The order book is a market design for the exchange of goods and assets. It dates back to the European coffee houses of the late 1600s. In London, Jonathan’s Coffee House was a significant meeting place for traders in London in the 1700s. It later became the site of the first London Stock Exchange.
In the late 1700s, in what later became known as New York City, Dutch traders met at a Buttonwood tree in lower Manhattan island to buy and sell goods. Now known as Wall Street, this location became the center of financial asset exchange in the United States.
Until the 1970s, stock exchanges were characterized by a market design involving traders gathering around pits with specialists manually matching bids and asks in paper order books (see below).
The great financial economist Fisher Black wrote a prophetic article in 1971 called “Toward A Fully Automatic Stock Exchange” where he laid out the implications of the coming automation of the manual order book. He speculated on what the computerization of the order book would mean for bidding mechanisms, liquidity and overall stock market efficiency.
Screenshot of Bid-Ask Order Book of Poloniex
Market design, indeed all design related to computers, is coupled tightly to the computer technology itself. Just because one design is associated with a particular computer technology does not mean that the same design should be mindlessly carried over when the computer technology changes.
We recall the mindless carry over of the 80 character line limit established by IBM for punch cards in the 1920s to cathode ray tube (CRT) terminals in the 1970s.
There is a whole host of other instances of mindless carry over of design when the technology changes. One notable example is the organization of the factory floor after the conversion of machine power from a centralized shaft driven by water to decentralized electric power.
In the last several years, there has emerged a new decentralized, serverless, peer-to-peer (p2p) paradigm in computer architecture propelled by several trends: Internet of Things (IoT), autonomous vehicle-to device communication (V2X), and crypto.
This technological change demands a rethinking of the appropriateness of the centralized client-server order book market design as the core of a transaction layer in a crypto-economic platform.
The trend away from client server architecture is driven by a need to do more raw compute “at the edge” before sending data to the server for storage and higher order analytics. This is known as “edge computing.” The use cases for edge computing are Internet of Things (IoT) and autonomous vehicle-to-device (V2X) communication.
The trend away from client server architecture is also driven by the tremendous interest in Bitcoin, Blockchain and Ethereum. Interest in crypto could be the start of a paradigm shift away from client server financial intermediaries earning opaque rents and toward decentralized, trustless p2p protocols for validating and accounting for the exchange of financial assets.
A true decentralized crypto-economy involves not only a DLT layer but also a high speed transaction layer.
The thesis of this paper is that the time is now to consider the possibility of pairing a transaction layer with a true decentralized market design with a decentralized distributed ledger technology (DLT).
We believe that publish-subscribe currently is the leading protocol for the transaction layer as it has already been deployed at scale an the platform behind several MMO games (from MZ) and chat platforms (WhatApp from Facebook, WeChat from TenCent).
We believe that MZ’s recently spun-off subsidiary Satori is leading the integration of a pub-sub transaction layer with a DLT called Hedera Hashgraph. The question is what will be the market design for the transaction layer?
Satori’s “AI Mesh network” transaction layer stats — 500 Million “messages” per second or 1 million publishers sending 100 bytes a second
Hedera Hashgraph’s DLT stats— 500,000 transactions per second with 100% consensus based on a “gossip of gossip protocol” and a consensus latency of a 3.5 seconds.
In a crypto-economic trading platform:
“The network becomes the exchange”
Snapchat (ephemeral) bid-asks
User-defined smart contracts
Value Proposition:
user-defined contracts ( e.g. options with odd expiration dates, long-short pairs, straddles)
tokens earned by peers supplying liquidity spread contracts
elimination of latency rents going to HFT and server co-location fees going to exchange
elimination of “data ownership” rents earned by exchange
Specification suggestions:
high frequency, many-to-many, pub-sub protocol
messages in form of Myerson “take it or leave it” (TIOLI) bid-asks
“serverless” with ephemeral matching with-in Redis-like in-memory data structure store, used as a database, cache and message broker.
ephemeral bid-ask data, only data “owned” is history of matches.
Posted on my www.nu-retail.com blog as .pdf on 12-1-2005
The roll out of the new Medicare drug benefit program has renewed the debate about who is best at negotiating drug rebates with pharmaceutical manufacturers (“Pharma”). Those who favor the Federal government point to statistics showing that Medicaid gets significantly better rebates than private sector entities called pharmacy benefit managers (“PBMs”).
The purpose of this paper is to provide new insights into this question by viewing pharmaceutical rebates through bargaining theory. The first insight is that rebates are paid selectively so that averages of rebate rates across all brands are a poor measure of negotiating power. PBMs’ ability to extract rebates from Pharma is much greater than previously realized when rebate averages are disaggregated.
The second insight derived from bargaining theory is that the rebate transaction is much more complex than the price theory conceptualization of rebates as volume discounts. Drug rebates are tariffs, or entrance fees, paid by Pharma to gatekeepers who promise access to markets with reduced competition.
Medicaid gets more than PBMs, but gives up more in terms of rights to impose co-payments, “prior authorization” restrictions and rights to switch on-patent brand prescriptions to lower cost therapeutic equivalents. PBMs receive less than Medicaid, but they give up less. Based on the overall goal of cost containment, there is evidence that private sector PBMs do a better job at rebate negotiation than Medicaid.
