Utility Maximization A Utility Function Mathematically Representing Preferences U U=U(x, y) U(A) Y U(B) Utility functions have U(A) U(B) iff A U(A) U(B) iff A B B A Indifference curves describe bundle-ordering preferences B X Consumption Opportunities: The Budget Constraint • Assume that an individual has I dollars to allocate between good x and good y pxx + pyy M y M py The individual can afford to choose only combinations of x and y in the shaded triangle x M px The Budget Constraint • M px y x py py • MC of consuming one more unit of x, the amount of y that must be foregone. • The slope of the budget line is this MC. y M py Slope p y x x py The slope is the change in y for a one unit increase in the consumption of x. If Px = 10 and Py = 5, then consuming one more x means consuming two less y. x M px Maximizing Utility • Keep buying x until the MB(x) = MC(x) • Interaction of… MB, MC Not an indifference curve! MC MB X X* – Preferences, diminishing MB because of diminishing MRS. MB = MRS • MB in terms of y willing to be given up • In dollars, MB = py*MRS – MC of x = px/py • MC in terms of Y given up • In dollars, MC = px. Optimization Principle • To maximize utility, given a fixed amount of income, an individual will buy the goods and services: – That exhaust total income • Savings or borrowing is allowed (if we modify the budget constraint to include a temporal component) – So long as MB(x) ≥ MC(x), MB(y) ≥ MC(y), etc. – Or, until MB(x) = MC(x), MB(y) = MC(y) Intuition • MRS is the maximum amount of y the person is willing to give up to consume more x; the definition of MB. • px/py tells us the number of units of y that must be given up to consume one more x; the definition of MC. y At “A”, MRS>Px/Py (MB > MC), At “B”, MRS<Px/Py, (MB < MC) You are willing to pay more than you have to, consumer surplus increases. Utility and consumer surplus can be increased by consuming less x. A Utility and consumer surplus can be increased by consuming more x. Px = 10 Py = 5 Slope of budget line = -2 B U0 x Intuition • At “C”, the MB = MC for the last unit of both goods consumed. • That is, at “C”, MRS = px/py, or U x px U y py y Ux U y px p y A C U1 B U0 x Optimization • Unconstrained optimization is a lot easier to solve than constrained optimization. – Substitution: maximize the cross section of U along the budget line – Lagrange method Substitution • This turns the constrained optimization into an unconstrained problem. • Find the equation for the cross section of the U=U(x,y) above the budget line and maximize it -- i.e. find the top of the parabola U y y* x* x Substitution • Substitute and maximize U x y; M=p x x+p y y dU Mx 1 dx py Mx 1 py Mx px x 1 x py py 1 px x 0 py And substitute again p M Ux x p y py 1 px x py Mp y x 1 x 1 px p y x M 1 px M M=p x +p y y 1 px M py y M 1 M 1 M py y 1 y M 1 p y Problem with this method • It can get very mathematically complicated very quickly. • Even U=xαyβ gets very tricky to solve. LaGrange Method • LaGrange knew that unconstrained optimization (like profit max) is relatively simple compared to constrained optimization. • Taking what he knew unconstrained optimization he attempted to simplify the constrained maximization problem by making it mimic the unconstrained problem. Unconstrained Optimization Example max v f ( x, y ) g ( x, y ) FOC v x f x ( x , y ) g x ( x, y ) 0 f x g x v y f