1 MGT 511: Hypothesis Testing and Regression Lecture 9: Applications K. Sudhir Yale SOM-EMBA.

1

MGT 511: Hypothesis Testing and RegressionLecture 9: Applications

K. SudhirYale SOM-EMBA

2

What we did… Hypothesis Testing

When you can’t do a census, you have to sample to get best estimates of the population. Systematic Approach to figure out if sample result is due to just chance occurrence or really true.

Regression Analysis To quantify relationships between variables.

How much does sales change with prices, advertising and sales force effort ?

How do long term interest rates (10 year treasury bonds) relate to short-term interest rates (overnight federal funds and 3 month treasury bills)?

Hypothesis Testing used to see if these relationships truly exist.

3

Hypothesis Testing: Key Issues

One Sample Means and Proportions Two Sample Difference – Paired Samples Two Sample Difference – Independent Samples

4

Hypothesis Generation and Testing

Freakonomics Abortion Debate Teacher Cheating

RBC--Mortgage Promotion

Capital One

5

Regression Applications

Prediction Forecast of Sales Forecasting who are likely prospects for catalog purchases Forecasts of who are likely to default on a bank loan Forecasts of who are likely to succeed in an MBA program

Benchmarking How much should a firm donate to charity? What should a newspaper charge for advertisements? How much should a realtor recommend as the selling price for a home? What should be a reasonable salary for an employee with certain qualifications

and job requirements Deciding on the quotas/bonuses for a salesperson What should be your insurance rates?

6

Regression Applications

Describe the relationships between variables Relative Volatility of a stock (beta) Relationship between EPS and Stock Prices Relationship between rate of return and maturity period for

short-term bond fund Using the estimated equations to help make decisions

What should be the optimal price and advertising given demand equation?

How should I price my product given the experience curve?

7

The Impact of AutobyTel on Car Pricing(Fiona Scott Morton)

Sample Research Questions How much do consumers who use AutobyTel gain in terms of

prices? Does the presence of an AutobyTel Franchise reduce prices

for consumers? Do “cowboys” or “cowards” save more by using AutobyTel?

8

Bunch of other control variables were used

Car Related Variables Car Model Dummies Month of Sale Dummies Region Dummies Model Year (To account for recency) Whether consumer traded-in a vehicle Dealer’s cost of car

Demographic Variables of individual Income, Education, Gender

Regression Equation

0 1 2ln(Price) AutobyTelUsed AutobyTelPresent+...

9

Main Result

Partial Regression Equation Ln(Price)= Intercept

-0.98*AutobyTelUsed -0.49*AutobyTelPresent+….

How much does a consumer who used AutobyTel gain? How much does the presence of AutobyTel in an area

help in terms of prices to all consumers

10

Some Additional Results involving more Complicated Analysis

Actually, here we only get the average relative gain for all AutobyTel consumers Would the gain be greater for people who do not like to

bargain? (“Cowards”) Would the gain be greater for people who like to haggle?

(“Cowboys”) Additional Analysis showed that it is the “cowards” who gain

more. Their gain is actually close to 2%. In additional analysis they show that African Americans and

Hispanics pay 2% more for cars, but with the use of Internet they pay roughly the same prices as Whites and Asians.

11

12

The “reality” check

Welcome Canine User 39… Mutt, mostly black lab, enjoys pepperoni, fetching, and sniffing other dogs’

heinies… Updating profile

13

Why do manufacturers pay slotting allowances to retailers? (Sudhir)

Slotting Allowances Lump-sum advance payments made by manufacturers to

retailers to stock new products A third of new product marketing budgets (about $ 9 billion)

Slotting allowances are controversial Subject to Congressional Investigation in 1999

Small business owners testified in masks for fear of retaliation with electronically altered voices

Bureau of Alcohol, Tobacco and Firearms banned slotting allowance

Federal Trade Commission allows slotting allowances

14

Sample Research Questions

Is it to compensate retailers for the opportunity cost of shelf space?

Do manufacturers pay this voluntarily or retailers force them to pay slotting allowances?

Are manufacturers using this to signal their product’s likelihood of success?

Slotting Allowances and Retail Competition Do slotting allowances decline when more retailers have accepted the

product? Do slotting allowances increase when more retailers have accepted

the product in order to reduce retail competition?

15

Results

Dependent Variable: Probability(Slotting Allowances)

Variable Coefficient p-value

Intercept -8.083 0.000 Opportunity Cost of Shelf Space Space 0.082 0.113 Competing Private Label 0.106 0.087 Who gives slotting allowances Manufacturer Reputation 1.957 0.000 Manufacturer Reputation2 -0.184 0.000 Effect of Information Test Market 0.814 0.002 Market Research 0.287 0.100 Retail Competition Competing Stores 0.281 0.000

Competing Store2 -0.017 0.000

16

Rating and Likelihood of Slotting

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

1 2 3 4 5 6 7 8 9 10

Rating

P(S

lott

ing

Allo

wan

ces)

17

Differences between Large and Small Manufacturers of the Effect of Competing Stores

0.000

0.100

0.200

0.300

0.400

0.500

0.600

1 2 3 4 5 6 7 8 9 10

Rating

Large Large*CS

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

1 2 3 4 5 6 7 8 9 10

RatingP

(Slo

ttin

g A

llow

ance

s)

Small Small*CS

18

Key Findings

Slotting allowances are offered More when the opportunity costs of shelf space is higher most by manufacturers with medium reputations; this is

where there is the greatest uncertainty more when manufacturers are sure about potential success

of product (they are signaling to retailers) More when more retailers have accepted the product… (Very

surprising to me) However differences between large and small retailers.

19

How does Word of Mouth affect TV Show Ratings? (Dina Mayzlin)

Word of Mouth is hard to measure Internet Chat Rooms provide records of a subset of

such conversations Can we develop measures of “buzz” to serve as a

leading indicator of TV Show Ratings?

20

Carefully controlled for all other types of effects such as show timings, network effects…

Regression Equation

0 1 t-1 2 t-1 3 t-1Rating Posts Dispersion ...tRating

21

Key Findings

Past Ratings are quite significant Surprise:

Past # of Posts are not significant Dispersion of Posts are very significant and have an

important impact on ratings Important Finding

Because people usually think of WOM as volume of Buzz Dina shows this is the wrong variable to focus on!!!

22

Conclusion

Regression can be used for a wide range of managerially useful applications Great at forecasting, benchmarking, description of

relationships and helps in managerial decision making Usefulness and Relevance of Course

Ready access to data with the availability of computers, the use of quantitative methods in decision making keeps rising

As a manager and decision maker, comfort with regression should help you appreciate quantitative results provided to you and ask the “right” questions

23

THANK YOU