Does surge pricing maximize consumer welfare?

Steve Randy Waldman makes one of the more persuasive arguments against surge pricing (a term which may have existed before Uber—the practice certainly did—but which now seems inextricably intertwined with their brand).

I don’t care all that much about Uber’s “surge pricing” — its practice of increasing its usual fare schedule by multiples during periods of high demand. I do, however, care about the damage done by a kind of idiot dogmatism that hijacks the name “economics”. Uber’s surge pricing may or may not serve Uber’s objectives of profit maximization and world domination. It may or may not increase “consumer welfare”. But it is not unambiguously a good practice, either from the perspective of the firm or as a matter of economic analysis. Its pricing practices impose tradeoffs that must be addressed with reference to actual, on-the-ground circumstances. Among prominent academic economists there may well be a (research-free) consensus that surge pricing promotes consumer welfare (ht Adam Ozimek), but that reflects the crude selection bias of the profession much more than actual analysis of the issue. The dogmatism which has arisen in support of Uber’s surge pricing is quite analogous to the case of urban rent regulation, a domain in which there is incredible heterogeneity across localities and nations, both of circumstance and policy, and a wide range of legitimate values that conflict and must be reconciled. (Here’s an interesting case in the news today, in Spain, ht Matt Yglesias.) Almost as a right of passage, economists drone in every intro course that rent controls are bad. By preventing price signals from working their magicks, they prevent the explosion of real-estate supply that a truly free market would deliver. This is stated as uncontroversial fact even while economists who research and opine prominently on housing policy have endlessly documented that housing supply is not in fact price-elastic in the prosperous cities where rent controls are typically imposed. None of this is to say that rent controls are good or bad, or that non-price barriers to construction are good or bad. These are complex questions involving competing values textured by local circumstance. They deserve bespoke analysis, not pat dogma imposed by distant central planners economics professors.
As in the case of rent control, the stereotyped economist’s case for surge pricing is based on a conjectured elasticity of supply. With higher prices, the reasoning goes, more drivers will hit the road, more customers will be served, and the world will be better off. And that’s a good case, as far as it goes. But it doesn’t go very far, without some empirical analysis. It doesn’t justify Uber’s actual practice of surge pricing, which is far from the transparent auction our stereotyped economist seems to imagine. It doesn’t account for the trade-offs imposed by price-rationing (as opposed to time- or lottery- rationing), both between customers and for the public at large.
First, how price elastic is driver supply? If we presume that Uber is a Walrasian auctioneer, a disinterested matchmaker of supply and demand, apparently supply is not very elastic. Uber surges prices by multiples, two, three, even four times “typical” pricing in periods of high demand. That’s extraordinary! If supply were in fact elastic, small increases in price would lead to large increases in supply. The supply-centered case for dynamic pricing is persuasive in direct proportion to actual elasticity of supply. Uber’s behavior suggests that the supply-based case is not so strong. Of course, we cannot make very strong inferences about driver supply from Uber’s behavior, because they are not in fact a disinterested Walrasian auctioneer. When Uber surges, it dramatically raises its own prices and earns a lot more money per ride, whether ride supply increases not at all, or whether it spikes so much that drivers end up competing heavily for riders and suffer long vacancies. As a profit maximizer, Uber’s incentives are to impose surges primarily as a function of demand, and say nice things about supply to con economists and journalists.

This is one of the paradoxes of surge pricing. It is supposed to attract more drivers to the road, but if you live in SF as I do and have tried to call an Uber at the end of the workday (say 5-8pm), or on a weekend night, you know that Uber will inevitably be in surge pricing. And yet it still happens every day. If surge pricing worked, you'd expect drivers to learn that those are great times to drive to maximize their earnings, and that as more drivers did so surge pricing would wane as supply matched demand.

Perhaps there are still not enough drivers in total to match these spikes in demand. Or perhaps driver supply isn't as price elastic as claimed.