y ( x, y ) g y ( x, y ) 0 f y g y Maximizing v f ( x, y ) g ( x, y ) means fx gx fy gy • Profit maximization is an example of this. We maximize the difference between two functions: π=R(q) – C(q). Lagrange Method • LaGrange wanted to find a way that was simpler than constrained optimization and more workable than simple substitution. – He wanted to make constrained optimization take the form of the simpler unconstrained problem. max v f ( x, y) g ( x, y) • First, let’s look at a simpler problem. max v f ( x) g ( x) Maximize Utility - Expenditure • Maximize utility minus the cost of buying bundles. Think about a one-good world. U U=U(x) slope = Ux Expenditure = E = pxx max v f ( x) g ( x) slope = Ex= px max v U ( x) px x x* x Problems: • Ux not measured in $ like E. • E is not constrained, we can spend as much as we like. Maximizing Utility - Expenditure • First change the expenditure function by multiplying px by λ. Now call that function EU. U U=U(x) slope = Ux Slope = EUx= λpx EU=λpxx max v U ( x) EU ( x) max v U ( x) px x x* x • We want λ to measure the marginal utility of $1. – In that case, units of x consumed would cost us utility and both U(x) and EU(x) would be measured in the same units. Maximizing Utility - Expenditure • Problem, an infinite number of λ choices that will each solve this with a different x* EUx = λ1 px U U=U(x) EUx= λ2 px EUx = λ3 px x* x* x* x • Now the slope of the expenditure function and expenditure are measured in utils, not dollars. But we are not constraining x yet. LaGrange Method • So first subtract λM from the expenditure function to get EL = λpxx - λM U U=U(x) slope = Ux EU = λpxx EL = λpxx - λM x* -λM slope = ELx= λpx x LaGrange Method • We know we want to find the x* such that that distance between U(x) and EL(x*) = U(x*). That is, where EL(x*) = 0 • So we maximize v = U(x) - 0 • Substitute λpxx – λM = 0 in for 0, to constrain x* to our budget. U U=U(x) slope = Ux max L U ( x) E ( x) L max L U ( x) ( px x M ) U=U(x*)-0 EL = λpxx - λM 0 x* -λI slope = ELx= λ px x LaGrange Method • Our optimization becomes an unconstrained problem by including the requirement that λpxx = λM. • λ is chosen along with x to maximize utility so that λ = the marginal utility of $1. That is, λpx = Ux. U U=U(x) slope = Ux max L U ( x) ( px x M ) U=U(x*)-0 EL = λ(pxx – M) 0 x* -λM slope = ELx= λ px x LaGrange Method max L U ( x) ( px x I ) or, equivalently max L U ( x) ( I px x) FOC Lx : U x px 0 L : I px x 0 L x is the condition that f x g x when we maximize v f ( x) - g ( x) L ensures that the solution satisfies px x I , i.e. that L( x*) U ( x*) Two Goods: Lagrange’s Manufactured Plane U • To maximize utility, maximize the height of the utility function above the plane EL = λpxx + λpyy – λM • Such that λpxx + λpyy – λM = 0 U = U(x,y) y When x = 0 and y = 0, U = - λM ELy= λ py ELx= λ px x LaGrange Plane EL=g(x,y) EL= λpxx+ λpyy- λM g’x=ELx= λ px g’y=ELy= λ py Lagrange Method U U = U(x,y) max v U ( x, y) 0, such that 0 = ( px x py y M ) max L U ( x, y ) ( px x p y y M ) max L U ( x, y ) ( M px x p y y) y x UL=g(x,y) = 0 λ(pxx+pyy – M) = 0 Basic Demand Analysis • Using Lagrangian to generate ordinary (Marshallian) demand curves. – FOCs necessary – SOCs sufficient (check that they hold) – Ordinary (Marshallian) demand curves – Inverse demand curves – Meaning of λ – Indirect Utility – Expenditure Function – Comparative Statics General Results Demand Functions using Lagrange’s Method • Set up and maximize: L (x, y) U(x, y) (M p x x p y y) FOC: necessary conditions for a maximum λ* chosen so that the constraint plane is L x U x p x 0 U x p x L y U y p y 0 U y p y parallel to the utility function. L M p x x p y y 0 M p x x p y y Solve to get two interesting results Ux px , tangency condition Uy py Any x* and y* that maximizes utility will also have to exhaust income. Ux Uy , bang for the buck the same for last unit px py FOCs for an Optimum • For utility to be maximized, it is necessary that the indifference curve is tangent to the budget constraint (as above). U x px Uy py • But it is not sufficient, we also need a diminishing MRS. y Utility Maximized y Utility Maximized y FOCs satisfied FOCs satisfied x x x SOCs for an Optimum • Sufficient condition for a maximum to exist – If the MRS is non-increasing (utility function quasi concave) for all x, that is sufficient for a maximum to exist – but it may not be unique. – If the MRS is diminishing (utility function strictly quasi concave) for all x, that is sufficient for a unique maximum to exist. Need this to satisfy second order conditions for maximization. SOCs do not hold y Utility Maximized y y SOCs satisfied x x x Expenditure Minimization: SOC • The FOC ensure that the optimal consumption bundle is at a tangency. • The SOC ensure that the tangency is a minimum, and not a maximum by ensuring that away from the tangency, along the budget line, utility falls. U*>U’ Y U=U* U=U’ X Checking SOC: utility function strictly quasi-concave • The second order conditions will hold if the utility function is strictly quasi-concave – A function is strictly quasi-concave if its bordered Hessian is negative definite. That is: H 0 Ux U x U xx 0 Ux Uy 0 and H U x U xx U xy 0 U y U yx U yy • A function is strictly quasi-concave if: 1. 2. -UxUx < 0 2UxUxyUy - Uy2Uxx - Ux2Uyy > 0 Checking SOC: Constrained Maximization • The second order condition for constrained maximization will hold if the following bordered Hessian matrix is negative definite: Note: let L( x, y ) U ( x, y ) g ( x, y ) L H L x L y H 0 L x Lxx Lyx px px U xx Uy Ux , py So this Hessian and The last only differ by px L y Lxy Lyy 0 px py 0 and H px U xx U xy 0 p y U yx U yy px 2 0, and 2px p yU xy px 2U yy p y 2U xx 0 Ordinary Demand Curves • And from the FOC: From these three equations and unknowns: Lx : U x px U p Ly : U y p y MRS x Uy x py , solve for y to get income consumption curve L : M px x p y y Solve to get the following: x* x( px , p y , M ), ordinary or Marshallian demand y* y ( px , p y , M ), ordinary or Marshallian demand Ux U y * ( px , p y , M ), = = , marginal utility of $1 px p y Inverse Demand Curves • Starting with the ordinary demand curves: Solve to get the following inverse demand equations: p*x px ( x, p y , M ) Recognize that at the optimal bundle p*x MRS * py And p*x p*y MRS So the inverse demand curve tells us the (willingness to give up y for another x) p y . I.e. the dollar value of the y that the individual is willing to give up for an x. Utility and Indirect Utility • Maximum Utility, a function of quantities U * U x* , y * • Indirect Utility a function of price and income Once we plug in x* x( px , p y , M ), y* y ( px , p y , M ) and get utility as a function of only prices and income V * V