Professional Documents
Culture Documents
DemandMang 411A
DemandMang 411A
DemandMang 411A
Overview
1. Demand curves 2. The multiplicative/log-linear demand model 3. Price elasticities 4. Demand for groups of items and cross price elasticities 5. Estimating the multiplicative/log-linear demand model 6. Examples of Base Pricing Analysis 7. Appendix: Logarithms and the exponential function
Demand curves
to various elements of the marketing mix and other variables inuencing consumer behavior
Q
sales units
Both own and the competitors marketing actions will typically affect
demand
Where do demand curves come from? Example: Survey data obtained from full-time MBA students in April 2010
How much are you willing to pay for one ticket for the Soccer World Cup nal 2010 (in U.S. dollars)?
3000
2500
2000
Price
1500
1000
500
10
20
30
40
50
60
70
80
90
100
110
Order willingness to pay in descending order and plot against the rank (highest
price rst, ) 10 students indicated a willingness to pay of $1,000 or higher
5
A demand curve shows the distribution of consumers willingness to pay How does this relate to what they said in micro?
The basic microeconomics model emphasizes a model in which consumers
choose a quantity demanded that is continuous (e.g. .5 units!!!) and they adjust demand downward as price increases. This is a representative consumer model. Not applicable to most products in the marketplace. The more general model emphasizes differences heterogeneity in consumers willingness to pay as a source of downward sloping demand
Modeling demand
Goal:
Infer the relationship between unit sales and prices, promotions, and other
variables from the data using regression analysis
This formulation should be exible and have a good chance of tting the demand
relationship present in common data sets
Formula:
Q = a bP
8 7
1.4
6 5 4 3 2
Example:
Q = 40 5 P
1.4
1.4
1 0 0 5 10 15 20 25 30 35 40
Formula:
Q=AP
8 7
1.4
6 5 4 3 2
Q = 100 P 2
1.4
Example:
Effect of a given price change is larger at low than at high price points
1 0
10
15
20
25
30
35
40
log log(x y) = log(x) + log(y) (L1) (L3) log(xa ) = a log(x) Dene = log(A)
We will therefore also refer to the multiplicative demand model as the Purpose of this transformation
The parameters and now enter the model linearly model has the form of a linear regression
10
Effect of parameters
6
5 4 3 2
A = 100 A = 150
10
20
30
40
Q
2.5
Parameter : slope increase in makes the demand curve steeper, i.e. more responsive to price
=2
log(P)
1 0 -1 -2 -4 0 4 8 12
=4
log(Q) 11
Price elasticities
%Q price elasticity = = %P
Q1 Q0 Q0 P1 P0 P0
Interpretation
Example: Price elasticity is -2.8 A 1% increase in price is associated with a 2.8% decrease in unit sales
Note: The price elasticity can be calculated for any demand curve, not
just the multiplicative demand model
12
Substitutability
Availability of substitutes Actual or perceived quality differences Awareness of substitutes Cost of nding (or switching to) substitutes
Private label orange juice Tropicana/Minute Maid orange juice Orange juice All juices All drinks
13
14
15
The property that one parameter directly measures the price elasticity is
very special and not true for other demand models
16
Use the log-linear representation of the demand model Change price from P0 to P1 sales will change from Q0 to Q1
log(Q1 ) = log(P1 ) log(Q0 ) = log(P0 )
take difference between equations
log(Q) = log(P )
log(Q1 ) log(Q0 ) = [ log(P1 )] [ log(P0 )] log(Q1 ) log(Q0 ) = [log(Q1 ) log(Q0 )] log(Q) = log(P ) = log(Q) %Q log(P ) %P
use logarithm property L*
17
= price elasticity
The multiplicative demand model discussed so far is too simple it does We now generalize the multiplicative demand model
Allow for the effect of competing product prices Predict demand for multiple products or groups of products
Qi = Ai P1 i1 P2 i2 PKiK
not account for the effect of the prices of competing products on demand
18
= log(Ai ) ):
19
Why are cross price elasticities typically positive? Do you expect that cross-price elasticities are typically symmetric, =
e.g.