The Variability of Pharmaceutical Rebates
Two recent studies provide new data on the variability of pharmaceutical rebates rates. Disaggregating this data by classes of drugs demonstrates the weakness of using broad averages to measure rebate-negotiating power. A recent letter from the Congressional Budget Office (CBO) provided an unprecedented breakdown of the often-quoted nineteen percent average rebate rate that Medicaid receives from Pharma.
The Medicaid deal is as follows: in return for being placed on the Medicaid formulary and sheltered from prescription switching and other restrictions, a drug’s manufacturer must rebate the government the greater of 15.1 percent of its average manufacturer’s price (AMP) or the difference between the “best price” given to a private sector reseller and the 15.1 percent minimum.
The CBO letter revealed that thirty-six percent of Medicaid brand drug spend received a “best price” rebate while sixty-four percent of brand drug spend received the minimum. In addition, the CBO letter provided evidence of the value of a little known clause in the Medicaid deal that gave the government inflation protection. The value of that clause amounted to a surprisingly large additional rebate of 11.7 percent of AMP in 2003.
Based on the above data, it is possible to disaggregate the 19.6 percent basic Medicaid average into a 27.6 percent average paid on “best price” drugs and the 15.1 percent minimum paid on the rest.
In September of 2005, the Federal Trade Commission (FTC) released a long awaited study of potential conflicts of interest by independent PBMs. The FTC obtained data on rebate receipts by drug for the “Big Three” independent PBMs—Caremark Rx, Medco Health Solutions, and Express Scripts. These three large PBMs manage about sixty percent of all outpatient prescriptions in the United States.
The FTC study confirmed that Pharma pays rebates only on a small portion of brand drugs. It does not pay on brand drugs with a monopoly position. Nor does Pharma pay on brand drugs in aging therapeutic classes where most of the competing brands have lost patent protection.
The FTC also confirmed that Pharma negotiates brand rebate deals only with PBMs, and not retail drugstore chains like Walgreen and CVS. Size does not matter on the buy-side if an entity does not also have the power to affect the demand for brand drugs through discretion in formulary design and compliance.
On the other hand, it is generic drug manufacturers that negotiate volume discount deals with drugstores because only dispensing pharmacies have the power to choose from an array of suppliers of perfect substitutes.
Unfortunately, the FTC failed, or was prohibited, from disclosing any detail about rebates rates by drug or therapeutic class. However, it is possible to perform a rough disaggregation of the disclosure that the Big Three PBMs received an average of $6.34 in rebates and data fees per on-patent drug in 2003.
The FTC disclosed that data collection fees paid by Pharma amounted to three percent of wholesale acquisition costs. Subtracting out these substantial fees, that some allege are masked rebates exempt from inclusion in Medicaid “best price” calculations, the estimate for the average rebate received per brand is $4.22.
The key to disaggregating this average was another disclosure that seventy-one percent of rebate receipts for each PBM were concentrated in the top twenty-five rebate receiving drugs. Assuming that these highly rebatable brands represent about twenty percent of brand drug prescription volume, it is possible to decompose the $4.22 broad average into rebate components of $14.98 for the top rebate-receiving drugs and $1.53 for the rest. For comparison purposes, Table 1 presents the conversion of rebate levels to rebate rates based on an assumed average wholesale cost (WAC) or AMP. While the Big Three PBMs received on average 8.0 percent of AMP across all brands, they receive 22.7 percent of AMP on rebatable drugs. This is comparable to the 27.6 percent that Medicaid receives as a “best price” rebate. Not too much should be made of differences as the ratio is sensitive to an assumed average for WAC and AMP.
Overall, Medicaid receives more, but this is mostly due to the inflation factor and government’s unique ability to confiscate a 15.1 percent minimum from the all powerful (monopolists) and the all powerless (many competitors).
Disaggregation presents a completely different perspective on the rebate negotiation ability of the Big Three PBMs. This ability is not something that they care to make transparent for strategic reasons. They also show no inclination to challenge the validity of using broad averages as measures of rebate negotiating ability.
There is much speculation over who actually triggers the Medicaid “best price”. Some believe that the Big Three PBMs get the “best price” based on their sheer size and a business model bias that favors abstaining from switching of a rebatable brand drug to lower cost generic that garners no rebates.
Others believe that HMOs with captive PBMs, like Kaiser Permanente, trigger the “best price”. These HMOs know that their members are willing to sacrifice some freedom of choice for competitive premiums. HMOs with captive PBM operations garner rebates through formulary design and compliance that advantages a single drug in a therapeutic class.
In contrast, the Big 3 PBMs tend not to play favorites, but extract rebates from competing manufacturers by promising each that they will abstain from disadvantageous restrictions and prescription switching. This passivity allows other forms of competition like advertising and physician “detailing” to take over.
The Pharmaceutical Rebate Bargain
Price theory has been misapplied to the case of pharmaceutical rebates. Rebates are not volume discounts. They are not evidence of price discrimination based on elasticity of demand. It may be incorrect to apply the analysis of “most favored nation” clauses to the Medicaid “best price” formula.
