The Fragility of Optimized Systems

Delta Air Lines designed and optimized a system to pack the seats on their flights and extract maximum revenue from passengers by charging for every element of an airline trip. The process is called revenue maximization. A senior Delta executive once told me, “We try to pull feathers until just before the goose honks.”

Delta’s system becomes very fragile when off-normal situations arise, potentially causing enormous grief for thousands of passengers. The optimized system has no slack, no resources for responding to off-normal situations.  If there were slack resources, Delta would have eliminated them to increase profits, the only metric that really matters to them.

Earlier this week, weather in Atlanta delayed flights, in my case returning from Los Angeles en route to Washington.  Delays added to delays.  Passengers just waited and waited. Upon landing, hours were consumed waiting for a gate. Whole days were lost, according to the Atlanta Journal-Constitution, due to Delta’s gross incompetence.

“Delta apologizes for your inconvenience. We appreciate your patience.”  This is Delta’s tag line. I hear it on every flight for one reason or another.  However, Delta talks service but does not deliver it. They are too focused on maximizing revenue to pay attention to service. Passengers become incensed and Delta just repeats their tag line.

As my flight was delayed by progressive half hours, a passenger asked the agent why the delay was always 30 minutes. She responded, “That is all we are allowed to tell you.” The passenger asked, “So when is this flight really likely to leave?”  She answered, “We have absolutely no idea. This could go on for several hours and then we may cancel the flight.”

It is rather amazing that many of my worst experiences in life are associated with Delta. This has fostered visceral negative emotions with anything I see or hear about Delta. There are very few, if any, other things that can prompt such negative emotional responses.

What happened to air travel?  When I first traveled with Delta in the 1980s, it was a quite enjoyable experience.  I fondly remember the ice cream sundaes on my frequent Atlanta-California trips.  I also remember my worldwide adventures, typically ending in London when I would relax with Delta who would “take me home.”

My million miler colleagues agree that those times are long gone.  Delta is now an adversary who is focused on extracting revenue from you.  Quality of service is now a joke, glittery words that mean absolutely nothing.  Delta and other airlines are totally focused on what we can do for them.

Beyond the Affordable Care Act

What are we trying to do by rethinking the ACA?  Perhaps we are seeking an ideologically acceptable ACA, one that the Republicans get credit for rather than the Democrats.   On the other hand, is insurance coverage really the ultimate goal?  I don’t think so.  We want a healthy and educated population that is competitive in the global marketplace.

Health insurance is one way to enable this, but so is a single payer system, e.g., Medicare for everyone.  Health insurance is our apparently preferred choice because that is the incumbent mechanism for paying for healthcare in the US.  A broader question is how do we pay for health, education, and social services that enable a healthy and educated population that is competitive in the global marketplace.

At one extreme, the answer could be that everyone pays for these services themselves.  In other words, everything is privatized and we have the world of Charles Dickens.  Of course, lots of people would not be able to afford health and education, so we would have a large class of people stuck in unhealthy, uneducated poverty.  They would eventually revolt, not just at ballot box, but also by physically attacking the privileged – this has happened many times before (Brook, 2013).

Despite the rhetoric of those who argue for a totally market-based approach, most people accept that governments – federal, state, and local – have to play some role in creating a more equitable situation.  And, at the other extreme, few argue that government should provide all health, education, and social services.  For example, there have been few, if any, calls for nationalizing private institutions of higher education, despite the financial bubble that has emerged in this arena (Rouse, 2016).

So either extreme has a rather limited constituency.  The answer is somewhere in the middle – a public-private “enterprise” that enables a healthy and educated population that is competitive in the global marketplace.  By enterprise, I don’t mean a single organizational entity.  I invoke the enterprise concept to motivate the need to look at the whole system that provides health, education, and social services.

Before suggesting how to find the middle ground needed, two phenomena should be explicitly addressed.  First, people both consume and generate resources, which includes money as well as food, housing, etc.  Health, education, and social services cost money, but people who are healthy, educated, and productive generate money and other resources, a portion of which is taxed by federal, state, and local governments to pay for services.