12 21 ?
20
Competitive price change will have a large impact on the demand for our product Strong competitive effect
21
In the case of two products we have two regression equation, one for each
product log(Qi ) is the dependent variable, yi log(P1 ) and log(P2 ) are the independent variables, x1 and x2
General approach
elasticities)
Obtain data on sales units and prices Estimate the regression equations to obtain the specic parameters (price
22
.4
log(P) .6
.8
9.5
10 log(Q)
10.5
11
11.5
23
42 studies from the marketing and economics literatures 367 price elasticity estimates Most studies are based on time series data, most use brands Average
Detergents
Durable goods
Food
Toiletries
Drugs
-1.76 -2.77
-2.03
-1.65 -1.38 -1.12
Elasticities
24
Usually ts the data better than the linear demand model Easy to generalize for groups of products
Own price and cross price elasticities measure: The log-linear demand model can be estimated as a regression model
25
Base-Pricing Analysis
1. Scanner data 2. Base pricing analysis 3. Examples
26
Scanner data
Timeline
code) level - Honey Nut Cheerios 25.25 oz size Prices and promotions Aggregation levels - Market (Raleigh-Durham) - Chain/account (Kroger) - Store Time - Weekly, monthly, ...
27
IRI
InfoScan 34,000+ stores (store census, not sample) Retail management system (ScanTrack) Scanner (retail audit) data available worldwide Wal-Mart had 14% U.S. grocery market share in 2004
Nielsen
Problem: Wal-Mart no longer shares their POS data (since 2001) Retailer Data Warehouses -- a new source of power in the channel
28
Base price = non-promoted price (everyday shelf price) Purpose of base pricing analysis
Understand competitive inuence of prices on sales Adjust / ne-tune base prices
29
Company sells a national brand Product line: 16 oz, 24 oz, and 32 oz bottle
size (32 oz size has recently been added)
30
Demand analysis
Cannibalization?
Price simulations
31
Note
16 oz
Note: IRI slides p. 11
24 oz
32 oz
standard errors of estimates Always question marketing consultants about statistical precision of results
32
16 oz
24 oz
32 oz
0.25 NA
Effect on:
16 oz
24 oz
32 oz
PL 24 oz
1
PL 32 oz
0.75
0.5
Effect on:
16 oz
24 oz
32 oz
Note: IRI slides p. 13
Goal: Predict prot for each package size k in the product line
protk = Qk [Pk (1 retail margin) V Ck ]
P . . . retail shelf price
V C . . . variable cost Total prot from product line (if we drop the 24 oz pack size):
5,672 5,771 5,866 5,957 6,046 6,131 6,213 6,292 6,369 6,443 6,515 5,756 5,856 5,952 6,045 6,135 6,221 6,304 6,385 6,462 6,538 6,610 5,838 5,939 6,037 6,131 6,222 6,310 6,394 6,476 6,555 6,631 6,705 5,918 6,021 6,121 6,216 6,308 6,397 6,482 6,565 6,645 6,722 6,797 5,998 6,102 6,203 6,299 6,393 6,482 6,569 6,653 6,734 6,812 6,888 6,076 6,181 6,283 6,381 6,476 6,567 6,655 6,740 6,822 6,901 6,978 6,152 6,260 6,363 6,462 6,558 6,650 6,739 6,825 6,908 6,988 7,066 6,228 6,336 6,441 6,541 6,638 6,732 6,822 6,909 6,993 7,074 7,153 6,302 6,412 6,518 6,619 6,718 6,812 6,903 6,992 7,077 7,159 7,239 6,376 6,487 6,594 6,697 6,796 6,892 6,984 7,073 7,160 7,243 7,323 6,448 6,561 6,668 6,773 6,873 6,970 7,064 7,154 7,241 7,325 7,407
Prots in $1,000
36
Prots highest if both prices are increased by 10% Why not increase prices even further?
