the two events). With two states of the world fair gamble satisfies the condition x1 p1 x 2 p2 0 , which implies that x2 dc 2 x 2 p p 1 1 . As slope of budget line is , the absolute value of x1 dc1 x1 p2 1 p1 the slope of fair odds line equals to the ratio of the probabilities dc 2 x 2 p 1 . dc1 x1 p2 Going back to our example of symmetric flipping coin we can illustrate the fair odds line as a p 1/ 2 1 . If 1 , the initial gamble is favourable straight line with the slope of 1 p2 1/ 2 and fair odds line would be flatter then budget line. Expected gain>0 Expected gain<0 Fair odds line c1 Note that expected value of consumption remains constant along fair odds line: EV c c1 p1 c 2 p2 w x1 z p1 w x 2 z p2 w zx1 p1 x 2 p2 w . Preferences Three types of attitude toward risk can be distinguished. 47 A.Friedman ICEF-2022 A person is said to be risk averse if he prefers a certain prospect with a particular expected value to an uncertain prospect with the same expected value. A person is said to be risk neutral if he is indifferent between a certain prospect with a particular expected value and an uncertain prospect with the same expected value. A person is said to be risk loving if he prefers an uncertain prospect with a particular expected value to a certain prospect with the same expected value. In order to illustrate certain prospects we will draw certainty line – the locus of all possible certain consumption levels (i.e. the line c 2 c1 ). Indifference curves for risk neutral agent are given by straight lines parallel to the fair odds line. Reason: any uncertain prospect for a risk neutral agent is equivalent to certain bundle with the same EV. Note that along FOL expected value of consumption is constant, thus all these points lie on the same indifference curve. As more is better, agent becomes better off while moving along certainty line further from the origin. Increase in utility FOL To illustrate indifference curves for a risk averse agent, let us take two points on fair odds line: certainty point A and some uncertain prospect B. FOL FOL (1) (2) As both bundles (A and B) have the same expected value of consumption but A is certain, then by definition risk averse agent would prefer certain bundle A to any uncertain prospect like B that gives the same EV of consumption. It means that all points on FOL would give lower utility than A. In other words, A belongs to indifference curve that lies further from the origin. As a result ICs cannot be bowed outward as in diagram (1). Otherwise point B would bring higher utility than A, which contradicts to risk aversion. 48 A.Friedman ICEF-2022 ICs cannot cross fair odds line at certainty line as in diagram (2). Otherwise risky prospect (D) would be equivalent to the certain one (A) with the same EV. Thus the IC of risk averse agent satisfies the following properties: absolute value of the slope (MRS) at certainty points is equal to the ratio of probabilities (absolute value of the slope of FOL); ICs are bowed in. c2 FOL c1 Exercise. Show that indifference curves of a risk lover are bowed out and at certainty points have slope that is the same as the slope of FOL. Optimal bet: the case of risk averse agent By definition, a risk-averse agent will never participate in fair game (i.e. will make zero bet) as his initial endowment lies on certainty line and is preferred to any risky prospect that belongs to FOL. Optimal choice FOL=Budget line Fair game If game is favourable (this is the case if 1 ), then risk averse agent will take some risk and optimal bet would be positive as risk is compensated by positive expected gain. 49 A.Friedman ICEF-2022 Budget line bet FOL Unfair favourable game If game is unfavourable (this is the case if 1 ), then risk averse agent will make zero bet as expected consumption at any point on budget line is less than at initial endowment and, in addition, the endowment point is certain. Optimal choice Budget line FOL Unfair unfavourable game 3.2 Expected utility In presence of uncertainty utility depends on the quantities of contingent commodities and corresponding probabilities. In principle, probabilities can enter utility function in quite complex ways. Under some additional requirements on preferences utility function takes the S form which is linear in probabilities: U c1 , c 2 ,, c S ; p1 , p2 ,, p S ps uc s . A utility s 1 function that takes this form is called a von Neumann-Morgenstern utility function or expected utility function (EUF). It is not entirely an ordinal function as only positive affine transformations are allowed: aU b , a 0 . 50 A.Friedman ICEF-2022 EUF and attitude toward risk Risk-averse person: U c1 , c 2 ; p,1 p puc1 1 puc 2 u pc1 1 pc 2 for any p 0, 1 and c1 c 2 . This is Jensen inequality which implies that uc is strictly concave. It implies that the marginal utility of a risk averse agent is decreasing in wealth. u Risk averse agent uc2 uc1 c 0 Risk-neutral person: U c1 , c 2 ; p,1 p puc1 1 puc 2 u pc1 1 pc 2 for any p 0, 1 , which implies that uc is linear. u Risk neutral agent uc2 uc1 0 c1 c1 c Risk-loving person: U c1 , c2 ; p, 1 p pu c1 1 p u c2 u pc1 1 p c2 for any p 0, 1 and c1 c 2 . This is a Jensen inequality which implies that utility function of a risk-loving person uc is strictly convex so that the marginal utility is an increasing function of wealth. 51 A.Friedman ICEF-2022 3.3 Willingness to pay to avoid risk Certainty equivalent (CE) of a gamble – the certain wealth that would make an agent indifferent b/w accepting the gamble and accepting the certain wealth: uCE puc1 1 puc2 Let us show that for risk-averse agent certainty equivalent is less than the expected value of a gamble. By definition of risk aversion: uCE puc1 1 puc2 u(EV ) . Since uс is increasing then CE EV . It means that a risk averse person is ready to pay a risk premium EV CV 0 to avoid the risk, that is, to exchange a gamble for its expected value. u uc2 uc1 Risk premium c1 0 c Example Consider risk averse individual with initial wealth W . With probability p she can incur losses of L , 0 L W . Individual is offered to purchase full insurance that will compensate all the loss in case of accident. What is the maximum premium that this agent is willing to pay for this insurance? The maximum premium should make this person indifferent between purchasing insurance and staying at initial endowment. The expected utility at initial endowment is EU NO _ INS puW L 1 puW while with full insurance under premium R his utility equals EU INS uW R . Thus the maximum premium could be derived from the following equation uW RMAX puW L 1 puW . Note that W RMAX CE , which implies that this premium is given by the difference between initial wealth and certainty equivalent: RMAX W CE . 52 A.Friedman ICEF-2022 u c Note: insurance premium risk premium since insurance premium is calculated as a change in wealth, not the change in expected wealth. u c What is the minimum premium that a risk-neutral insurance company is willing to accept? For a risk neutral agent expected utility is given by expected wealth. Thus, insurance company will offer insurance iff the resulting expected profit is nonnegative: R pL 0 . Under minimum premium 0 , which implies RMIN pL W EV . Application 1. Obtaining additional information Mary has a utility function EU economist 50 120 / c , where c is her consumption, measured in thousands of dollars. If Mary becomes an economist, she will make 30 thousand per year for certain. If she becomes a pediatrician, she will make $60 thousand if there is a baby boom and $12 thousand otherwise. The probability of a baby boom is p=0.5. 53 A.Friedman ICEF-2022 economist pediatrician boom p= ½ 30 no p= ½ 30 boom p= ½ 60 u(60)=50-120/60=48 no p= ½ 12 u(12)=50-120/12=40 By comparing utilities, we can find that Mary prefers to become an economist as EU economist 46 44 EU pediatrician . Now suppose a consulting firm has prepared demographic projections that indicate which event will occur. Will she purchase the projection at price of $6000? economist 30-6=24 pediatrician 60-6=54 boom p= ½ economist 30-6=24 no p= ½ pediatrician 12-6=6 As the maximum utility achieved without demographic projection was only 46 Mary would be willing to purchase this projection. Application 2. Demand for Insurance Re-consider an example with insurance but now assume that any amount of insurance (full or partial) might be purchased at insurance premium r per dollar of insurance coverage. Insurance is actuarially fair if EV $1 p $r 0 or r p . If r p , then insurance is unfair. The case of unfair favourable insurance ( r p ) is quite unrealistic as insurance company incurs loss under this price. Thus we will consider only fair and unfavourable insurance. Two states of the world: “loss” and “no loss”. 54 A.Friedman ICEF-2022 Two contingent commodities: c L - consumption in case of loss, c NL - consumption in case of no loss. Denote by x the quantity of insurance and assume that over-insurance is not allowed (you cannot insure more than L), then c L w L rx x c NL w rx 0 x L If we solve this system with respect to x , then we get the budget constraint of the form: c NL w r w L c L where w L c L w rL . 1 r The slope of budget line is r /1 r . endowment W budget line W-L The slope of fair odds line equals p . 1 p In case of fair insurance budget constraint coincides with the fair odds line and as a result utility is maximised at certainty point, where indifference curve is tangent to budget line, which means that person will purchase full insurance: c L w L x 1 r c NL w rx or x L . 55 A.Friedman ICEF-2022 Optimal choice Fair insurance Thus any risk averse agent under fair insurance purchases full insurance (insures all the loss). In case of unfair unfavourable insurance ( r p ) budget line is steeper than fair odds line. As a result risk averse person will never purchase full insurance. He will either purchase partial insurance ( 0 x L ) or no insurance at all. Optimal choice Optimal choice Unfavourable insurance These results could be also derived algebraically if we setup and solve the expected utility maximization problem: maxpuW L X rX 1 puW rX X 0 As EU is strictly concave because of the risk-aversion then the FOC is both necessary and sufficient. FOC: 1 r p u W L X 1 r r 1 p u W rX 0 0 if X 0 Let us start with the fair insurance r p and rearrange 56 A.Friedman ICEF-2022 u W L X 1 p u W pX 0 0 if X 0 Claim 1. X 0 . Suppose that this is not the case and X 0 . Then uW L uW 0 due to diminishing marginal utility. But this inequality violates FOC. Claim 2. X L . Since X 0 FOC could be restated as uW L X 1 p uW pX 0 . Due to diminishing marginal utility this requires equality of the levels of wealth W L X 1 p W pX , which implies X L . Now we proceed to unfair unfavourable insurance r p . In this case it might happen that no insurance is purchased: EU X 0 1 r p uW L r1 puW 0 . If X 0 then 1 r p uW L X 1 r r 1 puW rX . Rearranging we get: u W L X 1 r r 1 p 1 r p u W rX 1 r 1 p and 1 r p r1 p . Thus we obtain uW L X 1 r r 1 p 1 , that is uW L X 1 r uW rX . As u is 1 r p uW rX diminishing then W L X 1 r W rX and we get that only partial insurance can be As rp then purchased X L . 3.4 The Arrow–Pratt measure of risk aversion How do we measure the degree of risk aversion? Intuition: the degree of risk aversion should reflect the curvature of utility function, which means that we should use the second derivative: u . For risk-averse agent the second derivative is negative and it is more convenient to put minus sign before the derivative. But utility function is not unique: u and au b, a 0 represent the same preferences. Solution: divide by the first derivative. Arrow–Pratt coefficient of absolute risk aversion: u u Result: the larger the Arrow-Pratt measure of risk aversion, the smaller gambles an individual will take. 57 A.Friedman ICEF-2022 This coefficient may change with wealth. Example 1: u lnW u 1/ W and u 1/ W 2 imply that coefficient of risk aversion u 1 u W decreases as individual’s wealth goes up. Conclusion: with an increase in wealth this person will take higher risks (buy less insurance or invest more in risky asset). Example 2: u e aW , a 0 u ae a and u a 2 e a imply that coefficient of risk aversion u a u doesn’t change with an increase in wealth. Conclusion: with an increase in wealth this person will take the same risks (buy the same amount of insurance or invest the same sum in risky asset). This function is called CARA (constant-absolute-risk-aversion) utility function. 3.5 Reducing risk via diversification Diversification - the spreading of risk among several alternatives rather than choosing one Example 1. Consider two companies A and B and assume that their performance depends on the weather (see the table below). A B Sunny (p=1/2) 100 20 Rainy (p=1/2) 20 100 If he invests in A or B only then his expected utility is EU A EU B u20 u100/ 2 . u EU0.5A+0.5B EUA =EUB 20 60 100 c Suppose that he invests 50% in A and 50% in B, then he gets (20+120)/2=60 in each state and utility goes up: EU 0.5 A0.5 B u(60) u20 u100/ 2 . Conclusion: If returns are perfectly negatively correlated diversification eliminates the risk 58 A.Friedman ICEF-2022 Example 2. Assume that the shares of two companies (A and B) bring 100 or 20 with equal probabilities but now returns are independent. If a risk-averse agent invests in A or B only, he gets EU A EU B u20 u100/ 2 . If he invests 50% in A and 50% in B then the following outcomes may be observed Prob 1/4 Wealth 1/4 (20+20)/2=20 1/4 (20+100)/2=60 1/4 (100+20)/2=60 (100+100)/2=100 1 1 The resulting expected utility is EU 0.5 A 0.5 B u(60) u20 u100 . 2 4 Compare it with the initial one: 1 20 100 1 EU 0.5 A 0.5 B EU A u u20 u100 0 . 2 2 2 Graphical analysis for independent returns We will proceed in two steps. Step 1: EU A 1 u20 u100 2 u u(60) EU0.5A+0.5B EUA =EUB 20 Step 2: 100 60 c 1 1 EU A u60 . 2 2 3.6 Sample exercise with solution Susan has preferences, represented by the 12 12 1 p 5 EU (c1 , c 2 , p) p 5 1 c1 1 c2 following expected utility function , where c i stays for wealth in the state of the world i i 1, 2 and is the probability of state 1. Suppose that Susan has $7 and in addition owns a risky investment project that brings -$2 with probability 1/3 and $4 with probability 2/3. 59 A.Friedman ICEF-2022 (a) Find the minimum sum (Xmin) at which this individual is willing to sell this project? (b) Compare Xmin with the expected value of this project and explain the result. (c) Produce graphical solution for part (b) using contingent commodities diagram. (d) Dan is a risk-averse person with a smooth elementary utility function. He has the same initial wealth and the same risky project as Susan. Assume that he was offered to sell the project fully or partially at the price of $1 and he decided to sell 50% of the project. Is it true that he will sell greater share of the project if the probability of loss goes up? Solution (a) Suzan will sell iff her utility does not diminish. Xmin is the sum that makes Suzan indifferent between selling and keeping the project. Thus, in both cases EU should be the 1 same. EU project 5 3 EU sell 5 12 2 12 1 2 3 2 4 11 5 5 2 5 1 1 7 2 3 1 7 4 3 3 3 3 12 11 36 32 12 4 , 1. , X min 1 7 X min 3 8 X min 3 4 (b) EV project 1 2 2 4 2 X min 1 3 3 Explanation. If agent gets EVproject then he has the same expected consumption but if the project is sold then she has no risk. By definition for a risk averse person (verify that Susan is risk averse) this option is strictly preferred to the one with risk under the same EV, thus he will sell the project even if the price is a bit lower. (c) Contingent commodities graph with comments 11 E -1/2 7+EV 7+Xmin 7 EV Xmin W 5 Fair odds lines 7+Xmin7+EV c1- consumption if project is not successful c2- consumption if project is successful E- initial endowment if the person owns the project W-initial wealth (without project) Slope of FOL=-(1/3): (2/3)=-1/2 (d) Dan is a risk-averse person with a smooth elementary utility function. He has the same initial wealth and the same risky project as Susan. Assume that he was offered to sell the 60 A.Friedman ICEF-2022 project fully or partially at the price of $1 and he decided to sell 50% of the project. Is it true that he will sell greater share of the project if the probability of loss goes up? True. Optimal investment can be derived from max pu7 21 1 pu7 41 0, 1 FOC for interior solution: pu5 3 1 pu11 3 (*) Let us prove that >0. Assume 0 then c1 3 0 uc1 0 as MU is diminishing for a risk-averse person. Since p 0 then puc1 increases. c 2 3 0 uc 2 0 as MU for a risk-averse person Since 1 p 0 then 1 puc 2 decreases AS the LHS of (*) while the RHS then they cannot be equal and (*) is violated. This contradiction proves that >0. 61 A.Friedman ICEF-2022 4. GAME THEORY 4.1 Simultaneous-move games A situation of strategic interactions is called a game. To specify a simultaneous-move game we have to identify players(the decision makers): i 1, 2,, n ; the set of strategies of each player: S1 , S2 ,, Sn (strategy - complete description of the plan of actions of the player); payoffs for each strategy profile: ui s1 , s2 ,, sn , i 1, 2,, n Simultaneous-move game can be represented in a matrix form (known as the normal form). Strategy ~ s is a dominant strategy for player i if it performs better than any other strategy of i player i no matter what others are playing: u ~ s , s u s , s i i i i i i for all si ~ si and all s i . If each player has a dominant strategy, then we get a dominant strategy equilibrium. Example. Prisoners’ Dilemma. Two prisoners who were partners in a crime were being questioned in separate rooms. Each can either confess to the crime (and thereby implicate the other) or deny his participation in the crime. If only one prisoner confessed, then he would go for free and the other would spend 6 months in prison. If both denied, then each would spend one month in prison (time required for investigation) and if both confessed, then each would spend 3 months in a prison. Note that each player would be better off by choosing strategy ‘confess’ whatever is the strategy chosen by the other as 0 > -1 and -3 > -6. Thus in Prisoners Dilemma each player would choose strategy confess, that is (‘confess’, ‘confess’) is a dominant strategy equilibrium. Note, that the outcome of this equilibrium is inefficient as both agents would be better off by playing (Deny, Deny). But (Deny, Deny) outcome cannot be achieved as players are unable to cooperate. Player B Player A Deny Confess Deny -1,-1 -6, 0 Confess 0, -6 -3, -3 62 A.Friedman ICEF-2022 Iterative elimination of dominated strategies (IEDS) Even if a player does not have a dominant strategy, he might have one or more dominated strategies. Strategy si is a strictly dominated strategy for i if it yields a lower payoff compared to another strategy (say s i ) irrespective of what others are playing: ui si , si ui si , si for all s i Rational agent will never use a dominated strategy, thus we can eliminate it. If all strategies except one for each player can be eliminated by iteratively eliminating strictly dominated strategies, the game is said to be dominance solvable. Example. Consider the following simultaneous-move game with two players. Player 2 Player 1 A2 B2 C2 a1 2,2 4,2 0,4 b1 4,0 6,8 2,2 c1 6,4 4,0 0,6 Step 1: eliminate a1 as it is dominated by b1 (2 < 4, 4 < 6, 0 < 2) and eliminate A2 as it is dominated by C2 (2 < 4, 0 < 2, 4 < 6). Step 2: Remaining game Player 2 Player 1 B2 C2 b1 6,8 2,2 c1 4,0 0,6 Eliminate c1 as it is dominated by b1 (4<6, 0<2). Step 3: Remaining game Player 2 Player 1 b1 B2 C2 6,8 2,2 Eliminate C2 as it is dominated by B2 (2 < 8). Equilibrium obtained by IEDS: (b1, B2) 63 A.Friedman ICEF-2022 Nash equilibrium In some games IEDS do not produce a unique outcome. To solve the games of this sort we should increase our requirements to the rationality of the players. This bring us to the concept of Nash equilibrium. A strategy profile s1 , s2 ,, sn is a Nash equilibrium (NE) if for each player i his strategy si performs at least as well as any other strategy given si : ui si , si ui si , s i for all si S i . A dominant strategy equilibrium is also NE but NE does not require dominance. Greater scope of NE comes at a cost: it places greater rationality requirements on players: each player must correctly anticipate the strategies that the other players are going to play. Example Player B Left Right Top 2, 1 0, 0 Bottom 0, 0 1, 1 Player A In this game “Left” is best response for “Top” as 1>0 and “Top” is best response for “Left” as 2>0. Thus (Top, Left) is Nash equilibrium. Similarly “Right” is best response for “Bottom” as 1>0 and “Bottom” is best response for “Right” as 1>0. Thus (Bottom, Right) is another Nash equilibrium. Some games do not have pure strategy NE. Example: Matching coin game Player 2 Player 1 Heads Tails Heads +1, -1 -1, +1 Tails -1, +1 +1, -1 Mixed strategy Nash equilibrium A mixed strategy is a probability distribution over the set of pure strategies of the player. Example: play ‘Heads’ with probability p=1/3 and ‘Tails’ with probability 1-p=2/3. NE definition is the same: profile of mixed strategies that constitute a mutual best response (BR). Example. Derivation of mixed-strategy NE for Matching coin game. 64 A.Friedman ICEF-2022 Denote the mixed strategy of agent 1 by p, 1 p and the strategy of agent 2 – by q, 1 q . Derivation of BR for agent 1: EU1 p1 q 1 (1 q) 1 p 1 q 1 (1 q) 1 2q 2 p2q 1 max p0, 1 q 1/ 2 1, BR1 0, 1, q 1/ 2 0, q 1/ 2 Derivation of BR for agent 2: EU 2 q 1 p 1 (1 p) 1 q1 p 1 (1 p) 2 p 1 2q1 2 p max p0, 1 p 1/ 2 1, BR2 0, 1, p 1/ 2 0, p 1/ 2 q BR2 1 BR1 1/2 NE 0 1 1/2 p Nash Equilibrium (NE): p 1/ 2, q 1/ 2 Derivation of mixed-strategy NE. Useful property If in some NE two strategies (A and B) are played by agent i with positive probabilities they should bring the same utility to this agent: ui siA , si ui siB , si . If utilities are different ui siA , si ui siB , si then we can increase expected utility by reallocating the weight of low-utility strategy to the high-utility one EU i pui siA , si 1 pui siB , si pui siA , si 1 pui siA , si ui siA , s i . It means that the initial strategy of player i doesn’t bring the highest possible expected utility under given strategies of other players and so this strategy profile violates one of the requirements of NE. Let us demonstrate the application of this property for the Matching coin game. Let us take the first player and calculate his expected utility for each pure strategy: u1 H1 q 1(1 q) 2q 1 and u1 T1 q 1(1 q) 1 2q . This player will play both 65 A.Friedman ICEF-2022 positive probabilities iff he is indifferent: u1 H1 u1 T1 , that is 2q 1 1 2q , which happens if q 1/ 2 . Thus in mixed NE (with 2 players) each player's chosen probability distribution must make the other player indifferent between the strategies he is randomizing over. Existence of NE Existence theorem (Nash 1951). Every game with a finite number of players and finite strategy sets has at least one NE (in mixed or pure strategies). Existence result makes NE the main solution concept for simultaneous-move games. Proof is based on fixed point theorems (take Game Theory Course). Most of the models in Economics deal with continuous (rather than discrete) set of strategies: Bertrand, Cournot. Many of the results are still applicable to continuous case. 4.2 Sequential-move or extensive-form games Up until now we have been looking at games in which all players move simultaneously. But in many situations one player gets to move first and the other players responds. For sequential games we need to draw a game tree to depict the sequence of actions. We start with games with perfect information, where each player can observe the moves of players who act before them. In extensive form games we differentiate between actions and strategies [strategy indicates a complete plan of actions of the player]. Example: 1 L R 2 2 T (0,3) B (1,0) T (4,1) B (2,2) Player 1 has two strategies: {L, R } Player 2 has two actions {T, B} but four strategies: {TLTR, BLBR, BLTR, TLBR}, where the lower index is used to indicate the choice of player 1. Let us proceed from extensive to a normal form. 66 A.Friedman ICEF-2022 Player 2 Player 1 TLTR, BLBR TLBR BLTR L 0, 3 1, 0 0, 3 1, 0 R 4, 1 2, 2 2, 2 4, 1 Sequential-move games with imperfect information In imperfect information game at least one player is not perfectly informed about some of the (or all of the) previous moves. Reconsider the last example, assuming that now player 2 does not observe the action of player 1. To illustrate this lack of information we combine two decision nodes in one information set (dotted line). 1 R L 2 2 T (0,3) B T (1,0) (4,1) B (2,2) The strategy set of player 1 stays the same: { L, R } but the strategy set of player 2 is different { T, B }. As a result the normal form game looks like this: Player 2 Player 1 L T 0, 3 B 1, 0 R 4, 1 2, 2 Non-credible threats in NE and subgame perfection Threat is credible if it would be in the self-interest of the player to carry out the threat if called upon to do so. The following example demonstrates that NE may involve non-credible threats 67 A.Friedman ICEF-2022 Example 1 In Out 2 Fight (-1, 0) No (0,4) (1, 2) We can proceed to the normal form game: In Firm 1 Out Firm 2 Fight No -1, 0 1, 2 0, 4 0, 4 There are two pure strategy NE: {In, No} and {Out, Fight}. The last one {Out, Fight} includes non-credible threat as it is not profitable for firm 2 to fight if firm 1 enters as 0 < 2. To solve this problem we need a refinement of NE concept in sequential games, that is, we should impose additional conditions. Subgame perfect Nash equilibrium (SPNE) A subgame is a part of a game that starts from a node which is a singleton and includes all successors of that node [you cannot cut an information set so that only part of it belongs to a subgame!] A strategy combination is SPNE if it induces a NE in every subgame. Note: the whole game is also a subgame SPNE is a NE but not every NE is SPNE In finite games we can derive SPNE using backward induction. In our example with two firms there are just two subgames: the game itself and the part that starts from the decision node, where firm 1 goes in. If firm 1 plays ‘in’, the best choice for firm 2 is to play ’no’ as 2>1. Anticipating that firm 2 will respond by playing ‘no’, firm 1 chooses ‘in’ as it gives 1 instead of 0. Thus only {In, No} constitutes SPNE. 68 A.Friedman ICEF-2022 1 In Out 2 (0,4) No Fight (-1, 0) (1, 2) SPNE under imperfect information Now, let us consider another example of sequential game, where information is imperfect. 