Empirical IO

7/25/2019 Empirical IO

1/33

ECONG047: Empirical Industrial OrganisationLectures 1: Determinants of Market Outcomes

Lars Nesheim

University College London

January 2015

Nesheim (UCL) ECONG047: Lecture 1 2015 1 / 33
http://find/


2/33

Introduction

What is Empirical Industrial Organization?

Aim to develop an empirical understanding of how firms operate and

how markets operate

1 Firms

2 Consumers

3 Market structure

http://find/http://goback/


3/33

What is a firm

Technology, costs and contracts

Technology determines cost structure of firm and to some degree whatportfolio of products it will produce.Cost functions are crucial for determining optimal operating size,optimal portfolio of products, and the number of firms in a market.They also influence pricing through marginal cost. Key components ofthe cost function are the magnitude and nature of fixed costs (are the

fixed costs sunk or recoverable), the nature and magnitude of marginalcosts (are they constant or not), and the nature of economies of scale(are there economies of scale or scope and what is the minimumefficient scale. Production costs are crucial determinants ofcompetition conditional on entry and costs of entry are crucial for

determining the number of firms that enter the market.Contracting costs can affect the size of the firm, the extent of verticaland horizontal integration.

The key firm outcomes empirical IO is concerned with are: firmsdemand for inputs, prices, output quantities, revenues, profits,

externalities.Nesheim (UCL) ECONG047: Lecture 1 2015 3 / 33
http://find/


4/33

Demand

What is demand structure?

Who are the consumers?What is the nature of the product?

http://find/


5/33

Market structure

Define market

Which products? Apples or fruit?Location? Euston station, London, or UK?Time period/duration? Tuesday morning, First week of jan, month,

year?Number and type of competitors. Competitors might be distinguishedby size, quality, cost, etc.

Nature of competition

Compete in prices, quantities, quality, or in some other dimensionStatic vs. dynamic competition. Competition related to who enters orafter entry about prices or other outcomes?

http://find/


6/33

Market structure: other factors

Other factors are also important in shaping market structure. However,this course will not discuss them in much detail. Examples include

Role of regulation. In some markets like railroads, utilities,communications, airlines, regulation plays an important role

influencing market outcomes.Role of information. What firms know about their customers andtheir competitors plays an important role in influencing marketoutcomes. Does firm 1 know the the marginal cost of firm 2? Doesfirm 1 know the quality of the product of firm 2? Does firm 1 knowonly the aggregate demand elasticity or does it have informationabout the demand elasticities for different groups in the population.

http://find/


7/33

Outcomes: what measureable quantities is empirical IOinterested in?

1 Prices and quantities: demand

2 Consumer welfare3 Revenues and profits and costs

4 Types of products

5 Entry and exit

http://find/


8/33

Demand

1 Measurement

1 prices2 quantities

3 characteristics2 Demand functions

1 Examples: linear, log-linear, discrete choice2 Individual vs market3

Market demand function is aggregate, sum over all customers

http://find/


9/33

Linear demand

One common assumption is the demand is linear with demand qrelated toprice pby the linear relation

q = a bp (1)

p = a

b

1

bq

where a > 0, b> 0 and p ab

. See the figure on the next page.

http://find/


10/33

Linear inverse demand

quantity

price

http://find/


11/33

Linear demand: implications (1)

1 Impact of price (slope) is same regardless of whether price is low orhigh.

1 Do we expect the sensitivity of demand for petrol to be the same whenthe price is 1 per litre as when the price is 10 per litre?

2 No non-linear pricing. For example, there are no quantity discounts.3 In many cases, when goods are homogenous, the linear demand

model is a good approximation for small changes in p. Suppose thetrue demand relationship is

q=f (p) .

Then, if the demand function is differentiable, Taylors theoremimplies that the approximation

q=f (p0)+f (p0)

p (p p0)

is valid when pis close to p0

.



12/33

Linear demand: implications (2)

1 Model (1) ignores product differentiation. In reality, product

diff

erentiation is important. Characteristics such as quality, time ofday, pack size, day of week, season, etc. In practice, empirical analysisof a model such as (1) relies on data that has been aggregated acrosssimilar but differentiated products. For example, suppose totalquantity of coffee is the sum ofqH, demand for high quality coffee,

and qL, demand for low quality coffee,

q=qL+qH

and the price is

p=

qL

qL+qH

pL+

qH

qL+qH

pH.

Does analysis of demand for qand pprovide an accurate picture ofdemand for coffee?



13/33

Log-linear demand

An alternative demand model that is commonly employed is the log-lineardemand model in which

ln q = a bln p

q = ea

pb

.

For this model, the inverse demand relationship is

ln p= a

b

1

bln q.

See the figure on the next page.



14/33

Log-linear inverse demand

quantity

price



15/33

Log-linear demand: implications

1 Elasticity of demand with respect to price is constant

= p

q

q

p= b.

2 Goods are homogenous (as in linear demand case) and pricing islinear (that is there are no quantity discounts).

3 Both price and quantity must be strictly positive.


( )


16/33

Discrete choice demand (1)

In many case, a model that assumes homogenous goods is not adequatebecause product differentiation is important. In these cases, it is importantto understand not only how much people buy but which products they

buy. Suppose there are J+1 distinct products (or options) in a marketand index each product by j=0,1,..., J. Let option j=0 be the optionnot to buy.any product and let each option j> 0 be an option to buy onespecific product.


Di h i d d (2)


17/33


For example, in the market for travel to Paris, a consumer can choose notto travel at all, or can choose to fly, to take the train, or to drive and take

the ferry. Or, in the market for flights from London to New York, supposethere are 6 options, the option not to travel plus 5 distinct airline optionsthat can be chosen. Assume each consumer can choose only one.