37
IRIs client had very limited access to sales and price data and only a
limited understanding of the key pricing issues
Private label competition poses only a very limited threat to the brand There is cannibalization within the product line, but the main offender is not the
new 32 oz size but mainly the 24 oz size - Consider eliminating 24 oz size to save on costs (packaging, distribution, ) Base price points are sub-optimally low
38
Goal:
Examine cannibalization within the Tide product line Examine competitive threat from Wisk Make pricing recommendations Approach
Estimate log-linear demand model using scanner data
39
Store id number Week ACV (all commodity volume), in $1,000 = 1 if any product in the category was on promotion Tide 128 oz: unit sales Tide 128 oz: price ($) Tide 64 oz: unit sales Tide 64 oz: price ($) Wisk 64 oz: unit sales Wisk 64 oz: price ($)
40
Calculate:
41
Straightforward approach: Use log(Q) as dependent variable However: Stores differ in size and hence sales will differ across stores even if the
product prices are the same
Use ACV (all commodity volume) is a measure of store size, dened as the store
revenue from all products sold in $ million per year - Includes the sales of all products (produce, meat, detergents, milk, batteries, etc.), not just the products in the demand model Add log(ACV) as an independent variable in a multiple regression
42
Focus is on the effect of the everyday price, not on the effect of price
promotions Price promotions are typically associated with sales spikes - More details later
43
Create store xed effects and add them to the regression model
45
Own price elasticity looks much more reasonable Strong competitive effect from Wisk Note
Pricing simulations
Goal: Predict the impact on total (product line) prots if we change the
current base prices of one or more of our products model parameters (elasticities) in the product line
We make these predictions using the estimates of the log-linear demand To predict total prots we need to predict the prot for each product k
protk = Qk [Pk (1 retail margin) V Ck ]
V Ck variable cost
47
Pk is the price of one of the products before the change, Pk is the price after the change
Correspondingly Qk is demand before the price change, Q is demand k after the price change
48
Change the price of product 1 (Tide 128 oz) by 1 percentage points and
the price of product 2 (Tide 64 oz) by 2 percentage points. The price of product 3, Wisk 64 oz, is unchanged:
P1 = (1 + 1 ) P1 P2 = (1 + 2 ) P2 P3 = P3
49
Focus on the demand for product 1 (Tide 128 oz) Write down the log-linear demand equation after and before the price
change and take the difference:
log(Q ) = 1 + 11 log(P1 ) + 12 log(P2 ) + 13 log(P3 ) 1
5. The ratio of unit sales of product i after versus before the price changes is Q i = (1 + 1 )i1 (1 + K )iK Qi
51
Cut the base price of Tide 128 oz by 7% and increase the price of Tide 64
oz by 4%
Predicted Tide 128 oz sales ratio after vs before the price changes:
Q = (1 0.07)2.414 (1 + 0.04)0.1591 = 1.20 Q
Average Tide 128 oz unit sales are 55.25 at the week/store level
Implies a current annual base volume of 86 52 55.25 =
chain (Dominicks) level (86 stores in the chain)
Base volume after price changes: 1.20 247,078 = 296,494 Average Tide 128 oz price is $8.47 $7.88 after the 7% price cut Tide 128 oz prot after the price changes:
prot(Tide 128 oz) = 296, 494 [7.88 (1 0.25) 128 0.027]
unit sales price retail margin variable cost of 128 oz size
Note: Retail margin at Dominicks = 25% and the per-ounce cost of Tide is 2.7c
53
Understand how the prices of the products in the category inuence demand for
the products that we sell Do the prices of competing products inuence our own sales competition? Do the prices of other products in our product line inuence our own sales cannibalization?
Use data on past prices and sales Use regression analysis to estimate a demand model Can we improve pricing? How should we adjust the prices in our product line?
54