Rebates are the result of a bargain over the surplus that is created in a market characterized by a bilateral oligopoly. Secrecy and credible threats are central to the negotiation of pharmaceutical rebates.
There are several fundamental insights gained by viewing pharmaceutical rebates through bargaining theory.
First, the welfare implications of this deal are complex. PBMs may not be the populist countervailing force conceptualized by John Kenneth Galbraith. PBMs may be enabling and codependent, rather than countervailing.
The surplus that is being divided up here is not necessarily fixed. It is possible that Pharma and PBMs are tacitly colluding to increase the surplus in an intermediate market at the expense of excluded competitors on the sell side and powerless downstream consumers on the buy side. Evidence that PBMs and the government extract substantial rebates from Pharma is not sufficient proof that their actions improve consumer welfare.
Second, there are two sides to the rebate deal. The question of who is the best rebate negotiator involves an evaluation of what is received relative to what is given up. Private sector PBMs receive less than Medicaid, but they give up less in terms of rights to engage in therapeutic interchange, impose usage restrictions, and affect demand through co-payment differentials.
When the Medicaid deal was struck in 1990, generic drugs were not the competitive threat that they are today. In return for rebates, the government promised to place a brand drug on the Medicaid formulary and exempt it from any prescription switching and usage restrictions such “prior authorization”. In 1993, the deal was modified so that generic substitution—an off-patent brand switched to a generic equivalent—was allowed.
Companies paying Medicaid for formulary placement today are still protected from any type of therapeutic interchange—an on-patent brand switched to less costly generic that is a therapeutic equivalent. In addition, “prior authorization” restrictions cannot be imposed at the national level, but remain a local option that several states have seized as a bargaining chip to extract supplemental rebates from Pharma.
A recent study by the Lewin Group evaluated the performance of different Medicaid plan managers on the basis of overall drug spend delivered, and not narrowly on net unit prices after rebates.
One set of plans—known as Medicaid fee-for-service (FFS)—are eligible for “best price” rebates, but must adhere to Medicaid guidelines limiting therapeutic interchange and usage restriction. The other set of plans—known as Medicaid managed care organization (MCO)—are privately managed by PBMs and not eligible for Medicaid rebates. However, these PBMs have a lot more discretion in formulary design and compliance.
The Lewin Group study found that Medicaid FFS plans received rebates averaging fifteen percentage points higher than Medicaid MCO plans. However, the discretion allowed by PBMs who managed Medicaid MCO plans enabled them to deliver a fifty-nine percent generic dispensing rate—the number of generic prescriptions divided by the sum of all prescriptions —compared to a fifty percent rate found for Medicaid FFS plans.
Even more dramatic was difference in usage rates—number of prescriptions per member per month. PBMs were able to deliver a usage rate that was fifteen to twenty percentage points lower than Medicaid FFS plans.
Overall, the drug spend of privately managed plans was ten to fifteen percentage points lower than government run plans despite lagging in rebates received from Pharma.
The results of this study clearly demonstrate the weakness of evaluating bargaining agents solely on the basis of rebate rates. Based on overall goal of cost containment, this study provides evidence that private sector PBMs do a better job at rebate negotiation than Medicaid.
References and Further Reading
Beronja, Nancy et. al. 2003. Comparison of Medicaid Pharmacy Cost and Usage between Fee-for-Service and Capitated Settings. Resource Paper prepared by the Lewin Group for the Center for Health Care Strategies, Inc.
Congressional Budget Office. 2005. The Rebate Medicaid Receives on Brand-Name Prescription Drugs. Attachment to letter to Senator Charles Grassley dated June 22, 2005.
Federal Trade Commission. 2005. Pharmacy Benefit Managers: Ownership of Mail-Order Pharmacies.
Disclosures: I have not received any remuneration for this paper nor have I financial interest in any company cited in this working paper. I have a Ph.D. in Economics from Washington University in St. Louis and a B.A. in Economics from Amherst College. Other working papers on PBMs can be accessed at www.nu-retail.com.
Carpooling is now seen as last big opportunity to grow a shared mobility as a service (MaaS) business ahead of the arrival of autonomous vehicles (AVs).
We present the case that Waze’s altruistic vision of carpooling is insufficient to scale the business.
Our transactional vision of the business, requiring market pay rates to drivers, creates little incentive for people to choose carpooling over solo commuting.
We think that it will take a minimum of $4,000 in cost saving to motive a significant number of people to go carless. This implies that fares will have be reduced by an additional $1,583 a year to reach that level of cost savings.
The way to recoup this is by negotiating referral credits (dollar or accounting) with related units offering last-mile ride-sharing, delivery, and weekend car rentals.
Inquiries : Lawrence W. Abrams, labrams9@gmail.com, (cell) 831-254-7325
Our Vision of the Modern Carpooling Business
A Horizontally Integrated MaaS Business
The success of ride-hailing apps has given rise to the idea that app-enabled carpooling could be a scalable business. Plus, carpooling at scale could become a much needed poster-child of tech “public good” as it would be the first impactful solution to traffic congestion and automobile pollution in years.