The Centers for Medicare and Medicaid Services (CMS) work to control the costs of health services.  CMS does not tradeoff these costs versus the resources a healthy population generates.  For example, they have no mechanism to incentivize employers to invest in prevention and wellness programs that will result in healthier employees when they retire and enter the Medicare program, despite the fact that this impact has been repeatedly proven (Rouse & Serban, 2014).

The second phenomenon was identified in an analysis of how value can be estimated for investments in people’s training and education, safety and health, and work productivity (Rouse, 2010).  When the entity investing receives the subsequent returns on these investments, it is often rather straightforward to make the economic case compelling.  However, when the entity investing does not receive the returns, they tend to see the expenditures as costs and try to minimize them.

To deal with the above two phenomena, we need to address the overall enterprise that provides health, education, and social services.  How are delivery, payment, and regulation accomplished in each of these service domains?  The high level of fragmentation in the US across federal, state, and local governments has resulted in a large numbers of silos that see monies spent as costs rather than investments.

This fragmentation affects both the effectiveness and efficiency of these services.  We can use computational simulations with interactive visualizations (Rouse, 2015) to explore ways of breaking down the silos of delivery, payment, and regulation.  I propose that the interactive environment be immersive and enable key stakeholders, which includes almost everybody, to explore the complexity of the overall enterprise.  Ideas such as the sharing of returns on investments – across silos — can be investigated.

Creation of such an environment will require many types of data related to the efficacy and costs of health, education, and social services.  Fortunately, the increasing emphasis on evidence-based policy (Haskins & Margolis, 2015) should provide support for the needed data analytics.  This should, in turn, enable investments in achieving the overarching goal of a healthy and educated population that is competitive in the global marketplace.


Brook, D. (2013). A History of Future Cities. New York: Norton.

Haskins, R., & Margolis, G. (2015). Show Me the Evidence: Obama’s Fight for Rigor and Results in Social Policy. Washington, DC: Brookings.

Rouse, W.B. (Ed.).(2010). The Economics of Human Systems Integration: Valuation of Investments in People’s Training and Education, Safety and Health, and Work Productivity. New York: Wiley.

Rouse, W.B. (2015). Modeling and Visualization of Complex Systems and Enterprises: Explorations of Physical, Human, Economic, and Social Phenomena. Hoboken, NJ: John Wiley.

Rouse, W.B. (2016). Universities as Complex Enterprises: How Academia Works, Why It Works These Ways, and Where The University Enterprise Is Headed. Hoboken, NJ: Wiley.

Rouse, W.B., & Serban, N. (2014). Understanding and Managing the Complexity of Healthcare. Cambridge, MA: MIT Press.

Test Driving MOOCs

I have been researching Massive Open Online Courses (MOOCs), compiling best practices and other good ideas that I sought from a variety of colleagues.  I recently completed the first lessons of three courses on the best-known MOOC sites:

  • Coursera course: “Chicken Behavior & Welfare”
  • edX course: “Dinosaur Ecosystems”
  • Udacity course: “Design of Everyday Things”

All three courses provide lessons composed of a series of 1-3 minute video clips, interspersed with short exercises and a multiple choice quiz at the end.  All three have forums where students can interact with the instructor(s) and other students.  All three are reasonably engaging.

Forums are like streams of emails or texts, rather than real interactions.  It may be that younger students find this quite acceptable as their daily lives are laced with these forms of communications.  Something like a multi-person Skype might feel more personal, although that would be cumbersome for some courses where large numbers of students are enrolled.  I suppose Skype could be limited to student-teacher interactions, although instructors with hundreds or even thousands of students could be totally overwhelmed.