1 In Out 1 A C (3, 1) 2 B C D (2, 4) D (0, -1) (-1, 0) (1, 3) There are two subgames: the game itself and the part that starts from the decision node, where firm 1 plays ‘In’. Let us consider this subgame and produce a normal form game for it. Firm 2 Firm 1 A C 3, 1 D 0, -1 B -1, 0 1, 3 There are two pure strategy NE in this subgame: {A, C} and {B, D}. If firm 1 plays ‘In’ and in the following subgame firms play {A, C} then firm 1 gets 3, which is more than it would get staying out (3>2). Thus ‘In’ is firm’s 1 best response for {A, C} and the strategy profile {(In, A), C} constitutes SPNE. If firm 1 plays ‘In’ and in the subgame that follows firms play{B, D}then firm 1 gets 1, which is less than 2. Thus ‘Out’ is a best response for {B, D} and we get one more SPNE: {(Out, B), D}. Thus in this game there are two pure strategy SPNE: {(In, A), C} and {(Out, B), D}. 69 A.Friedman ICEF-2022 4.3 Repeated games In a repeated game the same stage game is played several or infinite number of times. In infinitely repeated game we discount the payoffs: the payoff of player i is given by ui a0 ui a1 2 ui a2 t ui at . In finitely repeated games there is no t 0 discounting. Finitely repeated game. Let us suppose that the both players know that the same (for example Bertrand) game will be played N times ( N is finite). What will the outcome be? As the game is dynamic, we look for the perfect Nash equilibrium using backward induction. Consider the last round. As this round is the final one and everybody knows it, then there is no incentive for cooperation and every player will choose the static game Nash equilibrium strategies by charging price equal to MC. Now consider what will happen on round N 1 . As at the last round there will be no cooperation there is no incentive to cooperate at this round as well. If one cooperates by charging the monopolist price the rival will find optimal to cheat by charging lower price and getting all the market. Each player has an incentive to deviate and as a result the only equilibrium at this subgame is given by a static Nash equilibrium, where each firm charges price equal to MC. The same logic proves that there would be no cooperation at each round and the only perfect equilibrium corresponds to prices equal to marginal cost. The result is not surprising as players cooperate only if there is a punishment for cheating. With finitely repeated game at the last round cheating cannot be punished and this creates incentive for deviation at each round. In finitely repeated game with one NE this NE is played at each stage. Infinitely repeated game. If the game is infinitely repeated, then the last round does not exist and as a result backward induction cannot be used. With infinite number of rounds a deviation at any point of time could be punished in the future. Trigger strategy – strategy in repeated game, where the player stops cooperating in order to punish other player’s break with cooperation Example. Infinitely repeated advertising game Firm 2 Firm 1 A A 2, 2 N 5, 1 N 1, 5 4, 4 70 A.Friedman ICEF-2022 The only NE in one-shot game is {A, A}. We can note that this game is similar to Prisoners dilemma: ‘A’ is a dominant strategy for each player but the payoffs of both firms could be higher if they stop advertising. Consider the following trigger strategy of player i (i=1,2): start by playing N (cooperate at t=0) in period t > 1: play N if (N, N) was played at t-1 play A otherwise. Let us find out, under which value of discount factor neither of the firms will deviate from cooperative strategy. Assume that one of the firms deviate at moment t. Then we can calculate the present value of the payoff starting from t and compare it with the payoff in the absence of deviation. Payoff in case of deviation at t (starting from t): 5 2 2 2 5 2 Payoff if tacit cooperation is sustained: 4 4 2 4 . 1 4 . 1 4 . Solving this inequality we can get 5 2 1 1 that if 1/ 3 (players are patient enough) cooperation can be sustained in infinitely repeated advertising game. There is no incentive for deviation if: Note that the threat of punishment is credible. The payoff of the firm that implements punishment (starting at t) is 1 2 2 2, while its payoff in the absence of punishment is 1 1 2 1 . Thus punishment is in the self-interest of the firm as it gives higher payoff: 1 2 2 2 1 1 2 1. As the payoffs are the same at every period it is enough to look at one period only: 2>1. Credibility is achieved due to the nature of the punishment: each firm uses Nash-reversion strategy. 4 for each player. To compare it with payoffs in one1 shot game we proceed to normalized payoff by multiplying by (1-). The resulting equilibrium payoff is Thus we demonstrated that (normalized) payoff (4, 4) can be sustained as SPNE in the infinitely repeated game if is high enough. But this is not the only possible payoff that could be sustained in SPNE. There are other SPNE as well. For example, playing {A, A} at every round is also SPNE. Thus we can sustain (2, 2). The result about the set of payoffs that can be sustained as equilibrium outcome is known as the `folk theorem'. To identify all the (normalized) payoffs that could be sustained in the infinitely repeated advertising game considered above, we will produce a graph. 71 A.Friedman ICEF-2022 First of all, we will find all feasible payoffs by illustrating the convex combinations of possible payoff profiles of the stage game (indicated by blue). Then we look only at those payoffs from this set that are individually rational. Individual rationality means that the equilibrium payoff of each player must be at least as large as the minmax payoff of that player as otherwise a player who gets less than the minmax payoff has incentive to deviate by playing the minmax strategy. Player i’s minmax payoff of player i (denote it by *i ) is the lowest payoff in the stage game that the rival can impose on him through his choice of a strategy s−i, given that player i chooses his own strategy to maximize his own payoff: *i min max ui si , si . s i si Since we deal with Prisoners’ dilemma type game, where every player has a dominant strategy, then maximum payoff is always provided by this trategy and, as, a result the minmax payoff for each player coincides with the payoff that he gets in the dominant strategies equilibrium (in our case it is 2 for each player). 2 5 4 2 1 1 2 4 5 1 Folk Theorem. Any feasible individually rational pair of normalized payoffs (1, 2) can be supported as SPNE for a high enough ( close to 1). 72 A.Friedman ICEF-2022 4.4 Sample exercise with solution Answer the following questions. (a) Consider the following extensive form game. 1 T 1 L B 2 R 2 C D C (0, 0) (3, 1) D E F (5, 4) (2, 5) (4, 0) (1, 2) (i) Identify all pure strategies of player 2 (ii) Identify all subgames and find all pure strategy SPNE (b) ’In any finite sequential game with perfect information if no player has the same payoffs at any two distinct terminal nodes then the SPNE is unique’. Is this true or false? Explain. (Note: you should take into account both pure and mixed strategies) Solution (ai) Pure strategies of player 2: CE, CF, DE, DF (aii) Three subgames should be identified Subgame 1 1 1 L Subgame 3 T B 2 R 2 C (3, 1) D (0, 0) E C (1, 2) D (5, 4) F Subgame 2 (2, 5) (4, 0) Backward induction must be used to derive SPNE. Subgame 1 NE: {L, C} (C is dominant strategy for player 2 as 1>0 and 2>0; L is BR for C as 3>1) Subgame 2 NE: F as 5>4 Subgame 3: T (if agent 1 plays T then his payoff is 3; if he plays B the his payoff is 2. Thus he is better off by playing T) SPNE: {T L, C F} (b) True. In finite game we find SPNE via backward induction by identifying NE in every subgame. In the last subgame the player who moves has a finite number of actions that result in different payoffs u1 , u2 , , un . Moreover, due to the perfect information the payoffs at this 73 A.Friedman ICEF-2022 subgame do not depend on the other players’ choice in this subgame as this is the only player that moves in the considered subgame. WLG we can assume that u1 u2 un . He is looking for p1 , p2 , , pn (where each probability is b/w 0 and 1 and the sum=1) that max EU, where EU p1 u1 p2u2 pn un . If at least one probability different from pn is positive then this is not the Max EU as EU p1 u1 p2u2 pn un p1un p2un pn un un . Thus max exists and it is unique. It means that the player that moves should never randomize and should choose the action that generates the highest payoff at this stage (i.e pn 1 ) and this action is unique. Then we go backward and deal with the player that moves before last. He also has finite number of actions and as he perfectly anticipates the action chosen by the last player every action generates some particular payoff and the payoffs are different. Thus we apply the result derived above that demonstrates that this agent should also play pure strategy in this subgame and choose the action that generates the highest payoff. This action exists and is unique. We continue backward and at every step identify unique action that generates the highest payoff. The combination of these actions forms the resulting unique SPNE. 74 A.Friedman ICEF-2022 5. THE FIRM 5.1 Modeling the firm’s technological opportunities Production function is the relationship between the quantities of various inputs used per period of time and the maximum quantity of output that can be produced per period of time: Q f z1 , z 2 ,, z n . We are going to concentrate on the case with two inputs: capital (K) and labour (L). Examples of production functions ( 0, 0 ): Fixed proportions or Leontieff technology minL, K , Linear technology L K , Cobb-Douglas production function AL K In case of two inputs we can represent production function graphically in terms of isoquants. Isoquant is the locus of all the (technically efficient) combinations of inputs for producing a given level of output. Question: illustrate isoquants for the Leontieff, linear and Cobb-Douglas production functions. Properties of the production functions. Marginal physical product of input i - the extra amount of output that can be produced when the firm uses additional unit of this input, holding the levels of other inputs constant: f . MPi z i A technology exhibits decreasing/increasing/constant returns to factor when the marginal physical product of an input falls/rises/stays constant as the amount of the input used increases. Question: characterize the three considered technologies (Leontieff, linear and CobbDouglas) with respect to returns to each factor of production Marginal rate of technical substitution of labour for capital MRTS LK measures the rate at which the firm has to substitute one input (L) for another (K) in order to keep output (Q) dK constant: MRTS LK . Thus MRTS equals to the absolute value of the slope of an dL Q const isoquant. Definition suggests that MRTS can be calculated as the ratio of the marginal products: 75 A.Friedman ICEF-2022 dQ f K dK f L dL 0 , which implies MRTS LK dK dL Q const f L MPL . f K MPK If MRTS LK is some positive constant, then the factors are perfect substitutes (linear technology) If MRTS LK is 0, then the factors are perfect complements (Leontieff. Note: MRTS is not defined at kink) If MRTS LK isn’t constant, then the factors are imperfect substitutes (Cobb-Douglas). In case of Cobb-Douglas technology we observe diminishing MRTS while moving along the isoquant. 0 Most of the production functions used in empirical analyses are homothetic functions. (Note: function f is homothetic if f hgz , where gz is homogeneous of degree 1 and h is monotone). For homothetic production functions the slopes of the isoquants are preserved along every ray through the origin, i.e. MRTS remains the same for any given K/L ratio whatever the level of output. Returns to scale If all inputs are changed by the same proportion this is referred to as a change in the scale of production. Decreasing returns to scale (DRS): a proportional increase in all inputs leads to less than proportionate increase in output: f L, K f L, K for all L, K and 1 . Constant returns to scale (CRS): a proportional increase in all inputs leads to proportionate increase in output: f L, K f L, K for all L, K and 0 . 76 A.Friedman ICEF-2022 Increasing returns to scale (IRS): a proportional increase in all inputs leads to greater than proportionate increase in output: f L, K f L, K for all L, K and 1 . Returns to scale for homogenous production functions Note: production function is homogeneous of degree t if f L, K t f L, K for all L, K and 0 . If t 1 , then we have CRS technology, If t 1 , then we have IRS technology, If t 1 , then we have DRS technology. 5.2 Profit maximization and Cost minimization Profit maximization problem m ax pQ wL rK K 0, L 0 f K , L Q Solution: demand for factors of production L p, w, r , K p, w, r and supply of output Q p, w, r . This problem can be divided into 2 sub-problems: 1) Cost minimization problem TC w, r , Q m in wL rK K 0, L 0 f K , L Q Solving this problem we obtain conditional demand for factors of production Lw, r, Q , K w, r, Q . Plugging the solution into the objective function we get the firm’s cost function: TC w, r, Q wLw, r, Q rK w, r, Q . 2) profit maximization with respect to output max pQ TC w, r, Q . Q 0 5.3 Cost minimization Cost minimization in the long run In the long-run all factors are variable TC LR w, r , Q m in wL rK K 0, L 0 F K , L Q Graphical solution of cost-minimization problem 77 A.Friedman ICEF-2022 isocost lines Isocost line - a line representing all input combinations that have the same cost for the firm. Equation of isocost: wL rK const At interior solution the slope of isoquant is equal to the slope of isocost: MRTS LK L , K w / r . This condition implies that if the firm uses both factors in equilibrium, then they bring the same marginal product per dollar spent: MPL L , K MPK L , K . w r Thus if the firm uses both factors, then the corresponding quantities demanded can be found from the following system F L , K Q MRTS LK L , K w / r This system could be derived algebraically. Lagrangean: ℒ wL rK Q f ( K , L) FOCs for interior solution: ℒ/L = w FL 0 ℒ/K = r FK 0 w FL r FK ℒ/ = Q F ( K , L) 0 From the first two conditions we get MRTS LK FL w . FK r Let Lw, r, Q and K w, r, Q be the solutions of long run cost minimization problem. These demand functions are called conditional demands for factors of production since the quantities depend on the level of output produced. We can look for the optimum input combinations in production for different levels of output; the resulting curve is known as expansion path. Expansion path allows for given factor prices to get the relationship between output and the long run cost. 78 A.Friedman ICEF-2022 Expansion path 0 Exercise. Illustrate expansion path for the homothetic production function. Analytically we get the long run cost function by plugging conditional demand functions into the objective function: TC LR w, r, Q wLw, r, Q rK w, r, Q . Properties of long run costs Long run marginal cost ( MC LR ) - the change in long-run total cost due to the production of additional unit of output: MC LR TC LR . Q Long run average cost ( AC LR ) - the long-run total cost divided by the number of units produced: AC LR TC LR . Q When long run average costs fall as output rises, costs are said to exhibit economies of scale. When long run average costs rise with the output level, costs are said to exhibit diseconomies of scale. Due to homotheticity, K/L ratio under CRS is not affected by the level of output. As a result, with CRS technology to increase output by times we increase the employment of each factor by times. Then cost of production also increases by times, while average costs stay constant: AC LR TC LR Q TC LR Q TC LR Q Q AC LR Q . Q Q Q That is, in case of CRS we have neither economies, nor diseconomies of scale. NOTE: As AC=const, then TC LR cQ , where c - cost of producing of one unit of output. It implies that AC LR MC LR c . 79 A.Friedman ICEF-2022 $ CRS technology With IRS we can produce Q by increasing the factors’ employment in a smaller proportion, which implies that total cost also increase by less than times1. As a result AC falls and we deal with economies of scale: AC LR Q TC LR Q TC LR Q TC LR Q AC LR Q . Q Q Q $ IRS technology In case of DRS under homothetic production function in order to increase output by times, we increase the factors’ employment in a greater proportion, which implies that total cost also increases by more than times. As a result AC goes up: AC LR TC LR Q TC LR Q TC LR Q Q AC LR Q . Q Q Q $ 1 DRS technology Note if production function isn’t homothetic then proportional factor increase does not necessarily correspond to cost minimizing bundle but this implies that cost minimizing bundle might cost even less. 80 A.Friedman ICEF-2022 That is, in case of DRS we have diseconomies of scale. Relationships between LRAC and LRMC TC Q MC Q lim MC 0 . Q 0 Q 1 If TC(0)=0, then AC 0 MC 0 . Proof: AC 0 lim If AC reaches minimum at Q 0, then AC Q MC Q , Q 0 Proof. If AC reaches minimum at Q 0, then AC Q 0 , which implies TC Q Q TC Q TC Q MC Q AC Q or AC Q 0 Q2 Q2 Q AC Q MC Q . If AC diminishes over some range of outputs, then AC Q MC Q for all Q from considered range, Proof. If AC diminishes over some range of Q , then AC Q 0 for each Q from given range. This implies that TC Q Q TC Q TC Q MC Q AC Q AC Q 0 or AC Q MCQ . Q2 Q2 Q If AC increases over some range of outputs, then AC Q MC Q for all Q from considered range. Proof. If AC increases over some range of Q , then AC Q 0 for each Q from given range. This implies that TC Q Q TC Q TC Q MC Q AC Q AC Q 0 or AC Q MC Q . Q2 Q2 Q Exercise. Suppose that AC and MC are U-shaped and TC(0)=0. Sketch AC and MC curves on the same graph. Cost minimization in the short-run In the short-run, capital is fixed and we choose only one variable factor - labour: TC SR w, r , Q, K m in wL r K L 0 F K , L Q , KK This problem can be restated as 81 A.Friedman ICEF-2022 TC SR w, r, Q, K m in wL r K L 0 F K , L Q If F K ,0 0 , then labour employment is given by the condition F K , L Q . Solving this equation, we get short run demand for labour L LQ, K . Plugging into the objective function we get short-run total cost as a sum of variable cost (VC) and fixed cost (FC): TC SR w, r, Q, K wLQ, K variable cost r K . fixed cost Properties of the short-run cost Short-run marginal cost ( MC SR ) - the change in short-run total cost due to the production of additional unit of output: MC SR NOTE: As TC SR TC SR . Q VC Q FC , then MC SR TC SR VC Q FC VC Q . Q Q Q Shape of the short-run MC is determined by return to variable factor. MC SR VC Q wLQ LQ w . w Q Q Q MPL Conclusion: MC=const if MPL const , MC increases if MPL diminishes in Q, MC decreases if MPL increases in Q, Short-run average cost ( AC SR ) - the short-run total cost divided by the number of unites TC SR VC Q FC AVC Q AFC , where AVC denotes average Q Q VC Q FC variable cost: AVC Q and AFC denotes average fixed cost AFC . Q Q produced: AC SR As FC const , then AFC is diminishing function of output $ 82 A.Friedman ICEF-2022 The shape of AVC depends on return to variable factor. VC Q wLQ QL Q wLQ w LQ w 1 1 AVC Q L Q w 2 Q Q Q MPL APL Q Q Q As we can see dynamics of AVC depends on the relationship between average and marginal products of labor. If MPL APL , then AVC Q 0 , i.e. AVC is constant. If MPL APL , then AVC Q 0 , i.e. AVC is increasing. If MPL APL , then AVC Q 0 , i.e. AVC is diminishing. Suppose that F K ,0 0 , i.e. labour is essential factor. The first case takes place if marginal product is constant. 0 The second case takes place if MPL is diminishing. 0 The last case takes place if MPL is increasing. 83 A.Friedman ICEF-2022 0 MPL AVC AFC AC SR constant constant decreasing decreasing diminishing increasing decreasing decreasing at small Q increasing at large Q increasing decreasing decreasing decreasing NOTE: U-shaped short-run AC appears if MPL is diminishing Relationships between AVC and short-run MC (the same as for long-run AC and MC) If VC(0)=0, then AVC 0 MC 0 . If AVC reaches minimum at Q 0, then AVC Q MC Q , If AVC diminishes over some range of outputs, then AVC Q MC Q for all Q from considered range, If AVC increases over some range of outputs, then AVC Q MC Q for all Q from considered range. Exercise. Suppose that AVC, AC and MC are U-shaped and TC(0)=0. Sketch AVC, short-run AC and MC curves on the same graph. Relationships between the SR and LR cost curves TC LR Q TC SR Q, K for any Q and K , TC LR Q TC SR Q , K , where K - the level of capital optimal for Q . Implication: AC LR Q AC SR Q, K for any Q and K , AC LR Q AC SR Q , K , where K is the level of capital optimal for Q and SMC Q , K LMC (Q ) . 84 A.Friedman ICEF-2022 Cost minimization with two plants Suppose firm has two plants that produce the same output with different technologies. The resulting cost functions are given by TC 1 q and TC 2 q . Question: what is the firm’s cost function? To find out the firm’s cost function we have to allocate any given output Q in a cost minimizing way between the two plants: TC firm Q min TC 1 q1 TC 2 q2 s.t. q1 q 2 Q, q1 0, q 2 0 As q2 Q q1 , then TC firm . Q qmin TC 1 q1 TC 2 Q q1 . 0, Q 1 If both plants are used in production (i.e. we deal with interior solution), then the FOC implies TC1 q1 TC 2 Q q1 , i.e. MC1 q1 MC 2 q2 . It means that in case when both plants are used, output is allocated in such a way that marginal costs of production are equalized. If for MC of production at one plant is always less than MC of production of the other plant, then only the plant with lowest marginal cost would be used. 5.4 Profit maximization in case of perfect competition Profit maximization problem max pQ TC w, r, Q . Q 0 Rules of profit maximization: 1. produce only if Q 0 or p TC / Q TC 0/ Q . 2. if Q 0 , then produce at a point, where p MC Q 0 (First order condition for interior solution); 3. produce at a point, where 0 (Second order condition); Implications for the long-run: 1. If TC LR 0 0 . then produce if p AC LR Q 0 , i.e. p AC LR Q . 2. if Q 0 , then produce at a point, where p MC LR Q 3. produce at a point, where MC LR Q Q 0 , i.e. at non-diminishing part of long-run MC. Implications for the short run: 1. produce only if p AC SR Q AFC , i.e. p AC SR Q AFC AVC Q as TC LR 0 FC . 85 A.Friedman ICEF-2022 2. if Q 0 , then produce at a point, where p MC SR Q 3. produce at a point, where MC SR Q Q 0 , i.e. at non-diminishing part of short-run MC. Example Suppose that AC and MC are U-shaped and TC(0)=0. Let us sketch the long run supply curve. NOTE: As AC is U-shaped, then p AC LR Q for all p min AC LR Q . As a result we get the following LR supply curve: Q LR LR 0, p m in AC Q p LR LR LR Q : MC Q p, MC Q 0, p m in AC Q $ Firm’s LR supply 0 Consider an example with U-shaped short-run AC, AVC and MC. As AVC is U-shaped, then p AVC Q for all p min AVC Q . As a result we get the following SR supply curve: 0, p m in AVC Q Q SR p Q : MC SR Q p, MC SR Q 0, p m in AVC Q 5.5 Sample exercise with solution Consider a perfectly competitive industry that produces good X. All firms in this industry have identical technologies with cost function с q , where с' 0 0 , с' q 0 , с q 0 for q 0 . Unfortunately a fraction of the output produced by each firm is defective and cannot be sold. Moreover firm experiences some utilization cost for unsold output and the corresponding cost function is given by l z , where z is the volume of utilized output, l 0 0 , l' z 0 and l z 0 for z 0 . Both production and utilization cost are zero if 86 A.Friedman ICEF-2022 output is zero. Suppose that improvement in management brings a reduction of while cost function stays the same. What is the impact on individual supply of each firm? Provide both graphical and analytical solution (Note: change in is not necessarily small). Solution Profit maximization problem 1 pq cq l q max q 0 Function is strictly concave (second derivative is negative due to the assumptions), thus FOC is both necessary and sufficient. 1 p c q l q 0 and 1 p c q l q 0 if q 0 . Note that q 0 for any p 0 as otherwise 1 p c 0 l 0 1 p 0 which violates FOC. Algebraic analysis Let us prove that reduction in for given price results in an increase in quantity supplied by any firm. The LHS that represent marginal revenues goes up and so should do the RHS. If q 0 then marginal production cost would fall or stay the same (as MC is increasing) and the second term (marginal utilization cost) definitely falls as: q 0 and MUC is increasing then l q 0 and with smaller we have l q 0 . Thus RHS goes down while LHS up and we get a contradiction. It means that q 0 . Graphical analysis with comments Supply is given by nondiminishing part of total MC per efficient unit (i.e. sold unit) that lie above AC. Here as cost function is convex and goes from the origin than at any point MC>AC and MC is increasing. Thus total MC per efficient unit represents supply curve. With reduced under the same output we utilize less and due to increasing marginal utilization cost assumption we get lower value of MUC. Thus total MC falls at every q. Moreover, as we sell more, the level TMC per efficient unit goes down which strengthens the effect of reduction of MUC. As TMC per efficient unit shift downward, it means that firm is willing to produce the same output under lower price (i.e. supply curve shifts down or to the right). 