Matching by Algorithm

· May 24, 2019

There was a great story in the NY Times about the craziness around taxi medallions in New York - the things that let you own a cab - and some of the trouble it got drivers into. The medallion system among NYC taxis put a sharp divide between those who drove taxis and who owned them, which lead to people making some bad decisions to get on the other side of that line, aided by intermediaries who were neither buyers or sellers, but profited off the transaction.  The ridehailing apps came along and basically removed the barrier, but the scheme was so shaky it was destined to fail. 

To meet the high cost of cabs, drivers needed to be efficient with their time. One of the other impacts of ridehailing apps in New York is that it’s easier to get a ride in the boroughs outside Manhattan. Despite attempts (green cabs) to provide better service, there was a lot of pent up demand. Offering rides in the very dense Manhattan meant that wait time between fares was low, which made for efficient days for the drivers. Longer wait times made the cost of transacting with a rider in Long Island  higher than in Manhattan for the driver. 

Ridehailing reduced that cost (as well as many of the other friction points). While other areas might not be as dense, they are dense enough to offer a sufficiently low wait time for drivers and passengers if you have a global view of the drivers and riders and can put them together effectively. 

There is another view of that change though: If you had been an outer-borough resident, and really wanted to know you could get a ride, you could buy a car, use a car service or even employ a driver. This was a larger outlay, but for a significantly lowered barriers to take a ride. Owning a car, hiring a driver, or even contracting with a regular car service are quite far from the experience of a taxi: you own this specific car, of this specific make and model, or you see this particular driver every day. These can be very nice things, but they are somewhat secondary to the core goal here - getting from A to B.

The kind of automated matching that makes ridehailing work requires a level of standardization - the product or service offered has to be interchangeable. Famously, standardizing wheat opened up competition across the world for a product that was previously seen as more locally differentiated, which benefited consumers of wheat (both businesses and individuals) but hurt farmers who had fixed cost structures for their products and were suddenly competing with other farmers with lower costs (though who also needed shipping). 

Ridehailing apps standardised getting from A to B at a different level than taxi cabs. The standard was in some ways higher - tracking and credit card payments - and in some ways lower - higher range of cars, shared rides. The net was similar - good for consumers, bad for producers, though specifically for producers who were the most down this line of differentiation outside of the core A to B. If you owned a medallion, you had a lot invested in a specific taxi. If you just drove for someone else, your harm was the difference between what you could get driving ridehailing and what you could get driving a yellowcab, after paying to rent it. If you no longer or had never had a job as a taxi driver, the difference was positive. 

More broadly, the more your work and work product is standardized, and the more it is easy to transact for that type of work in a marketplace, the less pricing power you will have: the ability to increase what you charge for your labor. The more your work is differentiated or hard to buy, the more pricing power you will have, in exchange for it being harder to find buyers. This is why some farmers travel long distances to city farmer’s markets where they supplement their income by highlighting the unusual aspects of their produce. 

A lot of people look for marketable skills when they are in education. They want the wide range of employers and opportunities given by skills for which there is a ready market. If that market crosses a tipping point  by which those skills are too readily available, or the output of deploying those skills too predictable, any premiums are likely to be competed away.