EMPIRICAL INDUSTRIAL ORGANIZATION

STATIC DEMAND ESTIMATION

For this project you will have the opportunity to apply one of the methods for estimating demand that was discussed in class: the simple logit model in which products are described by their characteristics (Berry (1994)).  The dataset for this assignment comes from a marketing study performed by the University of Chicago.  It includes information about cereal sales from all 83 outlets of Dominick's Finer Foods, a grocery chain in the Chicago area.  The raw data file is called "demand_data_undergrad_IO" and it can be downloaded from OWL.  This dataset contains all of the variables you will need to conduct your analysis.  A description of the variables is contained in the dataset. 

The goal of this assignment is to learn to apply the methods studied in class to real data and to be able to interpret your empirical results and convey this information in written form. Although the assignment looks like a problem set, you should treat this as if you are a consultant and a firm has approached you to perform a demand analysis on cereal.  Your write-up should read like a report, not like itemized responses to questions. Explain the analysis that you perform, the results that you obtain, and how these results should be interpreted.  Your ability to convey what you did in writing will be an important portion of your grade.

This assignment is due in class on March 6th.

1. DATA PREP

a. Market Size: For the simple logit discrete choice demand model (Berry 1994), you will need to account for the outside good.  In order to do this, you will need to know the size of the market.  Included in the dataset is a variable called "potential_consumers." The variable is an estimate of the total potential number of consumers.  This variable is computed as the average number of shoppers per store per week, multiplied by 3 to account for the assumption that each shopper is purchasing for 3 consumers, and multiplied by 3 again to account for the fact that Dominicks is not the only grocery chain in each market, and that some people (in this case two-thirds) shop at these other stores.  Since the quantity data is reported as total number of servings, to compute the market size, you will need to multiply potential_consumers by 7 (to create potential servings).

2. ESTIMATING DEMAND

a. Simple Logit Model

i. Estimate demand for all brands of cereal using the simple logit model.  For characteristics of the brands, use a constant, "fat calories", "sugar", "mushy",  "fiber", and of course the price.(Remember that shares in this model are shares of servings.)  Report coefficient estimates and standard errors.  Which coefficient estimates are statistically significant at the 5% level?

ii. Re-estimate the model using lagged wholesale costs to instrument for the potential endogeneity of prices.  Use lagged wholesale costs as your instrument.  Report coefficient estimates and standard errors.  Explain and interpret any differences in the price coefficient estimates that you find.

iii. Derive the analytical expression for both own-price and cross-price elasticities.  Show your work.

iv. Compute the matrix of own-price and cross-price elasticities based on your instrumental variables estimates.Summarize your results.  Do the results make sense?  Are there any elasticities that look strange?

3. INSTRUMENTS

a. Are the instruments of lagged wholesale costs good instruments in this scenario?  Are they likely to be valid instruments?  Discuss with reference to both of the assumptions underlying instrumental variables techniques.

b. In what direction does the estimate of the price coefficient change from OLS to IV?  Is this what you expected?  If so, explain why.  If not, explain why not, and discuss what could be causing this.

4. COUNTERFACTUAL EXPERIMENT--ACQUISITION OF POST RAISIN BRAN

a. Suppose that Kellogg wants to increase its presence in the raisin bran market and purchases Post's raisin bran production facility as well as the rights to the brand "Post Raisin Bran".  Kellogg realizes that since they now control a large percentage of the raisin bran market, they can increase profits by raising the price of both Kellogg's Raisin Bran and Post Raisin Bran.  Using your elasticity estimates, calculate the change in revenues Kellogg would experience from a 10% increase in the price of both brands of raisin bran.  (Assume for now that the prices of all other brands of cereal do not change in response.) 

b. In response to this increase in prices, it is likely that the prices of other brands will adjust as well.  Qualitatively analyze and explain what you expect will happen to the prices of other brands (both those owned by Kellogg and those not owned by Kellogg).

i. Is this change good for Kellogg?

ii. Is this change good for consumers?

Assignment Files - https://www.dropbox.com/s/6z9vhonc2j6biqr/Assignment%20Files.rar?dl=0