The question is: Why would any company want to enter the carpooling business today? What kinds of driver and passenger incentives would be required to scale this business?
The unexpected early success of autonomous vehicle (AV) R&D has given rise to the idea that automobile ownership will be replaced within a decade by companies offering shared mobility-as-a-service (MaaS).
Given this, why would any company want to enter the driver-centric carpooling business with its limited life expectancy and profit potential?
There have been at least a dozen carpooling startups trying to grow the business since 2010, but none have gained traction and most have closed down. Perhaps their timing was premature as the urban ride-hailing companies like Uber had not yet matured enough to provided acculturation spillover benefits or been in a position to partner with a carpooling company in a tight “mesh-transit” network. (see more on this later)
Today, the only commuter carpooling service with serious financing is Commute by Waze, a division of Alphabet (Google). But, Uber, Lyft, and Ford could enter this business easily by expanding their existing urban ride-sharing services to long-distance commuters.
uberPOOL and Lyft Line are urban shared-ride versions of their on-demand services. Ford’s new acquisition Chariot has recently rolled out an app-enabled, fixed-route vanpool service using Ford Transit vans and full time drivers. Lyft has just introduced a similar urban fixed route car jitney service called Lyft Shuttle.
Waze’s vision of of the modern carpooling business is in the spirit of altruistic carpooling among neighbors and coworkers. Everyone takes turns driving and chip in to cover out-of-pocket expenses if there is an imbalance.
Consistent with this vision, Waze has capped driver pay rates at the business mileage reimbursement rate of $.54 per-mile. It has set per-ride fares that, when aggregated, just cover driver pay. It also limits drivers to two-a-day rides, eliminating the possibility of full-time work.
Despite low pay, driver commitment is high because drivers have an altruistic “gift relationship” with passengers rather than a “transactional relationship.” (more on this later)
While Waze’s vision for modern carpooling is laudable, we will argue that scaling the business will require a business-like transactional approach, starting with driver pay rates on par with Uber.
We envision the modern carpooling business as a unit of horizontally integrated MaaS company that also offers ride-hailing and has third-party tie-ins with ecommerce delivery companies like Amazon and Walmart and rental car companies like Hertz and Avis.
The carpooling unit would be credited for dropping off passengers at transit hubs for last-mile ride-shares. It would be credited for delivery of meals, groceries, etc. It would be credited by rental car companies when carless carpoolers come in for cheap rentals on the weekends.
While never profitable on its own, the carpooling unit would be building brand-awareness and customer loyalty. It would be an important contributor to positioning for the biggest business opportunity of a lifetime — the AV MaaS business.
The carpooling unit would be accumulating valuable MaaS logistics data. It would play an important role in the acculturation of commuters to the shared-ride lifestyle much like AirBnB and WeWork are doing for the shared-living and the shared-work lifestyle, respectively.
More MaaS synergies originate from carpooling than any other mobility service. And those synergies magnify when carpoolers go carless.
The fundamental strategy of an integrated MaaS company today should be to reduce carpool fares to the point that passengers will go carless and unleash a demand for related services. Good accounting practices dictate that the carpooling unit get credit for these spillover benefits.
We think that it will take a minimum of $4,000 in cost saving to motivate a significant number of people to go carless. This implies that fares will have be reduced by an additional $1,583 a year to reach that level of cost savings.
The way to recoup fare reductions would be to negotiate referral credits (dollar or accounting) with related units offering last-mile ride-sharing, delivery, and weekend car rentals.
Next, we present a brief look at the carpooling business from the vantage point of specific companies — Lyft, Uber, Ford, GM, Google, and even Amazon, Walmart, Hertz and Avis.
Besides spillover benefits to related companies, carpooling at scale will have a significant impact on traffic congestion and automobile emissions. These “public goods” justifies government support. We discuss the merits of a few ways government can help scale the carpooling business with minimal expenditures.
A Sense of Carpooling at Scale
By quantifying “carpooling at scale”, we will show why Waze’s altruistic vision of carpooling with driver pay set at $ .54 / mile is insufficient to scale the business.
For this exercise, we chose California Highway 101, a.k.a. “The Bayshore Freeway” between San Jose and San Francisco (SF).
The Reverse Commute along Highway 101 — aka the Bayshore Freeway
There are number of reasons why Highway 101 would be good starting place to scale a carpooling business:
Significant carless commuters in SF
Significant reverse commute from SF to Peninsula
Ride-hailing at scale that facilitates a “mesh-transit” system
California highway not US Interstate
Home of Waze, Chariot, Uber, Lyft, Google, and Ford Smart Mobility
The question is how many carpool drivers would be needed to reduce the rush hour traffic along Highway 101 by, say 30%? How would that estimate compare with the number of Uber and Lyft drivers now working in SF?
Based on 2015 CalTrans data of vehicle traffic flow, we estimate that there are approximately 150,000 vehicles flowing both ways past a mid-peninsula point along Highway 101 (at Highway 92) during a typical weekday commute period of four hours (5-9 AM or 3-7 PM).
We derive the following table of driver requirements:
Driver Requirements for Scaling Carpooling Along Highway 101
We conclude that it would take one-third the scale of Uber’s and Lyft’s combined operation in San Francisco for a carpool service to impact commute congestion along Highway 101, assuming an average of 3 passengers per carpool. And this is just one highway in the Bay Area.
