My fellow bloggers seem to have the big picture in focus. Here's something a little smaller. In my Principles class, I have students read this clever Slate article on price discrimination, "The mystery of the rude waiter," by Tim Harford.
He suggests that a restaurant may deliberately employ a rude waiter in the bar, in order to discriminate, charging a high price in the dining room and a lower price in the bar for the same food. The rude waiter keeps the high rollers in the dining room - it's the hurdle that the low willingness-to-pay customers jump to get the lower price.
Here are some numbers that I use to illustrate Harford's point. Suppose there are two potential customers, rich and poor. The rich person will pay up to $24 for a meal with decent service, but only $11 for a meal served badly. The numbers for the poor person are $10 and $9, respectively. The cost of a meal is $5. The differential in each case is the Willingness-to-Pay for good service, which is higher for the rich than for the poor. Without the rude waiter, the seller would charge $24, sell to the rich only, and make profits of $19 (this beats the $10 profit he could make serving both at a price of $10). With the rude waiter, he will set the price in the bar at $9 and the price in the dining room at a little less than $22. The rich person then chooses the dining room, getting consumer surplus of a little bit more than $2 ($24 minus a little bit less than $22), as opposed to the $2 surplus ($11 -$9) he would get eating in the bar. The seller's profits are then $31 - $10 = $21.
Harford then discusses price discrimination through quality degradation generally (where the lower quality version of the good is the high quality version with resources spent to degrade it!). Do these possibilities worry Tim? Not a bit of it. In good "real is rational / what me worry?" fashion he asks "what are the alternatives?": the crux is that without price discrimination, the seller "could have passed up the opportunity to serve low-end customers altogether. That would have been no better."
Yes. Well, I chose numbers where that is true, but it needn't be. In my example, change the rich person's willingness to pay to $14. Now, if unable to discriminate, the seller does best to sell to both at $10 each, earning profits of $10. The discriminatory price scheme would set the bar price at $9 as before, and the dining room price at a bit less than $12, thus giving profits of (a bit less than) $11. Compared to the no-discrimination outcome, net benefits are obviously lower, since total value drops by $1 due to the degradation, while costs are identical. If the degradation costs something (and the seller would be willing to spend up to the $1 increase in profits to degrade if necessary) - say the waiter must get extra compensation to make up for the disutility of being mean-the efficiency loss is even greater. Funny how this second possibility gets overlooked. Dr. Pangloss, call your office!