Di h i d d (3)


18/33


A discrete choice model can be used to model this choice setting. For eachconsumer isuppose utility from purchasing product j is

uij=xj+pj+ ij

where pj is the price of product j, xjis a vector of (observable) productcharacteristics (for example, brand, quality, departure time, etc.), and ij isan unobserved demand shock. In discrete choice demand estimation, weassume that we have data on (xj, pj) for all j(we assume that

p0=x0=0) and that we assume which product was purchased by eachconsumer.



19/33

L it i d d


20/33

Logit inverse demand

quantity

price


L it d d i li ti


21/33

Logit demand: implications

1 Each consumer chooses only one option and conditional on choosing jonly buys one unit

1 Model can be extended to account for simultaneous purchase of 2 ormore options.

2 Model can be extended to account for purchase of more than one unitof quantity

2

Substitution patterns1 Simple model implies very restrictive substitution patterns across goods2 Alternative assumptions about unobserved shocksijare less restrictive

3 Number of goods

1 Do we always know what the rightJ is?2 Do we always know what the correct set of characteristicsxto consider

are?

4 Functional form

1 Utility function is linear in price and demand curve is convex. This willhave important implications for analysis of mergers.


Utility functions (1)


22/33


Demand function is derived from utility maximisation of consumer.Assume that the consumer solves

max{q} (u(q, x, ) subject to i piqi y)

where x is a vector of observable household demographics and is avector of parameters. This model implies that demand has the form

q=d(p, y, x, ) .

That is, demand is a function of prices, income, x and .




23/33


For example, suppose the consumer problem is

maxq

{xln q1+(1 x) ln q2 subject to p1q1+p2q2 y}

where 0 < x< 1. Then, the demand functions are

q1 = xy

p1

q2 = (1 x) y

p2.




24/33


This simple example imposes many important restrictions including:

1 0 < x< 1

2 Demand is linear in income

3 Expenditure share of each good i is a constant

s1 = p1q1y =x

s2 = p2q2

y =1 x

independent of price and income.4 Demand for good i is independent of the price of good j6=i.

5 Price elasticity is =1.


Aggregate demand


25/33

Aggregate demand

Aggregate demand is

Q= i

di(p, yi, xi, i)wi

where wi is a weight.

Is aggregate demand linear? No. But, as an approximation?Aggregate demand approximation

Q = d(p0, y0)+d

p(p p0)

+d

y (y y0)


Indirect utility


26/33

Indirect utility

In many cases, we are interested in the indirect utility function v(p, y).

This function describes the maximum utility attainable given prices pandincome y. The indirect utility function is defined by

v(p, y)=maxq

{u(q) subject to p q y} .

Equivalently, ifq =d(p, y)

is the optimal level of demand, then

v(p, y) = u(q)

= u(d(p, y)) .

It often proves useful in demand analysis. For example, it is used tocompute consumer welfare.



27/33

Cross-price elasticity


28/33

Cross price elasticity

1 The elasticity of demand for product iwith respect to the price ofproduct j is

ij= pj

qi

qi

pj.

2 Examples

1 When ij < 0, then i and jare often called substitutes.2 When ij > 0, they are often called complements.


Dynamics (1)


29/33

Dynamics (1)

When goods are durable or utility is influenced by some form of habitformation (marginal utility of consumption in period t+1 is affected by

quantity consumed in period t). Then the dynamics of consumption anddemand can be very important. For example, let q1 be consumption inperiod 1 and q2 be consumption in period 2, let (p1, p2) be thecorresponding prices, and let (y1, y2) be income in periods 1 and 2.Suppose consumers can borrow or lend and the interest rate is zero.Suppose consumers maximise utility of the form

u(q1, q2)= ln q1+ ln (q1+q2)

subject to the budget constraint

p1q1+p2q2=y1+y2.

The quantity purchased and consumed in period 1 impacts utility in period2 through the term q2.


Dynamics (2)


30/33

Dynamics (2)

Examples in which this type of durable goods or habits model mightapply include:

1 demand for soap powder

2 automobiles

3 houses

4 alcohol, cigarettes

In these cases, if demand is estimated while ignoring the dynamics, then

estimates of demand elasticities may be biased. See assignment 1 from2011.


Consumer welfare: consumer surplus


31/33

Consumer welfare: consumer surplus

One measure of consumer welfare is consumer surplus. It is defined to be

CS =

q0Z0

(p(x) p0) (2)

dx =

q0Z0

p(x) dx p0q0. (3)

When the marginal utility of income is a constant it measures the

diff

erence between what consumers are willing to pay and what they pay.When the marginal utility of income is not constant, consumer surplus isan approximate measure of welfare.


Consumer welfare: compensating variation


32/33

p g

Compensating variation measures how much money is required tocompensate a consumer for a price change. For example, suppose initial

prices are p0 and the consumer obtains utility u0. That is,u0 = v(y, p0)

= maxq

(u(q) subject to

j

p0jqj y

).

The cost of obtaining the utility level is

c(p0, u0)=minq

(j

p0jqj subject to u(q) u0

)

If prices increase to p1, then the compensating variation isCV=c(p1, u0) c(p0, u0) .

Equivalently, CV satisfies the equation

v(p0, y)=v(p1, y+CV) .


Consumer welfare: equivalent variation


33/33

q

A second exact measure of welfare is the equivalent variation. It is definedby

EV=c(p1, u1) c(p0, u1) .

In contrast, to CV it is measured at the new utility level u1.


Documents

Empirical IO