Attracting 15,000 new drivers would be a huge undertaking. But, if the business could use existing Uber and Lyft drivers during peak commute hours and allow them to do multiple commute loops, the task becomes much more manageable.
Reverse Commuters as Early Adopters of Carpooling
Highway 101 is especially attractive as a place to start a carpooling business because of a strong city-to-suburb reverse commute.
The two other areas with strong reverse commutes are in Washington, D.C area with reverse commutes to government complexes in suburban Maryland and Virginia and along Santa Monica Freeway from downtown Los Angeles to coastal Santa Monica.
There are several reasons why highways with reverse commutes should help. It may be that these corridors should be the only places targeted, given the limited lifespan and profit potential of the business.
Reverse commuters are city dwellers who do not need a car for running errands, going out to eat or seeing a show at night. Parking is expensive. They already sense tremendous value and little added inconvenience by going carless. They are primed to be early adopters of a well-run carpooling service.
Corridors with strong reverse commutes also are attractive to the carpooling business because companies can offer drivers full time work via a combination of multiple carpooling loops mixed with periods of ride-hailing work.
Finding other metropolitan areas with strong reverse commuting would be a high priority research project for any carpooling company.
The Rationale for Market Rate Pay for Carpool Drivers
Carpool drivers have to be on-time twice a day, five days a week, 250 days a year. After all, failure could cost passengers their jobs. Work-going carpoolers are “on-the-clock”. Bar-hopping ride-hailers are not.
The management of a carpool business has to demand a greater level commitment out of its drivers than Uber and Lyft now demand of their drivers. As independent contractors, Uber and Lyft drivers have a great deal of latitude in choosing work hours and routes.
Driver commitment isn’t an issue in Waze’s altruistic vision of the carpooling business because a driver is driving for neighbors and co-workers. The desire for continued respect is the prime motivator. Waze’s choice of limiting driver pay to the $.54 / mile is consistent with this vision.
But, we believe that carpooling at scale has to involve a vast majority of drivers working for strangers, not neighbors and co-workers. It is a transactional business where driver commitment is secured by market rates of pay and the threat of being fired.
Recognizing that performance is affected significantly by the type of relationship a driver has with his passengers is similar to what Richard Titmuss discovered in blood-giving as chronicled in The Gift Relationship, a social science classic.
Basically, Titmuss found that the quality of blood was much better when it was give freely by altruistic donors than when it was given in exchange for pay.
As a result, we firmly believe that a carpooling business has to pay drivers equal to what an Uber driver gets per-mile.
An Estimate of Driver Pay
Below is an estimate of an Uber fare and related driver pay rate on a per-mile basis for a 25 mile uberX ride from Redwood City to San Francisco taking 50 minutes during rush hour. We use 80/20 as an estimate of Uber’s current driver/company distribution ratio.
Estimate of Uber Fare and Driver Share for Typical Rush Hour Commute
We believe that a transactional carpool business has to match Uber all around in terms of gross fare rate of $1.55 / mile and driver pay at $1.18 / mile, which is set at 80% of gross fare less booking fees.
The idea of uniform fare rates and driver share across all mobility services is consistent with our vision of an horizontally integrated MaaS company and a “mesh-transit” system that seamlessly integrates carpooling with last-mile ride-sharing.
An Estimate of Passenger Cost-Saving Over Solo Commuting
We next estimate a passenger fare assuming Uber’s fare of $1.55 / mile shared by 3 passengers. We also derive the cost savings for switching from a solo commute to carpooling.
Estimate of Carpool Passenger Fare and Cost Saving Over Solo Commuting
Even with 3 passengers who share the fare, carpooling yields only a $2,417 cost saving over solo commuting. Even if passengers did commit to carpooling, we do not believe that this cost-saving would be enough incentive to “cut the cord” of car ownership and go carless.
Motivating Carpoolers to Go Carless
We were not surprised to see a lack of passenger incentive to choose carpooling over solo commuting assuming market rate pay for drivers.
Our initial thought was that government-mandated congestion pricing would be the only way the carpooling business could scale. Congestion pricing would force the cost of solo commuting even higher than the already high cost of carpooling.
We now envision carpooling as a unit of horizontally integrated mobility company. The business scales via reduced fares. These reductions are recouped by referral programs that offer credits and rewards coupons redeemable by passengers for using ride-hailing, delivery, and rental car services of related units.
In economic terms, the business scales via its “own elasticity of demand” through reduced fare prices rather than via its “cross-elasticity of demand” through raising the price of substitutes via congestion pricing.
The fundamental strategy of an integrated MaaS company today should be to reduce carpool fares to the point that passengers will go carless and unleash a demand for related services.
We think that it will take a minimum of $4,000 in cost saving to motivate a significant number of people to go carless. This implies that fares will have be reduced by an additional $1,583 a year to reach that level of cost savings.
The way to execute this strategy would be to build app payment algorithms that posts dollar credits and rewards coupons to passenger accounts that are redeemable at related companies.
