JUNE 2004

• • • • • • • • • • • • • • • • • • • • • • • • • •

Why High Demand Doesn't Always Mean High Price

The Importance of Loss Leaders in Pricing Decisions

Research by Anil Kashyap and Peter E. Rossi

Popular wisdom tells us that it's human nature to get while the getting is good. Popular wisdom tells us this is good financial practice as well. So when it comes to the world of retail, it's no surprise that textbook economic theory says that as demand for a product increases, so too should its price. The surprise is that textbook economic theory is wrong.

In fact, prices actually fall during periods of peak demand, say University of Chicago Graduate School of Business professors Anil Kashyap and Peter E. Rossi. In their recent study, "Why Don't Prices Rise During Periods of Peak Demand? Evidence From Scanner Data," Kashyap and Rossi, along with Judith Chevalier of the Yale School of Management, argue that retailer cuts to profit margins account for the drop in prices and that advertising plays an important role in pricing.

Kashyap, Rossi, and Chevalier's findings show that consumers get deals, for example, on tuna during Lent, beer over the Fourth of July holiday, and snack crackers at Christmas-time when those items are in high demand. Their findings also indicate that because of imperfect competition, as opposed to a textbook model, retailers follow a "loss leader" model of pricing. This means that retailers spend a fixed amount to advertise a select number of popular goods, lowering the prices for those goods regardless of the overall shopping climate.

Cheap Tuna

To find out if pricing is truly counter-cyclical, Kashyap, Rossi, and Chevalier drew on seven and a half years of weekly store-level scanner data from Dominick's Finer Foods. Data collection began in September 1989. Dominick's, the second-largest supermarket chain in the Chicago metropolitan area, has approximately
100 stores and a market share of about 25 percent. The authors chose to look at supermarket data because of the size of the grocery/retailing sector (where sales totaled $435 billion in 1999) and because the consumer-retailer interactions that occur in grocery stores are similar to those in many other types of purchases.

"The patterns we uncover are potentially relevant for other parts of the economy," says Kashyap. "The same model could apply to any industry that has high search costs relative to what people are buying."

The authors then used climatological data from the National Oceanic and Atmospheric Administration and the dates of holidays to choose periods in which demand for particular goods is likely to vary. The resulting shopping periods were "hot," "cold," Lent, Easter, Memorial Day, the Fourth of July, Labor Day, Thanksgiving, the week after Thanksgiving, and Christmas. Next, the authors grouped individual items into aggregates and grouped aggregates into categories. Choosing eleven categories, they decided to look at seasonal prices for beer, "eating" soup, soup used in cooking, oatmeal, tuna, snack crackers, regular crackers, cookies, cheese, analgesics, and dish detergent.

Discovering a "pronounced tendency for retail prices to be lower" at demand peaks for these categories of items, the authors came up with two possible explanations: marginal costs have seasonal patterns or markups move counter-cyclically over the seasonal cycle. They ruled out the first hypothesis by showing that while cooking soups and eating soups have similar marginal costs, their pricing patterns are different. They also found this to be true with regular crackers and snack crackers. Moving forward with their second hypothesis, the authors explored possible reasons for counter-cyclical pricing.

Three Models of Competition

Kashyap, Rossi, and Chevalier considered three models of imperfect competition that could generate counter-cyclical pricing. The first looks at the cyclical elasticity of demand. This theory claims that consumers decide which store to visit based on the total purchase price of everything in their carts, not the prices of individual items. Assuming that the cost of searching for the lowest total price is fixed, consumers are going to search more places when they are buying a large number of goods, i.e. when overall demand is high. Prices for consumer goods, then, should fall during high purchase periods such as Christmas and weekends, when intensive shopping takes place.

The second model suggests that firms tacitly collude when assigning prices to goods. While this implicit agreement works well under normal circumstances, temporary demand spikes tempt individual retailers to break the agreement by lowering their markups and dropping their prices. In doing so the retailers would hope to capture a larger market share. As with the first model, these price wars should occur when overall shopping volume is high, because these times are when the most customers are up for grabs.

Finally, the loss leader model gives advertising a role in pricing, working on the assumptions that retailers must advertise to inform customers about prices and that retailers pay a fixed cost per ad. According to this model, it is efficient for retailers to advertise (and commit to) a low price on high-demand items, whether or not the aggregate volume of shopping is high or low.

All three models predict that the prices of high-demand goods will fall during high-demand periods. The main difference is in their predictions for the behavior of retailer margins for high-demand goods during low-demand periods. The loss leader model is the only one to predict that retailer margins will fall for a high-demand good even if the demand for all other goods is low.

To decide among the theories, Kashyap, Rossi, and Chevalier began by looking at holiday-sensitive goods. Defining retail margins as price minus the wholesale cost, divided by price, the authors found that the retail margins for such goods fall during the holidays. Margins for cheese and snack crackers drop significantly at Thanksgiving and Christmas, while beer retail margins drop at Memorial Day, the Fourth of July, Labor Day, and Christmas.

As aggregate demand tends to increase at these holidays, however, the authors did not consider this evidence conclusive. They turned to Lent, which is not associated with any shopping surges. During Lent, retail margins on tuna declined 5 percent, further weakening the case for the first two models. So did results for the non-seasonal goods, such as cookies and analgesics. Six out of these seven categories showed higher retail margins at Christmas, despite the fact that Christmas is the peak shopping period of the year.

Seeking more direct proof for the loss leader model, Kashyap, Rossi, and Chevalier examined Dominick's advertising. The typical supermarket circular advertises prices for about 200 of the 25,000 items the store carries. The data from Dominick's showed a positive correlation between sales and reduced prices. Even more significantly, the data also showed that popular goods were the ones that were advertised the most, especially seasonally peaking items.

A Word to Manufacturers

While clearly concluding that retailer behavior accounts for most pricing changes, the authors acknowledge that manufacturer behavior also plays a role. Although Kashyap, Rossi, and Chevalier were able to study wholesale prices, they lacked data on manufacturing costs and therefore could not explain why wholesale prices showed a tendency to fall at seasonal demand peaks. "On balance, we read the evidence as saying that manufacturer behavior plays a more limited role in the counter-cyclicality of prices than retailer behavior," they write.

"Manufacturers can only encourage retailers to use whatever prices they think are optimal," explains Rossi. "Smart retailers are going to cut their markups-the difference between the retail and wholesale prices-at times of seasonal demand and expand their sales."

During demand peaks, advertising and other promotions become even more critical. Not only do goods receive greater exposure to an increased number of consumers, but consumers pay attention to the advertised prices and may switch brands. Knowing this, retailers are less likely to advertise multiple products.

"If consumers regard different kinds of beer as being roughly equivalent," explains Rossi, "why should a retailer chew up two slots of advertising space for beer?"

To avoid losing out in the "promotions sweepstakes," Rossi advises manufacturers to consider offering retailers more incentives to advertise or promote items directly rather than cutting wholesale costs.

"Instead of investing $10 million in cutting wholesale prices during the holidays," says Rossi, "maybe it would be more efficient to spend that money on various inducement incentives, such as lump-sum payments or funds for developing new markets."


• • • • • • • • • • • • • • • • • • • • • • • • • •

Anil Kashyap is Edward Eagle Brown Professor of Economics and Finance at the University of Chicago Graduate School of Business. Peter E. Rossi is Joseph T. Lewis Professor of Marketing and Statistics at the University of Chicago Graduate School of Business.

>> Request a copy of the research paper featured in this article