Econ 604 Advanced Microeconomics s1

Total Page:16

File Type:pdf, Size:1020Kb

Econ 604 Advanced Microeconomics s1

Econ 604 Advanced Microeconomics Davis Spring 2006

Reading. Chapter 3 (pp. 66-86) for today Chapter 4 (pp. 91 to 113) for next time Problems: To collect today Chapter 2, 2.3, 2.5,2.6, 2.9. For next time: Chapter 3, 3.2, 3.4, 3.5, 3.7 Lecture #3.

REVIEW

II. The Mathematics of Optimization

A. Maximization of a Function of One Variable. 1. Necessary and sufficient conditions for a maximum. Recall, our rule is that optimization requires that the first order condition equal zero, and that the second derivative be negative (for a maximum) or positive (for a minimum).

Example: f(x) = 150x – 5x2 f’(x) = 150 – 10x f’’(x) = -10

B. Functions of Several Variables

1. Partial Derivatives. Given y  f (x1 , x2 ,...xn ) . The partial derivative of y f y =f(x1, …., xn) with respect to x1 we denote as   f x1  f1 . x1 x1 (f ) x Second order conditions are i = fij = fji. Young’s Theorem assures

x j that the cross partial is independent of the order in which the derivatives were taken.

2. Maximizing Functions of Several Variables. a. Total differentiation. Given several variables, the total differential is dy = f1dx1+ f2dx2+… + fndxn

b. First-Order Condition. Critical points arise when dy=0, or when f1 = f2 =… = fn =0. c. Second -Order Condition . The same as the single variable case, except that there are “cross effects.” For a maximum or a minimum, own effects must dominate cross effects, or (in the 2 variable case) 2 f11(f22) –f12 >0

Note: the same second order condition implies a minimum as well as a maximum (intuitively the own effects must dominate the cross effects).

i. Implicit Functions. Note, we can write functions in “implicit form” (that is, an equality set equal to zero). If second order conditions hold, we can solve for one variable in terms of another. ii. The Envelope Theorem. A major application of the implicit function theorem that we will use frequently. Consider an implicit function f(x, y, a) = 0, where x and y and variables and a is a parameter. The envelope theorem states that for small changes in a, dy*/da can be computed by holding x constant at its optimal value x*, and simply calculating y/a {x = x*(a)}.

C. Constrained Maximization. A problem that arises frequently in practice (e.g., consumers maximize utility subject to an income constraint. 1. Lagrangian Multiplier Method. In general, we optimize a function f(x1, …, xn) in light of a series of constraints about those independent variables g(x1, …,

xn).=0. Write: L = f(x1, …, xn) +g(x1, …, xn) and take FONC, a series of n+1 equations in n+1 unknowns.

2. Interpretation of the Lagrangian Multiplier. Each of the first order conditions

in the above system may be solved for . That is f1/g1 = f2/g2 =, …., fn/gn = , or = marginal benefit of xi / marginal cost of xi

Observation: One way to better understand the role of  from an intuitive level would be to consider a constrained function of a single variable. Suppose f(x) = 10x – x2 and suppose that this was subjected to the constraint that x=3.

Then, L =10x- x2 +(3-x) From FONC

10 – 2x =  and 3 – x = 0

When x=3, 10-2(3) = 4. Suppose we relax the constraint, and increased x to 4. Then 10 – 2x =  and 4 – x = 0 and  = 2. In x, f(x) space, this is just the slope of the line tangent to f(x).

2 Reasoning similarly, when x=5 or more,  =0 and the constraint is no longer binding. So  is the marginal increase in the objective brought about by relaxing the constraint, as long as the constraint binds.

3. Duality. Observe that for every constrained maximization problem, there is an implied constrained minimization problem (this is conceptually no more complicated than looking at a simple sum, and realizing that there is an implied difference associated with it. Thus, for my 8 year old daughter, given the problem 24-17 = 7, it was much easier to solve dual problem 17+7 = 24.) We encounter the same problems in optimization, albeit they are more complex.

4. Envelope Theorem in Constrained Maximization Problems. The envelop theorem applies just as it did for unconstrained problems. Find optimal value of the dependent variable in terms of the parameters at the critical point. Then take the derivative w.r.t. the constant.