x( px , p y , M ), y ( px , p y , M ) V * V px , p y , M Optimization : Expenditure Function • Start with indirect utility V * V px , p y , M • Solve for M: M * E px , p y ,U • This equation determines the expenditure needed to generate Ū, the expenditure function: E * E px , p y ,U Digression: Envelope Theorem • Say we know that y = f(x; ω) – We find y is maximized at x* = x(ω) • So we know that y* = y(x*=x(ω),ω)). • Now say we want to find out * * * dy* d * dy dy dy dx * d d dx d • So when ω changes, the optimal x changes, which changes the y* function. • Two methods to solve this… Digression: Envelope Theorem • Start with: y = f(x; ω) and calculate x* = x(ω) • First option: • y = f(x; ω), substitute in x* = x(ω) to get y* = y(x(ω); ω): * dy x(), dy y* () d d * • Second option, turn it around: then substitute x* = x(ω) f x, y (x; ) • First, take y * dy x(), y* () into yω(x ; ω) to get dy d d * •And these two answers are equivalent: y* () y* () Envelope Result, 1st option to get U* * M • Plug the optimal values into the LaGrangian L* U x(p x , p y , M), y(p x , p y , M) (p x , p y , M) M p x x(p x , p y , M) p y y(p x , p y , M) Differentiate with respect to M L* x * y* * x * y* * Ux Uy 1 p x py M p x x * p y y * M M M M M M L* x * y* * * * * Ux px Uy py M p x x * p y y * M M M M And as we are at a maximum, the FOC get us: L* x * y* * L* * 0 0 0 * M M M M M U As L=U (because M - p x x - p y y 0), (p x , p y , M) M In other words, when income rises by $1, you gain $1 worth of utility. Optimization: Envelope Result U* *x* p x • Plug the optimal values into the LaGrangian L U x(p , p , M), y(p , p , M) (p , p , M) M p x(p , p , M) p y(p , p , M) * x y x y x y x x y y x y Differentiate with respect to px * * x* * y* * * p x x py M px x py y p x p x p x * L* x* y* * * * * * * Ux px Uy py x M px x py y p x p x p x p x L* x* y* Ux Uy p x p x p x And as we are at a maximum, the FOC get us: L* x* y* * * * L* 0 0 x 0 * x p x p x p x p x p x U x L* U U* As L=U and as , x , means that x x(p x , p y , M) p x p x p x px In other words, when the price of x rises by $1, you lose $1 worth of utility for every x bought. Optimization: Comparative Statics • If we have a specified utility function and we derive the equations for the demand functions, the comparative statics are easy. – Take the derivatives to calculate the changes in x and y when prices or income change. • However, what if all we know is U = U(x, y) and we feel safe only assuming: Ux > 0 Uy > 0 Uxx < 0 Uyy < 0 • Can we get anything from that? Optimization: Comparative Statics • Start with: L (x, y) U(x, y) (M p x x p y y) FOC: necessary conditions for a maximum L x U x (x, y) p x 0 L y U y (x, y) p y 0 L M p x x p y y 0 And the equations for utility maximizing x, y, x* x(p x , p y , M) y* y(p x , p y , M) * (p x , p y , M) Comparative Statics: Utility Maximizing x*, y*, λ* Substitute equations for x*, y* and * into the FOC (1) U x x(p x , p y , M), y(p x , p y , M) (p x , p y , M)p x 0 (2) U y x(p x , p y , M), y(p x , p y , M) (p x , p y , M)p y 0 (3) M p x x(p x , p y , M) p y y(p x , p y , M) 0 Whatever happens to prices or income, consumption will adjust to maximize utility. Comparative Statics: Effect of a change in M Differentiate (1), (2), (3) w.r.t. M y x 0 px U xy M M M y x 0 py U yy U yx M M M y x 0 py 1 px M M Rearrange y x 1 py px 0 M M M y x 0 U xy U xx p x M M M y x 0 U yy U yx p y M M M U xx Side note: x y px py 1 M M Tells us that if income increases by $1, so will total expenditure. Comparative Statics: Effect of a change in M Put in Matrix Notation • Solve for Mx 0 p x p x U xx p y U yx p y M 1 x U xy • 0 M 0 U yy y M Assuming H 2p x p y U xy p 2x U yy p 2y U xx 0 0 1 p y p x 0 U xy x p y M 0 U yy H ? p y U xy p x U yy () X could be either normal or inferior. 