Ashok Goel of Georgia Tech, working with IBM, created an AI teaching assistant, Jill Watson, to field the 10,000 student questions his MOOC receives each semester.  Obviously, there is much repetition in these questions, which greatly enhances the feasibility of this approach.  Students responded quite positively to Jill, not imagining she was other than human.  This kind of automation has been used in industry for some time to respond to customers’ questions about services being provided.

The production quality of the short videos varies greatly, some being very professional and others looking a bit like home movies.  Some are easy to consume, while others provide enormous detail.  I suppose I could have taken notes, but I would have had to repeatedly stop and restart the videos.  Navigation in each of the three MOOCs can be a bit confusing, but I expect one will quickly get over this.

My sense is that highly polished, well-done MOOCs will increasingly succeed.  Simply posting PowerPoint slides online, with recorded audio lectures, is not engaging, and will eventually disappear.   Such stale offerings do not leverage the engagement potential of online technologies.  Greater engagement can compensate for some of the limitations noted above.

An important hurdle that must be surmounted to succeed is the cost of highly polished, well-done MOOCs.  One very credible estimate is 1,000 hours of design and development time per course.  Those that can make such investments will attract thousands of online students.  Once the credentials associated with success in these online courses are acceptable to employers, it is easy to imagine a massive shift away from traditional classrooms.

Everyone will take the course on any particular topic from the very best instructor of that topic.  For example, everyone will take physics from Richard Feynman and economics from Paul Samuelson.  The fact that these luminaries are no longer with us will not be a hindrance.  Technology will enable them to teach new developments in their fields, despite never having heard of them during their lives.

Appearing In and Winning the Super Bowl

There have been 50 Super Bowls (SB). There have been 100 starting quarterbacks (QB). 62 of the 100 QBs have started more than one SB.  This 62 includes 20 individual QBs.  36 of the 62 QBs won the SB, a 58% winning percentage. 38 (100 – 62) QBs have started only one SB.  14 of these 38 QBs won the SB, a 37% winning percentage.

32 NFL teams times 50 years yields 1600 starting QB years. 4.4 years is the average career length of a QB.  Thus, there have been roughly 364 (1600/4.4) starting QBs.  58 (20 + 38) QBs have started a SB, yielding a 15.9% (58/364) starting percentage and a 9.3% (34/364) winning percentage.

32 NFL teams times 53 players per team times 50 years yields 84,800 player years.  3.3 years is the average career length of an NFL player. Thus, there have been roughly 25,697 (84,800/3.3) players.  There have been 5300 (53 x 2 x 50) players in one or more SB.  Thus, there is a 20.6% (5300/25697) appearance percentage.  However, this estimate is too high for the following reason.

For QBs, multiple appearances are known and included in calculations.  For players in general, this data is not readily available.  If this phenomenon were ignored for QBs, the appearance percentage would be 0.275 (100/364), an overestimate by a factor of 1.73 (.275/.159).  Adjusting the players in general percentage by this factor yields an 11.9% (.206/1.73) appearance percentage.

Professional Relationships

The wonders of the Internet and social media seem to have radically changed the nature of relationships.  This is perhaps most apparent in personal relationships where email, texting, Facebook, Twitter, and other offerings provide constant updates on what a vast network of family and friends are doing and thinking at the moment.  Many people spend a significant portion of their time generating and responding to this flow of messages.  I find this particularly disconcerting when teaching class and several students in the front row never look up from their devices, their fingers endlessly tapping on their smart screens.

As amazing as this all is, in this post I address professional relationships and how information technology has morphed the ways in which individuals seek opportunities, secure positions, and perform once in these positions.  My daughter pursued jobs a few years ago and my son more recently.  They prompted the observations that follow.  It struck me that the value of any advice I could offer was substantially offset by the fact that I have never applied for a job in the sense that this act is now construed.

