It’s New Years’ Eve. You’ve had a blast and you’ve partied till the wee hours of the new year. But then the time comes to figure out how to get home. You’re probably going to order a taxi. Along with hundreds of others. For the first time on NYE in Colombo, you might be checking for taxi availability on the apps of the new sharing economy services – PickMe, Uber, and Hire1. But if you opened the Uber app during around 1am to 4.30am, you’d probably see a screen like this one below. Ouch. Prices are going to be higher than normal – ‘surge pricing’ is in force. (Customers in Miami saw prices surge up to 9.9 times on NYE!)
Sri Lanka is not alone in this. Uber has seen this spike in cities around the world. This graphic from Uber shows what the typical spike in demand looks like on NYE globally, based on five years of data.
I found Uber’s ‘surge pricing’ really intriguing from a microeconomics point of view and thought about exploring it a little further, including wondering “how can I model it on demand and supply curves?”.
What is ‘surge pricing’ all about?
It’s simple demand and supply. At times of regular demand, the fare remains the same at all times, for everyone. But during periods of uber-busy demand something changes. With lots of riders trying to book a cab on Uber more than the availability of drivers (or ‘driver-partners’ as Uber refers to them) – Uber employs a ‘surge pricing’ algorithm to attempt to “equilibrate demand and supply”. The algorithm assigns a multiple that multiples the standard fare in order to derive the “surge price”. It appears that the degree of the surge (2x, 2.2x, 4x, etc) depends not only on the degree of demand-supply mismatch at a point in time, but also on hyper-local data on proximity of drivers. There isn’t someone in an Uber control room in Colombo turning surge pricing off and on – it happens with Uber’s pricing algorithm (patent-pending) that reacts automatically to demand.
Why does Uber employ ‘surge pricing’?
Given that driver-partners that work with Uber are free to enter and exit the market whenever they wish, there can be a fair amount of variability in supply. So when demand is high, Uber would want more and more drivers to come on board and start driving. Using prices to signal this to drivers is, according to economic theory too, a good way to influence supply by inducing more drivers to get on the road and meet the new demand. So, when ‘surge pricing’ kicks in, it signals to drivers that its now a valuable time to be on the road. Drivers that may have ‘clocked off’ by then or were not driving that day may be induced to enter. Meanwhile, on the riders side, more riders seeking Uber rides can get one – albeit at a higher price – and some riders may choose not to and seek other options instead.
It also demonstrates a fundamental economic tenet of ‘allocative efficiency‘, by allocating rides to those who value it the most – those people who place a higher monetary value on an Uber ride home after a NYE party (and willingness to pay).
“Our goal at Uber is to ensure you can push a button and get a ride within minutes – even on the busiest nights of the year. And due to surge pricing, that’s almost always possible.”
The company claims that surge pricing really improves their ability to deliver rides to those who request it. In a case study of NYE in 2014 in New York when the surge pricing system failed for about 26 minutes, research showed that less than 20% of ride requests were ‘completed’ or fulfilled, whereas when surge was active, 100% of ride requests were completed. This, they argue, is vindication of employing surge pricing – the market clears – i.e., all those requesting a ride get to take one.
By Uber’s own explanation, surge pricing has two effects: 1) people who can wait for a ride will decide to wait until the price falls or they choose another travel option, and 2) drivers in nearby areas with less demand head to those locations with higher demand or drivers that aren’t working decided to get on the road.
Why is this particularly important on a night like NYE?
Because more drivers (than normal) would be more reluctant to work because of the ‘value of the ‘next best alternative foregone’ which may be leisure – staying at home with family; seeing fireworks at Galle Face Green; going for a party with friends; watching Homeland and sleeping early; babysitting a firework-wary pet, etc. This, then, is a classic explanation of ‘marginal cost of production/supply’ or ‘opportunity cost of supply’ – the opportunity cost of driving on NYE is of staying home or going for a celebration. We must also consider the connection to another economic concept of ‘reservation wages‘ or ‘reservation prices‘ – that there is some level of wage (or price of a ride, in this case) below which a driver doesn’t think its worth it to forego whatever else he/she was doing, and get on the road and seek hires.
Now here’s my modelling of surge pricing and the effects described above with simple demand and supply diagrams.