5. Second Order Conditions with Constrained Optimization. A final observation. Optimize an objective f(x1, x2) subject to a linear constraint, g(x1, x2) =c - b1x1 - b2x2=0. Form the Lagrangian expression

L = f(x1, x2) + ( c - b1x1 - b2x2) Totally differentiating the objective twice, and inserting the constraint, we 2 2 2 2 2 find that d y = (f11 f2 - 2f12 f1 f2 + f22 f1 )(dx1 / f2 ) <0 as long as the bolded term is negative, or 2 2 (f11 f2 - 2f12 f1 f2 + f22 f1 )<0

This is a quasi concavity condition that we will use later. Intuitively, this indicates the minimum condition for an interior solution to a linearly constrained optimization problem. In contrast to the simple “mound” condition necessary in the case of unconstrained optimization, a sufficient condition for an optimum given a constraint is the weaker condition that the constraint be more convex than the objective function. As we will see below, this condition will be equivalent to asserting that any linear combination of two points on a curve be in the interior of the function.

Note: One way to view quasi concavity is as follows: A three dimensional “mound” is quasi concave if it can be completely represented in a two dimensional topographical map

PREVIEW III. Choice and Demand. A. Chapter 3. Preference and Utility. 1. Axioms of Rational Choice 2. Utility a. Nonuniqueness of Utility Measures b. The Ceteras Paribus Assumption

3 c. Utility form Consumption of Goods d. Arguments of Utility Functions e. Economic Goods 3. Trades and Substitution a. Indifference curves and the Marginal Rate of Substitution b. Indifference Curve Map c. Indifference Curves and Transitivity d. Convexity of Indifference Curves. e. Convexity and Balance in Consumption 4. An Alternative Derivation 5. Examples of Utility Functions a. Cobb-Douglas Utility b. Perfect Substitutes c. Perfect Complements d. CES Utility 6. Generalizations to More than Two Goods

LECTURE______

III. Choice and Demand. The purpose of the next several chapters is to develop the notion of market demand, starting from primitive assumptions about individual optimization. In the first chapter (chapter 3= we discuss the assumptions economists usually use to model individual behavior, and then use those assumptions to develop indifference curves

A. Chapter 3. Preferences and Utility. 1. Axioms of Rational Choice. One way to begin an analysis of individual behavior is to state a set of basic “axioms” that characterize rational behavior. We define 3. a. Completeness: For any two goods, services or situations A and B exactly one of the following conditions holds i. A is preferred to B (A ℏ B) ii. B is preferred to A (B ℏ A) iii. A and B are equally attractive. This condition implies that all items can be ranked.

b. Transitivity. If an individual reports A ℏ B and that Bℏ C then Aℏ C . This transitivity condition implies that choices are internally consistent.

c. Continuity If an individual reports A ℏ B then for a situation A’ sufficiently “close” to A it must be the case that A’ ℏ B. This largely technical assumption rules out “knife’s edge responses to small changes in income and prices.

2. Utility. If individuals follow these three assumptions, all situations must be rankable. Following Bentham, we label this preference ranking as “utility”. We will say

4 that preferred goods confer higher utility. That is, A ℏ B implies U(A) > U(B) (where the latter term is an inequality, and the former is a preference ranking.) Formally define the follow

Utility: Individual preferences are assumed to be represented by a utility function of the form utility = U(X1, X2, …, Xn) where X1, X2, …, Xn denote quantities of goods 1 to n consumed in a given time frame. Utility is unique up to an order-preserving constant.

Some comments on this definition:

a. Non-Uniqueness of Utility Measures. Importantly, the utility measure is only ordinal. We can order utility only up to an order-preserving transformation. For this reason interpersonal utility comparisons are not possible. b. The Ceteras Paribus Assumption Utility is a general measure of satisfaction. Clearly such a measure is affected by general factors, such as one’s outlook on life, or the opinions of one’s friends. However, economists typically hold these effects constant when discussing utility, and focus on the utility of quantifiable goods. c. Utility from Consumption of Goods One implication of the Ceteras Paribus condition is that we often simplify notation, and omit from models variables we hold constant. Thus, given a general utility function utility = U(X1, X2, …, Xn; other things) we often write utility = U(X1, X2, …, Xn) or, if we are concerned particularly about just two goods, we might write utility = U(X1, X2)

This is done with the understanding that the remaining goods, and preferences exist, but are held constant.

5 d. Arguments of Utility Functions. Economists often include rather strange arguments in utility functions. These actually reflect some implicit conventions. For example, we often write

utility = U(W) where W = wealth. Of course individuals do not derive utility directly from wealth, but rather the wealth spent on consumption. So this expression denotes the utility of consuming the most satisfying consumption bundle possible with wealth W.