Or perhaps the cost of maintaining even the normal supply of drivers on a Friday night is just higher. Driving in rush hour in San Francisco isn't exactly pleasant work, and more importantly, drivers want to go out on Friday nights, too. The price to get them on the road may just permanently be higher on that night, like how evening movie tickets cost more than matinee tickets, or dinner costs more than lunch at restaurants even if the food is the same, or how Monday morning flights with Friday evening returns cost more to target business travelers. Completely plausible, but different from the story that ride sharing PR departments continue to put out, which is that surge pricing attracts more drivers until supply matches demand.

If that were the case, from a brand perspective, it would be better to just put those higher rates into effect permanently and call them peak rates and call the pricing at other times off-peak rates. Surge pricing is already a dirty word. If for some reason demand was even higher than normal peak demand, then put in surge pricing, and if for some reason it happened to be lower, you could offer a discount off of peak rates (ebb pricing?) and gain some consumer goodwill.

The other argument for surge pricing is, of course, price rationing. That is, by raising prices, we ensure that the scarce resource of Uber drivers goes to those who most need it.

Unfortunately, the argument for price-rationing (as opposed to lottery-rationing, or queue-rationing) of goods as being welfare-maximizing depends (at the very least) upon a rough equality of wealth so that interpersonal dollar values can stand in for interpersonal welfare comparisons. In an unequal society, price rationing ensures disproportionate access by the rich, even when they value a good or service relatively little. There is no solid case that price-rationing is optimal or even remotely a good idea when dispersion of purchasing power is very large. I’ve written about this, as has Matt Yglesias very recently. Matt Bruenig has two excellent posts relating this point to Uber specifically (as well as another post on ethical claims about Uber’s pricing).

The service is still scaling (incredibly), so it may not be fair to judge the validity of the price-rationing argument. However, based on the three times I've seen crazy surge pricing multiples for Uber (I'm talking 8X to 9X, for example during a blizzard in NYC the day after the Super Bowl two years ago), price rationing meant Uber was only available to price-insensitive wealthy folks. Great for maximizing Uber's profits, but not exactly what people claim when they say the market is the best way to allocate scarce resources during times of peak demand, for example an emergency. Unless you want to argue that because the wealthy were willing to pay more, they deserved or needed the service more than poorer folks. Pursue that line and next thing you know, you're dancing with a woman in a mask at your masquerade ball and she's whispering in your ear:

There's a storm coming, Mr. Wayne. You and your friends better batten down the hatches, because when it hits, you're all gonna wonder how you ever thought you could live so large and leave so little for the rest of us.

Matt Bruenig explains this with a very clear example:

Suppose that, in a given location, 10 people will normally hail an Uber cab, and 10 drivers will normally be cruising about to accept them. Now suppose that, because of an emergency, the number of people trying to hail a cab shoots to 100 people. In response, Uber jacks up prices very high, which has the effect of bringing 10 additional drivers on to the road. That means there are now 20 drivers (a doubling of supply) and 20 of the 100 people trying to hail an Uber cab succeeds in doing so.
Under Econ 101 analysis, you say that there was a welfare increase here. See, there were 20 people who got Uber cabs rather than 10 people. But, as I keep pointing out, this argument is not determinative if we assume the 100 people vying for Uber cabs have unequal economic resources. Further, the more unequal the resources are among those people, the more likely using prices like this actually decreases aggregate utility.
To see why, consider these two scenarios:
  • Rider Demand: 100
  • Cab Supply: 10
  • Chance of Getting a Cab: 10% for all 100 riders
  • Rider Demand: 100
  • Cab Supply: 20
  • Chance of Getting a Cab: 100% for wealthiest 20 riders, 0% for other 80 riders
From a glance, you can immediately see that for the bottom 80 riders, the rational preference should be the Non-Surge. In Non-Surge, their odds of getting a cab are 10%. In Surge, their odds of getting a cab are 0%. Don’t let stupid journalists confuse you on this point.

Did more cars hit the road in response to the 8x or 9x surge pricing that snowy night in NYC? Without data from Uber, it's impossible to say. Given that Uber has been under a bit of a public relations siege, at least in the tech press and locally here in the Bay Area, if surge pricing increased supply of drivers in any meaningful manner in times where demand outstrips supply, I would've thought they would've released data proving that point.