Diagram 1 below shows a market clearing condition, when demand and supply are at regular levels. Lets say, the afternoon of the 31st December. All riders who open up the app and request a ride are able to get one within a reasonable time, as any other time. Equilibrium price and quantity is at e. Price is the same at P*.
Diagram 2 below is when there is a spike in demand after around 12.30am on NYE. The demand curve now shifts out from D to D’. The surge price kicks in and the price now rises from P* to P’. Drivers are induced to drive, there is more supply of drivers, and supply expands along the supply curve (S) and there is a new equilibrium at e‘. But there’s a catch – e‘ is unlikely to be a stable equilibria for two reasons. Some drivers may not respond to this price signal as they know that the surged price does not last for the whole night and don’t feel it is worth it – they won’t supply, or may only supply until the surge lasts. So supply expands only some way along the curve. Meanwhile, not all drivers will want to accept the surged price either, and they may opt out of demand and choose other options. So, the demand curve may be between D and D’. The combination of these means that the equilibria is unstable, and will hover between e and e’.
Diagram 3 below is when now there are more drivers on the road, the surge pricing ends and prices return to the earlier level of P*, but the level of supply does not fall back to the earlier level. The opportunity cost of driving is now reduced, and so drivers stay on the road. Recall the earlier discussion of ‘reservation prices’. Think of it like this, the driver has now already decided to forgo watching TV at home or going for a friend’s party and has decided to be on the road. So even after surge pricing ends, it makes more sense for the drive to stay in the market (i.e., remain on the road) than exit the market. So the new supply curve is at S’, and the new equilibrium is e”. Of course, over an extended time when demand and supply normalises once again, we are back to the previous equilibrium price (P*) and quantity (Q*) and equilibrium point e.
In extending this further, there are a couple of questions in my mind that may complicate the analysis above:
- There are now some clever ‘cheat’ apps that have emerged to help folks navigate surge pricing. One is called ‘SurgeProtector’ which interacts with the Uber app to tell riders if theres a cheaper cab within a short walk away. How does this influence the theoretical model?
- To what extent does surge pricing adversely affect demand, and therefore not work as well as the theory suggests? For instance, lets say you’re at Taj Samudra on NYE with dozens of others requesting an Uber, surge pricing kicks in, and if the surged price is so high that it kills demand, and many of you decide to try another service or simply wait it out? But by now the surge price has made a bunch of drivers flock to Galle Face Hotel, responding to the price signal (drivers can see a map of locations where surge is active). But with demand killed, the price they responded to is now no longer there and the surge ends. The drivers may have come for the surge, but are now stuck there even as surge has ended. Good for the next bunch of ride requestors, but not so much for the drivers that flocked there. (Maybe its a virtuous cycle though?)
- To what extent is surge pricing necessary / relevant in the Colombo context where information costs are lower – i.e., cab drivers can (more easily than in a bigger city perhaps) get to know of (‘learn’) locations that will have higher demand, for instance after concerts, sports events, parties, etc.? It’s not easy to measure if its surge pricing that got more drivers on the road or into a particular area or was it them learning about when to be and where. How much of a ‘signalling effect’ will surge pricing really play in Colombo? Yet, evidence from other markets apparently show that drivers actively ‘chase’ surge pricing locations – i.e., stay on alert for when surge kicks in in a certain area, and then leaves the area they were operating in and goes to the other one. Completely rational behaviour. Ironically, this ‘learning’ behaviour by drivers may, in the long term, go against them. Why? If more and more drivers ‘learn’ of events and locations that may have higher demand at a particular time, predict it and get there, then surge pricing won’t even need to kick in as demand and supply will clear.
- During surge pricing periods, there is no way for riders to know whether the driver he/she has just got has come to that location responding to the surge pricing or whether the driver was in the area and would have accepted the request anyway. So, this complicate the allocative gain argument. The rider pays more for a ride that was there anyway, rather than a ride that was ‘flocked’ in due to the surge pricing.
Besides the microeconomics of it all, the fact remains that Uber’s surge pricing approach is controversial. Some have argued that it come under legal scrutiny for ‘price gouging’. Others have said that it will hurt the brand in the long run, and so there could be some preferential rates for high frequency or loyal customers.