In an on-demand world, such as a cab aggregation service, an efficient ‘surge pricing’ algorithm has the potential to serve a public good – help overcome the anomaly of asymmetry of information, and, do so dynamically in real time. So you are less of a hostage to the whims of a regular cab.
In effect it is to continuously re-estimate the demand and supply adjusting in very short intervals, finally converging to a minimum equilibrium demand – supply benchmark. Or as Dr Subir Gokarn former Deputy Governor of the RBI (now on the Board of the IMF, Washington DC) calls it – a ” dynamic cobweb model ” .
A solvable and simple frictionless model – in the early stages there is ‘overshoot’, as seen in the Melbourne café hostage situation last year (4x hike), or the Arianna Grande Madison Square Concert in March 2015 (4x hike), as also ‘undershoots’, as exhibited during New Years eve in Times Square on 31st December 2014.
But since demand and supply of these cabs are known to be textbook elastic – on the demand side price-elastic, and on supply side revenue – elastic. What the geo-spatial nature of the black-box algorithm essentially does is just help in matching up supply with demand in a given ‘polygonic’ area with the objective of maximizing ‘completed rides’ within a waiting period (say ideally 3 – 5 minutes).
But, are these principles being compromised?
Whilst the surge is quick to move up, it would seem a tad slow to settle down back – thus making for unfair rent seeking practice.
My sense is that the algorithm is not that dynamic enough to make these instant adjustments and give the most efficient and equitable price for the maximum period of time. It seems to behave more like time of day (ToD) metering for public utilities viz. electricity. Predictable – office hours and such – so surge!
For e.g. during one long weekend I noticed the surge factor during peak period by default! It would seem that some form of rent seeking is predetermined, maybe based on post processed data sets on which the algorithm may be feeding off, rather than contemporaneous demand-supply considerations. Or perhaps the number of polygon fences in a city, or vertices within a fence is not adequately mapped or defined for the NCR or Greater Metropolitan Area of Mumbai to ensure computational efficiency.
Just to illustrate in New York City there are 167 neighbourhood, and 100 fence polygons. And in some neighbourhoods 100 vertices to a fence! Seems to be well mapped for quick (100 milliseconds response), and optimal pricing ( surge of upto 5 minutes ).
No doubt a good algorithm should be a constantly learning one — a high level of sophistication in design, tuning of the code and servers, CPU speed, good math, use of big data, etc. I am not sure whether the algorithm for cabs is a learning one. It should be.
For e.g. the ‘wait time’ displayed in the app seems to be way off, sometimes by a factor of two or even three during surge periods. Or how about the fact that whilst one app indicates surge, the other competing service has no sign of it in the same vertice ! Or the inverse – one indicates surge, and there are no cabs available on the other! Lets be fair an off surge period (90% of the time) does give you a cheaper fare to your kaali peeli too. Or that a surge period (10 % of time) is self equilibrating, as fares top off to match the next best alternative – radio cabs, and substitution effect kicks in. Or demand just drops off a cliff because of no takers.
So are we now being held hostages?
If there is a variance between price and value then one could say there is market failure and case for intervention. And argue for some sort of cap on account of inefficiency and collusion by cab companies and cab drivers (remember the Government contemplated price caps last year on airlines on some routes during the festive season, and is contemplating regional flights to have a maximum price cap as I write this).
As everyone knows most of the driver-partners and customers carry more than one competing app. Drivers also serve an offline loyal customer base on fixed rates too! And they can always check out the surge area & price multiple by opening a customer app on another phone to see when they should come online, with which cab aggregator and where! Customers too have a choice. They have close substitutes (auto, kaali peeli, Radio Cabs, A C Cabs, share cabs, share autos, share minis, shuttles, etc).
Though In principle one is not for caps, and for a full play for market determined surge). But in India it seems to trigger some unfair practices more often than not on account of just about any whack—a– mole rent seeking opportunity.
(A tip: If one hangs about a little bit or if you walk around to the boundaries of the surge area you get normal base fare or one with lower surge price)
Probir Roy Is Founder of Paymate and commentator on public policy, digital & FINTECH.