In labor economics, we often express utility as a function of consumption C, and non work hours, H or leisure

utility = U(C, H)

Finally, in some contexts were are interested in the intertemporal aspects of utility. Thus we express utility as a function of consumption today (time period 1) and tomorrow (time period 2). Or

utility = U(C1, C2)

e. Economic Goods. We frame the arguments in a utility function as economic “goods”. That is items that have the property that more is preferred to less. These goods may include arguments as simple as hot dogs and pizza, or may include complex composite goods such as leisure. Given that more is preferred to less, we may view movements away from the origin (e.g., the point of no consumption) as utility enhancing. In a two-dimensional space, for example. Y

More preferred than X', Y'

Y' Less preferred than X', Y' X' X

3. Trades and Substitution. Most economic activity involves trading between one good and other. If you purchase a soda to bring to class, for example, you exchange money for something you value more highly, soda. In this section we develop an apparatus for modeling this trading activity.

6 a. Indifference curves and the Marginal Rate of Substitution. Generally an individual will be indifferent to several combinations of a particular combination of goods. Utility from increased consumption of one good can just offset utility losses from decreased consumption of another. We gather these points of indifference together as an “indifference curve”

Indifference Curve: A set of consumption bundles among which the individual is indifferent. That is, the bundles all provide the same level of utility. Y

Indifference Curve

Marginal Rates of Substituion Y1

Y2

U1

X X1 X2

Notice that the curve is bowed inward. This reflects the assumption that individuals generally prefer balanced combinations of goods to one or the other.

One way to analytically characterize the change in substitutability between goods is with the slope of the line tangent to utility curve U1. This is termed the Marginal Rate of Substitution

MRS = -[dY/dX ] |U=U1

Notice that the slope of the line tangent to U1 “flattens” as consumption bundles move from those comprised mostly of Y to bundles consisting mostly of X. This flattening reflects that starting from an evenly split allocation of Y and X, , increasing amounts of X must be given in exchange for units of Y in order to maintain constant utility. This reflects the standard assumption that goods substitute imperfectly for each other. (This, of course, is the opposite of the concavity assumption we impose on production possibilities curves. There, starting from a center allocation, decreasing amounts of X must be given up to increase Y and still maintain constant total production)

7 Of course, for any individual the MRS (and the Indifference curve in general) is determined by individual choices. b. Indifference Curve Maps Now in the above illustration only a single indifference curve, U1 was drawn. More generally, we may conceive of the X Y quadrant as being packed with indifference curves, each reflecting a higher level of utility, as it radiates out further from the origin Y

Increasing Utility

U2

U1

X

Now we explore several properties of these indifference curves. c. Indifference Curves and Transitivity: One implication of our axioms of individual choice is that indifference curves cannot intersect. To see this, consider the illustration below. Y

Increasing Utility

C D

E A

U1

U2 B X

Is such a relationship possible? In fact, it is not. Notice that Aℏ B and C ℏ D. However, the consumer is indifferent between B and C. Thus A ℏ C and Cℏ D. But A and D are on the same indifference curve, a contradiction.

Result: Indifference curves cannot intersect

8 d. Convexity of Indifference Curves. Another way to illustrate a declining MRS is to observe that indifference curves are convex. Convexity implies that any line drawn between two points on an indifference curve will lie above the curve. Any indifference curve that did not satisfy convexity would have a ‘wave’ in it, as illustrated below Y Nonconvex Y Convex

X X

e. ConvexiConvexity and Balance in Consumption: Comparing the above charts, it is also evident that one can see that convexity implies a preference for balance in consumption. With convexity (as seen in the left panel, for any pair of bundles (X1, Y1) and (X2, Y2), a point halfway between ((X1,+X2)/2 , (Y1+ Y2)/2) and (X2, Y2), renders higher utility (with quasi-convexity, it is possible that a the point falls on a line and that consumers are indifferent between the three bundles) This is not true if convexity is violated, as illustrated in the right panel.

Example. Suppose a person’s utility function for hamburgers and soft- drinks is given by

Utility = (XY)1/2

An indifference curve is developed by picking a utility level, and examining all consumption bundles consistent with that level. For example, if U=10

10 = (XY)1/2

The slope of the indifference curve is found by solving for Y

Y= 100/X

Taking the derivative

dY/dX = -100/X2

9 Thus, at X=5, Y=20, the MRS = 100/25 =4. At Y=5, X=20, MRS = -1/4.

4. An Alternative Derivation. A somewhat more mathematical derivation proceeds directly from the definition of the utility function. Given

utility = U(X1, X2, …, Xn)

as defined above, we can define marginal utility of any good as the first derivative. For example, for X1,

marginal utility ==MU1 =U/X1

Notice that, as before MU1is dependent on the other things held constant, as well as subjective components. Thus, it is measurable only up to a monotonic transformation. Taking the total differential of utility

dU = (U/X1) dX1 + (U/X2) dX2 + …+ (U/Xn) dXn.

= MUx1 dX1 + MUx2dX2 + …+ MUxndXn.

