Unit 2: Consumer Theory

Prof. Antonio Rangel December 12, 2013

1
1.1

Consumer demand
The experienced utility function
Suppose that the consumer only cares about two goods: x, good of interest, measured in units m, a measure of all other consumption in $. Think of it as amount of $ spent on all other goods except for x. Assume x 0, but can have m T 0. Experienced utility function: U (x, m) = level of experienced utility (= satisfaction or happiness) from consumption bundle x, m. Assume U is quasi-linear: U (x, m) = B (x) + m, B 0 > 0, B 00 < 0. Remarks: B (x) = willingness to pay for x units of good x B (0) need not be zero = 1, i.e. constant marginal utility; marginal utility since B 00 < 0.
U m U x

= B 0 (x), i.e. decreasing

U (x, m) measured in dollars

1.2

Utility maximization problem

Assume consumer is rational; i.e. chooses action to maximize experienced utility Consumers problem, given wealth W and a price p per unit of good x: max B (x) + W xp max B (x) xp
x 0 x 0 x0,m

max U (x, W xp) by QL assumption by dropping constant W

Familiar structure: benet minus cost Notation: x (p) = optimal choice = demand for x at p. It follows that m (p) = W px (p) Solution: Concavity conditions satised Case 1: Interior (x (p) > 0): when B 0 (0) > p and B 0 crosses p . Case 2: Corner solution at x (p) = 0: when B 0 (0) < p. Case 3: x = : when B 0 (x) > p for all x. Not very interesting, since crossing conditions usually satised via B 0 () = 0

Economic intuition: starting at zero, keep buying as long as MB > MC

1.3

The demand function

Demand curve for individual consumer Demand function: x (p), dened for all p > 0 In economics, we always graph price on the vertical axis, and quantity on the horizontal axis Key result: Demand function MB curve 2

 With x as independent variable, graph B 0 Taking p as independent variable, same locus of points gives the demand function.

1.4

Example

U (x, m) = A ln x + m Observe that all solutions must be interior, since B 0 (0) = > p. Solving B 0 (x) = p yields x (p) =
A p

A is a taste parameter. As A increases the demand curve shifts right (i.e., the amound demanded increases at any p).

1.5

Example

U (x, m) = A ln x + m What is the demand function if there is a mandatory minimum purchase of at least 1 unit of x ? For p A, demand curve the same as in previous example, x bef (p) = xaf ter (p)
For p > A, demand curve now is x af ter (p) = 1 > xbef (p). In this case, since the consumers are forced to buy more than they want, they are made worse o by the policy.

1.6

Properties of the demand function

RESULT: Law of Demand: Outline of proof:

0, w/ inequality if x > 0.
dx dp

For prices in the range of corner solutions: x (p) = 0 =

=0

x (p) > 0 = interior solution = B 0 (x (p)) = p Taking derivatives of both sides w.r.t. p yields B 00 dx =1 dp dx 1 = dp = B 00 Therefore B 00 < 0 = dx <0 dp 3

 RESULT: No income eects:

dx dW

=0

Why? With quasilinear preferences, the FOC that characterizes x does not depend on how much of the good m is consumed Both results approximately hold for goods which account for a small fraction of consumers overall spending Both can fail when U is non-quasilinear

2
2.1

Consumer surplus
Consumer Surplus: Basics
How do we measure well-being of consumer? Consumer surplus = NET benet of buying optimally at price p, measured in dollars Mathematically, it is dened as follows: CS (p) = U (x (p), W px (p)) U (0, W ) = B (x (p)) B (0) px (p) by quasilinearity x ( p ) x ( p ) = B 0 (x)dx p dx the Fundamental Theorem of Calculus 0 0 x ( p ) = (B 0 (x) p) dx by linearity of the integral
0

Denition in the rst line says CS (p) equals experienced utility of buying x (p) minus experienced utility of buying zero. Graphical description: CS = area between B 0 and p, from x = 0 to x = x Properties of CS: CS 0 CS > 0 i x (p) > 0 4

 Can see both from the graphical description Can also see it from a revealed preference argument: (0, W ) is always feasible, i.e. the consumer can always buy nothing. Therefore CS cant be negative: if the consumer buys anything, it must make her better o.

2.2

Example
B0 =
A x A2 p2

Suppose U (x, m) = 2A x + m x ( p) =

Consider two dierent ways of computing CS (p). Direct method: CS (p) = U (x (p)) Uno trade h p i = 2A x (p) + W px (p) [0 + W ] 2A2 A2 2p p p 2 A = p =

Graphical/integral method: CS (p) =

A p dx = x 0 2 A2 = 2A x px 0p A2 = p
A2 p2

(B 0 (x) p) dx

2.3

Eect of price changes on consumer surplus

Consider eect on CS of price decrease from p0 to p1 It is given by: CSp0 p1 = CS (p1 ) CS (p0 ) = Change in experienced utility following price change from p0 to p1 x 1 = x0 p + (B 0 p1 ) dx
x 0

where 4p = p0 p1 Fist term = value of buying the old x 0 units at lower price p1 Second term = value of buying additional units at price p1