This is due to the simple fact that all my opportunities over 50+ years have started with relationships, not websites, electronic documents, etc.   During my senior year in college, I interviewed with several companies on campus and took trips to GE (railroad engines), IBM (computers), Pratt & Whitney (aircraft engines), Raytheon (submarine systems), and US Steel (railway cars), and received offers from all five companies. I applied for graduate school at MIT, RPI, and URI and was accepted by all three universities.  Thus, I had eight opportunities that were linked to people I had met and talked with along the way.  Of course, in those days, there was no other way to do this.

Once I finished my PhD at MIT, I took a visiting position at Tufts University on the advice of my advisor.  I then pursued a single alternative, the University of Illinois at Urbana-Champaign, after communicating with the department head.  I spent a year at Delft University of Technology, invited by a colleague with common interests.  The school chair at Georgia Tech contacted me, convinced me that I would find a visit interesting, and subsequently made an offer too good to refuse.  I called the dean at Stevens informing him that I would soon retire from Georgia Tech and I was offered a chaired position there within a few weeks.  I am currently considering a few alternatives that have emerged from a range of professional relationships.

My first company, Search Technology, was founded with a single customer where a former graduate student worked.  The company grew via major contracts with companies where colleagues worked.  The next company, Enterprise Support Systems, emerged when Search Technology customers asked for products and services that were not feasible within the older company’s costs structure.  Specifically, a company used to multi-million dollar contracts can find it quite difficult to create software products selling for $1,000 per copy.  Enterprise Support Systems grew by selling to other divisions of existing customers.  Eighty percent of revenues came from twenty Fortune 500 companies.

One or more of these companies served as lead customers for each new software tool.  They would buy a corporate license, at a big discount, for a product that did not yet exist.  Their users became members of the design and development team, assuring that they were pleased with how well the product met their needs.  Our close relationships enabled their trusting us and investing in yet-to-be-defined solutions.  A side benefit was that no other companies were able to bid against us.

What is different now?  One of my PhD students recently applied for almost 40 faculty positions.  He did this online.  The web-based system immediately requested that I provide a reference letter, with a two-week deadline.  Tailoring the letters a bit to each institution, I managed to almost meet the deadline for the 40 letters.

He will not end up interviewing with even half of these institutions.  However, it was convenient for them – not for me – to request these letters just in case they later needed them.  There were no humans involved – just me and a website.  There were no business relationships – just an IT system executing a workflow.

Lots of things work this way now.  When you submit articles to professional journals, the interactions are all automated.  A year or so ago, I submitted an article and was requested to choose key words from a fixed set.  None of the words matched the journal’s topical areas.

I emailed the editor, asking how to respond.  He said that I had to choose among the key words provided.  The vendor of the platform did not allow changing the choices.  It was rather difficult to map my article on healthcare to reinforced concrete and welding.  It certainly did feel that I was serving the platform rather than it providing services to me.

It used to be that people in your organization were experts in benefits, contracts, purchasing, etc.  You knew them and could call on them for help as needed.  Now there are IT systems, often multiple IT systems that require multiple entries of the same user names and passwords to access a single function.  If, for example, you want to change your mailing address, you have to do this in each system because databases are not integrated.  You need to keep track of what each system knows.

As machine learning increasingly becomes the underpinnings of such systems, they will know a lot about you – but they won’t know you.  We will work remotely, interact through various IT systems, submit work products electronically, and the balance of your bank account will occasionally be incremented.  In the “gig economy” we will bid on opportunities to create work products, compete with untold other bidders, sometimes get selected by the deep learning vendor selection system, and use various online sources to create and deliver the promised outcomes.  We won’t really know anybody professionally, although your buddies at the local pub may argue the strengths and weaknesses of the next generation IT platform.


The Academic Job Market

Engineering and science account for roughly three quarters of all PhD graduates, with half of these degrees awarded to US students and the other half to international students. Many of these graduates aspire to tenure-track faculty positions at universities. However, the percentage of faculty openings that are tenure track has been steadily decreasing for quite some time. Universities have found that non-tenure track faculty members, as well as post-docs and adjuncts, are much less expensive, which helps to compensate for strong growth of administrative costs at many institutions.