0 Comparative Statics: Effect of a change in I Put in Matrix Notation • Solve for ¶y ¶M 0 p x p x U xx p y U yx p y M 1 x U xy • 0 M 0 U yy y M Assuming H 2p x p y U xy p 2x U yy p 2y U xx 0 0 p x p x U xx 1 0 y p y M U yx 0 H ? p x U yx p y U xx () Y could be either normal or inferior. 0 Comparative Statics: Effect of a change in px Differentiate (1), (2), (3) w.r.t. px U xx x y U xy px 0 p x p x p x U yx x y U yy py 0 p x p x p x p x x y x py 0 p x p x Rearrange x y 0 p p x x y p x p x p x x y p x U xx U xy p x p x p x p y x y U yx U yy 0 p x p x p x Comparative Statics: Effect of a change in px Put in Matrix Notation • Solve for px x 0 p x p y p x U xx U xy p y U yx U yy p x x x p x 0 y p x Assuming H 2p x p y U xy p 2x U yy p 2y U xx 0 x p x 0 x p y p x U xy p y 0 U yy H X could be giffen. ? x U xy p y x p x U yy p y 2 () 0 Comparative Statics: Effect of a change in px Put in Matrix Notation… AGAIN • Solve for ¶y ¶ px 0 p x p y p x U xx U xy p y U yx U yy p x x x p x 0 y p x Assuming H 2p x p y U xy p 2x U yy p 2y U xx 0 0 p x x p x U xx p y U yx y p x H 0 ? p x p y x p x U yx x p y U xx 0 () X and y could be compliments or substitutes. Comparative Statics: Preview of income and substitution effects Rearrange these x p x () 2 U p y x p y U xy p x yy () ? ? x p y U xy p x U yy 0; M () x p x ? x p y 2 x M () ? p x p y x p x U yx p y U xx y p x p y x p x U yx x p y U xx p x () () Income effect matters x U xy p y x p x U yy p y 2 Sub in these ? ? y p x U yx p y U xx 0 I () ? y p x p y x ; y M p x ( ) Specific Utility Functions • Cobb-Douglas • CES • Perfect Compliments Cobb-Douglas: Utility Max • Problem: U(x, y) x y , s.t. M-p x x-p y y • Set up the LaGrangian L=x y + (M-p x x-p y y) • FOC L x : x 1 y p x 0 L y : x y 1 p y 0 L : M-p x x-p y y=0 U x x 1 y U y x y1 x 1 y y MRS 1 x y x Cobb-Douglas: Demand • FOC Imply, to maximize utility, these must hold. yp y xp x x ; y p x p y • Plug into the budget constraint to get the ordinary (Marshallian) demand functions: M M x* ; y* px py • Note, demand only a function of own price changes (one Cobb-Douglas weakness) Cobb-Douglas: Demand • Preferences are homothetic (only a function of the ratio of y:x). When income rises, optimal bundle along a ray from the origin. – Expenditure a constant proportion of income px x* M; p y y* M – Income elasticities are = 1 dx M eM dM x ( )p x M ( )p x M 1 Cobb-Douglas: Indirect Utility • Plug x* and y* into the utility function U(x, y) x y M M x* ; y* ( )p x ( )p y M M V ( )p x ( )p y Cobb-Douglas: Expenditure Function • Start with indirect utility function M M V ( )p x ( )p y • Solve for M, and then rename it E M ( )V px 1 py E ( )V px 1 py CES: Utility Max • Problem: U(x, y) x y , s.t. M - p x x - p y y • Set up the Lagrangian L x y (M - p x x - p y y) • FOC L x : x 1 p x 0 L y : y 1 p y 0 L : M - p x x - p y y 0 x MRS y px x py y 1 1 CES: Demand • FOC Imply p x y 1 x p y 1 1 ; p y x 1 y p x 1 1 • Plug into the budget constraint