The dollar value of these credits and coupons are set at the discretion of the related companies. Separately, the carpooling company negotiates payments with related companies for setting all of this up. Payments would accrue as these credits and coupons are redeemed.
In the case of tie-ins with third-party delivery and car rental companies, the carpooling company receives cash. If the carpooling business and the ride-hailing business are owned jointly, say in the case of Uber or Lyft, the carpooling unit earns accounting credits offset by debits to an intercompany clearing account.
Below is an illustration of a series of credits earned from referral programs that would recoup a $1,583 per passenger fare reduction. The distribution of the credits among the related companies is based mostly on a qualitative ordering of “spillover benefits” generated by carpoolers.
We believe that a ride-hailing partner would get the most benefit by far. The expected benefit values to delivery and rental cars companies are about equal, but far behind.
Recouping Reduced Carpool Fares
Surge Pricing Would Kill the Carpooling Business
The ride-hailing business is “on-demand” with no set commitments made by drivers or passengers. Peak-load pricing, or surge pricing, is used to balance out supply and demand.
Commuter carpooling is not “on demand”. Passengers rely on the service to get to and from work. They risk firing if late. The business depends critically on gaining customer confidence through reliability and predictability. This can be achieved by paying drivers market rates and require them to meet precise pick-up times. Surge pricing would kill the carpooling business.
Given the level of commitment by drivers, it would be reasonable to ask customers also to make weekly or monthly commitments in return for a set fare rate.
Why Would A Company Enter the Carpooling Business Today?
The success of ride-hailing apps like Uber and Lyft plus the unexpected early pace of autonomous vehicle (AV) research and development has given rise to the idea that shared mobility-as-a-service (MaaS) may be here sooner than later.
Most agree that so-called Level 4 AVs — no steering wheel or accelerator, but location-constrained — might start appearing by 2021. But, there is widespread disagreement as to when the ultimate Level 5 AVs (hereafter just AVs) will appear.
Also, there is widespread disagreement as the length of time it will take to scale AV production. For example, there are a number of optimistic predictions that mobility-as-a-service (MaaS) using AVs will start appearing around 2020 or 2021.
On the other hand, The Alliance of Automobile Manufactures, a trade group that represents Ford, General Motors, Fiat-Chrysler, BMW and more, has estimated that AVs won’t be available for sale before 2025 and it might take another three decades until 2055 when AVs represent a majority of vehicles in use.
Our view splits the difference between these two extremes. Namely, we start with the view that AVs first appear in a decade, say around 2027, with another three year to congestion-ending scale by 2030.
Given the driver-centric carpooling business has a short life expectancy and limited profit-potential, why would a company want to enter the carpool business in 2017?
Traditional Automobile Companies
We think that the fundamental reason for entering the carpooling business today is to establish a consumer-facing MaaS brand ahead of the biggest business opportunity of a lifetime — the AV MaaS business.
The only companies that NEED to enter this business are traditional automobile companies. Executives in the automobile industry knows that MaaS and AV are existential threats as they could end their 110+ year history as consumer-facing brand. Auto companies fear becoming the “Intel Inside” of the MaaS business.
We expect that both GM and Ford will seize this opportunity with the goal of scaling the carpooling business over the next decade. They are also in the unique position of subsidizing this business by using their own vehicles.
To have any success at scaling they business, they will have to partner either with Uber or Lyft to share drivers and mesh their branded carpooling with last-mile ride-sharing services offered by Uber or Lyft.
A Diagram of a “Mesh Transit” Sytem
Ford has entered the ride-sharing business by acquiring Chariot. Chariot is a modern day urban jitney service using 15 seater Ford Transit vans and full time drivers. It has plans to expand to eight cities in 2017.
The picture below illustrates what we mean by using carpooling to establish a consumer-face MaaS brand. Ford is much further along than GM in establishing a MaaS brand. It has brought all of its AV and MaaS efforts under one division located in Silicon Valley called “Ford Smart Mobility.” Ford also has promoted the head of this division Jim Hackett to CEO of the whole company, a huge indicator of Ford’s priorities.
Example of MaaS Branding Ahead of AV Era
GM has just begun to roll out a niche MaaS service called Maven, an hourly car-rental service. GM is way ahead of Ford in partnering with a ride-hailing company as it has a 9% stake in Lyft.
We expect GM to enter the commute carpooling business shortly with its own consumer-facing brand and partner with Lyft. But, Lyft and GM are “frenemies”. Both want be a consumer-facing MaaS brand. Lyft might consent to a carpooling service branded as “GM Mobility powered by Lyft”. The only question is whether Lyft will enter the business with its own brand similar to Lyft Line or Lyft Shuttle.
Ride-Hailing Companies
Uber and Lyft have established MaaS brands at great cost over the last 7 years. Their existential threat is from the AV supply side not from the branding side. Uber and Lyft might enter the carpool business, but they don’t need to. On the other hand, Ford and GM need to partner with Lyft or Uber as a source of shared-drivers and to mesh a carpooling service with last-mile ride-sharing service.
Google
Alphabet (Google) has chosen to enter the carpooling business via its Waze Division’s Carpool service. As we have mentioned earlier, Waze’s altruistic vision for the carpooling business doesn’t scale.