87 A.Friedman ICEF-2022 p q 88 A.Friedman ICEF-2022 6. PERFECT COMPETITION 6.1 Perfect competition Fundamental assumptions of perfect competition Buyers and sellers are price takers (each agent chooses its actions under the assumption that it cannot influence the prices of the goods) Entry into the market is free (new suppliers can enter the market without any restrictions on the process of entry) We will look at equilibrium in some particular market ignoring all other markets, that is we will perform partial equilibrium analysis. This analysis is valid under quite restrictive assumptions: absence of the feedback effects and negligible income effects. To find equilibrium market price and quantity we need information about market demand and market supply curves. Market demand could be obtained by summing up (horizontally) the individual’s demand curves. p p p 20 20 Market demand q1(p)+ q2(p) 10 10 q1(p) q2(p) q1 10 Q q2 10 20 30 Market supply is different in the SR and in the LR as in the SR new firms cannot enter the market as they cannot obtain the needed fixed inputs. In the LR new firms can enter and existing firms can exit. The short run In the SR the number of firms in the industry is fixed and we get SR industry supply by N summing up the firms SR supply curves: Q S p qiS p . i 1 As each firm produce at non-diminishing part of its MC the resulting SR industry supply is upward sloping. The result does not depend on whether suppliers are homogeneous (have the same cost functions) or heterogeneous (have different cost functions). 89 A.Friedman ICEF-2022 p p p SR Industry supply Q q2 q1 Market is in equilibrium if: (1) buyers are choosing their optimal purchase levels, given the prevailing market price; (2) sellers are choosing their optimal output levels, given the prevailing market price; (3) suppliers are willing to produce as much as buyers wish to purchase. SR equilibrium corresponds to the intersection of market demand with SR industry supply. Equilibrium price p is a solution of the equation Q D p Q S p . Industry supply Market demand The long run In the LR new suppliers can enter the market and old suppliers can exit. To find the market quantity supplied at a given price, we need to find both the quantities supplied by each firm in the market and the number of suppliers who choose to be in the market at that price. LR equilibrium is given by p , Q , N such that: D N p Q p , where Q p q p ; S S S i Q p q C q 0 i but it becomes negative for N 1 i i 1 i 90 A.Friedman ICEF-2022 (a) Homogenous firms We start with homogenous firms case, i.e. the case, where all firms have identical cost functions. Consider the constant cost industry - industry, where individual firm’s cost function remains unchanged as industry output expands. Note that as all firms have the same technology, nobody is willing to produce at price that is below the minimum of AC. Thus quantity supplies is zero at any price below p min AC . Now, let us take any price above min AC . At this price p a AC q a and firm gets positive profit. New firms will be attracted to the industry by the prospect of earning positive economic profit (due to assumption of constant cost industry the expansion of industry output has no effect on cost curves of individual firm). Thus industry supply is unlimited at this price. The same argument applies for any other price that exceeds p . If p min AC , then each firm produces q and gets zero economic profit. Thus firm is indifferent between being in and out of the market, there is no incentive to enter or exit the market. p p Constant cost industry LR Supply Q q Firm Market Conclusion: in case of constant-cost industry LR industry supply is zero for p min AC and the curve is horizontal at p min AC . As a result the quantity produced by the industry is determined by the market demand and the number of firms in the industry N is obtained as a ratio of industry and individual output: p min AC , Q Q D p , q argmin AC q , N Q / q . So far we assumed that an increase in the industry output has no effect on individual firm cost function. It may not be the case. We can observe both external diseconomies (an increase in industry output brings an increase in LRAC) or external economies (an increase in industry output results in a reduction in LRAC). There are two reasons for the presence of external economies/diseconomies: (1) pecuniary external economy/diseconomy is the result of interaction between industry output and firm’s cost function through the changes in the market prices of inputs; 91 A.Friedman ICEF-2022 (2) technological external economy/diseconomy is the result of interaction between industry output and firm’s cost function through the physical possibilities of production (i.e. the production function). Examples: pecuniary diseconomy may result from increased competition for specific factor; technological economy may come from innovations that represent a by-product of an increased industry output. Increasing cost industry - an industry in which external diseconomies take place, i.e. LRAC rise with the industry output level. The LR industry supply will be upward sloping as min AC LR rises with expansion in industry output. Note: although min AC LR rise with increase in Q , the minimum efficient scale may stay constant, fall or rise. Decreasing cost industry - an industry in which external economies take place, i.e. LRAC fall as industry output rises. The LR industry supply will be downward sloping as min AC LR falls with expansion in industry output. Note: although min AC LR falls with increase in Q , the minimum efficient scale may stay constant, fall or rise. p p LR Supply LR Supply Q Increasing cost industry Q Decreasing cost industry (b) Heterogeneous firms Let us assume that industry output has no effect on the firms cost functions, i.e. we deal with constant cost industry. Suppose that there are two types of producers of some good. Producers of type 1 have lower costs but the number of type 1 producers is limited (for example they use specific resource with restricted access). Assume for simplicity that there is only one firm of type 1. Producers of type 2 have higher costs of production but any firm can become a type 2 producer. 92 A.Friedman ICEF-2022 p p p LR Industry supply Q q2 q1 Type 1 firm Type 2 firm Market If price is below the minimum of LRAC of the first firm (firm with the lowest cost of production), then nobody is willing to produce at all. As a result industry supply is zero for any p p1 min AC1LR . If price is below the min AC 2LR but above min AC 1LR other firms would be willing to produce had they access to the technology used by firm 1, but the access is restricted so that firm1 will be the only producer and the industry supply corresponds to the part of MC of firm 1 (if there are more firms of type 1, we should sum up their supply curves for any p p1 , p2 . As price reach the level of min AC 2LR , then any type 2 firm is able to produce but its profit would be 0. Thus it is indifferent between coming in and staying out. As a result industry supply is horizontal at p2 min AC 2LR . Conclusion: industry supply can be upward sloping even in case of constant cost industry if firms have different cost functions. 6.2 Equilibrium and efficiency We want to know not only how a competitive market works but also whether the resulting equilibrium allocation is ‘desirable’ for the society. We will use the concept of total surplus (TS) to measure the social welfare. TS is a difference between social benefit (measured by total willingness to pay) and social cost: Q ~ ~ TS Q TB Q TC Q P D Q dQ C Q . 0 Total willingness to pay (or gross consumers’ surplus) gives total benefit from consumption of given quantity of a good. Graphically it can be represented as the area below the market demand function. For example if Q Q0 , then total benefit is represented by grey area: CS gross Q 0 A B . Total cost can be represented as the area below the market supply curve for given number of firms in the industry (integral from MC). If Q Q0 , the cost of production are given by dashed area (B). As a result TS=(A+B)-B=A. 93 A.Friedman ICEF-2022 From TS maximization we get the socially efficient output: MBQ MC Q . This condition is known as condition of allocative efficiency. In addition to this condition there are two more requirements. When we calculate TS we should be sure that TC Q represents the minimum cost of production for given output Q [i.e. total output should be allocated efficiently between the firms] and TB Q represents the maximum benefit from consumption [i.e. Q should be allocated efficiently between the consumers]. p Supply A Demand B Q Another approach to TS Note that in the absence of government intervention TS can also be calculated as a sum of net consumers’ surplus (CS) and producers’ surplus (PS): Q Q ~ ~ ~ ~ TS Q CS Q PS Q P D Q dQ QP Q QP Q C Q P D Q dQ C Q . 0 0 PS is the revenue the agent receives in excess of what he would require to produce a given quantity. If additional unit of output is produced then costs increase by the value of MC, thus the difference between the market price and MC corresponds to a net gain of producer from this additional unit. By summing up over all units produced, we get the PS. Note that in the absence of fixed cost PS is equal to profit. If we implement this approach, then consumers will purchase Q0 units of output at price p 0 (note that at this price producers are willing to supply more than Q0 so that p 0 does not correspond to equilibrium). The resulting CS is the area below the demand function above the market price (grey) and together with PS we get exactly the same value of TS as before. 94 A.Friedman ICEF-2022 p Supply CS PS Demand Q The key question is whether TS is maximized at competitive equilibrium. If some other allocation had a higher TS, then output would be called inefficient because it would be possible to make society better off. Claim: competitive equilibrium is efficient. To prove this claim let us look at output level, Qa , which is less than competitive. Then by moving from Qa to Q , total benefit increase by (C+D), while costs increase only by D and TS goes up by C, which implies that initial allocation was inefficient. Similarly if we look at output level that exceeds equilibrium one, Qb Q , then we can increase TS by moving from Qb to Q as total benefit falls by F, while costs go up by (E+F). As a result TS is increased by E. p Supply C E D F Demand Q The same result we can derive algebraically. Profit-maximization implies p s MCQ , while utility-maximization MBQ MCQ . requires MBQ pd . Since in equilibrium p s pd then 95 A.Friedman ICEF-2022 Implication: as competitive equilibrium output is efficient, then any government interventions that result in deviation from competitive output would reduce total surplus, i.e. the corresponding output would be inefficient. The corresponding reduction in TS is called deadweight losses. Question: Prove algebraically that perfectly competitive outcome is efficient Application 1: per unit tax analysis Suppose that per unit tax with tax rate of t on sales of good X is introduced. Let demand and supply curves be linear. Let us denote by p the price paid by consumers and suppose that tax is paid by producers, then the price received by producers is p t . As a result the FOC in profit maximization problem looks like p t MC q or p MC q t . Thus this tax can be treated as an increase in MC. As this increase is the same at every level of output, the MC shifts up parallel, which implies that industry supply curve shifts up by the value of tax. As a result consumers’ price goes up and quantity produced falls. p t A B E C D G F H Demand Q Initially With tax Change CS A+B+C+D A -(B+C+D) PS E+F+G+H B+E=H -(E+F+G) Gov.Rev.= t Qt 0 B+C+E+F B+C+E+F TS=CS+PS+GR A+B+C+D+ E+F+G+H A+B+C+ E+F+H -(D+G) Conclusion: this policy results in DWL D G Alternative approach to calculation of DWL: Total benefit is reduced by (D+G+I), while cost goes down only by I and as a result TS ( D G I ) I ( D G) 96 A.Friedman ICEF-2022 p t DWL D G I Demand Q Note that both producers and consumers are worse off due to the tax. Distribution of the tax burden between consumers and producers depends on relative slopes of demand and supply curves. Note that relationship b/w slopes implies the corresponding relationships of price elasticities. In equilibrium under per-unit tax: QS pS t QD pD t , where pS t pD t t . Consider a tax rate change dt assuming that initially t 0 : QS dpS QD dpD . Then we obtain the relative tax burden Q D dp S and proceed to elasticity (multiply and divide by before-tax dp D Q S price and quantity): D Q D p / Q p dpS Q p Q p S , where Dp D D and Sp S S . dp D Q S p/Q p Q Q Conclusion: greater share of tax burden corresponds to market side with lower (in terms of absolute value) price elasticity. p Loss in CS t t Loss in CS Loss in PS Demand Loss in PS Demand Q Q 97 A.Friedman ICEF-2022 Application 2: price ceiling Under effective price ceiling there is a gap between quantity demanded and quantity supplied. The quantity sold is determined by the short side of the market (in this case given by the quantity supplied) but different rationing mechanisms could be used. We will consider two extreme cases: in the first buyers with highest valuation obtain the good and in the second case the good produced is sold to the buyers with lowest valuation. A B A C A E D C B D F B E D D Q Q Case 1 Case 2 Initially Case 1 (High-valuation) Case 2 (low-valuation) With price ceiling Change With price ceiling Change CS A+B A+C CB F F(A+B) PS C+D+E E (D+C) E (D+C) TS A+B+C+D+ E A+C+E (B+D) F+E F(A+B+C+D) Conclusion: the change in CS and the resulting DWL crucially depends on the rationing mechanism used. If the good is sold to the high-valuation consumers (case 1) then the value of DWL is minimized and equal to A+D. This mechanism is called the efficient rationing as it allocates the good efficiently between the consumers. In case of inefficient allocation of the good the values of social loss may increase significantly and the upper bound of DWL is demonstrated by case 2, where DWL=A+B+C+DF. Application 3: price support program Suppose that the government wants to help the producers of the competitively supplied product. Government sets the support price above the equilibrium one and buys up any excess supply at the support price. 98 A.Friedman ICEF-2022 A B E C A F D H G E I D Q Initially With price support Change CS A+B+C A (B+C) PS E+F E+F+B+C+D B+C+D Government Surplus 0 (C+D+H+F+G+I) (C+D+H+F+G+I) TS=CS+PS+GS A+B+C+E+F A+B+C+ E+F+H (C+H+F+G+I) This program increases PS but requires huge government spending and finally results in tremendous value of DWL=C+H+F+G+I. This is due to the wrong incentive created by the policy: with high price producers increase output while demand is reduced. Government spends a lot of resources for price support and most part of the resources is wasteful since government cannot resale the good without dropping the price. Alternative policy in the form of lump sum subsidy equal to PS will generate that same increase in producers’ welfare without DWL. 6.3 Sample exercise with solution Consider perfectly competitive constant cost industry with identical firms. Suppose that, a per-unit sales subsidy is replaced by a lump sum subsidy that every active firm gets. The lump-sum subsidy leave the equilibrium price the same as it was under the per-unit subsidy scheme. Compare the total government expenditures under per unit and lump-sum subsidy schemes: (i) graphically assuming U-shaped AC and (ii) analytically (i.e. for any type of AC consistent with the conditions of the problem). Solution Graphical analysis 99 A.Friedman ICEF-2022 p p LS subsidy LR Supply per unit subsidy Q q Firm Market Analytical approach p LR AC q s AC q LS AC q LS S q LS 2 S , q LS AC q 0 AC q LS S q LS 2 , which implies 0 , i.e. AC is not minimized at q q LS but AC is minimized at q q . Thus AC q AC q LS and S s AC q LS AC q 0 . q LS Cost of subsidy comparison: S n LS S Q LS Q S s sQS . q LS q LS 100 A.Friedman ICEF-2022 7. GENERAL EQUILIBRIUM AND WELFARE ECONOMICS 7.1 General equilibrium in exchange economy Consider an economy, where there are K consumers k 1, 2,, K and N consumption goods: i 1, 2,, N . consumption Assume that every consumer k has an initial endowment of e k e1k , e2k ,, e Nk goods that can be traded at market prices p p1 , p2 ,, pN . We assume that there is no production so that this is an exchange economy. Denoting the consumption bundle of agent k by x k x1k , x2k ,, x Nk we get the following budget constraint p1 x1k p2 x2k p N x Nk p1 e1k p2e2k p N e Nk . Prices p p1 , p2 ,, pN and allocation x 1 , x 2 ,, x K constitute general equilibrium in exchange economy if: constraint under given p p1 , p2 ,, pN , K for every consumer k bundle x k x1k , x2k ,, x Nk maximizes uk subject to his budget all markets clear: K x e k 1 k i k 1 k i for every i 1, 2,, N . Example. General equilibrium in 2×2 economy with Cobb-Douglas preferences. Consider an exchange economy with two goods ( x and y ) and two consumers ( A and B ) with identical preferences u( x, y xy . We start with derivation of individual demands for agent A: m ax x A y A px x A py y A p x e xA py e yA x A 0, y A 0 We can notice that there are no corner solutions for positive income. The explanation is simple: under zero quantity of some good the resulting utility is zero, while it could be positive if, for example, we spend half of the income on each commodity. Thus zero consumption of any of the good doesn’t maximize utility. In interior solution MRS xy y A px , which implies equal spending on each good x A py py y A px x A . Plugging into the budget constraint and rearranging we get the demand functions: 101 A.Friedman xA ICEF-2022 px e xA py e yA 2 px 0.5e xA 0.5 py px e yA and y A 0.5 px A e x 0.5eyA . py In the similar way we derive demand of agent B from his utility maximization problem: max x B y B px x B py y B px e xB py e yB x B 0, y B 0 Due to the symmetry we get: x B 0.5e xB 0.5 py px e yB and yB 0.5 px B e x 0.5e yB . py To get equilibrium prices we should find total demand for every good and equate with total supply given by the total endowment. For x - market we get the following equation: x A x B 0.5e xA 0.5 py px e yA 0.5e xB 0.5 Solving this equation we get the following price ratio: py px py px e yB e xA e xB ./// exA exB . It could be easily eyA eyB verified that this price ratio will also equilibrate the second market y A yB 0.5e yA 0.5 px A p e x 0.5e yB 0.5 x e xB eyA e yB . py py py e xA e xB Plugging the equilibrium price ratio into demands we get the equilibrium p x e yA e yB allocation A B 1 A e xA e xB A 1 e y e y A x A ex A e x e yA ey , y A A B B 2 2 e x e x ey ey and A B 1 B B ex ex x B ex ey A 2 e y e yB e A e yB B B , yB 1 y . e e x y 2 e xA e xB We can note that equilibrium prices are not unique. If price vector p p1 , p2 ,, p N and allocation x x 1 , x 2 ,, x K constitute a general equilibrium in some economy then price vector p and allocation x is also a general equilibrium in this economy. It means that prices are relative and we can normalize one price to 1. From the previous analysis we can derive another important result. It is enough to equilibrate all markets but one and the last market will be automatically in equilibrium. This is an implication of the Walras law that we discuss below. 102 A.Friedman ICEF-2022 Walras law For any price vector p p1 , p2 ,, p N the total value of excess demands summed up over N all markets is always zero: p ED i 1 i i 0 , where EDi stays for excess demand for good i. K K k 1 k 1 In case of exchange economy: EDi x ik ( p) eik Let us prove Walras law for the case of exchange economy. Since demand of agent k is derived from utility maximization problem it should satisfy the consumer’s k budget constraint. Moreover, due to non-satiation all income is exhausted: p1 x1k p2 x 2k p N x Nk p1e1k p2 e2k p N e Nk . Summing up over all consumers and re-arranging we get: M k M k M k M k M k M k p1 x1 e1 p2 x 2 e 2 p N x N e N 0 . k 1 k 1 k 1 k 1 k 1 k 1 Edgeworth box We can illustrate the equilibrium in exchange economy graphically. We are interested in feasible allocations, that is, those allocations, where consumption of each commodity equals to the amount available in the economy. As total supply of each commodity is fixed then all feasible allocations in exchange economy can be illustrated graphically by Edgeworth box. OA We can also illustrate the consumers’ preferences by indifference curves. 103 A.Friedman ICEF-2022 OB Increase in utility of A Increase in utility of В OA To find consumer’s choice we illustrate the budget set and then indicate the bundle that gives the highest utility. The budget line should go through initial endowment. At the figure below we illustrate a general equilibrium. OB endowment Budget set B equilibrium Budget set A OA Existence and uniqueness of equilirium It is important to establish the existence if GE before we proceed to the discussion of its properties. The existence theorem was proved by Kenneth Arrow, Gerard Debreu and Lionel McKenzie in 1951. The proof is based on a fixed point theorem applied to mapping of a set of prices to itself (it requires restrictions on preferences & endowments). 104 A.Friedman ICEF-2022 Equilibrium is not necessarily unique. The following example demonstrates that we might have multiple equilibrium allocations and price vectors. Consider an exchange economy with two goods ( x and y ) and two consumers ( A and B ). Assume that these consumers have identical preferences and treat the two goods as perfect substitutes: ux, y minx, y . Suppose that there are 3 units of each good in the economy. Agent A has 3 units of good x and one unit of good y and the remaining units belong to agent B. OB D C endowment OA In this example any allocation from CD segment is an equilibrium and the corresponding prices are given by the slope of the budget line that connects this allocation with the endowment point. Thus for every allocation we have a unique (normalized) price vector but we have continuum of equilibrium allocations and prices: x A 1, 2 , y A x A , x B yB 3 x A and px / py y A 1/x A 2 . 7.2 Pareto optimum in exchange economy An allocation of resources is Pareto efficient (Pareto optimal) if it is not possible to reallocate resources to make one person better off without making someone else worse off. Pareto improvement is a reallocation of resources that makes at least one person better off without making anyone else worse off. Pareto optimal allocations (PO) in exchange economy with 2 goods (x and y) and 2 agents (A and B) could be derived from the utility maximization problem of one consumer2 under fixed utility of the other and feasibility constraints: 2 This is the case with well-behaved preferences that are represented by continuous utility function that increases in each good. 105 A.Friedman ICEF-2022 u A (x A , y A ) m ax x A 0, y A 0, x B 0, y B 0 u B (x B , y B ) u x A x B e xA e xB y A y B e yA e yB To solve the problem we write down the Lagrangean: ℒ u A ( x A , y A ) u B ( x B , y B ) u x exA exB x A x B y eyA eyB y A y B . FOCs for interior solution: u A x 0 , ℒ/x = x A A u A ℒ/y = A y 0 , y A u B ℒ/x = B x 0 , x B u B ℒ/y = B y 0 . y B A From the first two conditions we get MRS xy B MRS xy x and from the next two conditions we get y x . Thus in case of interior solution, in addition to feasibility conditions y x A x B exA exB , y A y B eyA eyB , the following requirement should be satisfied: A B MRS xy MRS xy . This condition could be also derived graphically. Consider allocation C. It is not efficient since both persons would be better off by moving to any point that lies above IC A and below IC B . 106 A.Friedman ICEF-2022 Pareto improvement over C OB Preferred by A Preferred by В OA A B Efficiency is achieved if the two IC are tangent, that is, MRS XY . If it is not the MRS XY A B case, and in some interior allocation, for example, MRS XY , then a Pareto MRS XY improvement is possible. As agent A values good X more relative to agent B, we should reallocate good X in favour of agent A. Suppose we take small unit of good X from B and give it to A. Then A is willing to sacrifice up to units of good Y. Let us take only ( ) / 2 units of good Y from agent A. As a result he would be better off as ( ) / 2 . If we give ( ) / 2 units of good Y to agent B, he would be better off as well since he would be as well off by getting units of good Y instead of one unit of X but gets more: ( ) / 2 . Thus by reallocation of resources we were able to improve the position of both agents, which means that initial allocation was not Pareto efficient. The set of all interior PO allocations in exchange economy is given by the locus of tangency of agents’ indifference curves and is called contract curve in consumption. 107 A.Friedman ICEF-2022 OB Contract curve OA Tangency condition can be violated in case of corner solution. For example, a feasible A B allocation with MRS xy MRS xy can be optimal if y A 0 and an allocation with A B MRS xy MRS xy can be optimal if x A 0 as it is illustrated below for the case of perfect substitutes. 