With more PhD graduates chasing fewer tenure track positions, universities have steadily increased their criteria for hiring. This can include 10-20 published journal articles, extensive teaching performance with good teacher ratings, and professional thought leaders who will write letters extolling your intellectual and social virtues. A fresh PhD graduate cannot possibly satisfy these criteria.

Consequently, especially in the sciences, new PhD graduates seek post-doc positions. These positions pay roughly twice what PhD assistantships pay, i.e., $4,000 per month rather than $2,000 per month, but much less than the $8,000-$10,000 per month paid to tenure track assistant professors. With an average of seven years to earn a PhD and perhaps three years as a post-doc, the candidates are now in their mid 30s before they are ready to compete for coveted tenure track positions.

The competition is fierce. Each position draws hundreds of applications or more. Consequently, people may apply for 50 or more positions. If they win a tenure track position, they now have 7-10 years to earn tenure. During this time, they need to publish 2-4 journal articles per year in top journals as defined by their subdisciplines.  To hit these numbers, they focus on brief incremental contributions that comfortably fit in reigning paradigms. They often get really good at this and will continue in this mode for the rest of their careers. Any effort that is more complicated or takes considerably more time will be shunned, as it will slow them down on the path to full professor in their mid to late 40s.

The process is further complicated for international PhD students. The income they receive in graduate school may be greater than their income would have been in their home countries. Thus, the international student may gain a couple of hundred thousand dollars during the ten years of PhD study plus post-doc.  In contrast, an American PhD student may forgo up to a million dollars of income over the ten years.

The overwhelming problem for international PhD students is the likelihood of being deported immediately after graduation. Federal agencies and other sponsors will have invested perhaps three hundred thousand dollars in creating a top expert, and they then force this expert to leave, to go home and compete against us. It makes no sense.

International students are quite creative in identifying training opportunities and internships that enhance their credentials beyond their degrees. Hoards of lawyers specialize in helping these graduates jump immigration hurdles. A significant number make it and are increasingly filling the ranks of science and engineering faculties across the US. Tenure-track faculty members born in the US are disappearing with retirements, slowed by the Great Recession, but inevitable nonetheless.

Many of the grads from MIT, Stanford, Berkeley, etc. return to China, India, Korea, Singapore, and elsewhere to become faculty members at their best local institutions of science and engineering.  With their countries making much greater investments in these institutions, compared to trends in the US, the numbers of students applying to US institutions are declining, in some cases rather significantly, e.g., Korea.

Combining the inevitable decline in international PhD students at US institutions with the steadily decreasing value proposition for US born PhD students, the future looks rather bleak for US PhD programs.  However, we could choose to change the value proposition for US students.  We would need to (at least) triple PhD students’ stipends and waive tuition.  There is a variety of ways this could be approached.

When I spent a year at Delft University of Technology, all the PhD students were full-time staff members with regular research and teaching responsibilities.  They had reasonable salaries, paid no tuition, and progressed as part of the intellectual fabric of the institution.  This was great for them and great for the university. There are, of course, quite a few implications of this idea.

How would this be funded?  In the current system, the university receives overhead on student’s stipends plus tuition.  Of the $80-100,000 that it annually costs for a half-time graduate student in a private university, only roughly 25% goes to the student.  Tripling the stipend would increase overall costs by at least 50%.  Would research sponsors accept $160-180,000 as the cost of a half time student?

The central balancing factor is that these PhD students would be full-time employees and have substantial research and teaching responsibilities.  These PhD students would be US born with great English skills, reasonable compensation, and aspirations to become faculty members.  It would make enormous sense to invest in enhancing their research and teaching knowledge and skills, particularly since they would not be deported upon graduation.

The availability of these personnel would allow significant reductions in the numbers of tenure-track faculty members.  Decreasing the number of tenure-track faculty members would steepen the promotion pyramid, likely decreasing the chances of becoming full professor.  It would certainly decrease the annual rate of faculty hiring, potentially making the competition even fiercer.