This isn't to say I'm not a fan of Uber and other ride-sharing services. I love ride sharing, I use Uber and Lyft all the time. They've undoubtedly produced a great deal of surplus consumer and societal welfare, especially in this time of a heavily subsidized price war for market share. What taxi drivers and the government are doing in Paris to fight off Uber is just one more reason I've fallen deeply out of love with what was once one of my favorite places in the world.

And I don't doubt some of the grumbling about surge pricing is just the usual consumer noise greeting any price increases, however reasonable. It would be more bizarre if consumers didn't complain simply as another signal to suppliers of their preferences.

But the argument that surge pricing always maximizes consumer welfare is a more complex one, and not a premise that should be accepted at face value.

When is everyday low pricing the right tactic?

When stores like Wal-Mart, Sam's Club, and Costco began their rapid expansion in the 1990s, supermarkets were thrown for a loop. The limited service, thinner assortments, and “everyday low pricing” of items in these “supercenters” — including foodstuffs — created enormous cost savings and increased credibility with consumers. What was a Safeway or a Stop & Shop to do in the face of such brutal competition?

A new paper from Stanford GSB looks at the strategic pricing decisions made by grocery firms during that period in response to the shock to their local market positions by the entry of Wal-Mart. The paper answers the age-old question in the supermarket industry: Is “everyday low pricing” (EDLP) better than promotional (PROMO) pricing that attempts to attract consumers through periodic sales on specific items? Investigators find that while EDLP has lower fixed costs, PROMO results in higher revenues — which is why it is the preferred marketing strategy of many stores.

The research is also the first to provide econometric evidence that repositioning firms’ marketing approaches can be quite costly. Switching from PROMO to EDLP is six times more expensive than migrating the other way around — which explains why supermarkets did not shift en masse to an “everyday low pricing” format as predicted when Wal-Mart entered the game.

From this article from Stanford's GSB. Ex-Apple exec Ron Johnson can attest to the switching costs of going from EDLP to PROMO pricing; it cost him his job at J. C. Penney.

I'm a Costco regular, but I'll buy groceries at Safeway or other grocery stores sometimes just because or geographic convenience and longer shopping hours. If you have proprietary products, that also allows you to sidestep, to some extent, the EDLP war of attrition.

For commodity products, however, the more retail moves online, the less tenable it is for stores to rely on the sheer convenience of physical store proximity to bypass the EDLP game. The paper above looked at the entry of Wal-Mart, but of course the modern day successor to Wal-Mart as an e-commerce gorilla is Amazon. If you are selling the same commodities as Amazon, it's a brutal game, especially as the eventual customer expectation will likely be same-day delivery AND every day low prices for most retail goods. In that scenario, the findings of the researchers above would not hold. 

Big data and price discrimination

Adam Ozimek speculates that Big Data might bring about more price discrimination.  First degree price discrimination has always been a sort of business holy grail, but it was too difficult to get enough information on the shape of the price-demand curve to make it so.

For some time now, though, this has no longer been the case for many companies, and in fact one company did try to capitalize on this: I know because I was there, and the reason that was a short-lived experiment is a real world case study of how the internet both enables and then kneecaps this type of price discrimination.

Amazon, until then, had one price for all customers on books, CDs, DVDs (this was the age before those products had been digitized for retail sale). A test was undertaken to vary the discount on hot DVDs for each customer visiting the website. By varying the discount from 10% up to, say, 40%, then tracking purchase volume, you could theoretically draw the price-demand curve with beautiful empirical accuracy. 

Just one catch: some customers noticed. At that time, DVDs were immensely popular, selling like hotcakes, and the most dedicated of DVD shoppers perused all the online retail sites religiously for the best deals, posting links to hot deals on forums. One customer posted a great deal on a hot DVD on such a forum, and immediately some other respondents replied saying they weren't seeing the discount.

The internet giveth, the internet taketh away. The resulting PR firestorm resulted in the experiment being cancelled right away. Theoretically, the additional margin you could make over such price discrimination is attractive. But the idea that different customers would be charged different prices would cause such distrust in Amazon's low price promise that any such margin gains would more than offset by the volume of customers hesitating to hit the buy button.