Google is probably in a better position for the coming era of AV MaaS than any other company on the planet. It’s Waymo division has a 5 year lead on AV R&D. With their Waze and Google Maps real time traffic monitoring apps, Google has established a brand awareness with commuters that is second to none.
But, Google is on the verge of applying the same muddled strategy to the carpooling business as it did with Android and the smartphone business.
Without an existential crisis to focus its thinking, it seems that Google is about to compete with itself once again. Namely, Goggle sold their own Nexus brand of smartphones. At the same time, they licensed Android to countless Asian manufacturers who turned around and competed with Nexus.
It is not clear what Google’s ultimate goal is. Does it want to become a consumer-facing MaaS brand with Waze taking the point? Or does it want to become an “Intel Inside” AV OEM to automobile brands like Chrysler-Fiat and a host of European and Japanese auto companies?
Delivery Services
Uber has begun to capitalize on the synergies between its ride-hailing business and the delivery of food, groceries, and other goods. These synergies would be even greater in the carpooling business.
Look for Amazon and Walmart to seek tie-ins with carpooling companies. This could include partial financing of transit hubs where ride-hailing, carpooling, and e-commerce delivery services meet and re-distribute people and goods. Commuter favorites like Starbucks and McDonalds might also want to lease space there.
Car Rental Companies
The stock prices of Hertz and Avis shot up by double digit percentages recently when Apple and Waymo announced that they had contracted with these two car rental companies to maintain their fleet of prototype AVs.
Suddenly, there was a recognition by investors that car rental companies might not be wiped out when the era of AV MaaS arrives.
We can see another reason why Hertz and Avis might want develop an association with a carpooling company. Carless carpoolers have a need to rent a traditional car for weekends and vacations where getting a car “on-demand” just isn’t good enough.
It is a natural fit for rental car companies as most of their cars are used for business purposes and sit idle on weekends. Indeed, they currently offer such steep discounts for weekend rental that we have observed their offices jammed on Friday afternoon with carless families jumping at the chance to get a cheap rental for the weekend.
Example of MaaS Tie-in
The Rationale for Public Support
If the carpooling business could scale, it would provide significant “public goods” via reduced traffic congestion and reduced automobile pollution.
This would justify public support via congestion pricing, increasing the minimum requirement to use the HOV lane, and building transfer hubs where carpools and last-mile ride-shares could redistribute passengers.
The unexpected early success of AV R&D has given rise to the idea that automobile ownership will be replaced by MaaS within a decade. This realization will actually make congestion and pollution worse in the meantime.
The reason why is that AV forecasts are starting to be used to persuade government authorities rightfully so to kill off plans for expensive, long lasting infrastructure projects like new highway lanes, light rail extensions, and bus terminals. The only positive environmental benefit of AV hype would be if it was used to kill off plans for new city-center parking structures.
Our initial thought was that government-mandated congestion pricing would be the only way the carpooling business could scale. Congestion pricing would force the cost of solo commuting ever higher than the high cost of carpooling.
Now, we see congestion pricing as a first option primarily in Asia and Europe. At one time, the technology necessary to implement congestion pricing was crude. But now, real-time pricing is possible via “connected cars” and real-time cloud-based pricing platforms using an architecture similar to MZ’s (formerly Machine Zone) RTplatform.
In the United States, we now view congestion pricing as a “doomsday” solution to be deployed a decade from now in the event that AVs show little promise in solving the congestion problem.
And, given that there are about 263 Million passengers vehicles registered in the United States, with about 17 million vehicles sold a year, it might take another ten years, or until 2037, until AV carpooling has scaled enough to end congestion.
In the meantime, scaling the carpooling business is one of the best options we have for reducing traffic congestion and automobile pollution before the era of AVs. And, support for carpooling won’t cost the government trillions of tax dollars. It may just take a boost in the HOV lane minimum from 2+ to 3+, which the State of California is considering for Highway 101 . If a carpooling company could show some success on its own in reducing congestion along 101, this could accelerate the State’s own plans to improve management of HOV lanes.
The order book is a market design for the exchange of goods and assets. It dates back to the European coffee houses of the late 1600s. In London, Jonathan’s Coffee House was a significant meeting place for traders in London in the 1700s. It later became the site of the first London Stock Exchange.
In the late 1700s, in what later became known as New York City, Dutch traders met at a Buttonwood tree in lower Manhattan island to buy and sell goods coming into the port. Now know as Wall Street, this location became the center of financial asset exchange in the United States.
Until the 1970s, stock exchanges were characterized by a market design involving traders gathering around pits with specialists manually matching bids and asks in paper order books (see below).
The great financial economist Fisher Black wrote a prophetic article in 1971 called “Toward A Fully Automatic Stock Exchange” where he laid out the implications of the coming automation of the manual order book. He speculated on what the computerization of the order book would mean for bidding mechanisms, liquidity and overall stock market efficiency.
Screenshot of Bid-Ask Order Book of Poloniex
Market design, indeed all design related to computers, is coupled tightly to the computer technology itself. Just because one design is associated with a particular computer technology does not mean that the same design should be mindlessly carried over when the computer technology changes.
