I don't think this is a statistical question, but rather a psychological one. It's a well known choice phenomenon that adding a third item to a mix can shift the buying response. In this case $399 price by itself is paired with a "no buy" response; however, by adding an artificial third response of $599, the proportion of buyers in the "no buy" position decreases to the benefit of the $399 position. It is sometimes called the Williams-Sonoma response, because Williams-Sonoma added an expensive alternative to an item that was not selling and were surprised at the increase in sales of the base item.
There are some fairly satisfactory theories that explain this in terms of the status-quo response and the human loss function. Dennis Roberts wrote: > I had sent this to the APSTAT list but thought it "might" have some > general interest in how research/evaluation methods might be brought to > bear on some issue of "importance" (??) > > -------------- > The note I sent re: low Dell computer price leads to a possible question > that APstat teachers and students might find interesting to pursue: > > DO REBATES REALLY MAKE A DIFFERENCE? > > In terms of consumer spending? > In terms of profits? > In terms of volume of sales? > In terms of consumer watching for ads in the newspapers and other > outlets that HAVE rebates attached (and glossing over those that don't)? > > For example, the Dell ad I sent (which had been shown in yesterday's > USATODAY), showed an "outrageous deal" price of $399 AFTER a $200 > rebate. Online (not in the print version) they gave an additional $50 > off for ordering online. > > Do these sorts of "rebate deals" really make a difference compared to > what Dell's sales might have been if the actual price listed in the > newspapers had been $399 or $349 directly, without having the $200 > rebate attached? Or w/o having to go through the trouble of filling out > a rebate form with UPC code attached and also the original sales receipt? > > What sorts of data would it take to make a strong case one way or the > other? What kinds of statistical analyses might be useful in this regard? > > Would some experiment be possible that might shed some light on this? > > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > ================================================================= -- Bob Wheeler --- http://www.bobwheeler.com/ ECHIP, Inc. --- Randomness comes in bunches. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