Ozimek notes this: "The headwind leaning against this trend is fairness norms." What's key to this is that the internet is the world's most efficient transmitter of information, and while it enables a greater degree of measurement that might enable first degree price discrimination, it also enables consumers to more easily share prices with each other. This greater transparency rewards the single low price strategy.

It's not a coincidence, in my mind, that Apple fought for a standard $0.99 per track pricing scheme with the music labels while Amazon fought the publishers for a standard $9.99 pricing for Kindle ebooks. Neither Amazon or Apple was trying to profit on the actual ebooks or digital music retail sales (in fact many were likely sold at break-even or a loss), they were building businesses off of the sale of complementary goods. In the case of Amazon, which is always thinking of the very long game, there are plenty of products it does make a healthy profit off of when customers come to its site, and getting users to invest heavily into building a Kindle library acted as a mild form of system lock-in. In the case of Apple, it was profiting off of iPod sales.

In the meantime, second and third order price discrimination continues to exist and thrive even with the advent of the internet so it's not as if the pricing playbook has dried up.

A skeptic might counter: didn't Ron Johnson get fired from J. C. Penney for switching them over to an everyday low price model? Didn't their customers revolt against the switch from sales and coupons and deals you had to hunt down? 

Yes, but everyday low pricing isn't a one-size-fits-all pricing panacea (as I wrote about in reference to the Johnson pricing debate at J.C. Penney). For one thing, there is path dependence. Once you go with a regular discount/deal scheme, customers create a mental price anchor that centers on that discount percentage and absolute price. It's hard to lift an anchor.

J. C. Penney was trying to go from a heavy sale-driven pricing scheme to an everyday low pricing model, and that's an uphill, unmarked path. Only the reverse path is paved. It's not clear whether the switch would have worked in the long run. Johnson ran out of runway from his Board soon after he made the switch and revenues declined. 

Everyday low pricing tends to work best when you're selling commodities since those items are ones your customers can purchase many places online. At Amazon we were far more interested in dominating one crucial bit of mental math: what website do I load up first if I want to buy something? We were obsessed with being the site of first resort in a consumer's mind, it was the core reason we were obsessed with being the world's most customer-centric company. Anything that might stand in the way of someone making a purchase, whether it be prices, return policy, shipping fees, speed of delivery, was an obstacle we assaulted with a relentless focus. On each of those dimensions, I don't think you'll find a company that is as customer-friendly as

Ultimately, customers have a hard time figuring out intrinsic value of products, they're constantly using cues to establish a sense of what fair value is. Companies can choose to play the pricing game any number of ways, but I highly doubt Netflix and Amazon will choose to make their stand on the first order price discrimination game. There are many other ways they can win that are more suited to their brand and temperament.

Still, the peanut gallery loves to speculate that Amazon's long term plan is to take out all of its competitors and then to start jacking up prices. A flurry of speculation that the price hikes had begun spun up in July this year after an article in the NYTimes: As Competition Wanes, Amazon Cuts Back Discounts. After the NYTimes article hit, many jumped on the bandwagon with articles with titles like  Monopoly Achieved: An invincible Amazon begins raising prices.

If you read the NYTimes article, however, the author admits "It is difficult to comprehensively track the movement of prices on Amazon, so the evidence is anecdotal and fragmentary." But the article proceeds onward anyhow using exactly such anecdotal and fragmentary evidence to support its much more certain headline. 

Even back when I was at Amazon years ago we had some longer tail items discounted less heavily than bestsellers. However, pricing the long tail of books efficiently is not as easy as it sounds, there are millions of book titles, and most of the bandwidth the team had for managing prices was spent on frontlist titles where there was the most competitive pressure. All the titles listed in the NYTimes article sound to me like examples of long tail titles that were discounted too aggressively for a long period due to limited pricing management bandwidth and are finally being priced based on the real market price of such books. Where in the real world can you find scholarly titles at much of a discount?