and solve: x M py px p x 1 1 ; y M 1 px py 1 p y CES: Indirect Utility • Plug x* and y* into the utility function U(x, y) x y M x ; 1 p px y 1 p x y M px py p y 1 1 M M V p 1 1 px y 1 p y p x 1 p p x y CES: Expenditure Function • Solve for M, then rename E. M V p y 1 px 1 px M 1 p px 1 y p y 1 V p y 1 px 1 px 1 2M 1 p px 1 y py CES: Expenditure Function (cont) • Solve for M, then rename E. V M 1 2 p y 1 px 1 p x E 1 1 p px 1 y p y V 1 1 1 2 p y 1 p x 1 1 py 1 px p p y x 1 Perfect Compliments: Utility Max • Problem: U(x, y) min(x, y), s.t. M-p x x-p y y • No Lagrangian, just exhaust income such that: x y • So plug this condition into the budget equation – Essentially, we substitute the expansion path into the budget line equation. Perfect Compliments: Demand • Demand equations y M=p x +p y y M= p x +p y y M y p x +p y x M=p x x+p y M= p x + p y x M x px + py Perfect Compliments: Indirect Utility M M V(p x , p y , M) min , p x +p y p x +p y • Since utility from x = utility from y at utility max: M M V(p x , p y , M) or V(p x , p y , M) p x +p y p x +p y Perfect Compliments: Expenditure Function • Since the utility from consumption of each must be equal, p x +p y p x +p y V E V, or E Bonus Topics • • • • • Money Metric Utility Function Homogeneity Corner Solutions (Kuhn-Tucker) Lump-Sum Principle MRS and MRT (Marginal Rate of Transformation – slope of PPF) Money Metric Utility Function • Start with an expenditure function and replace with Ū with U=U(x,y) E e px , py , U E e p x , p y , U(x, y) • Now we know the minimum expenditure to get the same utility as the bundle x’, y’. U E U E p x , p y , x, y • That is, for an x’, y’, this function tells us the cost of being at the tangency on the same indifference curve. As all bundles on the same curve get the same min. expenditure and prefernece ordering is preserved, this is a utility function. Evaluating Housing Policy • Assume you find the poor generally expend 1/3 of income on housing. • The government wants to double the quality of housing the poor consume at the same 1/3 their income. • How to evaluate? Pre-Public Housing If expenditure on housing is generally 1/3 of income, assume U=h1/3y2/3 Y ph=$1 px=$1 M=$1,000 667 M 2M h* ; y* 3p x 3p y 333 667 H Public Housing ph=$1 px=$1 M=$1,000 Qualified citizens get 667 housing units for 1/3 of income ($333) Y V E 1 3 1 2 3p 3p x y 1 U E 3 h y 1 3 2 2 3 667 3 1 2 3 3 2 3 333 667 H Public Housing: Money Metric Utility V E 1 3 1 2 3p x 3p y 1 U E 3 h y 1 3 2 ph=$1 px=$1 M=$1,000 Qualified citizens get 667 housing units for 1/3 of income ($333) Y 2 3 3 1 2 3 3 2 3 667 UEPH=1,261 UE=1,000 333 667 The extra housing has a value to the poor of $261. Depending on the cost to the government of providing the housing, the program can be evaluated. H Homogeneity • If all prices and income were doubled, the optimal quantities demanded will not change – the budget constraint is unchanged xi* = xi(p1,p2,…,pn,M) = xi(tp1,tp2,…,tpn,tM) • Individual demand functions are homogeneous of degree zero in all prices and income • To test, replace all prices and income with t*p and t*M. The quantity demanded should be unaffected (all the “t” should cancel out). Corner Solutions Y Non-corner solutions: Optimal bundle will be where x > 0, y > 0 and MRS = px/py Two corner solutions: Optimal bundle will be where x =0 X Corner Solutions Y A X • At “A”, MRS = px/py, but the optimal quantity of X = 0 Corner Solutions Y B X • At “B”, the tangency condition holds where x* < 0. Given the budget line, the optimal feasible x is where x = 0 and MRS < px/py Corner Solution • To develop these conditions as part of a Lagrangian equation, we add non-negativity constraints: 𝑥 = 𝑠 2 (we could add 𝑦 = 𝑡 2 if we wanted to be really thorough). • Lagrangian: – L = 𝑈(𝑥, 𝑦) + 𝜆(M − 𝑝𝑥 𝑥 − 𝑝𝑦 𝑦) + 𝜇(𝑥 − 𝑠 2 ) – Requiring x = s2 is simply a way of ensuring that x ≥ 0. Corner Solutions • FOC L x U x p x 0 L y U y p y 0 L M p x x p y y 0 L x s 0 2 Ls 2s 0 Corner Solutions • Kuhn-Tucker Condition Ls 2s 0 • This tell us that either: • μ =0 (the optimum is at a tangency) • s = 0 (the optimum is where x = 0) • μ =0 and s = 0 (the optimum is at a tangency where x = 0) Corner Solutions Y If s > 0 and μ = 0 L x U x p x 0 L y U y p y 0 Ux px Uy py • The usual assumption is that the optimal px bundle will be where x>0, y>0 and MRS py X Corner Solutions If s = 0 and μ = 0 Y L x U x p x 0 A L y U y p y 0 Ux px Uy py X • At “A”, MRS px py , but the optimal quantity of x = 0 Corner Solutions If s = 0 and μ > 0 Y L x U x p x 0 L y U y p y 0 Ux px Uy py B Ux Ux px MRS Uy Uy py B’ px MRS , at B py X • At B’, the tangency condition holds where x* < 0. px • The optimum is where x=0 and MRS py Kuhn-Tucker Example • Utility: U=xy+20y, M = 40, px = $4 and py = 1. L xy 20y (40 4x y) L x : y 4; L y : x 20 L : 40 4x y y Gets 4x 80 y, x 20 4 Solve x 5, y 60 Oops. Looks Like • Tangency where x=-5, y=60 Y y MRS = 4 x 20 X How about a feasible optimum? • Tangency where x=-5, y=60 Y MRS = 40 2 0 20 px 4 60 MRS = 5 20 p y 1 Slope = px 4 py 1 X Kuhn-Tucker Set-up • Utility: U=xy+20y, M = 40, px = $4 and py = 1 L xy 20y (40 4x y) (x s 2 ) L x : y 4 0 L y : x 20 0 L : 40 4x y 0 L : x s 2 0 Ls : 2s 0 Kuhn-Tucker Condition Kuhn-Tucker Result Kuhn-Tucker: 2s 0, so or s or both = 0 y px Use L x and L y : x 20 p y px y If μ=0, optimum is a tangency where x 20 p y If s 0, optimal x 0. If 0, and s 0 optimum at corner and: px y y MRS p y x 20 x 20 If 0, and s 0 optimum at corner but tangency where x 0. If 0, and s 0 optimum at interior. Kuhn-Tucker Result • In this example, at x = 0, y = 40, 0 : y At corner, this condition holds: =4 x 20 So 4x 80 y and 40 y 40 At the corner, MRS = 2 x 20 0 20 If p x 2, then the optimum is at x 0, y 40, 0 and s 0. How about a feasible optimum? • Optimum where x=0, y=40 Y At optimum: px y y MRS = p y x 20 x 20 p x 40 40 40 MRS = py 20 20 px y y 4 MRS = 2 p y x 20 x 20 X Solving Kuhn-Tucker • If you find that the optimal bundle is not on the budget constraint, check all corners for a maximum utility. Lump Sum Tax • Taxing a good vs taxing income – Tax on x only px x py y M M p which is y= - x x py py x * and y * are optimal bundles so p x x * p y y* M , and R* = x * – Lump sum tax (income tax) p x x p y y M x * M x * p x y x py py Lump Sum Tax • Difference in the budget lines: sales tax Without a tax, M px x * y* py py With the unit tax on x, at any x*, y* M p x x * x * py py py M p x x * M x * p x x * At any x*, y*-y py py p y p y p y * y*-y* x * py Lump Sum Principle • A tax on x rotates the budget line to have: Y M py y* x * py yτ* slope px py x* M px M px X Lump Sum Tax • Difference in the budget lines: income tax Without a tax, M px x * y* py py With an income tax R*= x * y*R M x * p x x * py py