Transcript Data Analysis for Competition Policy: an Introduction to Econometrics

Data Analysis for
Competition Policy: an
Introduction to
Econometrics
Katie Curry
Senior Economist, OFT
1
Outline
● What is econometrics?
● Why is it useful?
● Example
- Defining the relevant market in a recent merger case
2
What is econometrics?
● Statistical modelling of economic
relationships
- e.g. how consumers respond to price changes
● A commonly used technique is called
regression analysis
● Regression analysis uses actual data to
assess how one variable of interest (e.g.
quantity consumed) is related to possible
explanatory factors (e.g. price)
3
What is econometrics? ctd.
● Allows for the fact we don’t have data on all
possible influences by including an error term
● Use data to estimate equation of the form:
Y = a + bX + e
Variable trying
to explain.
Error term.
Average value
of Y.
X is the explanatory
factor. b measures the
average increase in Y
caused by an
increase in X.
4
What is econometrics? ctd.
Y
a
b
X
5
Why use econometrics?
● It allows us to analyse a large amount of information
in a systematic way
-
-
Suppose we have data on the price of ice cream and the quantity
of ice cream eaten and we are interested in how consumer
demand responds to changes in price
We could look at what happened to ice cream sales every time
the price changed
But if there are lots of observations this could take a long timeand if we only look at one or two instances our conclusions could
be unrepresentative
Regression analysis can handle millions of observations easily
6
Why use econometrics? ctd.
● Can give more meaningful results than simple
statistics
- We could look at the average price of ice cream and
the average quantity eaten in a large dataset very
easily
- Knowing that a 5% price increase leads to a 1%
reduction in demand could be much more helpful than
simply knowing the average level of each
7
Why use econometrics? ctd.
● Controls for effect of more than one variable
- We could use our data to make a chart, showing how
quantity is related to price
- However there may be many other factors affecting
quantity demanded that obscure this relationship (e.g.
more ice cream in the summer)
- Econometrics can control for these other factors,
allowing us to focus on the relationship we are
interested in
8
Example of econometrics in practice
9
Example of econometrics in practice
● Used to define relevant market in a merger
● Merging parties overlapped in manufacture
and supply of specific types of soft cheese
(Brie, Camembert and goat’s cheese)
● Parties submitted that British cheese was in a
separate market from French cheese so there
was no overlap
10
Econometrics results used for market
definition
● Used an econometric model to estimate how demand
for one cheese changes when its own price and the
price of a potential substitute changes
● The elasticity estimates obtained in this way were
used to conduct a critical loss analysis
● Critical loss analysis uses data on profit margins and
elasticities to calculate whether a 5% price increase
would be profitable
● Starts with narrowest market- if not profitable, then
move to broader market
11
Almost Ideal Demand System
● Useful way of estimating demand when there
are many brands that can be grouped together
in segments
● Assumes customers decide how much to
spend on one segment (e.g. camembert) and
then decide which brand to buy
● Demand for each brand can be estimated as a
function of the prices of all of the brands in
the same segment
12
The ideal model
SOFT CHEESE
Other soft
cheeses
Brie
Camembert
Somerset
Somerset
Goat’s cheese
French
branded
French
private
label
French
branded
French
private
label
Mozzarella
Etc.
13
What we actually did
SOFT CHEESE
Brie
(All types)
Camembert
Somerset
French
branded
French
private
label
14
Illustrative results
Following a 1% change in the price of:
% change in
demand for:
Somerset
Camembert
French
private-label
Camembert
French
branded
Camembert
Brie
Somerset
Camembert
French private-label
Camembert
French branded
Camembert
Brie
15
Questions?
16