The irony is that the authors cited in the article complain their titles aren't discounted enough, while publishers ended up in court with Amazon over Amazon discounting Kindle titles too much. This is to say nothing of the bizarre nature of book pricing in general, in which books seem to be assigned retail prices all over the map, with the most tenuous ties to any intuitive intrinsic value. The publishers set the retail price, then Amazon sets a price off of the retail price. If the publishers wants the discount on their books to be greater they could just increase the retail price and voila! The discount would be larger.

To take another category of products, DVDs, soon after we first launched the DVD store, long tail title like Criterion Collection DVDs were reduced from a 30% discount to a 10% to 15% discount. But just now, I checked Amazon, and most of its Criterion DVDs are discounted 25% or more. If I'd taken just that sample set I could easily write an article saying Amazon had generously decided to discount more heavily as part of its continued drive to return value from its supply chain to customers.

Could the net prices on Amazon be increasing across the board? I suppose it's possible, but I highly doubt that Amazon would pursue such a strategy, and any article that wanted to convince me that Amazon was seeking to boost its gross margins through systematic price hikes would need to cite more than just a few anecdotes from authors of really long tail books. 

It will remain a tempting narrative, however, because most observers think it's the only way for Amazon to turn a profit in the long run.

However, that's not to say big data hasn't benefitted them both in extraordinary ways. Companies like Amazon and Netflix know far more about each of its customers than any traditional retailer, especially offline ones, because their customers transact with them on an authenticated basis, with credit cards. Based on their customers' purchase and viewing habits, both companies recommend, better than their competitors, products their customers will want.

Offline retailers now all want the same type of data on their customers, so everyone from your local drugstore or grocery store to clothing retailers and furniture stores try to get you to sign up for an account of some sort, often by offering discounts if you carry a free membership card of some sort.

Your lobster roll is overpriced

A glut of lobster in the ocean has driven its wholesale price down, yet restaurants have kept lobster dishes on the menu at historical high price points. James Surowiecki explains the many reasons why, including this:

Lobster hasn’t always been a high-end product. In Colonial New England, it was a low-class food, in part because it was so abundant: servants, as a condition of their employment, insisted on not being fed lobster more than three times a week. In the nineteenth century, it became generally popular, but then, as overharvesting depleted supplies, it got to be associated with the wealthy (who could afford it). In the process, high prices became an important part of lobster’s image. And, as with many luxury goods, expense is closely linked to enjoyment. Studies have shown that people prefer inexpensive wines in blind taste tests, but that they actually get more pleasure from drinking wine they are told is expensive. If lobster were priced like chicken, we might enjoy it less.

Restaurants also worry about the message that discounting sends. Studies dating back to the nineteen-forties show that when people can’t objectively evaluate a product before they buy it (as is the case with a meal) they often assume a correlation between price and quality. Since most customers don’t know what’s been happening to the wholesale price of lobster, cutting the price could send the wrong signal: people might think your lobster is inferior to that of your competitors. A 1996 study found that restaurants wouldn’t place more orders with wholesalers even if lobster prices fell twenty-five per cent. As the study’s authors put it, “A low price creates suspicion.” This helps explain one of the interesting strategies that restaurants have adopted to take advantage of the lower price for lobster: they keep the price of lobster entrées high, but add lower-priced items—lobster bisque, lobster mac-and-cheese, a lobster B.L.T—to the menu. That way, they can generate more business without endangering lobster’s exclusive image.

I wonder if lobster prices remain high at seafood markets, or even Costco. They don't have to engage is menu price portfolio construction, but the perceptual illusion that higher priced lobster tastes better still exists.

The price game

When Ron Johnson took over as CEO of J. C. Penney, one of his most sweeping changes was to move away from the constant sales and coupons to a more straightforward pricing model. Not surprising considering he came from Apple where they hold one sale a year, on Black Friday, and not even a great one at that.

But J. C. Penney is not Apple, and the price game each is playing is different.

Mr. Johnson explained a similar logic when he moved the chain toward simplified pricing. In January 2012, while introducing his new plan to investors, the press and vendors, Mr. Johnson said that in the previous year, the company held 590 sales events; almost three-quarters of the stuff it sold was marked down 50 percent or more.

But here’s the thing: customers weren’t actually paying less. The chain just kept raising the prices that customers saw on the racks, and then discounted those prices during promotions. Why keep playing a game that is expensive and troublesome for the seller and a mirage for the consumer?