7.3 Welfare theorems for exchange economy The First Fundamental Theorem of Welfare Economics If there is a market for every commodity, optimizing behaviour of consumers and firms under perfect competition leads to a PO allocation. This theorem is also known as `Invisible hand' theorem since the market price mechanism ensuring that each agent is acting in pure self-interest results in a Pareto optimal outcome. 108 A.Friedman ICEF-2022 Informal proof for 22 economy. If an equilibrium is not PO then there exists some other allocation in the Edgeworth box that makes at least one agent better off. Since equilibrium is the best point in the budget set, to make the agent better off we must choose a bundle outside of his budget set. To make the other agent at least as well off we should either move along the budget line or outside of his budget set. There are no feasible allocations that satisfy the last two requirements. Implications of the FFWT If we have complete system of perfectly competitive markets then there is no case for `efficiency-enhancing' government intervention. Reasons for the market failure that explains the government: Incomplete market system due to asymmetry of information or externalities/public goods Absence of perfect competition. In such cases, government intervention is desirable (to reduce or eliminate the market inefficiency). Second Fundamental Theorem of Welfare Economics Usually in the economy we have many different PO allocations. Can any arbitrary PO allocation be attained by some set of competitive prices, assuming there is a suitable assignment of initial endowments? The answer to this question is provided by the second welfare theorem. If all agents have convex preferences, any efficient allocation can be obtained as an equilibrium under suitable redistribution of incomes. Proof for differentiable case. ~ p A ~A ~A x ,y Choose prices ~x MRS xy p A px ~ xA ~ py ~ yA ~ px exA ~ py eyA , and transfers: T ~ y T ~ px ~ xB ~ py ~ yB ~ px exB ~ py eyB . Due to the given definitions, the financial balance is A B p ~ xA ~ x B e A eB ~ p ~ yA ~ yB eA eA 0 . satisfied: T T ~ B x x x y x y A B ~B ~B x A,~ y A MRS xy x , y , therefore Since in PO indifference curves are tangent MRS xy ~ ~ p B ~B ~B MRS xy x , y ~x , which together with T B ~ px ~ xB ~ py ~ yB ~ px exB ~ py eyB under the p y convexity of preferences guarantees that ~ x B ,~ y B maximizes utility of consumer B under 109 A.Friedman ICEF-2022 ~ ~px , ~py . Similarly, as ~px MRS xyA ~x A , ~y A p and TA ~ px ~ xA ~ py ~ yA ~ px exA ~ py eyA then y ~ x A,~ y A maximizes utility of consumer A under ~ px , ~ py . ~ A ~B Both markets are in equilibrium as ~ xA ~ x B e x and y y e y due to the feasibility of PO allocation. Importance of convexity and other problems The following example demonstrates that the theorem might fail if convexity assumption is violated. At the following graph point C stays for PO allocation but it cannot be decentralized (cannot be achieved as GE) since the only price vector at which agent B will choose C is given by the tangency line but agent B will choose D rather than C at these prices. x B yA OB D C OA xA yB Problems with implementation: lump-sum taxes are difficult to impose in practice; people might not fully trust the government to make use of the proceeds in the right manner. 7.4 Production economy Consider a simplest version of production economy with one consumer. This economy is also called a Robinson Crusoe Economy. Assume that this representative agent consumes two goods and his preferences are represented by utility function u( x, y) . Production opportunities are represented by production possibilities frontier (PPF) that is also called transformation curve: T ( x, y) 0 . Any point on PPF shows the maximum amount of one good attainable under any given output of the other good for given technologies and factors’ availability. 110 A.Friedman ICEF-2022 By definition all points on PPF are technically efficient. The absolute value of PPF slope shows the marginal rate of transformation (MRT). MRT - the minimum quantity of one good that must be sacrificed to produce an additional (small) unit of the other good: MRT xy dy T / x . MRT xy shows the rate at which dx PPF T / y the economy can transform one output into another by shifting its resources, i.e. the opportunity cost of additional (differentially small) unit of x in terms of forgone y . y PPF x Pareto optimum in Robinson Crusoe economy To find PO allocation we should maximize utility of Robinson subject to the PPF: max u( x , y) x 0, y 0 T ( x, y) 0 . To solve this problem we setup a Lagrangean: ℒ u( x, y) 0 T ( x, y) . FOCs for interior solution: u T 0, x x u T ℒ/y = 0 . y y ℒ/x = By re-arranging we get u / x T / x , that is PO is given by the point on PPF, where u / y T / y indifference curve is tangent to the PPF: MRS xy MRTxy and T ( x, y) 0 . 111 A.Friedman ICEF-2022 y PPF PO x GE in Robinson Crusoe economy To find equilibrium prices in Robinson Crusoe Economy we should derive supply of the goods from profit maximization problem and equilibrate it with demand that comes from utility maximization. Note that along PPF total costs are fixed (all the resources are used). Thus profit maximization could be equivalently restated as revenue maximization under given costs: max px x py y x, y s.t. T ( x, y) 0 . Setup a Lagrangean: ℒ px x py y 0 T ( x, y) . FOCs for interior solution: ℒ/x = px Rearranging we get T 0 and x ℒ/y = py T 0. y px T / x . Thus the profit maximizing bundle is given by a tangency py T / y point of PPF and isoprofit line given by px x py y const . y PPF Production x 112 A.Friedman ICEF-2022 Now, we proceed to the derivation of demand. Robinson as a consumer tries to maximize his utility subject to the budget constraint: max u( x , y) x, y p x x py y M Since Robinson owns all the resources and the firm his income M is equal to the sum of the value of initial endowment and profit, that is px x py y , where ( y , x ) is the firms’ production plan. Thus the budget constraint of Robinson coincides with the iso-profit line that goes through the profit-maximizing production plan. The point of tangency of this budget line p with indifference curve indicates the utility-maximizing consumption bundle: MRS xy x py and px x py y px x py y . If we look at the figure below, then we can see that the given prices do not equilibrate the markets: there is excess demand at y market and excess supply at x market, that is the relative price of x is too high. y Consumption excess supply excess demand of y PPF Production x Economy is in equilibrium at a lower relative price of x , at which optimal production and consumption decisions coincide. 113 A.Friedman ICEF-2022 y PPF x Note that in equilibrium Robinson attains the highest indifference curve on his PPF, which implies that GE is PO. 7.5 Pareto efficiency in economy with production Any interior Pareto efficient allocation has to satisfy the following efficiency conditions: efficiency in consumption, efficiency in production, efficiency in output mix. Consider an economy with two consumption goods x, y , two factors of production ( L , K ) in fixed supply and two individuals ( A, B ). Efficiency in consumption An allocation of commodities is efficient in consumption if given the total amount of each commodity available for consumption, the only way to make one person better off is to make another person worse off. This condition was extensively discussed in the section on exchange economy. The key A B conclusion is that efficiency in consumption requires equality of MRS: MRS xy MRS xy . Efficiency in production An allocation of inputs is efficient in production if, given the total amount of each input available for production, the only way to increase the output of one commodity is to decrease the output of another commodity. Note that this condition is necessary for Pareto efficiency. Indeed, if this condition is violated, then by reallocating inputs we can produce more of one good under the same quantity of the other. Let us distribute this additional quantity produced equally between all consumers and each one would be better off, which means that initial allocation was not Pareto efficient. 114 A.Friedman ICEF-2022 Let us denote the fixed supply of capital and labour by K and L . Then all feasible allocations could be represented by production Edgeworth box. Contract curve in production Following the same logic as in the analysis of efficiency in consumption we can show that efficiency in production takes place if the two isoquants are tangent, i.e. x y . The locus of all points that are efficient in production is called MRTS LK MRTS LK production contract curve. Efficiency in production and PPF Once the economy produces efficiently, the only way to produce more of x is to give up some y . Thus we can represent efficient production in terms of produced outputs. Production possibility frontier (PPF) is derived from the production contract curve. Any point on PPF shows the maximum amount of one good attainable under any given output of the other good for given technologies and factors’ availability. qy 0 Production possibility frontier qx 115 A.Friedman ICEF-2022 The absolute value of its slope MRT xy can be calculated as a ratio of marginal products of the same factor in production of different outputs: MRT xy MPLy MPKy MRT or . Due xy MPLx MPKx MPKy MPLy to efficiency in production . MPKx MPLx This result is obtained immediately for one factor economy. Suppose production requires only x Q L x x dy one input - labour, then y Q y L y and MRT xy dx L x L y L PPC MPLy dLy MPLx dL x PPC Alternatively MRT xy could be calculated as a ratio of marginal costs: MRT xy MPLy . MPLx MC x . MC y Efficiency in output mix (allocation efficiency) For allocation efficiency the rate at which producers can convert y into x has to be equal to the rate at which consumers are willing to sacrifice y for x : MRTxy MRS xy . Suppose this is not the case and MRS xy MRTxy , then we should reallocate resources in favor of good x production as consumers value additional unit of good x more than it costs. Indeed, if we produce additional unit of x , then we have to reduce production of good y by . Let us change the bundle of consumer A leaving bundle of agent B unchanged. Then A will increase his consumption of x by one unit and reduce consumption of y by ; he was willing to sacrifice up to units of good Y for additional unit of x but the actual reduction in consumption of y was less, which means that he is better off, while welfare of B is unaffected. Thus, the initial allocation was not PO. qY PPF Pareto efficient allocation 0 qX General equilibrium with production 116 A.Friedman ICEF-2022 A set of prices constitutes general equilibrium if (1) every firm maximizes profits given its technology, (2) every consumer maximizes utility subject to his/her budget constraint, (3) demand equals supply for each commodity. The First Fundamental Theorem of Welfare Economics If there is a market for every commodity, optimizing behaviour of consumers and firms under perfect competition leads to a Pareto optimal allocation of resources. Let us show that interior GE is efficient in consumption, production and output mix. Equilibrium allocation is feasible as demand for each commodity equals supply. In equilibrium, the MRS between the two goods equals the price ratio. As both consumers face the same prices we get equality of MRS in equilibrium. As each consumer maximizes his B A px / py and MRS xy px / py , which utility subject to his budget constraint, then MRS xy A B implies MRS xy MRS xy . It means that the equilibrium allocation is efficient in consumption. Each firm maximizes profit taking prices as given. Cost minimization is a necessary condition of profit maximization. Cost minimization implies that MRTS between capital and labour in x production of each good equals the ratio of factors’ prices: MRTS LK w / r and x y y , i.e. efficiency in production takes MRTS LK MRTS LK w / r , which implies MRTS LK place. Then, from the consumer’s utility maximization problem MRS xy px / py . The profit maximization by firm x implies p x MC x and profit maximization by firm y implies py MC y . MRT xy As a result, efficiency in product mix takes place: MC x p x MRS xy . MC y p y Thus, the price system allows Pareto efficiency to be achieved in a decentralized setting. Nobody directs agent to choose particular combinations of inputs or consumption goods. Efficiency arises automatically as the outcome of a process in which each consumer and each producer observes prices and privately makes the decisions that maximize his or her wellbeing. This is the outcome of the ’invisible hand’ of the market. The Second Fundamental Theorem of Welfare Economics Provided that all indifference curves and isoquants are convex to the origin, for each Pareto efficient allocation of resources there is a set of prices that can attain that allocation as a general competitive equilibrium. 117 A.Friedman ICEF-2022 ~ ~ If (~ x, ~ y, K , L ) is Pareto efficient allocation, then it can be achieved as Walrasian equilibrium ~ and ~ with consumption good prices ~ r. p ,~ p and input prices w x y ~ p A B MRS xy 1) Choose ~x MRS xy . Then the FOC for UMP of each consumer is p y px ~ xA ~ py ~ y A and I B ~ px ~ xB ~ py ~ yB. satisfied. Required incomes: I A ~ ~~ r ~ py MPKy . Then FOC for firm y are satisfied. py MPLy and ~ 2) Choose w 3) A xy Efficiency in output mix implies MRS Thus MRTS ~ ~ px MPLx ~ py MPLy w x LK MRTS y LK and due ~ p p MPLy ~ ~x or MRT xy ~x . x py py MPL MRT xy to efficiency in production we have MPLy MPKy px MPKx ~ py MPKy ~ r . Thus we get or . It means that ~ x x MPL MPK FOC for profit maximization by firm y as well. 7.6 Sample exercise with solution Consider an economy with two agents A and B. The diagram below shows the indifference curves for the two agents in an Edgeworth box and the endowment point (E). In this case, the result of the Second Welfare Theorem does not hold. Is this true or false? Explain your answer. A B x2 OB x1 E ICA ICB OA G x1A x2B True. SFWT G is PO (preferred bundles for A lie above ICA and preferred bundles for B lie below ICB and these two sets do not have common points). 118 A.Friedman ICEF-2022 xB yA OB E ICA F ICB G O A x1A x2B Consumer B will max his utility in G iff BL is tangent to his IC at G, i.e. under BL given by GF. It means that if G can be decentralized then it can be done only under the prices given by this tangency line. At the same time under GF consumer B maximizes his utility at F, which results in excess demand for good x (and excess supply of y). Thus it is impossible to find the prices that will decentralize allocation G. The failure is due to nonconvexity of A’s preferences. Preferences are convex if for any u the set of weakly preferred bundles B(u ) x : ux u is convex, i.e. for any x B(u ) and x B(u ) we have x (1 )x B(u) 0, 1 . x2 H (H+G)/2 Consider B(u ) x : ux u - blue IC and above G x1 H B(u) and G B(u ) but H G / 2 B(u ) NOTE: convexity is a sufficient (not necessary) condition, that is, even in the absence of convexity some PO might be decentralized. 119 A.Friedman ICEF-2022 8. MONOPOLY 8.1 Pure monopoly Fundamental assumptions of pure monopoly Single firm is a price maker (affects the price by changing its output level) Buyers are price takers Price discrimination is impossible (each consumer pays the same price) Entry is blocked Profit maximization problem of a monopolist Let P Q represent inverse market demand function, then monopolists’ problem can be written as maxTR Q TC Q , where TR stays for total revenue: TRQ P Q Q . Q 0 The necessary condition for interior solution is given by MR(Q ) TR(Q) Q and the SOC: MR(Q ) MC Q . MC Q Q Q MRQ - marginal revenue, i.e. revenue that the firm gets from additional unit sold: MR(Q) TR(Q) P (Q) Q P(Q) P (Q)Q . Q Q Relationship between MR and inverse demand function: if Q 0 , then MR0 p0 p0 0 p0 as demand is diminishing, then p(Q) 0 and MR(Q) P(Q) P (Q)Q P(Q) , i.e. MR curve lies below the inverse market demand curve. Special case of linear demand function pQ A bQ . In this case MR curve is also linear and two times steeper than inverse market demand MRQ A bQ bQ A 2bQ . $ MRQ Q 120 A.Friedman MR(Q) ICEF-2022 TR(Q) P (Q) Q P (Q) P (Q)Q Q Q P (Q)Q 1 P (Q) 1 P (Q) 1 d (Q) P (Q) p , where dp denotes price elasticity of demand. As profit maximization implies MR(Q ) MC (Q ) , we can write down the following formula for profit maximizing monopolist price: 1 P (Q ) MC (Q ) 1 d . (Q ) p Note: MC 0 , which implies that monopolist will produce at a point, where 1 MC (Q ) MR(Q ) P (Q ) 1 d 0 . As price is positive, this inequality requires (Q ) p monopolist to produce where 1 d 1 0 , i.e. d 1 1 . Solving this inequality we get p (Q ) p (Q ) dp (Q ) 1 . As price elasticity is negative this is equivalent to dp (Q ) 1 . Conclusion 1: monopolist produces only at (price) elastic part of market demand. Conclusion 2: monopolist’s price is a markup over the marginal cost: d 1 1 P (Q ) MC (Q ) 1 d MC (Q ) as p (Q ) 0 , which implies that 1 d 1 . (Q ) p (Q ) p Equilibrium in case of monopoly: $ MC p profit B A Q P(Q) Q eff Q MR 121 A.Friedman ICEF-2022 Inefficiency of monopoly Note: monopolist’s output does not coincide with the competitive one as pQ MC Q , while at competitive output pQ comp MC Q comp . Monopolist solution results in DWL due to underproduction relative to the efficient level (consumers are prevented from buying units of the good that they value more than it costs to produce them). By moving from Q to Q eff Q comp gross CS increases by area A+B, while costs rise only by A, so TS increases by B. It means that equilibrium is inefficient: DWL=TS=B. Besides DWL, we may observe additional loss that comes from rent-seeking activity. As monopoly gets abnormal profit, each producer is willing to become a monopolist. As a result several firms compete for this monopoly rent and this rivalry involves resource costs that must be added to the value of DWL. 8.2 Sources of monopoly and regulatory responses: government franchise monopoly resource-based monopoly patent monopoly technological or natural monopoly monopoly by good management A franchise monopoly arises when a government grants the exclusive right to do business in a specified market to some firm. Resource-based monopoly power comes from the exclusive ownership of a natural resource essential in a particular production process. Regulator can force the monopolist to sell off some portion of essential resource, which makes competition a feasible market alternative to monopoly. Patent monopoly arises from the government action to enable inventors and authors to gain the exclusive right to their respective discoveries and writings by means of patent or copyright. Patents serve a potentially useful economic purpose by stimulating the invention and developments of new products. Eliminating patenting rights does not necessarily increase efficiency because the economic benefit to consumers of patented products might outweigh the economic cost of monopolistically exploited innovations. Regulation requires the optimal length of patent. Natural monopoly is usually defined as an industry where economies of scale make it cheaper to produce when there is one firm rather than several. (Example: public utilities) 122 A.Friedman ICEF-2022 Monopoly by good management occurs when the monopolist can deter entry and finds it profitable to do so. In fact this type of monopoly results from oligopolistic strategic interactions between the established monopoly and a potential entrant. In this case the monopolist will not produce the output that corresponds to equality of MR and MC. The firm would maximize the profit taking into account additional constraint that comes from the possibility of entry that results in oligopolistic competition. To deter entry firm may decide to produce output that exceeds the standard monopoly output. Regulatory responses to monopoly 1) Average cost pricing Objective: to limit monopoly to a fair, or normal, rate of return Problem: even though average cost pricing (if successful) does eliminate monopoly profit, it does not induce the monopolist to produce the efficient level of output. Both overproduction (a) and underproduction (b) relative to efficient output may take place. $ $ MC MC P(Q) P(Q) AC AC Q Q (a) (b) Moreover, the average cost pricing does not provide incentive to minimize the cost of production as under this policy monopoly covers the cost but does not get any profit. Finally, knowing that the regulator is going to use AC information to set price, a rational firm would misreport its AC curve in order to get positive profit. As regulation is based on private information, monopoly will use this information to its own advantage and report AC that induces the regulator to set monopoly price. 2) Rate of return regulation Rate of return regulation is aimed at limiting the rate of return a regulated monopoly can earn on its invested capital. Suppose that monopolist produces output with 2 factors of production, capital owned by the firm and labour hired by the firm. The monopolist’s total return on its capital is given by the difference between total revenue and labour costs: TRQ wLQ . Rate of return regulation imposes the following constraint on the firm’s behaviour: 123 A.Friedman ICEF-2022 TR( K , L) wL K , where is the allowed rate of return on capital Regulatory agency will choose that is not less than r as otherwise the monopolist’s profit would be negative and it will be forced out of business: K , L TR( K , L) wL rK r K 0 if r . If r , the monopolist will chose the combination of inputs that is not cost minimizing: it has an incentive to use too much capital and too little labour as with increase in the quantity of capital used the monopolist can get higher profit. 3) Marginal cost pricing combined with lump sum subsidy Regulator may set efficient price and if AC is above this level (it might be the case with natural monopoly due to diminishing AC) the policy could be combined with the lump-sum subsidy that allows to cover the gap. $ MC $ P(Q) MC P(Q) AC AC Lump sum subsidy Q (a) Q (b) Implementation problem is similar to the one discussed with respect to AC pricing: knowing that the regulator is going to use private information on MC to set price, a rational firm would misreport its MC and induce regulator to set monopoly price. 4) Per unit subsidy As unregulated monopoly under-produces relative to efficient output, government can create an incentive for output expansion by offering per unit subsidy. 124 A.Friedman ICEF-2022 $ MC P(Q) MC-s Q MR But the problem with private information is not eliminated. The efficient subsidy rate is given by s eff MC Q eff MRQ eff and Q eff is a solution of the equation pQ eff MC Q eff . Thus we need information on private marginal cost to find efficient subsidy rate, which creates an incentive for misreporting. 5) Two-tier pricing scheme Produce Q 0 and incur per unit costs AC 0 . Sell in two blocks: Q M at per unit price p M and Q 0 Q M at per unit price p0 such that profit from Q M balances the losses. p D MR pM AC 0 AC MC p0 QM Q0 Q 6) Efficient regulatory mechanism based on market demand only (optional) The scheme proposed by Loeb and Magat (1979)3 does not require private information to attain efficient output. The idea is to give monopoly a subsidy, equal to the value of CS. In this case, monopolist’s profit will be given by the sum of profit from sales (PS) and subsidy (CS). Thus, profit maximization would be equivalent to the maximization of TS. By definition, the output that brings maximum TS is efficient. Thus monopolist will choose 3 Loeb M. and W.A.Magat (1979) A Decentralized Method for Utility Regulation. The Journal of Law and Economics 22 (2), 399-404. 125 A.Friedman ICEF-2022 efficient level of output. To implement this policy regulator needs information on demand curve (which is not related to monopolist), so we eliminate the problem of misreporting. 