Joris Pinkse and Margaret Slade, "Semi-Structural Models of Advertising Competition", Journal of Applied Econometrics, Vol. 22, No. 7, 2007, pp. 1227-1246. Description of IRI Crackers Data Most of the data used in this article are the property of Information Resources Inc. (IRI), a marketing firm in Chicago. See http://us.infores.com/page/about/contact\_us They are available for a fee. Since users must sign a form agreeing not to transmit the data to a third party, the data cannot be housed here. Inquiries should be addressed to IRI. The IRI data consist of four files. The first contains a dictionary of alphabetic descriptions and brand volumes for the UPC codes in the category, the second contains demographic variables for each household, the third has household purchase records, and the fourth contains store data. The first data set was used to create the weights that we employed to create average rival prices. In other words, those weights are proportional to the market shares of each brand. We do not use the demographic data here. In addition, we use the purchase data only to attribute household purchases and coupon redemptions to brands, stores, and weeks. In particular, every time a household purchased a box of saltine crackers, we augmented the sales of that brand in that store and that week by one unit. When this had been done, we multiplied the sales variable by 20 to account for the fact that we observe only about 5% of the purchases. The variable that is constructed in this way is denoted Q. We generated the number of coupons redeemed (COUP) by brand, store, and week in a similar manner. The store data are already organized by brand, store, and week. We obtained the price of a two-pound box of saltine crackers (PRICE), the advertising dummy (AD), and the display dummy (DISP) from that data set. There are 4160 = 4 brands times 10 stores times 104 weeks in this data. The cost data are publicly available and may be found in the file cost-data.txt, which is zipped in the file cost-data.zip. This file is an ASCII file in DOS format. Unix users should use "unzip -a". The cost data, which vary only by week, consist of six variables: the real price of intermediate foods and feeds (PIF), the real wage (WAGE), the real cost of advertising in newspapers (CNAD), the real wholesale price of saltine crackers (PWHOLE), the real income per capita (EARN), and the US consumer--price index, all items (CPI). There are 104 observations in this file, with each line containing the six variables for a given week. All other variables (e.g., the number of rival brands on display) were constructed from the raw data as described in the paper.