J. C. Penney was not the first retailer to be astonished by the brilliance of this realization. In 2006, Macy’s had a similar idea after acquiring the coupon-happy May department stores. It decided to “retrain” those customers, as its chief financial officer put it at the time, by drastically cutting coupons. By 2007, it had abandoned that strategy. Its chief executive acknowledged that pulling back on coupons was Macy’s biggest mistake in its acquisition.

Even Walmart, which actually does pull off the trick of “everyday low prices” in its domestic stores, is finding it hard to convert consumers to a single-price model in countries like Brazil and China, where retailers give deep discounts on a few main products, then mark up the rest, said Mark Wiltamuth, an analyst at Morgan Stanley.

The problem, economists and marketing experts say, is that consumers are conditioned to wait for deals and sales, partly because they do not have a good sense of how much an item should be worth to them and need cues to figure that out.

Just having a generically fair or low price, as Penney did, said Alexander Chernev, a marketing professor at the Kellogg School of Management at Northwestern University, assumes that consumers have some context for how much items should cost. But they don’t.

Price strategy has to be supported throughout the organization. For Apple to have one price for its items means they must enforce that price through all of its distribution partners, and it must also create advertising that reinforces the premium quality of the goods. And of course, the products must be good enough to justify a no discount policy.

One thing is for certain: once you go sale, it's tough to go back (once you go red, it's hard to go black?). Companies that consider a sale or discount strategy should do so carefully. Once customers expect a regular cadence of sales or discounts (e.g. Restoration Hardware's bi-annual bath sales, or Bed Bath and Beyond's ubiquitous 20% off coupons) they orient their entire behavior around that pattern and won't easily be persuaded to buy at full price ever again.

Purple Pricing

It shouldn't be surprising, perhaps, that one of the more innovative sports ticket pricing schemes to be put into practice comes out of Northwestern University, not traditionally a sports powerhouse. Last month, they launched Purple Pricing for select men's basketball games for the rest of this season.

Essentially, Purple Pricing is a form of Dutch Auction, in which the prices of the item being sold are lowered until you hit the price at which all the items can be sold at that price. Then everyone who has bid above that price pays the lower price, the same as everyone else. The incentive, in the case of Northwestern men's basketball tickets, is to find that price at which they can sell out the game and also to capture some revenue back from the secondary market, from companies like Stubhub.

For the buyer, as soon as the price reaches one you're willing to pay, there's not reason to wait. The price can only go lower from there, it will never rise.

Purple Pricing is a joint effort between the Northwestern athletic department and economists Jeff Ely and Sandeep Baliga. The experiment will provide Northwestern with a ton of data to maximize revenue.

Theirs is not a perfect Dutch Auction, however, as there is an artificial price floor. Northwestern will not let the Dutch Auction prices fall below the price that season ticket holders paid. In the future, they could do away with season tickets altogether and use Dutch Auctions to price every game.

The first two games Northwestern tried this with offered a good divergence in demand to test out the system: a desirable game against #16 ranked Ohio State, and a much less desirable game against Penn State. The attendance against Ohio State was 7,036, while against Penn State it was just 5,517. It would have been interesting to see where the price would have landed for the Penn State game had there been no price floor in effect.

Alas, Northwestern lost both games. Some problems can't be so easily solved with economics.

Pricing anomalies

There are many positives to living in San Francisco, but parking is not one of them. It's the hardest city to find parking in that I've ever lived in. Someone wrote an entire book on the subject.

I went to Amazon to buy a copy of said book. Amazon didn't have a copy, so I checked the 3rd party seller tab for a new copy. The first seller offered it for $9.97 a copy, but the other two sellers wanted $75 and $92.20 plus shipping! This for a book that's sold off of its website for $12.95 plus shipping.

Sellers are always looking for out of print situations to try to shift the price of media up. I once sold a copy of the Criterion Collection DVD of Salo, which was out of print at the time, for $220 on Someone should start a Tumblr of products which have gone into short supply and thus have exorbitant prices from Amazon's 3rd party seller community.