8.3 Monopolistic price discrimination Price discrimination - selling different units at two or more different prices for the reasons not associated with differences in costs. Necessary conditions for profitable price discrimination: the firm must be a price maker; the firm must be able to identify which consumer is which (i.e. identify willingness to pay); the firm must be able to prevent consumers from engaging in arbitrage (arbitrage is the process whereby customers whom the firm charges low prices make purchases that they then resell to customers who would otherwise have to pay high prices); the transaction costs (the costs of meeting the second and third requirements) must be less than the benefits. Perfect or first degree price discrimination - the practice of selling each unit of output at a price just equal to the buyer’s maximum willingness to pay for that unit. Under first degree price discrimination monopolist leave consumers with zero CS as each unit is sold at a price equal to maximum willingness to pay. Thus monopolist gets TS and, trying to maximize it, he will end up with efficient level of output. Example. Suppose that the good is discrete, MC=const and there are two consumers with different demand functions. $ Profit from sales to agent A $ Profit from sales to agent B MC MC 5 3 126 A.Friedman ICEF-2022 Monopolist will sell 2 units to agent A at prices 14 and 8, which bring profit of 22-10=12 and 3 units to agent B at prices 12, 9 and 6, that brings profit of 27-18=9. Thus total profit is 21. He is not willing to produce the sixth unit, as additional cost of 5 can’t be covered since the maximum willingness to pay for additional unit is only 4. In case of continuous demand the same outcome could be achieved via two-part tariffs scheme, or bundling, that will be considered at the end of the chapter. Market separation (third-degree) price discrimination Assumptions: seller can observe the consumer’s willingness to pay and can charge different prices (based on differences in willingness to pay) but he is able to discriminate only between the groups of consumers; within the group each unit is sold at the same price. Third-degree price discrimination - the practice of identifying separate groups of buyers of a good and charging different per unit prices to these groups. Suppose there are two groups of consumers with inverse demand functions p1 q and p2 q , correspondingly. Note that if we know the price charged, then we can find the corresponding quantity sold to this group and vice versa. That is why we can maximize monopolist’s profit either with respect to prices or with respect to quantities. Let us follow the second approach, then the monopolist solves the following problem max TR1 q1 TR2 q2 TC q1 q2 , where TRi qi qi pi and i 1,2 . q1 0,q2 0 MR1 (q1 ) MC (q1 q2 ) The FOCs for interior solution: . MR2 (q 2 ) MC (q1 q2 ) Conclusion 1. If both groups are served then marginal revenues have to be equal: MR1 (q1 ) MC (q1 q2 ) MR2 (q2 ) . Conclusion 2. The segment with more elastic demand will be charged a lower price. q p 1 MRi qi pi qi pi pi 1 i i pi 1 pi i elasticity of group i demand. Proof. As MR1 (q1 ) MR2 (q2 ) , then p1 1 1 price elastic 1 2 , then 1 1 1 1 pi 1 , where i -price i 1 1 p 1 . Thus if group 1 demand is more 2 1 2 1 , which implies p1 p2 . 2 127 A.Friedman ICEF-2022 Graphical presentation of third-degree price discrimination $ $ $ p2(q) MR1 = MR2 MC p1(q) MR1 group 1 Q q2 q1 MR2 group 2 Market Welfare effects of third degree price discrimination The output produced under market segmentation is inefficient as at any segment pi qi MRi qi MC . Let us compare TS under pure monopoly and under market segmentation. The total output produced under market segmentation may rise, fall or stay the same as under pure monopoly. But we should take into account that under market segmentation, the output is inefficiently allocated among consumers. Conclusion 1. If price discrimination does not bring an increase in output, then TS segm TS monopoly. Conclusion 2. If price discrimination results in an increase in output, the change in TS is ambiguous: it may rise, fall or stay the same. Special case of linear demand functions and constant MC. Let us consider a special case of third-degree price discrimination, when demand curves are linear and marginal costs are constant. Claim 1. If all markets are served under uniform price then discrimination lowers (total) welfare in comparison with pure monopoly. Proof is based on direct calculation of total quantities produced in both cases. Let demand function at segment i ( i 1,, N ) be given by qi p ai bi p and MC c , where c ai / bi . To find output under pure monopoly, we need to find aggregate demand, derive the corresponding marginal revenue function and then equate MR with MC. 128 A.Friedman ICEF-2022 We look only at the segment of market demand where sales are positive at all markets, thus i ai Q d , inverse market demand is given by . As inverse Q ai p bi pQ b i i i i demand function is linear, MR would be also linear and two times steeper i ai 2Q . Monopolist’s output could be found from the equation MRQ bi i a 2Q b i MR Q М i М c . Thus we get Q М a i i c bi i . 2 i i Now let us find the sales under market segmentation. As we know, at each segment marginal revenue has to be equal marginal cost. Inverse demand function at segment i is given by pi q ai q/ bi . Thus marginal revenue at segment i is also linear and two times steeper MRi q ai 2q / bi . Profit maximizing sales corresponds to equality of group i marginal revenue q discr i and monopolists’ marginal ai cbi / 2 . Thus Q q discr i i a c b MRi qidiscr ai 2qidiscr / bi c cost: ai cbi / 2 i i i i 2 i or QM . As in both cases output is the same, TS discr TS M since under market segmentation the output is inefficiently allocated among consumers. Graphical proof for the case with 2 groups p p p p2(q) MR1=MR2 p1(q) Market demand MR MR2 MC q1 Q q2 Claim 2. If one group of consumers is not served under uniform price but is served under discrimination, then TS discr TS M . As Q M q2M , then MRtotal Q M MR2 q2M c . At the same time MR2 q2d c , which implies that q2M q2d but Q M q2M q2d q2d q1d Q d as by assumption q1d 0 . 129 A.Friedman ICEF-2022 Market separation Pure monopoly Comments q1d > q1M q 2d = q 2M Qd > QM Pd < PM due to diminishing demand CS1d > CS1M As q1d >0 and q1M =0 CS 2d = CS 2M As q2d q2M > PS d > PS M Monopolist under segmentation could charge the same price as pure monopolist but chooses different pricing policy TS d > TS M p q1d >0 and q1M =0 p p p2(q) MR1=MR2 MC MR p1(q) MR1 MR2 q1 q2 Market demand Q Multi-part pricing: two-part tariff In case of two-part tariff each consumer pays a fixed fee ( F ) for consuming any amount of the good plus a price per unit ( p ) for each unit consumed. If there is only one consumer, then for each per unit price p the maximum fixed fee equals to his CS. Thus firm is able to appropriate TS. As a result, firm finds optimal to set linear price p MC q and F CS q . As a result firm’s output is efficient. 130 A.Friedman ICEF-2022 P(q) F MC q In case of several homogenous consumers the optimal tariff is similar: the per unit price should be set at the level of marginal cost while the fixed fee should be equal to the individual’s CS. We can derive it algebraically. With N identical consumers market demand is given by N q d ( p) Q p . Then the profit-maximization problem of the monopolist could be written as: maxN F pQ( p) C (Q( p)) p,F s.t . F CS ind p q d ~ p d~ p. p It could be shown that the optimal contract will satisfy this constraint as equality, which results in the following equivalent setup ~ ~ d ~ ~ max Nq p dp pQ( p) C (Q( p)) max Q p dp pQ( p) C (Q( p)) . p p p p From the FOC Q( p) Q( p) pQ( p) C Q Q( p) 0 we get the per unit price p CQ and F CS ind p . Non-linear pricing in case of heterogeneous consumers Assume that monopolist can offer only one two-part tariff (alternative case with the menu of tariffs is considered later). In case of heterogeneous customers and unique two-part tariff, the monopolist may still find it optimal to set linear part of the tariff equal to MC. For example, this is the case if monopolist faces two (equal) groups of customers with large difference in the marginal willingness to pay. Then it is more profitable to deal with the high valuation group only (see figure below). 131 A.Friedman ICEF-2022 p p2(q) F p1(q) MC q1 q2 Another possibility corresponds to the situation, when the difference in marginal willingness to pay is not too big, but the number of customers with higher marginal willingness to pay is much bigger than the number of buyers with lower willingness to pay. Now, let us assume that we deal with two groups of equal size and the difference in willingness to pay is not too high. Let us illustrate that monopolist will benefit from charging linear price above marginal cost. $ P2(q) P1(q) A C B E D F MC q First of all we can show that it is more profitable to deal with both groups of consumers rather than only with the high valuation one. The optimal tariff in case of selling to the highest valuation (i.e. second) group only is to set linear price equal to MC (as a result the profit from sales is zero) and the fee equal to the corresponding value of consumer surplus: p0 MC and F0 A B C D F E . Resulting profit equals to the value of fee: 0 F0 . If the monopolist will decrease fee up to the value of CS of group 1 ( F A B D) under the same linear price, its profit will increase: 2F 2( A B D) 0 as A B D C E F . So it will sell to both groups. Now, we can illustrate that it is not optimal to set price equal to MC. If monopolist increases ~ ~p linear price to and sets then its profit equals F A, ~ 2 A B B D E ( E D) . So, it is profitable to raise linear price above MC. 132 A.Friedman ICEF-2022 Optimal 2-part tariff in case of heterogeneous consumers There are two options that must be considered. Option 1: sell to high-valuation group by charging p MC and setting fee equal to the CS of the high-valuation group F CS2 p MC . Option 2: sell to both groups by charging per-unit price above MC and setting fee equal to the CS of the lowest valuation group: m ax2 F p q1 p q2 p C q1 p q2 p p, F F m inCS1 p , CS2 p Finally, we should compare the resulting profits and choose the option that generates higher profit. We can note that in any case the total output produced under this scheme is below the socially efficient one. Second-degree price discrimination In reality no firm can observe every individual buyer’s willingness to pay for its product. In this case a monopolist can use the consumer’s own actions as a basis of discrimination. Second-degree price discrimination - is the practice in which the same price schedule is offered to all buyers but they sort themselves through self-selection. Example with bundling. Suppose we have two groups with linear demands but the second group has higher willingness to pay (see graph below). For simplicity let’s assume that we have one agent in each group and MC=0. Each consumer knows his/her willingness to pay but the monopolist cannot identify the type of each customer. Had the types been observable, then as we know from the previous discussion, the monopolist would sell under linear price equal to MC (zero in this case) and charge fees equal to values of individual’s CS: Fee1 A and Fee 2 A B С . In this case monopolist’s profit is 2 A B С . In fact it can achieve the same profit by bundling: if it offers to the first group the bundle with q1 units at price of A and to the second group bundle q 2 units at price of A B С , the outcome would be the same. 133 A.Friedman ICEF-2022 $ P2(q) P1(q) A B C q But as the types are unobservable, the monopolist is unable to implement this scheme. Now it has to offer the same menu of bundles to all customers. If it offers the two bundles found above, the high valuation agent (agent 2) would definitely take the bundle designed for group 1 as it gives him net CS of B , that is, full information bundles are not incentive compatible. We say that a contract is incentive compatible if it gets the agent to make intended choice. As a result, the monopolist’s profit would go down to 2 A . In fact, the monopolist can earn more by changing the contract designed for consumer 2 in such a way that consumer would find it optimal to choose it (that is, by proceeding to incentive compatible contracts). Incentive compatible contract should provide agent 2 the net CS not less than B . This can be done by reducing the price for the large package with q 2 units up to A C . In this case each consumer self-selects by choosing the contract designed for his/her type and monopolist’s profit would be 2 A C . Is it the maximum possible profit? No, we can change the package designed for group 1 to make it less attractive for the second group. In order to do this we reduce the quantity included in this package and the price: q 1 q1 , A A . Now, if agent 2 takes this contract, his CS will be equal to B B . Thus this contract is less attractive that before. This gives a possibility to increase the price of the large package by B : q2 , A C B . The resulting profit is 2 A C B A . Thus, the monopolists’ profit was increased by B A 0 . 134 A.Friedman ICEF-2022 $ P2(q) P1(q) B B A A C q Now, we can generalize the result of this exercise. At the optimum, the marginal profit should be zero. When q1 0 then A P1 q1 and B P2 q1 P1 q1 . Thus, the optimal ~ P q ~ P q ~ . quantity in the small package should solve the following equation: P q 2 1 1 1 1 1 Now, let us consider a more general case, when the two groups may differ in size. Denote the share of high-valuation group by , then the share of low-valuation group is 1 and the total number of customers – by N. The optimal contracts could be derived from the following profit maximization problem: max N 1 T1 T2 T1 ,q1 ,T2 ,q2 s.t. CS1 q1 T1 0 (PC1) CS2 q2 T2 0 (PC2) CS1 q1 T1 CS1 q2 T2 (IC1) CS2 q2 T2 CS2 q1 T1 (IC2) The first two constraints guarantee that agents will purchase the good since the CS from purchasing is non-negative. In addition to participation constraints we need incentive compatibility constraints (IC) that guarantee that agents cannot gain from purchasing the ‘wrong’ bundle, that is, the bundle that was designed for the consumer of other type. Our previous analysis suggests that (PC1) and (IC2) are binding. From (PC1) we get Plug it into (IC2) to obtain T1 CS1 q1 . T2 CS2 q2 CS2 q1 T1 CS2 q2 CS2 q1 CS1 q1 . Finally we plug both T1 and T2 into objective function and rearrange: max CS1 q1 CS 2 q2 CS 2 q1 q1 ,q2 The FOCs are given by: CS1 q1 CS 2 q1 0 and CS 2 q2 0 . 135 A.Friedman ICEF-2022 As a result we get the following equations for q1 and q 2 : P1 q1 P2 q1 and P2 q2 0 . If we compare these bundles with the first-best ones (i.e. the contracts obtained under full information) we can see that the quantity for high-valuation consumers is the same, while the quantity designed for the low-valuation is reduced below the socially efficient level. 8.4 Sample problem with solution A price discriminating monopolist allocates its output between domestic market and foreign market. The monopolist’s marginal cost schedule is rising and marginal revenue curves are declining on each market and current domestic price is below the foreign price. Suppose that the government decided to tax export by introducing a per-unit export duty equal to t . Assume that sales at both markets are positive both before and after the introduction of export duty. (a) Analyze graphically the impact of this policy on domestic sales assuming that all curves are linear (you are expected to provide some comments to the graphs). (b) Derive the analytical solution for general (non-linear) case for the impact of the policy on domestic sales. [Differentiability of MR and MC is not assumed]. Solution $ Foreign $ Home $ MRF = MRd PF(q) MC Pd(q) qF MRF-t MRF qd MRF –t = MRd Q MRd Step 1. We equalize the values of MR across the two markets (monopolist will sell at both only if MR is the same). Step 2. Find the optimal production by intersecting equalized MR with MC. Per unit tax reduces MR from export. As a result it affects the locus of points where MRs are equalized and this new (blue) line intersects increasing MC at lower value of MC. Domestic MR curve is unaffected and we need lower value of MR which happens at increased domestic sales. (b) Domestic sales will go up. 136 A.Friedman ICEF-2022 MR F (q F ) t MC (q F q d ) Proof. In equilibrium with export duty we have . MRd (q d ) MC (q F q d ) Note that initially t 0 and then it increases by t 0 . Let us prove that q d 0 . Suppose that this is not the case and q d 0 , then MRd 0 as MR is diminishing. In equilibrium MC MRd 0 . (**) As MC is increasing, then Q qd q F 0 . Thus q F Q qd 0 . MR F is diminishing then MRF t 0 . This implies MC MRF t 0 which contradicts to (**). Contradiction proves that q F 0 . 137 A.Friedman ICEF-2022 9. OLIGOPOLY We start our analysis of oligopolistic markets by looking at firms that compete with homogeneous products. 9.1 Cournot model with N firms Assumptions: firms compete by setting quantities, firms make their choice simultaneously, firms produce homogeneous product, entry into the market is completely blocked. We are going to look at the simplest example of Cournot model with N firms and assume that both firms have identical CRS technologies (cost function TCq cq ) and market demand is linear pQ A Q , where A c . To find Nash equilibrium we need to derive the best response function for each firm. But as the firms have identical technologies, the corresponding best response functions would be symmetric. The best response function of firm i could be derived as a solution of its profit maximisation problem, where rivals’ outputs are treated as exogenous parameters: m ax A qi q j qi cqi . qi 0 j i Note that this function is strictly concave in q i , thus FOC is both necessary and sufficient. FOC: A 2qi q j c 0 . j i Note that we can rewrite it as qi A Q c . As the RHS is the same for any i then qi q for all i 1,, N and we get q A Nq c , which gives q Q N A c 1 N i ( P c )q and price equals P A Q A Nc . 1 N Ac . Industry’s output is 1 N Profit of firm i is A c 2 . 1 N 2 Note that perfect competition can be treated as a limiting case of Cournot oligopoly when N A Nc A / N c goes to infinity: P A Q c . 1 N 1 / N 1 N The case of N 2 could be illustrated graphically as an intersection of the two best response functions. 138 A.Friedman ICEF-2022 Cartel NE Both firms get q1 higher profit Isoprofit lines of firm 1. Equation A q1 q2 q1 cq1 1 or q 2 A c q1 straight line 1 q1 hyperbola q2 A-c A-c-q1 A c q1 / q1 q1 / q1 Note, that profit of firm 1 increases as we move down along its reaction curve (i.e. when competitor’s output falls and as a result market price goes up). Point of maximum of iso-profit curve always lies on the reaction function. The reason is simple: for any output of the competitor (firm 2) the profit maximising output of firm 1 is given by the corresponding point on reaction function and any other output results in lower profit and so should lie on the higher iso-profit line. As we can see both firms can get higher profits by producing less (shaded area). Thus we have a problem similar to prisoners’ dilemma: if firms could collude, they would benefit from the collusion. 139 A.Friedman ICEF-2022 Collusion If the total output is chosen collectively by the two firms (rather than individually), i.e. firms create a cartel and choose the output by maximising joint profit, then max A Q Q cQ . Q 0 The first order condition requires the equality of market marginal revenue to marginal cost, which implies MRQ A 2Q c and Q cartel A c / 2 A c / 3 Q Cournot . Note that as both firms have identical CRS technologies the total output can be produced at any plant. Suppose the cartel’s profit is split equally and q1Cartel q2Cartel A c / 4 . This point is illustrated on the above diagram. Note that cartel solution is characterized by a tangency of the firms’ isoprofit lines. Comparison of Cournot oligopoly, perfect competition and monopoly Cournot Duopoly: Q Cournot A Nc N A c , p Cournot A Q . 1 N 1 N Monopoly (collusion): Q M A c / 2 , p M A c / 2 . Perfect competition: Q comp A c , p comp c . QM A c Q Cournot N A c A c Q comp 2 1 N and p M p Cournot p comp . As a result TS M TS Cournot TS comp since all the market structures are efficient in production (marginal costs are equalized) and the only reason for efficiency loss comes from the violation of allocative efficiency in case of monopoly and oligopoly. 9.2 The Stackelberg model Assumptions: firms compete by setting quantities, firms make their choice sequentially, firms produce homogeneous product, entry into the market is completely blocked. In the Stackelberg model one firm, the leader, sets its output first and the other firm (or firms) reacts. This is an example of sequential game and we will look for the perfect Nash equilibrium that requires using backward induction. The leader uses its first mover advantage by forcing the rival to produce the quantity that, combined with the leader’s output, results in the most profitable outcome for the leader. 140 A.Friedman ICEF-2022 To find the equilibrium, we need to derive the best response function of the follower. Let firm 1 be the leader and firm 2 – the follower. If we use the same assumptions about demand function and technologies as in Cournot model ( P Q A Q, AC i MC i c A ), then the reaction function of the follower is exactly the same is the one in Cournot model: A c q1 q 2 q1 . 2 2 Now let us look at the leader’s problem. He is trying to maximize his profit subject to best response function of the follower: m ax A q1 q 2 q1 cq1 q1 s.t . q 2 . A c q1 2 2 By plugging the expression for q 2 into objective function we get A c q1 A c q1 max A q1 q1 cq1 max q1 cq1 . q1 q1 2 2 2 2 Ac Ac . Plugging q1 c , which gives q1St 2 2 into the best response function of firm 2 we obtain the output of the follower A c q 2St . q 2St 2 2 The first order condition takes the form Comparison with Cournot We can see that the leader expands his output in comparison with Cournot case Ac Ac q1St q1Cournot , which induces the follower to produce less than in 2 3 A c q 2St A c A c simultaneous game q 2St q 2Cournot . Total output increases 2 2 4 3 3( A c ) 2( A c ) and price goes down Q St q1St q2St Q Cournot 4 3 A 3c A 2c P St P Cournot . 4 3 Profit of the leader goes up leader A c 2 goes down follower 16 A c 2 8 A c A c 2 9 1Cournot and profit of the follower 2 9 . Cournot 2 Question. Explain intuitively, why the leader can increase the profit. We can illustrate equilibrium graphically as a point of tangency of isoprofit line of firm 1 (leader) with reaction function of firm 2 (follower). 141 A.Friedman ICEF-2022 q2 Cornout Stackelberg q1 9.3 Price-setting oligopolists: Bertrand model with homogenous good Assumptions: firms compete by setting prices, firms make their choice simultaneously, firms produce homogeneous product, entry into the market is completely blocked, each firm has CRS technology, i.e. MC=AC=c. We start with a classical duopoly model with symmetric firms, that is, the case where MC1 MC 2 c . As a result the best response functions will be symmetric. 0 if pi p j Residual demand of firm i = Q pi / 2 if pi p j . Q pi if pi p j Let us consider the best response of firm i to any price chosen by firm j : If firm j charges price less or equal to MC , than the best response is to charge price, equal to MC (or higher). If firm i responds by lowering the price it will keep all the market and incur losses as the price is below its marginal costs. Monopoly If firm j charges price above MC but less (or equal) to P , then it is optimal to undercut this price by 4 and serve all the market. 4 Here we implicitly assume that prices are measured in discrete units and is the smallest unit like one cent. 142 A.Friedman ICEF-2022 If firm j charges price that exceeds the monopolist one, then firm i responds by setting the monopolist price. This policy attracts all the customers and results in the maximum possible (monopolistic) profit. These best response functions are illustrated below. p2 MC NE 45 MC p1 Nash equilibrium: ~ p1 ~ p2 MC . As a result each firm serves one half of the market ~ ~ ~ q q Qp/ 2 and profit of each firm equals zero. 1 2 Conclusion: price competition is more severe than quantity competition. The reason is that with identical products the firm that charges lower price takes all the market (not just part of the market). Bertrand paradox: if MС1 MC 2 с then p1 p2 с and firms do not get any profit. Solution: firms may have capacity constraints firms interact repeatedly and collusive outcome can be sustained in the infinitely repeated game if firms are patient enough. 9.4 Price leadership or Dominant firm model (optional) Assumptions: firms make their choice sequentially, leader sets the price and competitive fringe responds by choosing quantity, firms produce homogeneous product, entry into the market is completely blocked. Instead of setting the quantity, the leader may set the price. To find the optimal price the leader has to take into account the response of other firms (competitive fringe) that take this price as given and choose the quantity produced. Output produced by competitive firms determines the residual demand that could be served by the leader. 