py M p x x * M x * p x x * At any x*, y*-y py py p p p y y y * R x * y*-y py * R Lump Sum Principle • Tax paid = x*τ. Alternatively, an income tax of that same amount would shift the budget line so that the consumer can just afford the same bundle they chose under the tax on x. Y x * py M py M x * py x * py y* x* M px M x * px M px X Lump Sum Principle • When U xy.5 2M M x ; y 3p x 3p y • Indirect Utility function is: .5 2M M V(p x , p y , M) 3p x 3p y Lump Sum Principle • Set, I=60, py = 2, px=1 2M M x 40; y 10 3p x 3p y • Utility is: .5 2M M V(p x , p y , M) 126.49 3p x 3p y Lump Sum Principle • With a $1 tax on x, x 2M M 20; y 10 3 p x 1 3p y • Utility is: .5 M 2M V V(p x , p y , M) 63.24 3 p x 1 Px 3p y * • And tax revenue is $20. Lump Sum Principle • With a $20 income tax, x* 2(M 20) M 20 =26.667; y* 6.667 3Px 3Py • Utility is: .5 2(M 20) M 20 V V(p x , p y , M) 68.85 3Px 3Py * • 68.85 > 63.24 The Lump Sum Principle: Perfect Compliments • If the utility function is U = Min(x,4y) then plug x=4y into the budget equation to get: M x p x 0.25p y * M y 4p x p y * • The indirect utility function is then: M M V(p x , p y , M) Min ,4 p x 0.25p y 4p x p y The Lump Sum Principle: Perfect Compliments • Set, I=60, py = 2, px=1 M x 40 p x 0.25p y * M y 10 4p x p y * • Utility is: M M V(p x , p y , M) Min , 4 40 p x 0.25p y 4p p x y The Lump Sum Principle: Perfect Compliments • With a $1 tax on x, M x 24 (p x 1) 0.25p y * M y 6 4(p x 1) p y * • Utility is: M M V(p x , p y , I) Min ,4 24 (p x 1) 0.25p y 4(p 1) p x y The Lump Sum Principle: Perfect Compliments • With a $24 income tax, M - 24 x 24 p x 0.25p y * M - 24 y 6 4p x p y * • Utility is: M - 24 M - 24 V(p x , p y , M) Min ,4 24 p x 0.25p y 4p p x y • Since preferences do not allow substitution, the consumer makes the same choice either way and the tax does not affect utility. MRS=MRT • Using Varian’s example about milk and butter – B* and M* may be different for all consumers. – However, the MRS at the tangency is the same for ALL consumers, no matter the income or preferences. Milk Pb=3; Pm=1 MRS=3 MRS=3 MRS=3 Butter Marginal Rate of Transformation (a.k.a Rate of Product Transformation) • Varian is sort of implicitly assuming the PPF is linear. So there is a constant trade of in producing butter or milk as resources are reallocated. Milk MRS=3 MRT=3 Social Indifference curve Butter Marginal Rate of Transformation • With a more realistic PPF, the MRT rises as more butter and less milk is produced. MRT Milk MRT=3 Cb q m Cm q b Assuming competative firms producing b and m C p MRT b b Cm p m or Cb Cm , marginal cost p the same at pb pm the margin for all goods. Butter MRS=MRT • In the long run (π=0), the cost of producing butter must be 3 times the cost of producing milk. That is, the tradeoffs in consumption = the tradeoff in production Milk MRT Cb p b U b MRS Cm p m U m px/py=3 Butter Back to Varian’s treatment • A new technology that allows you to produce butter with 4 gallons of milk is not going to be a winner as everyone would choose the cheaper butter previously offered. Milk MRT=4 Pb=4; Pm=1 Butter • However, a new technology that allows you to produce 1 pound of butter with 2 gallons of milk IS going to be a winner! Milk MRS=MRT=2 Pb=2; Pm=1 Butter MRS=MRT • Here is what improved butter making technology does with a more standard PPF. Milk px/py=3 px/py=2 Butter