2.4

More on consumer surplus

Lets extend the notion of consumer surplus to more general situations Let = complete description of consumers problem x () = optimal choice at situation Now, CS () = U (x ()) Uno trade Example: U (x, m) = A x + m : price = p, mandatory minimum purchase = 1 unit Now, we have x ( ) = and CS () = ( 1
A2 p2

if p A if p < A

2A p if p A A2 if p < A p

Important lesson from example: There are situations in which CS can be negative. They often involve situations in which consumers are forced to make purchases that they would not make volutarily. 6

3
3.1

Recovering preferences from data

3.2

How to compute CS using only behavioral data

Given observations of several price/quantity pairs, can estimate the function x (p) using statistics Then, construct the inverse demand function p (x) = x1 (p) But we know that at an interior allocation, p (x) = B 0 (x), as long as the consumer makes decisions by maximizing her experienced utility function It follows that
x ( p )

CS (p) =

(p (x) p) dx

This gives CS as a function of x (p), p (), and p, all of which are observable Can use this to estimate the change in consumer surplus that would follow from an unobserved change in prices, the introduction of a new tax, etc.

3.3

Recovering the utility function from observed behavior

Suppose x () observed. We recover U using the following steps: Assume U is quasi-linear, i.e. U (x, m) = B (x) + m Then p (x) = B 0 (x) x So B (x) = B (0) + 0 p (u)du 7

 U (x, m) =

x
0

p (u)du + m + constant

So we can recover U (x.m) up to a constant, which is equal to B (0). Constant typically unimportant in most applications. What if U is not quasilinear? OK if U quasi-linear, even if not exactly so Method generalizes (taught in advanced courses, requires substantially more math)

3.4

Example
10 p

Consider an example of how to recover U (x, m) from x (p) Suppose we observe that x (p) = Then p =
10 x+1

So B (x) = 10 ln(x + 1), and U (x, m) = 10 ln(x + 1) + m + constant

3.5

Example
9 p

0 = initial situation at which consumer buys freely at price p At initial situation observe x (p) = 1. 1 = new situation at which individuals buy freely, but also get 2 free units (regardless of how much they buy) Question: Predict whatt is CS (1 )? Solve in three steps. Step 1: Recover U (x, m) from the initial observed demand. Get U (x, m) = 9 + ln(x + 1) + m (similar to previous example) Step 2: Predict x 1 (p) by maximizing the recovered prefences 8

 Important: x 1 (p) denotes the amount bought, not the amoung consumed which is equal to x 1 (p) + 2. Get x 1 ( p) = Step 3: Predict CS (1 ) Get CS (1 ) = ( 9ln3 if p 3 9 9ln p + 9 3p if p < 3 ( 0
9 p

if p 3 3 if p < 3

3.6

Application: Valuing new products

Hypothetical example: compute the value of introducing household robots for a typical US consumer Data: 2020 2025 p 9000 1000 x 1 9

Graph the data, and t a line through them (trivial in this example, since only two points) Get x (p) = max{0, 10
p } 1000

Value of introducing robots = CS (p) CS (p = ) = CS (p) . Using graphical method: - CS2020 = 1 1 1000 = $500 2 - CS2025 = 1 9 9000 = $40, 500 2 Lesson: Value of a new product depends strongly on the price at which its sold.

4
4.1

Consumer mistakes
Decision mistakes
A simple model of decision mistakes: 1. Decision Utility vs. Experienced Utility - Experienced utility: describes well-being/hedonics - Decision utility: describes objective that is maximized at decision time 2. x () = maxx U DU (x) s.t. feasibility constraints in xopt () = maxx U EU (x) s.t. feasibility constraints in 3. Rational behavior: U DU = U EU = x = xopt 4. Mistakes: U DU 6= U EU = x 6= xopt Example: U EU = 2A x + m U DU = 4A x + m
A So xopt (p) = A , x (p) = 4p 2 . Note the gap between the consumer p2 choice and the optimal choice.
2 2

Remarks: 1. B DU (x) 6= B EU (x) + constant = mistakes arise 2. If mistakes, then: Utility function recovered from behavior 6= U EU CS calculted with the recovered utility function is incorrect 3. Thus, critical to know if mistakes likely in a given context Example (continued): The CS estimated under the assumption that the consumer is 2 rational is CS est (p) = 4A . p But the true CS, given the consumers true EU function, is CS true (p) = 0. 10

4.2

Application: Addiction

Here is a simple model to think about the consumption of addictive substances: U EU (x, m) = B (x) + m U DU (x, m) = B (x) + m, 1 Can have CS true (p) < 0, in which case prohibition can be welfare improving!

Final remarks
Key ideas: 1. Modeling behavior: x () given by maxx0 U DU (x) s.t. constraints in 2. Measuring well-being: Consumer surplus, CS () = U EU (x ()) EU Uno trade 3. Given rationality, can recover B (x) from x (p) and measure CS from p (x) Tips on problem solving: Must specicy maximization problem correctly Consumer behavior (x ()) follows from maximizing U DU given the constraints in situation Optimal level of consumption (xopt ()) follows from maximizing U EU given the constraints in situation Dont mix up demand x (p) and inverse demand p (x) Careful when computing CS using the notion of area under the curve

11