143 A.Friedman ICEF-2022 As usually we look for the perfect equilibrium by solving this model backward. The problem of competitive fringe: max pq f TC qf f q . The first order condition is f p MC f q f . Thus, the non-diminishing part of MC f gives the supply of competitive fringe S p . Now, we can find the residual demand as the difference between market demand and supply of competitive fringe: Q res p maxQ p S p,0. Dominant firm will choose price by maximising its profit subject to the residual demand, that is, it acts like monopolist on residual demand. Problem of dominant firm: max pQ res p TC dom Q res p p or it could be equivalently rewritten as max QP res Q TC dom Q . Q The first order condition requires equality of MR for the residual demand and dominant firm’s marginal cost. Let us illustrate equilibrium graphically for the case of linear demand and marginal cost curves. P P(Q) MCfringe Pres(Q) Qfringe MCdom Qdom Q res MR 9.5 Repeated interactions Bertrand paradox: in a symmetric Bertrand game each firm gets zero profit, while had both firms agreed to charge the monopolist price, each would get half of the monopolist profit. This problem is similar to prisoners’ dilemma. In fact, firms set their prices each period, that is, they interact repeatedly. With repeated decision making firms can base their decisions at a given time on the actions that have been taken in the past, that is, firms can react to their rivals’ behaviour. This might help to achieve the tacit collusion as, if one firm cheats on the agreement, another firm may be able to punish it in the future. Whether this kind of strategy will be viable depends on whether the game is 144 A.Friedman ICEF-2022 going to be played a fixed (known) number of times or indefinite (or unknown) number of times. Finitely repeated game Let us suppose that both players know that the same (for example Bertrand) game will be played N times ( N is finite). What will be the outcome? As the game is dynamic, we look for the perfect Nash equilibrium using backward induction. Consider the last round. As this round is the final one and everybody knows it, then there is no incentive for cooperation and every player will choose the static game Nash equilibrium strategies by charging price equal to MC. Now consider what will happen on round N 1 . As at the last round there will be no cooperation there is no incentive to cooperate at this round as well. If one cooperates by charging the monopolist price the rival will find it optimal to cheat by charging lower price and getting all the market. Each player has the incentive to deviate and as a result the only equilibrium of this subgame is given by a static Nash equilibrium, where each firm charges price equal to MC. The same logic proves that there would be no cooperation at each round and the only perfect equilibrium corresponds to prices equal to marginal cost. The result is not surprising as players cooperate only if there is a punishment for cheating. With finitely repeated game at the last round cheating cannot be punished and this creates incentive for deviation at each round. Infinitely repeated game If the game is infinitely repeated, then the last round does not exist and as a result deviation at any point of time could be punished in the future. For punishment strategies to be effective, they must be both severe and credible. The punishment is severe enough to deter cheating by a firm if the cost of cheating (i.e. loss from punishment) outweighs the benefit. The cost is associated with the loss in profits due to the punishment. As profit could be reduced in a number of periods, we should look for the total reduction in profit, taking into account that profits of different periods have to be summed up with discount factor (discount factor reflects the current value of future profit). Thus, cost of punishment is the present value of the stream of forgone profits that results when cheating is detected and punishment is implemented. Forgone profit in each period is calculated as a difference between the profit that is earned under collusion and profit that is earned if punishment takes place. For example, if cheating takes place at period t and punishment starts from t 1, then the cost of punishment is given by collusion punish Loss from punishment where t 1 t 2 collusion punish t 1 1 , 1 and r stays for the market interest rate. 1 r 145 A.Friedman ICEF-2022 The benefit of cheating is the present value of the stream of additional profits that are enjoyed while cheating goes undetected. Each period benefit is given by a difference between the profit earned while cheating and the profit under collusion. In our example cheating is detected in the next period, so the benefit from cheating is derived only in one period: dev collusion t . Benefit from cheating Thus the punishment is severe enough to deter cheating if dev collusion t dev punish Solving this inequality, we get t 1 or dev collusion collusion punish . 1 1 dev collusion . Note that this condition has to be satisfied dev punish for each player. The punishment is credible if it is in the non-cheating firms’ self interest to implement the with pumishment no pumishment punishment when cheating is detected, that is, nondeviating firm nondeviating firm . Infinitely repeated Bertrand duopoly (with symmetric MC) Each firm uses the following grim-trigger strategy: it charges monopolist price in period t if there were no deviations in previous periods and switches to one-period Nash equilibrium (i.e. sets price equal to MC) otherwise: m p , t 0 or pi t c otherwise pi t 1 p j t 1 pm The strategy of this sort is called Nash-reversion strategy. Let us check whether these strategies are severe and credible. Assume that deviation is revealed in the next period. Optimal deviation corresponds to price which is a bit lower than monopolist and as a result deviating firm will serve all the market and gets almost the monopolist profit dev mon . Under tacit collusion each firm gets half of the monopolist profit collusion mon / 2 . Thus, the benefit from deviation is dev collusion mon mon / 2 mon / 2 . Had the punishment been implemented, each firm gets zero profit and as a result cost of mon cheating equals mon / 2 0 2 /2. 1 1 1/2. mon / 2 which is equivalent to 1r 1 This inequality is satisfied for any interest rate below 100%: r 1 . No deviation condition: mon / 2 Check credibility: if punishment is implemented, then profit of each firm becomes zero. If non-deviating firm does not punish the cheater and continues to set monopolist price, then nobody will purchase from this firm and its profit will be zero. Thus punishment does not decrease the profit of the punisher, which implies that it is credible. 146 A.Friedman ICEF-2022 Conclusion: under r 1 Nash reversion strategies constitutes perfect Nash equilibrium in repeated Bertrand game. Question. In the framework of infinitely repeated Bertrand model analyse how the likelihood of collusion is affected by the number of firms in the industry. 9.6 Bertrand model with differentiated goods So far we looked at the strategic interactions between firms that produce identical products. Now we are going to look at product differentiation. Differentiated products are goods that satisfy a particular need but differ in their individual specifications. Assumptions: firms compete by setting prices, firms make their choice simultaneously, firms produce differentiated goods, entry into the market is restricted by positive product development costs. Suppose that firms 1 and 2 are duopolists that compete by setting prices simultaneously. They have identical cost functions with MC=AC=c. Firms face the following demand functions Q1 A p1 p 2 and Q2 A p2 p1 , where A c . Thus, demand for the good produced by firm i depends negatively on its own price and positively on the price of its close substitute. As the two firms act simultaneously, we look for the Nash equilibrium that can be represented as an intersection of the two best response functions. Let us find the best response of firm i given that firm j charges p j by solving the following profit maximisation problem: max( pi c)( A pi p j ) . pi 0 The first order condition gives: A pi p j ( pi c) 0 or pi 0.5 A c 0.5 p j . Thus, the best response functions are upward sloping. Since demand functions are symmetric and cost functions are identical, the resulting prices would be equal: p1 p2 A c c . The corresponding outputs and profits are given by: Qi A and 1 2 A2 0 . As we can see, symmetric price competition may result in positive profits if firms produce differentiated products. Note: under sufficiently high product development costs (F), firms earn positive profits even in the long run. 147 A.Friedman ICEF-2022 NE Sequential pricing with differentiated goods Suppose now that firms interact sequentially. Let firm 1 move first and choose p1 , then firm 2 observes p1 and chooses p2 . We solve this game using backward induction. Best response function of firm 2: p2 A c p1 . 2 2 Firm 1 will maximize its profit taking into account the best response of the second firm: max( p1 c)( A p1 p2 ) max( p1 c)1.5 A 0.5c 0.5 p1 p1 p1 s.t. p2 0.5 A 0.5c 0.5 p1 FOC: 1.5 A 0.5c 0.5 p1 0.5 p1 c 0 . Thus p1 c 1.5 A and p2 0.5 A 0.5c 0.5c 0.75 A c 1.25 A . Plugging into demand functions we get the following equilibrium outputs: Q1 A c 1.5 A c 1.25 A 0.75 A , Q2 A c 1.25 A c 1.5 A 1.25 A 2 1.25 A 1.25 A and profits 9 1 1.5 A 0.75 A A2 8 and 25 2 A . 16 We can see that profit of each firm was increased in comparison with the simultaneous game and the firm that moves second has gained more than the first-moving firm: 25 2 9 2 2 A A 1 . 16 8 148 A.Friedman ICEF-2022 To illustrate the sequential equilibrium graphically we need isoprofit lines. Isoprofit curve for firm 1 is given by ( p1 c)( A p1 p2 ) , which can be restated as p2 ( p1 A). When profit increases the isoprofit line shifts upward. p1 c p2 p1 c Equilibrium corresponds to the point of tangency of isoprofit line of the leader (firm 1) and reaction function of the follower (firm 2). Simultaneous Sequential 9.7 Sample exercise with solution Consider a repeated Bertrand model with two risk-neutral firms that have identical technologies. Suppose that at each period there is a probability 0, 1 that the firms keep competing next period and with probability 1 market can seize to exist in any future period. Firms discount the future, so that, for each firm, a payoff of $1 received t periods from 149 A.Friedman ICEF-2022 today is worth $ t today, where 0 1 .How an increase in the probability affects the likelihood of tacit collusion? Your answer must specify the strategies that the firms should follow to sustain tacit collusion. Solution Punishment strategies (i=1,2): m m p , t 0 or pi t 1 p j t 1 p pi t c otherwise Optimal deviation: pm brings profit m and the other (non-deviating) firm gets zero profit as nobody is willing to purchase at higher price. Credibility: non-deviating firm gets zero profit in both cases with punishment (as it sells at p=AC) and without punishment (in this case it has zero sales). Thus punishment is credible as doesn’t reduce its expected profit. No deviation condition: m i 0 2 i m / 2 1 m 0 , 1 2 2 , 1/2 . i i 1 If goes up then the minimum level of discount factor required to sustain collusion will be lower so that the interval expands. Thus it would be more likely that the existing discount rate belong to this interval. Intuition is straightforward: deviating firm gain some profit today but is punished by the lost future expected profits. An increase in increases the expected value of these future profits and make punishment more severe, as a result firm has lower incentive for deviation. 150 A.Friedman ICEF-2022 10. FACTOR MARKETS 5 10.1 Demand for factors We get individual firm’s demand by solving its profit maximization problem max TRQK , L TC K , L K 0, L 0 FOC for interior solution: TR QK , L TC K , L TR QK , L TC K , L and , L L K K MRPL MFC L MRPK MFC K That is, the marginal revenue product of a factor (MRP) should be equal to the marginal factor cost (MFC). Marginal revenue product of factor i : MRPi TR Qx TR Qx Qx MR MPi . x i Q x i If output market is perfectly competitive, then MR p and MRPi pMPi . If factor i market is competitive, then MFC i TC x wi . x i Thus, if all (input and output) markets are perfectly competitive, then factor demands can be found from the solution of the system pMPi wi . Consider a model with two factors of production, capital and labour. Assuming diminishing marginal product of each factor and fixing the amount of capital used, the value of marginal product of labour gives the inverse labour demand curve (see the figure below). 0 5 In the new syllabus there is no separate topic on factor markets: demand is studied in the theory of firm, supply is studied in consumer theory and equilibrium is studied in different topics on market structure. 151 A.Friedman ICEF-2022 Firms’ demand curve for a factor and industry demand curve. To get the industry demand curve for a factor (in case of perfectly competitive industry) we should take into account the product-price effect. To simplify analysis let us assume that labour is the only factor of production. As individual firm’s demand for an input is given by its MRP curve, then we should sum up the individual demands by taking the horizontal sum of MRP curves. But the resulting curve does not correspond to industry demand. Suppose initially industry economy was in the equilibrium with output price p0 , input price w0 and market labour demand L0 . If wage rate falls, then each firm will demand more labour and we move to L . But as employment goes up, the quantity produced increases, which brings a fall in output price from p0 to p1 . Thus, employment rises only till L1 . This analysis suggests that the industry supply curve is steeper (less price elastic) than the horizontal sum of the firms’ demand curves. Industry demand horizontal sum of horizontal sum of 0 Conclusion 1. The lower the elasticity of demand for the product, the lower the elasticity of demand for a factor. Supply (more price elastic) 0 0 Explanation. With the more price elastic demand for the final product (flatter output demand curve), the price fall would be smaller and the resulting downward shift of the MRP sum would be less, which results in larger employment and flatter (more price elastic industry factor demand). 152 A.Friedman ICEF-2022 Firms’ demand curve for a factor in the SR and in the LR In the short run capital is fixed and MRPL K 0 gives the firm’s demand for labour. But in the LR capital adjusts and as a result MRPL curve shifts. Two inputs are said to be complementary (anticomplementary) if the increased use of one input raises (reduces) the marginal product of the other. If one input has no effect on the marginal product of the other input, then factors are said to be independent. Let the firm’s long run quantities demanded be given by L0 , K 0 and the initial wage rate is w w0 . Suppose wage rate falls till w1 w0 . In the SR firm will demand more labour, which can be illustrated by movement to the right along the SR demand. If factors are not independent, then MPK would be affected and capital will adjust in the LR, which, in its turn, would affect the position of MRPL curve. Here, two cases should be considered separately. (a) K and L are complementary. As labour increases in the SR, the marginal product of capital increases and as a result MRPK r . Thus firm will decide to use more capital, that is, K 1 K 0 . As a result marginal product of labour goes up and brings an upward shift of MRPL curve (since MRPL K 1 MRPL K 0 ). As marginal productivity of labour goes up firm finds it profitable to use more labour, thus in the LR labour increases more than in the SR (b) K and L are anticomplementary. LR demand 0 LR demand 0 (a) complementary factors (b) anticomplementary factors As labour increases in the SR, then marginal product of capital falls and as a result MRPK r . Thus firm will use less capital, that is K1 K 0 . As a result marginal product of labour goes up and brings an upward shift of MRPL curve (since MRPL K1 MRPL K 0 ). 153 A.Friedman ICEF-2022 As marginal productivity of labour goes up, the firm finds it profitable to use more labour, thus in the LR labour increases more than in the SR Summary: demand curve is flatter (more price elastic) in the LR than in the SR. Conclusion 2. Given either complementarity or anticomplementarity between inputs, the demand curve for an input is flatter (more price elastic) in the LR than in the SR (when some factors are fixed). Conclusion 3. Factor demand curves always have negative slope (i.e. there is no such thing as a “Giffen factor”. Analytical proof. Let firm use two factors of production L, K , with corresponding prices w and r . Let us fix the output price p and consider two different vectors of factors prices (w 0 , r ) .and (w, r ) Denote by ( L0 , K 0 , Q 0 ) and ( L, K , Q) the solutions of profit maximisation problems under (w 0 , r ) and (w, r ) , correspondingly. If ( L, K ) brings maximum profit under given prices, then any other combination of factors cannot result in higher profit and similarly for ( L0 , K 0 ) : TR(Q ) wL rK TR(Q 0 ) wL0 rK 0 TR(Q 0 ) w 0 L0 rK 0 TR(Q ) w 0 L rK . Adding up these inequalities we get: wL w 0 L0 wL0 w 0 L or 0 w( L L0 ) w0 ( L L0 ) (w w0 )(L L0 ) . Last inequality means, that if w w 0 under constant price of other input, then demand for labour either fall or stay the same: L L0 . The difference with consumer theory: when price of a factor goes up, the initial combination of factors is still affordable for the producer as he does not face financial constraint. Demand for a factor and supply of other factors. To simplify the analysis let us assume that product-price effect is negligible. Let the industry be in the long-run equilibrium initially with w0 , r0 and factors’ employment of L0 and K 0 . Suppose that wage rate falls till w1 w0 . In the SR industry labour employment goes up from L0 to LSR 1 . If factors are not independent, then MPK would be affected and capital will adjust in the LR, which, in its turn, would affect the position of the industry labour demand curve. Here, two cases should be considered separately. (a) K and L are complementary. As labour increases in the SR, the marginal product of capital increases and as a result MRPK r . Thus, the demand for capital goes up, which results in an increase in capital 154 A.Friedman ICEF-2022 employment: K 1 K 0 . As a result marginal product of labour goes up and brings an upward shift of MRPL curve for each firm (since MRPL K 1 MRPL K 0 ). As marginal productivity of labour goes up each firm will use more labour and industry demand curve shifts to the right. Finally, the labour increases from L0 to LLR 1 . complementary factors case (more price elastic) 0 0 Labour market Capital market (b) K and L are anticomplementary: This case brings the same result in terms of labour but K1 K 0 (FOR SELF-STUDY). Let us analyze the role of elasticity of capital supply curve. With flatter (more price elastic) supply of other factor its employment will change more in response to the increased demand, which results in greater shift in the SR labour demand curve. As a result, LR industry demand for labour would be flatter (more price elastic). Conclusion 4. An industry demand for a factor is less elastic the less elastic is the supply of other factors. 10.2 The supply of factors and competitive equilibrium Individual supply curves could be derived from consumer’s utility maximisation problem (it was studied in the theory of consumption). Although individual labour supply could be backward bending, the market supply curve is likely to be upward sloping. Explanation: even if some workers that are currently in the industry may prefer to work fewer hours as the wage rises, new workers will be attracted (from other industries or from leisure). Elasticity of supply of a factor to a particular use depends: on the degree of specificity of the factor to this use; 155 A.Friedman ICEF-2022 on the length of time allowed for the factor to be reallocated to or away from that use. Factor market equilibrium Economic rent Transfer earnings L Input earnings can be divided into two components: transfer earnings and economic rent. Transfer earnings represent the amount that any unit of a factor must earn in order to prevent from being transferred to another use. Economic rent is any excess over transfer earnings that a unit actually earns. Factor payments which are economic rent in the SR and transfer earnings in the LR are called quasi-rents. 10.3 Monopsony and monopoly in factor markets Monopoly on the supply side of the input market. By combining into a trade union, workers (with no market power individually) may collectively be restricting the supply of labour, raising their wages. Examples of trade union objectives: i. the rent maximization, ii. the total wage bill maximization. Consider labour market with linear demand curve for labor LD A bw and linear labour supply curve: LS cw , where A c . Suppose that the labor market is controlled by monopolistic labour union that maximizes economic rent. Economic rent is any excess over transfer earnings that a unit actually earns max L 0 L ~ ~ w D ( L) L w S ( L )dL , 0 where w S ( L) L / c and w D ( L) (a L) / b . 156 A.Friedman ICEF-2022 The first order condition for interior solution gives: or MRL MC L w D ( L) Lw D / L w S ( L) . Plugging expressions for w S () and w D () , we get ac ab 2ac ac a(b c ) and w w D ( L) . (a 2L) / b L / c . Finally we get: L b 2c b(b 2c ) b(b 2c ) L Now, suppose instead of rent trade union maximizes total wage bill: max L0 w D ( L) L Total wage bill maximization results in choosing employment level, where marginal revenue equals 0: MRL ( A 2L) / b 0 or L A / 2 and w 0.5 A / b . Note: employment is greater in case of wage bill maximization model: Lwage bill A A Lrent . 2 2b/c The problem with wage bill maximization model: it might happen that the wage rate does not cover the opportunity unit labour cost. This is the case if 0.5 A / b Lw.bill / c 0.5 A / c , i.e. b c. Monopsony on the demand side of the input market. Monopsony is a market situation in which there are many sellers but only one buyer that exercises the market power. Consider monopsonistic labour market. Monopsonist will MFC with p MPL ( L ) MFC ( L ) , where MFC ( L w ( L))L w ( L) Lw( L) . 0 choose 0 employment that S equates MRP: S If labour supply curve is upward sloping them MFC curve will lie above the inverse labour supply at any L 0 and MFC 0 w S (0) . If labour supply is linear, then MFC is also linear and two times steeper. 157 A.Friedman ICEF-2022 L 158 A.Friedman ICEF-2022 11. ASYMMETRIC INFORMATION 11.1 Types of asymmetric information problems Asymmetric Information HIDDEN INFORMATION One side knows some characteristics of itself that the other does not HIDDEN ACTION One side can take an action that affects the other side but is not observable ADVERSE SELECTION uninformed side ends up trading only with the wrong types, as the good types are driven out of the market MORAL HAZARD a party to a contract engages in post-contractual opportunistic behaviour (takes the wrong action) 11.2 Adverse selection and the market for lemons George A. Akerlof (1970) `The market for `lemons': quality uncertainty and the market mechanism', Quarterly Journal of Economics 84, 488-500 Consider a market for second-hand cars. There are some high-quality cars (rarely breaks down) and some low-quality cars. Assume that there are many more buyers (M) than sellers (N) in the market: M >> N. Each buyer is willing to purchase, at most, one car. Half of the sellers own high-quality cars and another half – low quality ones. A high quality car is worth $1000 to the seller, while a bad quality one is worth only $300 to the seller. A buyer is willing to pay $1300 for a high quality car, and $400 for a low quality one. All agents are riskneutral. Assume that markets are perfectly competitive. Sellers get the entire surplus from trade (as N<<M) Valuation ($) Seller Buyer H type 1000 1300 L type 300 400 First of all, let us derive the efficient allocation of cars. Efficient allocation should maximize total surplus: denoting by q H the number of high quality cars transferred to buyers and by q L - the number of low quality cars transferred from the sellers to buyers, we get 159 A.Friedman ICEF-2022 1300q H 1000 0.5 N q H 400q L 300 0.5 N q L ) 300q H 100q L 650 N max 0 q H q L M 0 q H N / 2 0 q L N / 2 Solving this problem we get: n H N / 2, n L N / 2 , that is, all cars should be transferred from sellers to buyers. Symmetric Information Case Under symmetric information we have to separate the markets: one for low-quality cars and another one – for high-quality cars. The supply of each type is zero if the price is below the sellers’ valuation and supply equals N/2 for the prices above the sellers’ valuation. If the price is equal to the sellers’ valuation, then each seller is indifferent and the market supply is any number from 0 up to N/2. In a similar way, the demand at every market is zero if the market price is below the buyers’ valuation and demand equals M for the prices below the sellers valuation. If the price equals to the buyers’ valuation then every buyer is indifferent and total quantity demanded is any number between 0 and M. $ $ SH 1300 SL DH 1000 DL 400 300 N/2 L market M QL N/2 M QH H market As a result in equilibrium all cars are sold. High-quality cars are sold at the price equal to $1300, while low quality cars are sold at $400 per car. As we have a complete system of perfectly competitive markets it is not surprising that the resulting allocation is PO. Asymmetric Information Case Now suppose that each seller knows the quality of his car but buyers are not able to distinguish a high-quality car from a low-quality one. Then, instead of two markets, we have only one. We get the market supply by summing up horizontally the supply curves for high- and lowquality cars: 160 A.Friedman ICEF-2022 N , p 1000 N / 2, N , p 1000 Q S p N / 2, 300 p 1000 0, N / 2, p 300 0, p 300 We assume that buyers rationally expect only low-quality cars to be sold under price less than $1000. As each buyer is willing to pay up to $400 for low-quality car, then the quantity demanded is M if price is below $400. If the price equals $400 then every buyer is indifferent and total quantity demanded is any number between 0 and M. At price between $400 and $1000, the quantity demanded is zero as buyers expect only lowquality cars and they are not willing to buy these cars at price above $400. Buyers expect average quality at price of $1000 and higher. For the average quality each buyer is willing to pay $13001/2 + $4001/2 = $850. But at price of $850 only low-quality cars appear at the market. Thus, the quantity demanded stays zero at price above $1000: 0, p 400 Q D p 0, M , p 400 . M , p 400 S $ 1000 400 D 300 N/2 Q M The intersection of demand and supply appears at p=400, where only low-quality cars are sold. We observe an adverse selection as high-quality sellers leave the market. This allocation is inefficient as the potential gain from high-quality cars is lost: DWL=(1300 – 1000)N/2=150N. Solution of the problem: separation through voluntary refunds One of the solutions of the adverse selection problem at the used cars market is given by the voluntary refund policy. Suppose that sellers can promise a refund of R if the car breaks down. Assume that low-quality cars break down with probability 0.8, while high-quality cars break down with probability 0.1. Can this policy lead to a separating equilibrium? 