PhD students as full-time professionals would substantially decrease the number of such students needed.  If we were to explore this in more detail, I expect we would find:

  • Decreased numbers of PhD students with substantially increased percentages of US born students
  • Decreased numbers of tenure-track faculty, particularly as inevitable retirements increase
  • Decreased university revenues, but also decreased costs, except perhaps for administrative overheads that seem immune to cost pressures
  • Decreased numbers of journal articles published by PhD students whose full-time responsibilities would not allow the traditional focus on publications

Regarding this last observation, the consequences include the journal article “Laminar Flow Over an Inclined Plate at 17.5 Degrees” never appearing. (17.4 and 17.6 degrees had been addressed in two papers in earlier issues of the journal.)  Is this a loss?  This is a quite complicated question with many implications.  I will return to this question at a later time.

Complexity: Absolute or Relative?

I spent the last few days in Santa Fe, absorbed in discussions of complexity, with particular emphasis on healthcare delivery.  I have delved into this topic for quite some time. Three decades ago, we published our studies on the complexity of troubleshooting – figuring out the source of unfortunate symptoms, e.g., why your car won’t start.

Sponsors of our research asked us to devise a metric for the complexity of a troubleshooting task, which they intended to use to match to the complexity processing abilities of maintenance personnel.  Pursuit of this goal led us to conclude that complexity is related to the intent of the person asking the question or performing the task, as well as the knowledge and skills of this person.

To illustrate, let’s say you purchased a Boeing 747 to use as a paperweight.  From this perspective, this complicated airplane is just a large mass, pretty useful for keeping errant papers on a very large and structurally sufficient desk.  In contrast, if you made this purchase with the intent of operating and maintaining the aircraft, the Boeing 747 is much more complex than your unwieldy paperweight.

This insight leads to a fundamental conclusion.  Complexity has to be defined in terms of a relationship between an observer and an entity.  The observer’s intentions, knowledge, and skills frame the assessment of the complexity of the entity.  Thus, complexity is relative rather than absolute.  Consequently, for example, we can only assess the complexity of a troubleshooting task relative to the personnel involved in the task.

I have discussed this conclusion in many of my talks over the past ten years or so.  Roughly 90% of the people with some level of expertise in the topic agree with me.  The other 10% say something like, “What you are saying makes sense, but what about real complexity?”  These people are usually physicists who firmly believe in the absolute nature of complexity.

Many of those researching complexity construct network diagrams of the elements and relationships among elements of engineered, organizational, and natural systems of interest.  They calculate various metrics associated with these network diagrams and then argue that these metrics reflect the inherent complexity of the systems of interest.  I have done this as well, with the explicit acknowledgement that these network models reflect my intentions, for instance, to predict the difficulty of driving in different urban environments.

There are no intention-free models.  Every model is constructed with the intent to analyze, assess, or predict some set of phenomena.  Any properties of these models used as complexity metrics reflect the intentions of the modeler(s).  This is as essential today as it was for Newton, Darwin, and Einstein in past centuries.  Absolute complexity is a chimera.

Why You Hate Your Airline

The October issue of Consumer Reports outlines “Secrets to Stress-Free Flying.”  This 14-page article provides an interesting history of the airline industry, including the forces that drove your once loved airline to become an object of intense scorn and hatred for most passengers.

Over recent years, the airlines have refined their strategy for making record profits. Charge as much as possible, squeeze passengers into smaller and smaller spaces — which poses medical risks (see Consumer Reports article) — provide as little service as possible, make passengers pay for almost every breath they take in flight, and smile while they say they care about passengers.

I don’t think we should re-regulate the airlines, but they should be forced to pay for the problems they impose on passengers.  For example, they should pay you when they waste your time.  How about $100 per passenger for every hour they are late. This includes delays for mechanical problems, crew complications, and inclement weather.