To derive the MRS, set the differential to zero, and change only the level of two goods.

dU = (U/X) dX + (U/Y dY = 0 . = MUx dX + MUydY = 0

Solving -dY/dX |U=constant = MUx/MUy.

Notice that the absolute measures of utility drop out when constructing the MRS (e.g., the MRS of, a particular consumption bundles is 4 utils from soda, and 2 utils from pizza. “Utils” cancel out in the ratio). Thus, the MRS is a well defined concept, even when ‘utils’ are not observable.

Diminishing Marginal Utility and the MRS. Recall in the introduction that we used the notion of diminish marginal utility to resolve the Diamond/Water Paradox. Intuitively, it would seem that there is a relationship between diminishing marginal utility and a diminishing MRS, since both ideas refer to the idea that a consumer becomes relatively satiated in one good as they consume more of it. The two concepts, however, turn out to be distinct. The MRS assumption requires that the utility function be quasi-concave. This is related, in a rather complex way to the assumption that marginal utility diminishes (fii<0). The following example illustrates.

Suppose

10 utility = (XY)1/2

Then marginal utilities for X and Y are

1/2 1/2 MUx = .5(Y/X) MUy = .5(X/Y)

Notice in each case, that MUi is inversely related to consumption of i.

Now the MRS = -dY/dX |U=constant =

1/2 MUx/MUy = .5(Y/X) = Y/X .5(X/Y)1/2 Notice that this illustrates the Standard diminishing MRS relationship. That is, starting from an allocation of Y=5 and X=10, MRS = .5 (1/2 a unit of Y trades for a unit of X. But when Y=10 and X=5, MRS = 2, that is, 2 units of Y trade for a single unit of X.

Notice also that a monotonic transformation of the utility function does not affect the MRS. For example, take the natural logarithm of the above utility function

ln(U) = .5ln(X) + .5ln(Y)

Then MUx = .5/X and MUy = .5/Y

And MRS = MUX/MUy = Y/X

as before. Because utility is insensitive to monotonic transformations, we will often make such transformations, when convenient.

5. Examples of Utility Functions. Although the form of an individual’s utility function is not observable, specific functional form are often imposed on utility functions. We consider some of these here.

a. Cobb-Douglas Utility. This is a general case of the quadratic utility function used above for purposes of illustration. It is named after a pair of investigators who used it to investigate production relationships in the U.S.

utility = XY

where  and  are positive constants. In general the constants reflect the importance of the goods to the consumer. Because the function is unique only to a monotonic transformation, it is often convenient to impose the restriction that  +  =1.

11 b. Perfect Substitutes. When goods are perfect substitutes the MRS is constant. That is, a consumer remains on a level curve via a fixed trade-off between two products

utility = U(X,Y) =X+ Y

This relationship might describe consumer preferences between different brands of the same product. (Exxon gasoline substitutes 1 for 1 for Shell gasoline)

c. Perfect Complements On the other hand, some goods generate utility only if consumed in fixed proportion with another. Peanut-butter and bread, or coffee and cream are examples. Such a utility function is represented analytically as follows

utility = U(X,Y) =min(X, Y)

In this case, the goods are consumed in fixed proportion

Y/X = (/) Any deviation will result in waste

d. CES Utility The three functions mentioned are special cases of the more general constant elasticity of substitution function

utility = U(X,Y) = X/+ Y/

when >0 and

utility = U(X,Y) = lnX+lnY when =0.

It can be shown that for such functions, the elasticity of substitution is  = 1/(1- ) (we will develop this more later)

For the case  = -1, the function becomes

utility = U(X,Y) = -X-1 - Y-1

The negative entries make this a curious utility scale. It is, however, perfectly acceptable, as can be verified by checking first derivatives.

Example. Homothetic preferences.

All of the above utility functions share the property of being “homothetic” In other words, the MRS for these functions depends only on the ratio of the two

12 goods consumed, and not the absolute levels. For example, in the Cobb-Douglas function discussed above,

MRS = MUx/MUY = Y/X

a relation which depends only on Y/X. The analytical advantage of homothetic preferences is that indifference curves are similar. Indifference curves for higher utility levels are simply substitutes for those at lower levels of utility.

As a counter example the utility function utility = U(X,Y) = X + ln Y

is nonhomothetic. To see this observe

MUx = 1 MUy = 1/Y

Thus MRS MUx/MUy = Y.

So the MRS changes as the level of consumption increases.

6. Generalizations to More than Two Goods The Cobb-Douglas and CES utility functions can be readily generalized to the case of n goods.

1 2 n utility = U(X1,X2,…Xn)= X1 X2 … Xn

or / / / utility = U(X1,X2,…Xn)= X1  + X2  + … + Xn  These functions have the same properties as those developed above for the 2 good case.

13

Recommended publications