161 A.Friedman ICEF-2022 In separating equilibrium, high-quality cars are sold at price of $1300 with refund while lowquality ones are sold at price of $400 without refund. In equilibrium the two types of constraints should be satisfies. First of all, sales should be profitable for both types, that is, participation constraint (PC) should take place for every seller. PC for high quality 1300 – 0.1R 1000 is satisfied if R 3000, while PC for low-quality is satisfied automatically as 400 > 300. In addition to PC, we need incentive compatibility constraint (IC): it should be profitable to sell with R for H-type and without R for L-type: 1300 – 0.1R 400 R 9000, 400 1300 – 0.8R R 9000/8=1125. Finally, we get a separating equilibrium if 1125 R 3000. Forced refunds policy Now, assume that the government forces each seller to offer a full refund if the car breaks down. As the policy is mandatory it cannot result in a separating equilibrium where both highquality and low-quality cares are sold. Either all sellers stay at the market and offer full refund or some sellers leave the market. If both types stay at the market, then they sell at the same price as consumers cannot identify the car type. As the probability of breakdown is higher for a low-quality car seller, then the expected profit of this seller is lower and we should analyze his participation constraint. Sales are profitable for L-type sellers if P – 0.8P 300, that is, PC is satisfied for P300/0.2=1500 but nobody is willing to purchase a car at price above $1300. It means that only the sellers with high-quality cars stay at the market and sell at price equal to $1300. For these sellers PC is satisfied as 1300 – 0.11300=1170 > 1000. The resulting outcome is inefficient as potential gain from low-quality cars is not realized: DWL=(400 – 300)N/2 = 50N. Thus, obligatory refund destroys the separating role of this policy. 11.3 Adverse selection at insurance market Assumptions: price taking, two groups of customers: high-risk with probability of loss H and low-risk with probability of loss L L H , the share of high risk 0, 1 , customers have the same initial wealth W and have the same potential loss L , customers are risk averse, preferences are represented by EUFs with the same elementary utility functions but different probabilities of loss, insurance companies are risk-neutral, 162 A.Friedman ICEF-2022 there are no operation costs. Equilibrium under symmetric information As there are two different types of risk, identified by each agent, then there would be two different insurance markets: market for high risk insurance and market for low risk one. Consider insurance market of type t H , L. Demand side analysis. If insurance is fair ( rt t ), then each agent wants to purchase full insurance as agents are risk averse. If insurance is favourable ( rt t ), then agents want to over-insure (if it is possible) or purchase full insurance (if overinsurance is not allowed). Finally, if insurance is unfavourable ( rt t ), then less than full insurance would be demanded. Supply side analysis. As insurance companies are risk neutral, then each company would maximize expected profit. If insurance is fair ( rt t ), then insurance company gets zero profit per dollar of insurance sold and is willing to sell any quantity. If insurance is favourable ( rt t ), then insurance company gets negative expected profit and is not willing to supply insurance at all. Finally, if insurance is unfavourable ( rt t ), then each unit of insurance sold brings positive profit and the solution of profit maximization problem does not exist (supply is unlimited). Thus the case of unfavourable insurance can never be observed in equilibrium. Favourable price does not equilibrate the market as well, since supply is zero while demand is positive. It means that equilibrium price has to be fair, i.e. each agent would be insured completely but prices are different. 163 A.Friedman ICEF-2022 L-type H-type Equilibrium under asymmetric information Suppose that each customer knows his probability of loss, but insurance companies do not. As insurance companies are not allowed to differentiate between the two types of customers, there would be one market with uniform price. Note that price which is equal to the average probability of loss will never equilibrate the market. Indeed, suppose r AV H 1 L . This price is in between L and H . As r H , then insurance is favourable for high-risk agents and they will purchase full insurance (assuming that over-insurance is not allowed), i.e. x H L . Low-risk agents find this price too high and will purchase less than full insurance, i.e. x L L . As the quantity demanded by good customers (low risk) is less that quantity demanded by high-risk, the resulting expected profit would be negative: AV H L 1 AV L x L AV H 1 AV L L AV H 1 L L 0 Thus, the insurance company will never sell insurance under this price. As a result, the price of insurance goes up and two types of equilibrium may take place. Note that, in general, equilibrium price should bring zero expected profit: negative expected profit results in zero supply, while positive expected profit leads to unlimited supply. Equilibrium of type 1 corresponds to r H . If low risk agents find this price too high and prefer to stay at initial endowment, then only high risk stay at the market and purchase full insurance at fair price. Thus, only high risk agents may stay at the market, i.e., we deal with adverse selection problem. Equilibrium of type 2 corresponds to AV r H . At this price, high risk agents purchase full insurance (as price is favourable) while low risk find the price unfavourable and purchase 164 A.Friedman ICEF-2022 only partial insurance. Moreover, the losses of insurance company from high-risk agents are compensated by profits from low-risk so that total expected profit is zero. cNL cNL Low-risk W Lowrisk High-risk High-risk cL cL W-L Whatever is the equilibrium, the risk is allocated inefficiently. As firms are risk neutral while agents are risk averse, at any efficient allocation insurance companies should bear all the risk, while agents have to be insured completely. But in equilibrium only high-risk agents are insured completely, while low risk take at least part of the risk. 11.4 Private and Government Response to Adverse Selection Problem Private response Screening: uninformed party moves first and offers a menu of contracts Example: Monopolist offers menu of bundles to consumers that differ in WTP (see second degree price discrimination) Signaling: informed party moves first and signals its type Student gets education to signal his talent to potential employer Seller may offer a guarantee to signal high quality On-line seller may invest in professionally-looking site to signal customers that the business is not a fly-by-night operation Group insurance plans (employers offer health insurance as part of benefits. Insurance company deals with a group of people and so it faces average risk since agents can’t refuse from purchasing insurance) Targeted insurance rates (insurance premium is based on some observable characteristic that is correlated with unobservable one) 165 A.Friedman ICEF-2022 Government response: compulsory health insurance, compulsory public pension programs, OSAGO (as program is compulsory, insurance company deals with average risk) information policies: prohibiting false advertising, setting quality standards, mandatory disclosure of certain facts (used car dealers are supposed to disclose known defects) 11.5 Spence model of job market signaling Assume that there are two groups of workers that differ in productivity MPL L MPL H . The share of high-productive group is . There is no possibility for partial employment: either worker gets a contract that guarantees annual wage w for n years or he stays unemployed and gets 0 reservation utility. The utility of worker of type t t L, H increases in his income n w but decreases in education ( y years of education): ut , y ct y , c L c H and costs of education are higher for low-productive worker. As a result indifference curves of a low-productive and a high productive agent are upward sloping and indifference curves of L-type agent are steeper (this property is known in economic literature as single-crossing property). $ ICL ICH y Firms compete by offering wages (Bertrand-type price competition) and as a result in equilibrium under symmetric information each worker is paid in accordance with wH MVPL H and wL MVPL L Assume that each worker knows his/her productivity but firms cannot differentiate highproductive workers from the low-productive ones. To solve the problem workers acquire education to signal their productivity. Game timing: workers choose education, firms observe education and offer wage schedule, workers decide whether to take the offer or not. 166 A.Friedman ICEF-2022 Separating equilibrium In separating equilibrium, each type chooses a different action, that is, in our case highproductive workers signal their productivity by acquiring education while low-productive workers do not get education. There is a threshold number y of years of education such that if y y , then the worker is considered to be high-productive and otherwise (if y y ) worker is considered to be a lowproductive one. Price competition between firms results in the following compensation offered: L , y y , where L wL n and H wH n , y H , y y Separating equilibrium exists if costs of signalling are less than benefits for high-productive workers but higher for low-productive ones. Benefit from signalling is equal to the wage premium B H L wH wL n . Thus, the following conditions should take place: B H L c L y and B H L c H y . Solving this system we get: H L L y H . cL cH So far, we ignored participation constraints. These constraints are trivially satisfied due to the assumption of zero reservation utility: for a low-productive worker PC requires L 0 and for a high-productive worker we need H c H y 0 , which implies y H / c H . This condition is satisfied as y H L H . cH cH Let us analyze the efficiency of separating equilibrium. It could be noticed that we have many equilibria that differ in the level of signal of high-productive workers. In any equilibrium, the expected profit of each firm is zero and utility of low-productive worker equals to L but utility of high-productive agent is different and it is the highest in the equilibrium with the L lowest level of signal, that is, the equilibrium with y H is a Pareto superior one. cL But even in this equilibrium, the utility of high-productive workers is below the one observed under symmetric information due to unproductive signaling costs that constitutes DWL. $ H (y) L y 167 A.Friedman ICEF-2022 Pooling Equilibrium (both types acquire education) Now we proceed to the analysis of pooling equilibrium. In pooling equilibrium all types choose the same action, that is, in our game both types get education and as a result compensation is based on average productivity: H wH 1 H wL n . There is a threshold number y of years of education such that if y y , then worker is considered to have average productivity and otherwise (if y y ) worker is considered to be a low-productive. Price competition between firms results in the following compensation offered: L , y y y . , y y Pooling equilibrium exists if costs of signalling are less than benefit for both types. Benefit from signalling is equal to wage premium: B L H wH wL n . Thus, the following conditions should take place: B L c L y and B L c H y . Deviation is more profitable for low-productive than for high-productive worker, that is why only Ltype constraint is binding: L 0 y . cL $ H (y) L y Now we move to the analysis of participation constraints: с L y 0 and c H y 0 . Since y L cL then both constraints are satisfied automatically: y L / c L / c L / c H . 11.6 Screening Screening deals with uninformed party’s attempt to sort the informed parties by offering special pricing schemes. Screening requires some self-selection device: the uninformed party offers a set of options, and the choice made by the informed party reveals his/her hidden characteristic. 168 A.Friedman ICEF-2022 Example: low insurance premium in case of deductibles and high insurance rates for full insurance. Deductibles: if damage is below certain limit, than agent pays for the bill himself, if damage is above certain limit, then only damage in excess of this limit is compensated. Low risk agents prefer insurance with deductibles with low insurance premium, while highrisk purchase expensive full insurance. Another example of screening policy is given by the second degree price discrimination. Below we consider a modified model of price discrimination, where the monopolist offers quality-price instead of quantity-price contracts. Assume that the cost of producing quality q is cq q 2 . The seller sells a unit of quality q at price t and gets the profit t , q t q 2 . His iso-profit lines t q 2 are illustrated below. t q Each buyer purchases one unit of the good at most. Reservation utility of each buyer equals 0. Buyers utility (consumer’s surplus) is increasing in quality and decreasing in price: u, t , q q t , where reflects quality preferences. There are two groups of consumers that differ in quality preference parameter: H L 0 . Indifference curve of the type i buyer is given by: i q t u , which can be restated as t i q u . Thus, indifference curves are straight lines with the slope dt / dq i . It implies that the group with higher quality preferences has steeper indifference curves. t uH uL q Optimal Contracts under Full Information 169 A.Friedman ICEF-2022 Under full information, different contract are offered to different types. Denote by qi , t i a i contract offered to consumer of type . This contract should satisfy participation constraint i could be found from the following profit i qi t i 0 . Optimal contract for type max t i qi2 maximization problem: t i , qi s.t. i qi t i 0 . Participation constraint must be binding as otherwise we can increase the price and the resulting profit will go up. Thus we can express the price from the constraint and plug it into the objective function: max qi qi2 . qi t q FOC: i 2qi Optimal contract for type i : qi i / 2 and t i i2 / 2 . Optimal Contracts under Asymmetric Information Now assume that monopolist cannot observe the preference parameter of the customers and ~ ~ ~ ~ offers a menu of contracts to each customer: q H , t H and q L , t L . Optimal menu of contracts could be derived from the following profit maximization problem that takes into account both participation constraints (PC) and incentive compatibility constraints (IC): max t L q L2 1 t H q H2 t i ,qi s.t. L qL t L 0 (PCL) H qH t H 0 (PCH) L qL t L L qH t H (ICL) H qH t H H qL t L (ICH) 170 A.Friedman ICEF-2022 PCL and ICH will be binding (this point is discussed below). From PCL we can find t L L qL . Plugging it into ICH we obtain t H H qH qL t L H qH qL L qL . Substituting prices into the objective function we can rewrite the profit maximization problem as max qL ,qH FOCs: L q L q L2 1 H q H q L L q L q H2 1 H 2q~H 0 and L 2q~L 1 H L 0 . 1 ~ ~ Solving this system we get q H H / 2 and q L L H L L . Plugging back 2 2 2 into PCL and ICH we obtain the prices: ~ ~ ~ ~ ~ ~ t L L q L and t H H q H q L t L . Note that the quality level for the consumer with high quality preferences stays the same as ~ under symmetric information q H q H while the quality for the consumer with low ~ L 1 L q to make the contract preference parameter was reduced q L H L L 2 2 2 unattractive to H-type. This is because full information contracts are not incentive compatible. qi i / 2 , t i i2 / 2 Indeed, if both full information contracts are offered under the asymmetry of information then L-type will take the one that was designed for this type q L , t L since u L q L , t L 0 u L q H , t H L H 2H H L H 2 2 2 , but H-type takes q L , t L as H uH q , t H 0 u q H L ,t L L 2L H H L L , 2 2 2 that is, ICH is violated in this case. Incentive Compatible Contracts If we fix q L , t L then ICH should lie along indifference curve that goes through q L , t L or below. 171 A.Friedman ICEF-2022 t q Contracts that lie below this indifference curve are not profit-maximizing for seller as he can increase t L . Conclusion: ICH is binding. Optimal Contracts Optimal contracts calculated above could be illustrated graphically. t q ~ ~ Optimal contracts are given by q H q H H / 2 , and we find tH from incentive ~ ~ ~ ~ compatibily constraint of H-type which implies H q tH H q tL , H L ~ ~ ~ q ~ H L t H t L H q H L 2 . ~ H - type gets surplus equal to t H t H , which is called information rent. ~ we should solve profit-max problem with two binding constraints: PC and IC . To find q L H L 172 A.Friedman ICEF-2022 11.7 Asymmetric information: hidden actions In situations of hidden action one side of a transaction can take an action that affects the other side but is not observable Hidden actions may lead to moral hazard problem. Moral hazard problem - a party to a contract engages in post-contractual opportunistic behaviour, i.e. takes the wrong action. Examples: worker may choose to shirk on a job; an insurance policy holder may fail to take enough care to prevent an accident. The principal-agent model gives another example of moral hazard problem. Agent is hired by principle to perform certain tasks. Moral hazard problem arises when principle and agent have different goals and actions of agent can’t be monitored by the principle. As a result, the agent may take actions that are not in the interests of the principle. Suppose that a principal hires an agent to carry out a particular project that might be successful and bring H or unsuccessful and bring L , where L H . Once hired, the agent chooses an effort level: either high eH or low eL , e L e H . Effort affects the outcome of the project: higher effort leads to a higher probability of success P H eH P H eL 0 , where P e is the probability of obtaining gross profit under effort level e . The following numerical example illustrates this assumption: H=100 L=20 eH 2/3 1/3 eL 1/4 3/4 We assume that principal is risk-neutral and maximizes net expected profit. Agent maximizes expected utility, his vw, e uw ge , where w is wage and ge stays for costs of efforts. We assume that utility is increasing in wage and marginal utility is non-increasing: uw 0 , uw 0 . We will start with the case of a risk-averse agent u 0 and then proceed to the case of a risk neutral one u 0 . We also assume that costs are increasing in effort level ge ge L . The reservation utility is denoted by u 0 . The principal pays the agent a profit-contingent wage schedule: w H if profit is H and w L if profit equals L . Full Information Case: Observable Efforts Under full information case, the contract specifies the level of efforts and the corresponding profit-contingent compensation wH , wL , e . 173 A.Friedman ICEF-2022 Let us derive the optimal contract for a risk-averse agent. We will proceed in two steps. First we will find the wage schedule that implements the effort level e i , where i L, H . Then we will compare the resulting net profits and choose the effort level that generates higher profit. To implement effort level e i , the principal should maximize his net profit subject to participation constraint: max P H ei H wHi 1 P H ei L wLi wHi , wLi s.t. P H ei uwHi 1 P H ei uwLi gei u0 (PC) Note that PC is binding (explain!). Let us form a Lagrangean: ℒ = PHi H wHi 1 PHi L wLi i PHi uwHi 1 PHi uwLi gei u0 . FOC: ℒ/ wHi PHi i PHi uwHi 0 and ℒ/ wLi 1 PHi i 1 PHi uwLi 0 From this system we get uwHi 1 / i and uwLi 1 / i , which implies that marginal utilities are equal in the two states: uwHi uwLi . Since u is diminishing this implies that wHi wLi , that is, the principal should offer full insurance to a risk-averse agent. This result could be explained intuitively. If wage schedule is risky wHi wLi , then a riskneutral principal can increase his expected profit by replacing this risky payment with a fixed wage. If the fixed wage gives the same expected value, then a risk-averse agent will be better off. Thus, the principal can replace a risky payment with a fixed wage with lower expected value and make risk-averse agent as well off. At the same time, the expected profit will go up due to cost economy. We can conclude that under risky wage schedule there is inefficient risk sharing. Putting risk on risk-averse agent is costly for the principal as risk should be compensated by higher expected wage. Thus, we should finish our calculation and find out the value of this fixed wage wHH wLH w and wHL wLL w . To do this, we plug fixed wage into participation constraint: uw 1 P uw ge uw ge u PHH u w 1 PHH u w ge H u w ge H u0 and PHL HL L L 0 . Thus we get the following equations for wages under high and low efforts: uw ge H u0 and uw ge L u0 . Note that as ge H ge L then uw ge H u0 ge L u0 uw . Since utility is increasing in wage it implies that higher efforts require higher compensation: w w . 174 A.Friedman ICEF-2022 Thus, we cannot predict unambiguously whether high effort will be more profitable as, on the one hand, it brings higher gross profit: H P H e H L P L e H H P H e L L P L e L but, on the other hand, it requires higher costs w w . Principal should compare: e H H P H e H L P L e H w and e L H P H e L L P L e L w and implement the effort that gives a higher net profit. Inducing e H is profitable if e H e L 0 , that is . H L P H e H P H e L w w exp ected gain from exp ected effort cos t Unobservable Efforts Under unobservable efforts we cannot offer a contract contingent on effort level as monitoring is impossible. Thus, the contract includes only wage scheme wH , wL . This wage scheme should be designed in such a way that makes desirable level of effort incentive compatible. We will derive optimal contract in three steps. We start with a derivation of the contract that implements low effort e L (requires PC but IC is satisfied automatically), then proceed to the analysis of the contract that implements high effort e H (requires both PC and IC) and finally compare the expected net profits to choose the profit-maximizing contract. Contract that implements low effort Let us start with the contract that implements low effort level. The principal maximizes expected profit-maximization subject to participation and incentive compatibility (IC) constraints: max P H e L H wH 1 P H e L L wL wH , w L s.t. P H e L uwH 1 P H e L uwL ge L u0 P H e L uwH 1 P H e L uwL ge L P H e H uwH 1 P H e H uwL ge H We can verify that the contract that was optimal under full information satisfies IC constraint. uw ge If wH wL w then uwH uwL u w . Plug into IC and get u w ge L H since ge L ge H . 175 A.Friedman ICEF-2022 Conclusion: the principal does not need to provide incentives not to work hard. As a result offer the same contract as under observable efforts: fixed wage wH wL w , where u w ge L u0 . Contract that implements high effort Now we proceed to the analysis of the contract that implements high level of effort. The principal maximizes expected profit-maximization subject to participation and incentive compatibility (IC) constraints: max P H e H H wH 1 P H e H L wL wH , w L s.t. P H e H uwH 1 P H e H uwL ge H u0 P H e H uwH 1 P H e H uwL ge H P H e L uwH 1 P H e L uwL ge L Now fixed wage doesn’t satisfy the incentive compatibility constraint. If wH wL w then uwH u wL u w and (ICH) is violated: uw ge H uw ge L . Conclusion: Putting some risk on the agent is unavoidable in providing the incentive for high effort. Thus we should find two different wages wH wL and we have two binding constraints. The optimal contract is derived from the system given by participation and incentive compatibility constraints. Optimal contract under unobservable efforts We demonstrated that the full insurance contract implements low effort level: wH wL w , where uw ge L u0 . The resulting expected profit is UNOBS e L OBS e L H P H e L L P L e L w . To implement high level of effort, we need a state dependent payment scheme wH wL , where wH , wL is a solution of the system given by participation and incentive compatibility constraints. The resulting expected compensation exceeds the one that corresponds to the full information case: w wH P H e H wL P L e H w since uw U w ge H u0 u w . Putting risk on the agent is good for incentives but bad for risk-sharing, and bad risk-sharing is costly for the principal: UNOBS e H H P H e H L P L e H w OBS e H . To find the optimal effort level, the principal should compare UNOBS e L and UNOBS e H and choose the one that generates higher expected profit. 