Airline executives will complain that delays are seldom their fault. This is akin to shipping executives complaining about all the water, or trucking executives complaining about the traffic.  My answer is simple. If you don’t know how to run an airline, get out of the business. Flip burgers. Mow grass. But stay away from airports.

Cultures of Compliance

I have encountered many organizations, mainly in government and academia, where compliance with policies, procedures, and norms became the primary organizational objective. Producing useful outcomes became secondary, almost a nuisance because production took resources away from compliance.

This becomes an almost insurmountable problem when the organization is laced with administrative incompetence. Perhaps well-intended but fundamentally incompetent administrators force compliance on those who would have been producing useful outcomes.

This is further complicated by fragmented and antiquated information systems. One measure of this is the number of times you have to enter your user name and password to accomplish one task. Another measure is the number of times you have to start all over because the system does not recognize the computer they bought for you and told you to use.

The ultimate complication is when the legal function is in charge.  They want to make sure that the organization cannot be blamed and held accountable for anything.   This objective is, of course, much easier if the organization avoids doing anything.

I once asked a Chief Legal Counsel if her compliance job would not be easier if the organization provided no services, accepted no monies from sponsors, and created nothing of value. She replied, “It certainly would minimize our risks.”

I then asked, “How could the organization survive if it provided no value to anyone?”  She responded, “That not my responsibility.  My job is to maximize compliance so as to minimize risks. You need to talk to the President if you are concerned about the value we provide.”

She was right. Her function was risk management, not value creation. I talked to the President, but he was all hype and slogans. His dominant goal was assuring a financial surplus each year that got bigger the following year.

I then talked to the organization’s equivalent of production workers. They were frustrated by increasingly tight budgets, driven by the goal for surpluses.  They were angry about all the time they had to devote to compliance paperwork, often entering the same information into multiple information systems.

Morale was abysmal across the organization. In the executive suite, however, everything was upbeat. All the slogans were prominent. Glossy brochures touted the smoothly running organization.   Everything was aligned for an unfortunate surprise.

The Disruption of Autonomous Vehicles

Many pundits argue that driverless cars will soon be here.  You can argue with the timelines they articulate, but it is difficult to disagree with the distinct possibility of the technology eventually maturing and becoming an increasing portion of the vehicles on the road.  This technology will be truly disruptive.

There will be the benefits of a more efficient transportation system, dramatically fewer accidents, and commute times spent being productive or at least being more relaxing.  There will, however, be costs associated with the technology and infrastructure needed to support it.  There will still be some accidents, although the technology, unlike human drivers, will be continually improved as lessons are learned.

Many of the disruptions will be byproducts of these innovations.  As accidents disappear, the most profitable segment of the insurance industry will wither.  As car and truck services replace individual ownership, vehicles will be used 24 x 7 and the number of vehicles will steadily decrease.  Used cars will disappear, eliminating roughly three quarters of the car loan business.  The after market for vehicle add-ons will disappear.

Truck drivers will become rare, as will drivers of taxis, limos, and other car services.  The autonomous car service industry will have their own service operations, replacing corner filling stations and car washes.  The need for parking places will plummet, proving real estate for other purposes but also significantly reducing municipal revenues.  The need for traffic police and the issuing of speeding tickets will disappear, also reducing municipal revenues.

I have read that as many as 5,000,000 jobs will be eliminated.  At the same time, millions of new jobs will be created, but probably not for the same people.  This has happened before.  Electricity disrupted the marketplace in the late 19th century and automobiles dramatically disrupted horse-drawn transportation in the early 20th century.  The acreage and labor associated with feeding and caring for horses plummeted.  The pollution of horse manure did as well.

The process of one or more technologies disrupting a market is often termed “creative destruction.”  The creation of innovative new ways of doing things results in destroying the old ways.  This process can be very painful for those skilled in the old ways.  Efforts and resources have to be devoted to gaining new skills.  Over time, the new ways flourish and the overall economy greatly benefits.