Dynamic pricing has long been a topic of great interest here, in large part because digital technology is increasingly making it feasible to implement dynamic pricing in retail electricity markets in ways that can be acceptable to consumers. But dynamic pricing is fraught with challenges, and not just in retail electricity markets. Dynamic pricing is a form of price discrimination, and as such can improve efficiency; prices also are knowledge surrogates, communicating diffuse private knowledge about the relative scarcity of that good in that place at that time. Dynamic pricing is also most applicable in markets in which demand varies over time, and if you are going to implement dynamic pricing without annoying and aggravating consumers, consumers have to have access to the prices in advance.
These general principles showed up recently in a kerfuffle over dynamic pricing of taxi services by a new firm called Uber:
On New Year’s Eve, Uber, a start-up in the city, adopted a feature it called “surge pricing,” which increases the price of rides as more people request them.
Although New Year’s Eve was very profitable for Uber, customers were not happy. Many felt the pricing was exorbitant and they took to Twitter and the Web to complain. Some people said that at certain times in the evening, rides had spiked to as high as seven times the usual price, and they called it highway robbery. Uber’s goal is to make the experience as simple as possible, so customers are not shown their fare until the end of the ride, when it is automatically charged to their credit card. While the app does not show the total fare in dollars when customers book a ride, Uber did show a “surge pricing” multiple to customers booking rides for New Year’s Eve.
So what’s the underlying economics here? Jodi Beggs comments on the kerfuffle starting from first principles, pointing out that when demand increases, consumers are not likely to be able to get the quantity they want at the price to which they may have gotten habituated as “the price”. She also points out that the dynamic pricing is what keeps shortages from occurring — think about it: would you rather pay 7 times the base fare to have an immediate ride home after your New Year’s Eve party, or would you rather wait in line for the next available car? Either way, you pay; opportunity cost matters. In other words, as my colleague Jeff Ely observes, variable pricing means that prices go up and down, and generally will be higher when more people want the good (due both to higher and more inelastic demand at that time and to higher relative scarcity). Note that these observations also apply to retail electricity pricing — market demand varies over time, and prices can signal relative scarcity, if regulators allow them to.
The relative scarcity is another aspect of the economics here, because in the immediate run the firm can’t go out and scare up more cars and drivers; in other words, supply is not going to increase. Here we see the analogue to other industries that use dynamic pricing, such as airlines and hotels and car rentals — they have a pretty fixed supply, so as demand rises and falls the price to the consumer will rise and fall accordingly, because the supply response at the time is so limited. Over time profit signals will indicate to them whether or not to invest in more cars, planes, hotels, but if you’re trying to get home on that New Year’s Eve that’s not going to kick in that quickly. Thus prices adjust to communicate relative scarcity.
But notice another aspect of the story of Uber’s pricing: although they told the customer what the “surge multiple” was when they called the car, the customer doesn’t know the fare until after the transaction has occurred. Here I concur entirely with the NY Times blog post, Jodi, and Jeff, that not informing customers ex ante about at least an estimate of the fare is a bad way to implement dynamic pricing! Especially for flesh-and-blood humans who are more than calculating machines, and are likely to be royally ticked off when they are charged 7 times base fare for such a short ride. The NY Times blog post quotes Yale economist Dirk Bergemann, saying that consumers prefer price predictability, which is true as a very broad claim … but if I draw an analogy from regulated retail pricing in electricity (and using a rhetorical trope of Jeff’s), consider the equilibrium. If instead they kept their fares constant, it’s entirely possible that the average fare could be higher in the single fare market design than in the dynamic pricing market design. That’s one of the anxious concerns that regulated electricity firms have about dynamic pricing — what if our revenue falls because a large enough share of demand ends up happening in low price periods (i.e., more demand is more elastic)? In any case, not giving customers at least an estimated fare before they commit to the order is bad business, and it should be easy to communicate that estimate, because customers are all using smart phones to order the cars.
I’d like to suggest a couple of alternatives that my colleagues did not. The first alternative is inspired by time-of-use pricing as used in electricity, or by the types of dynamic pricing contracts used in car rentals. If I know well enough in advance that I want a car at 2AM on New Year’s Eve, why not offer me a contract in which I can make a reservation at a firm price, albeit one that is higher than the base price? Then Uber could, say, take reservations for 50% of their fleet, and leave the other 50% open for spot-market transactions. With that model, those customers who are risk averse and want to make sure to have a car at a particular time at a reasonable price will have an option. But if they can’t commit to a time for a pickup, then they suck it up and deal with the spot market.
The second alternative is less relevant to the Uber example, but in lots of markets that could have dynamic pricing, we can use technology to automate our responses to the price. I’ve gone on ad nauseam about the potential for transactive technology in retail electricity markets, and it’s applicable in other markets too — automated reservation bots for making a flight reservation if the price on your preferred itinerary on your dates goes below a trigger price that you set, or a device in your car or an app on your phone that receives the current toll level and tells you whether or not to take the toll road, wait to go home, etc. Transactive technolgies reduce the cognitive barriers associated with price uncertainty, as well as reducing the transaction costs of using dynamic pricing in the first place.