176 A.Friedman ICEF-2022 Case of risk-neutral agent In case of a risk-averse agent there is a trade-off between incentives and risk sharing when high level of effort is implemented. There is no such trade-off if the agent is risk-neutral, that is, has a utility function vw, e w ge , as in this case putting risk on the agent is not costly and the principal can generate incentives for free. Participation constraint takes the form: P H ei wHi 1 P H ei wLi gei u0 . It implies that expected compensation must be equal to gei u0 . Then the optimal effort level under observable efforts could be derived from the following problem: max P H ei H wHi 1 P H ei L wLi ei eH ,eL max P H ei H 1 P H ei L gei u0 . ei eH ,eL Since constant u 0 doesn’t affect the solution, the problem could be restated as max P H ei H 1 P H ei L gei ei eH ,eL First best contract can be implemented even under unobservable efforts since the principal can put all the risk on the agent and ask for a fixed fee F (agent becomes a residual claimant). Agent as a residual claimant will choose e i by maximizing the expected utility: max P H ei H 1 P H ei L gei F . As constant F ei eH ,eL doesn’t affect the solution this problem is equivalent to the one obtained under the full information case. 11.7 Sample exercise with solution Consider a foreign firm that is willing to supply a unique product to country A. Suppose that the firm’s cost function is C(Q)=cQ. The demand for its product is given by Q(p)=A-p. Government of country A is going to propose a contract to this firm that will specify the quantity of the good (Q) that this firm must sell at country A at a constant per-unit price chosen by the firm and the fixed sum (T) that it should pay to (or receive from) the government of country A. Government is looking for the contract that maximizes domestic welfare (W) equal to the sum of consumers surplus and government budget surplus. (a) Setup government optimization problem and derive the optimal tariffs for the two cases (1) when c=A/2 and (2) c=A/4. (b) Suppose that the potential entrant might be a high cost firm with c=A/2 or a low cost firm with c=A/4 with equal probabilities. Government cannot identify whether the firm is highcost or low-cost but the firm learns its type before it chooses the contract. Assume that the government is risk neutral. (i) Suppose that government simultaneously offers two contracts identified in (a). Show using your one graph that these contracts are not incentive compatible. Illustrate the resulting expected welfare [do not make any numerical calculations]. 177 A.Friedman ICEF-2022 (ii) Find graphically the payment for L-type firm that will make the contracts found in (a) incentive compatible and will provide the maximum possible expected welfare. What is the resulting increase in expected welfare? (iii) Reproduce your graph from (ii) and change the contract for the high-cost firm by reducing the output a bit (by q) but keep the output for low-cost firm unchanged. Illustrate the corresponding changes in fixed payments. What is the resulting change in expected welfare. (iv) Starting from the contracts represented in (iii) consider a differentially small reduction in the output of high-cost firm. Identify (using your graph from (iii)) the resulting change in expected welfare. (v) Produce a new graph with demand and marginal cost curves. Based on the analysis performed in (iv) show graphically the two contracts that maximize the expected welfare under given cost uncertainty. Explain carefully. Solution (a) If the contract is accepted then the firm will charge the maximum price that consumers are willing to pay for given Q, i.e. p=A-Q . Thus PC is given by: (A-Q-c)Q-T0. Under linear demand CS(Q)=Q2/2. Government problem: Q2 / 2 T max Q ,T A Q cQ T 0 PC is binding (proof) Q2 / 2 A Q c Q max , Solution: Q A c , T 0 Q 0 (1) QH=A-A/4=3A/4; TH=0 (2) QL=A-A/2=A/2; TL=0 (b) As the government is risk neutral it maximizes expected welfare. (i) High cost firm gets negative profit (loss equal B+2F) if it takes the contract designed for low cost firm and it gets 0 profit from the contract designed for high cost firm, so it will take the right contract. Low cost firm gets positive profit (B) if it takes the contract designed for the high cost firm and it gets zero profit from the contract designed for the low cost firm, thus it takes the wrong contract. 178 A.Friedman ICEF-2022 $ A D A/2 B F A/4 Q QH A QL Expected wealth=D as CS=D in any case (firm of any type takes H-type contract) and T=0 (ii) TL=-B (government should increase the profit of L-type firm by providing lump sum subsidy equal to B to make low cost firm indifferent b/w the two contracts) CS in case of L-type firm increases by B+F but government pays B as a subsidy, thus EW=F/2 since we have L-type firm with probability ½ (iii) Contracts: QH-q; TH=G Now if L-cost firm takes this contract its profit = B-B Thus government should reduce the subsidy by B QL; TL=-(B-B) $ Gain Loss A A/2 G D B F A/4 QH-q QH QL A Q EW=-D+B >0 (iv) EW=Gain –Loss (v) Contracts: QH= Q* should result in zero marginal expected welfare: Gain=Loss; TH=N QL; =A-A/4=3A/4; TL=-M 179 A.Friedman ICEF-2022 $ A = N A/2 M A/4 q* QL A Q 180 A.Friedman ICEF-2022 12. EXTERNALITIES AND PUBLIC GOODS An externality refers to the effect when one economic agent’s action directly confers benefit or imposes a cost on some other agent without that consequence being reflected directly in market prices and exchange transactions. Externalities can be positive or negative. A positive externality occurs when one economic agent creates a benefit for another (without compensation). A negative externality occurs when one economic agent imposes a cost on another (without compensation). Externalities can be divided into 4 categories according to the source-recipient principle: Consumption-consumption (smoker sitting next to a table of non-smokers in a restaurant), Consumption-production (healthy lifestyle increases productivity), Production-consumption (factory pollutes the air in the area where people live), Production-production (factory pollutes the water used by another firm). 12.1 Simple Model of Consumption Externalities Consider an economy with two agents ( i 1, 2 ) with quasilinear utility: ui x, mi i x mi , where m i - money and x is a costless action taken by agent 1 that affects utility of agent 2. Assume that utility of agent 1 increases in x but marginal utility is diminishing: 1 x 0, 1x 0 . Depending on the sign of the marginal utility of agent 2 we deal with positive 2 x 0 or negative 2 x 0 external effect. Marginal utility is still assumed to be diminishing: 2 x 0 . Further we consider the case of negative external effect, and the case of positive external effect is described in the Subject Guide. Equilibrium and Efficiency Externalities can adversely affect economic efficiency as complete market system assumption or the assumptions of the FFWT is not satisfied, i.e., externalities results in the missing market problem. In unregulated economy agent 1 decides on the desirable level x by maximizing his utility: max 1 x m1 x Since the objective function is strictly concave, the FOC is both necessary and sufficient. Thus, the equilibrium level of x is given by the following equation: 1 x 0 . 181 A.Friedman ICEF-2022 To derive Pareto optimum, we should maximize the utility of one agent subject to a fixed level of utility of the other under given resource constraint: max m1 0,m2 0, x x m 1 1 s.t. 2 x m2 u m1 m2 em We can express m 2 from the first constraint and plug it into the second one to find m1 : m1 em m2 em u 2 x 2 x const . Now we substitute it into the objective function: max 1 x 2 x const . As we assumed diminishing marginal utilities for both x agents, this function is strictly concave and the optimum level x could be derived from the FOC: 1 x 2 x 0 . Now we could compare equilibrium level x with efficient one x . Since 2 x 0 then 1 x 2 x 0 1 x . As 1 is diminishing, this inequality implies that x o x , that is, in equilibrium x is overproduced since agent 1 doesn’t take into account the negative external effect imposed on agent 2. This result could be illustrated graphically. As x is costless then PMC 0 and agent 1 equates PMB 1 x with zero MC when he chooses x . Equilibrium: 1 x 0 $ PO B A Equilibrium x To find PO we should solve the following equation: 1 x 2 x 0 that could be equivalently restated as 1 x 2 x . MSB MSC The LHS gives the marginal social benefit that coincides with the private social benefit while the RHS gives the marginal social costs that reflects disutility of agent 2 and exceeds zero 182 A.Friedman ICEF-2022 private marginal cost of agent 1. Thus, as we can see from the graph, the resulting PO level of activity is below the equilibrium one. The blue triangle illustrates DWL resulting from negative externality. If x decreases from x to x , then social benefit goes up by area A, while social cost rises by (A+B). As cost increases more than benefit, total surplus is reduced by area of B, corresponding to deadweight loss. 10.2 Private responses to externalities Possible responses to externality problem could be separated into the government and private response. Some examples of private response: Internalization via firms’ merger Social conventions Bargaining (Coase theorem). Internalization An externality can be internalized by combining the involved parties, i.e. putting the problem in the hands of a single decision maker. If firm A pollutes the water while firm B suffers from pollution, then the combined firm (if merger takes place) will take into account the negative impact of water pollution as it would maximize the total profit. Unfortunately this method doesn’t work in case of consumption externalities. Social conventions Individuals cannot merge in order to internalize the externalities. Instead social conventions are used to force people to take into account the externalities they generate. Some moral norms induce individuals to internalize the externalities that they might create. Property Rights Solution (Coase Theorem) Suppose that property rights are assigned. Then parties may bargain and if bargaining is costless they will reach the mutually beneficial solution. This is the essence of the Coase theorem6: 6 Coase, R.H. (1960) The problem of social cost, Journal of Law and Economics 3, 1-44 183 A.Friedman ICEF-2022 Under perfect competition (i.e. if there are no transaction costs), if property rights are welldefined, bargaining would lead to the efficient outcome irrespective of who the property rights are allocated to (to the party generating the externality or to the party affected). Let us demonstrate how this theorem works for our example of negative externality in consumption. We start with the case of permissive law, when property rights are given to agent 1 that generates this external effect. It means that in case of disagreement x x . Agent 1 asks agent 2 for a payment T to reduce x to some level below x . Then optimal contract is derived from the following optimization problem: max 1 x T x ,T s.t. . 2 x T 2 x You should be able to demonstrate that participation constraint is binding. Then we can find T from this constraint: T 2 x 2 x and plug it into the objective function: max 1 x 2 x 2 x x As objective function is strictly concave FOC gives the optimal level of x : 1 x 2 x 0 . We can note that this equation coincides with the one for PO allocation. Now, let us move to the case of restrictive law, when property rights are given to agent 2 that suffers from the external effect. In this case disagreement results in x 0 . Agent 2 asks for a payment T to increase x to some level above 0 and the optimal contract is derived from the following problem: max 2 x T x ,T s.t. . 1 x T 1 0 . Similarly, it can be shown that the constraint is binding and T 1 x 1 0 . Substituting into the objective function we get max 2 x 1 x 1 0. x The solution should solve the following equation: 1 x 2 x 0 , which coincides with the one derived for permissive law case. Thus in both cases we have x x . Problems with implementation: preferences – private information even with just two parties, one or both could comprise of many individuals (many fishermen may suffer from water pollution) they should delegate decision to one leader 184 A.Friedman ICEF-2022 a free rider problem may arise if committing to a negotiation is costly in terms of time and effort difficult to start negotiations if many parties are involved Importance of the Coase ideas: create the field of law and economics (contract law) have influenced the creation of pollution rights markets 12.3 Government Regulation Some examples of government response: Direct regulation (for example, via emission quota in case of negative external effect) Corrective (Pigouvian) taxes Cap-and-trade policy Direct regulation: quota To correct negative externalities, the government can impose some maximum limit on pollution that corresponds to the efficient level ( x x ) or require installing special equipment that reduces pollution. To correct positive externality, the government can introduce the minimum requirement on the level of output. Problems. To make direct regulation efficient, government needs private information on costs of production, which is problematic. Corrective tax Inefficiency can be eliminated if per unit tax (in case of negative externality) or subsidy (in case of negative externality) is introduced. This tax/subsidy is called Pigouvian tax (Pigou 1932). This per unit tax/subsidy should be equal to the value of MEC (or MEB) evaluated at the efficient level of output. With a per unit tax this activity becomes costly for agent 1. Thus, he will take into account these costs deciding on the level of activity: max 1 x m1 t . From FOC we get x 1 x t . If t 1 x then x x . The figure below illustrates how the Pigouvian tax works. It shifts the net private marginal benefit curve downward exactly by the value of tax and, as a result, the equilibrium level of x falls. This solution still requires private information about preferences to find the right value of the tax rate. 185 A.Friedman ICEF-2022 $ x Cap-and-Trade Policy Government sets an overall quota (a cap) for pollution and then allows trade in pollution rights. Firms with a lower MC of abatement would then be net sellers in the emissions trading market, firms with higher MAC would be net buyers. This policy achieves the same reduction of pollution with lower costs that uniform quota. Examples: EU Emission trading system is the largest greenhouse gas emission trading system in the world US: SO2 emissions trading system Since 1995, trade in an SO2 allowance reduced emissions by half, saving $1 billion per year compared with a conventional approach. 12.4 Efficient Provision of Public Good Public good represents a special type of positive externality in consumption. A good is called public good if it is non-rival in consumption, i.e. consumption by any one person does not reduce the amount available for others. Examples: national defence, TV broadcasting, roads. Public goods may be non-excludable (if it is very costly to exclude non-payers from consumption of the good) or excludable. Non-excludable good is called pure public good. Examples: national defense, roads, basic public health, basic education, protection against natural catastrophes. Consider a model with N consumers and two goods: X – public good and m -private good. Assume that all agents have quasilinear utility functions: ui X mi with positive but diminishing marginal utility of public good ui X 0 , ui X 0 . 186 A.Friedman ICEF-2022 Assume that there is no initial endowment of the public good and each individual has some i endowment of the private good e m so that the total endowment of the private good is N em emi . Suppose that public good could be produced from the numerior private good i 1 with CRS technology that results in linear cost function: C X cX . To find the efficient amount of the public good produced we should maximize utility of one of the agents under fixed utilities of others and resource constraint: maxu1 X m1 s.t. ui X mi ui , i 2, , N N m i 1 em cX . i We can express m i from utility constraint of each agent mi ui ui X for i 2, , N and plug into the resource constraint: m1 N m i 2 N i m1 ui ui X em cX . Then we i 2 N N i 2 i 2 find m1 from this equation: m1 em cX ui ui X ui X cX const and N max u X ui X cX const . As constant substitute into the objective function: 1 i 2 does not affect the solution we proceed to the following optimization problem: N max ui X cX . i 1 The objective function is strictly concave as we assumed diminishing marginal utilities and so we can restrict our attention to the FOC that takes the form: u X c . N i 1 i This condition suggests that at the PO the total willingness to pay (the sum of MRS over all consumers of the public good) should be equal to the marginal cost. This condition is known as Samuelson equation7. 12.5 Private Provision of Public Good Consider a voluntary contribution mechanism of private provision of public good, where each individual finances some fraction of the total amount of the public good consumed so that 7 Samuelson P.A. (1954) The Theory of Public Expenditure. Review of Economics and Statistics 36, 386–389. 187 A.Friedman ICEF-2022 X x i x i , where x i - contribution of agent i to the public good provision and x i x j - total contribution of other agents. j i Assume that both markets are perfectly competitive markets and denote prices by p x p and pm 1. Suppose that all agents simultaneously and independently decide on x i . Then the utility maximization problem of agent i is given by: max ui x i x i mi xi ,mi 0 s.t. px i mi emi . Due to non-satiation, the budget constraint will be satisfied as equality and we can find consumption of the private good mi emi px i and plug it into the objective function: max ui x i x i px i . xi 0 Taking into account the possibility of corner solution, we get the following FOC: ui X p 0 and ui X p 0 if x i 0 . Suppose that the public good is produced in equilibrium X 0 . Then there exists at least one agent that contributes to the production of this public good x j 0 . For this agent, we have interior solution and so FOC takes the form of equality uj X p . Since marginal utility of public good is positive u X u X u X p . N i 1 i j i j N implies p c . Then i ui X 0 for every consumer i then Under perfect competition, the profit maximization N ui X p c ui X . Since ui is diminishing for every i i 1 i 1 then X X , that is, the public good is underproduced in comparison to the socially efficient level. Example Consider an example with N consumers that have quasilinear utility functions ui X mi and u1 X u2 X uN X . Then socially efficient amount of the public good is given u X c . N by: i 1 i Let us show that in the equilibrium only agent N that has the highest valuation of the good will contribute to the public good production. Suppose that this is not the case and there exists an agent j N that makes some positive contribution x j 0 . Then FOC should be satisfied 188 A.Friedman ICEF-2022 as equality uj X p . Due to our assumption uN X uj X p , which implies that FOC is violated for agent N. This contradiction proves the claim. Thus, in the equilibrium x1 x 2 x N 1 0 and x N X . The equilibrium quantity is derived from the FOC for agent N: uN X p c . Below we illustrate PO and equilibrium levels of production for the case with 3 consumers. MC X 12.6 The Commons Problem Finally we move to a special case of mutual external effect that deals with common property resources that represent rival but not excludable goods. The resulting negative external effect results in over-exploitation problem known as ‘tragedy of commons’ Examples: water, fishery, pastures Evidence: Stavins, R.N. (2011) The problem of the commons: Still unsettled after 100 years. AER 101, 1-108 New England lobster fishery (1966): efficient number of traps - about 450,000 but actual number is nearly 1 mln of traps; Two lobster fisheries in eastern Canada (1979): losses due to unrestricted entry amounted to 25% of market value of harvests, fishery effort exceeds the efficient level by 350% 189 A.Friedman ICEF-2022 Model of Resource Extraction Consider a simple model of resource extraction with N users that simultaneously and N independently choose efforts. Production function is given by Q L , where L Li and i 1 L i stays for the effort of user i . Assume that the price of final good is normalized to 1: pQ 1. Cost of extraction (cost of efforts) for user i is given by C i Li cLi . The socially optimal level of efforts could be found from the maximization of total net benefit: max L cL . L 0 FOC takes the following form 0.5 / L c and we get the socially optimal total level of efforts L 1 / 2c . 2 Now we proceed to the analysis of equilibrium. In the equilibrium each user decides on his effort level taking the effort levels of other users Li L j as given: j i Li max Li Li L i Li Li cLi max Li The FOC for user i is given by 1 / L 0.5Li / L 3/ 2 Li cLi . Li Li L 0.5Li / L c 0 , which can be restated as L c . Summing up over all users we get N / L 0.5L / L . Solving these equation we find the equilibrium level of total efforts: L 3/ 2 Nc 1 1 1 /2 N 2 . 2 c 1 1 1 /2 N 2 2 c 1 L N 1 2 L we can 4c Let us compare it with the socially optimal one. We can note that L increases when N goes up. Since L N 1 1 1 /2 N 2 2 c conclude that the equilibrium level of efforts exceeds the socially efficient level. Solutions to commons problem The first solution deals with the assignment of property rights, that is, privatization. If property rights are assigned to one person, than the equilibrium will result in socially efficient 1 level of efforts L N 1 2 L . Unfortunately, for many resources (for example, 4c fishing) it is impossible to assign the rights just to one person. 190 A.Friedman ICEF-2022 Another solution deals with the creation of the market for extraction rights via the introduction of individual transferable quotas (ITQ). In this case the government sets the overall quota and issues individual quotas that can be traded. Due to trade, quotas will be used by the users with the lowest cost of exploitation. Stavins (2011) mentioned that ITQ were successfully used to regulate fisheries in 17 countries (Australia, Canada, New Zealand, USA, and others). But this solution is only a partial remedy since monitoring is costly and not all countries stick to the quotas. Another approach deals with taxation. Suppose that a tax t is introduced per unit of effort which increases tax-inclusive marginal cost to c t . As a result, the equilibrium level of efforts becomes L t 1 c t 2 1 1 /2 N 2 . We should choose the tax rate in such a way that the resulting level of efforts is Pareto optimal: L t Solving this equation we get t c1 1 / N . 1 c t 2 1 1 /2 N 2 L 1 . 4c 2 This solution is not perfect as well, since the calculation of optimal tax rate requires private information that the policymaker might not have. Now we proceed from the government to private response. Tragedy of commons problem is similar to Prisoners Dilemma as all users are better off if extraction is reduced. It means that there is a scope for informal cooperation in equilibrium if the game is repeated. As we know Folk theorem suggests that cooperative outcome can be sustained as SPNE for high . Assume that every user applies the following trigger strategy. He starts by playing Li L / N (cooperate at t=0) and in period t > 1 plays L / N if L / N was played by every user in the past and plays L / N otherwise. , where is user’s profit from DEV 1 1 cooperation and is user’s profit in NE (one-shot game), where . Thus, informal cooperation is sustained if DEV 1 . Solving this equation with respect to There is no incentive to deviate if discount factor we get DEV . DEV 12.7 Sample exercise with solution Consider two neighboring regions of the same economy that decide simultaneously and independently on the own level of environment protection expenditures s i , i 1, 2 . The resulting net benefit of region i is given by ki s1 s2 s i . Assume that 0 k1 k2 . (a) Find the equilibrium. (b) Demonstrate that the equilibrium outcome found in (a) is inefficient. Explain the reason for inefficiency. 191 A.Friedman ICEF-2022 (c) Is it possible to eliminate the loss from inefficiency via taxes/subsidies? Find the required taxes/subsidies or prove that it is impossible. Solution (a) NE is given by the intersection of best response functions. Here we should take into account the possibility of corner solution as region might find zero contribution to the provision of the public good profitable under high contribution of the other region. maxki s1 s2 s i . [The function is strictly concave, so we can restrict our analysis to the si 0 FOC only] FOC ki 2 s1 s 2 1 0 and ki 2 s1 s 2 1 0 if s i 0 . Starting from this point equilibrium could be derived graphically OR algebraically. Graphical analysis: BR curve of region i: 2 2 ki / 4 s j , s j ki / 4 . si 2 0, s j ki / 4 We can find graphically (graph is skipped) that the only intersection is the one where s1 0 and s2 k2 2 / 4 . Alternatively equilibrium could be obtained algebraically. Let us prove that region 1 will never make a contribution. k k k Suppose that s1 0 and s2 0 then s1 s2 1 and 1 s1 s2 2 which implies 2 2 2 k1 k2 but it contradicts to the assumption k1 k2 . We can also note that s1 s2 0 as net benefit at s1 s2 0 goes to infinity. Thus in equilibrium s1 0 and s2 k2 2 / 4 . (b) The equilibrium allocation found in (a) results in SMB that exceeds SMC: SMB MB1 MB2 k1 k 2 2 s1 s 2 k1 k 2 k2 1 MC . This happens because each agent bases his/her decision on private MB that doesn’t take into account the positive external effect (i.e. additional benefit enjoyed by the other region) and equates it with MC. As a result, the social marginal benefit exceeds the private one and thus exceeds the private marginal cost that results in insufficient of the public good. (c) Is it possible to eliminate the loss from inefficiency via taxes/subsidies? Find the required taxes/subsidies or prove that it is impossible. k k2 2 Efficiency requires SMB MB1 MB2 1 MC 1 , i.e. s1 s2 k1 k2 / 4 . 2 s1 s 2 Thus any allocation that results in total S k1 k2 / 4 is PO. To eliminate the efficiency loss the government may subsidies the environmental spending. Moreover we need region-specific subsidy as the two regions create different external effects. 2 192 A.Friedman ICEF-2022 Let i stay for the subsidy rate for region i . Then At PO level of protection ki 2 s1 s 2 ki k1 k2 ki 2 s1 s 2 1 i and 1 i if s i 0 . . Thus we can take i 1 ki k1 k2 kj k1 k2 to guarantee efficiency of equilibrium allocation. 193