We recall the mindless carry over of the 80 character line limit established by IBM for punch cards in the 1920s to cathode ray tube (CRT) terminals in the 1970s.
There is a whole host of other instances of mindless carry over of designs when the technology changes. One notable example is the organization of the factory floor after the conversion of machine power from a centralized shaft driven by water to decentralized electric power.
In the last several years, there has emerged a new decentralized, peer-to-peer (p2p) paradigm in computer architecture propelled by several trends — Internet of Things (IoT), autonomous vehicle-to-device (v2x) communication, and crypto.
This change demands a rethinking of the appropriateness of the centralized client-server order book market design in a crypto-economic platform.
The trend away from client server architecture is driven by a need to do more raw compute “at the edge” before sending data to the server for storage and higher order analytics. This is known as “edge computing.” The use cases for edge computing are Internet of Things (IoT) and autonomous vehicle-to-device (V2X) communication.
The trend away from client server architecture is also driven by the tremendous interest in Bitcoin, Blockchain and Ethereum. Interest in crypto could be the start of a paradigm shift away client server financial intermediaries earning opaque rents and toward decentralized, trustless p2p protocols for validating and accounting for the exchange of financial assets.
A true true, decentralized crypto-economy involves not only a DLT layer but also high speed transaction layer.
The thesis of this paper is that the time is now to consider a transaction layer with a true decentralized market design.
We believe that publish-subscribe will be the leading protocol of the transaction layer as it has already been deployed at scale an the platform behind several MMO games (from MZ) and chat platforms (WhatApp from Facebook, WeChat from TenCent).
What is needed is an innovative p2p market design. It could be along the lines a many-to-many, high frequency “take it or leave it” (TIOLI) publish-subscribe mechanism which could also be described as a discrete time, many-to-many, high frequency version of the Myerson auction.
Value Proposition:
user-defined contracts ( e.g. options with odd expiration dates, long-short pairs, straddles)
tokens earned by peers supplying liquidity spread contracts
elimination of latency rents going to HFT and server co-location fees going to exchange
elimination of “data ownership” rents earned by exchange
Specification suggestions:
high frequency, many-to-many, pub-sub protocol
messages in form of Myerson “take it or leave it” (TIOLI) bid-asks
“serverless” with ephemeral matching with-in Redis-like in-memory data structure store, used as a database, cache and message broker.
ephemeral bid-ask data, only data “owned” is history of matches.
Gabe Leydon, CEO Satori (MZ) TokenPost Interview During Korea Blockchain Open Forum, July, 2018 https://www.youtube.com/watch?
Satori’s “AI Mesh network” transaction layer stats — 500 Million “messages” per second or 1 million publishers sending 100 bytes a second
Hadera Hashgraph’s DLT stats — 500,000 transactions per second with 100% consensus based on a “gossip of gossip protocol” and a consensus latency of 3.5 seconds.
Eric Budish, The Design of Financial Exchange, Some Open Questions at the Intersection of Econ and CS. Simons Institute of Computing UCB, November 2015 https://www.youtube.com/watch?v=Rilv2AJ1TWM
Exchanges are regulated by the SEC and take years to gain approval. Recently, the SEC has announced that all crypto exchanges are illegal unless they register with SEC.
There are two key design principles informing a market design presented below for a crypto-economy platform involving the exchange of digital assets including cryptocurrency deemed securities by SEC.
Eliminate enormous multi-million dollar rents captured by exchange intermediaries and front running HFT
Accept regulation by the SEC, but as an Electronic Communications Network (ECN) not an exchange.
Traditional currency or asset exchange involve a two-sided auction market design better known as an order book. Currently, all crypto exchanges whether custodial, so-called “decentralized” exchanges (DEX), or relays with 0x smart contracts , still feature order books as a market design.
What we propose is a market design where “the network is the exchange”. We strongly believe that this design would allow for registration with the SEC as a broker-dealer running an Electronic Communication Network (ECN) which is a subset of a Alternative Trading System (ATS) . Getting approval for an ECN would be must faster than getting approval as an exchange.
Value Proposition:
user-defined contracts ( e.g. options with odd expiration dates, long-short pairs, straddles)
tokens earned by peers supplying liquidity spread contracts
elimination of latency rents going to HFT and server co-location fees going to exchange
elimination of “data ownership” rents earned by exchange
Specification suggestions:
high frequency, many-to-many, pub-sub protocol
messages in form of Myerson “take it or leave it” (TIOLI) bid-asks
“serverless” with ephemeral matching with-in Redis-like in-memory data structure store, used as a database, cache and message broker.
ephemeral bid-ask data, only data “owned” is history of matches.
Satori’s “AI Mesh network” transaction layer stats — 500 Million “messages” per second or 1 million publishers sending 100 bytes a second
Hedera Hashgraph’s DLT stats— 500,000 transactions per second with less than a second to 100% consensus based on a “gossip of gossip protocol”
Some URLs relevant to stock and asset market design choices:
Eric Budish, The Design of Financial Exchange, Some Open Questions at the Intersection of Econ and CS. Simons Institute of Computing UCB, November 2015 https://www.youtube.com/watch?v=Rilv2AJ1TWM
