I underwent a period of intense harassment by Spreitzer and Rothery. The obvious goal was to get me to quit, and it worked. I could do nothing right, according to my bosses, verbally or technically. All my previous research was dredged up and criticized. Sadly, some of this criticism bled over to other researchers. I was spending a disproportionate amount of time defending myself and my colleagues.
Most of the criticism was highly subjective. For example, I was told my writing had too much jargon and too much mathematics. But one key criticism revolved around an important technical point in an older published paper. Rothery accused me of miscalculating speeds in a study I jointly did with a researcher from the old Traffic Science Department. That paper is:
Man-Feng Chang and Alan J. Horowitz, “Estimates of Fuel Savings Through Improved Traffic Flow in Seven U. S. Cities”, Traffic Engineering and Control, Vol. 20, No. 2, February 1979, pp. 62-66.
Chang did all the traffic analysis, and I did all the urban analysis. If there were wrong speeds, it was Chang’s doing. Except that I did not believe Chang could have miscalculated speeds. He was just too good of a researcher. Rothery said we used “time mean speed” rather than the preferred “space mean speed”. Don’t worry if you don’t know the distinction between these two. I did not know about either of these speeds at the time. That’s why I selected a traffic expert as my co-author. I did not have Chang’s original calculations, but I did have Chang’s data. I recalculated the speeds with Rothery’s space-mean speed formula and got the same speed numbers as Chang in every case to within three significant digits. The speeds were correct for all practical purposes. Rothery was not impressed. Chang was slandered, and I received a big black mark on my record. It was not until I started teaching a traffic engineering class many years later that I realized why Chang was 100% correct all along. See the note below for the technical details.
My health had deteriorated, I had accepted a job offer, and it was time to leave. I asked Spreitzer for the dataset collected at great expense for the subjective value of time study. He said I could take it. There were two additional high-quality publications in that dataset, and it gave me a solid head start on my academic career. I was grateful.
The Department at GM soon withered and died. Its researchers scattered. Rothery retired early and followed Robert Herman to the University of Texas. Tragically, Rothery died in an accident at a boat launch not long afterwards. Spreitzer remained at GM. He regained some prominence as GM’s point-person on intelligent transportation systems. He died several days ago.
I once asked a group of Tau Beta Pi students whether they would be willing to risk their job to do as I had done. They all said they would. Maybe, but I sincerely hope they never have to.
My going away party was attended by virtually everyone in the department, except Spreitzer. There was joy for me and more than a small amount of jealously. I had survived GM and was moving on with my life. Those remaining had no idea whether they would be able to make a clean transition to a new career path. Vogler’s card was given to me at the party. For the last half of my tenure at UWM, I hung the card on my office wall where I could see it from my desk. The card held a message for me and any of my students who asked — at one moment in my life I had the guts to follow my convictions against seemingly insurmountable odds.
Occasionally, I have wondered why Rothery chose to slander Chang over a simple speed calculation that Chang had correctly done. Rothery had the reputation of being a good traffic theorist, so this blunder made no sense. The only plausible explanation I can surmise is Rothery was so blinded by the political imperative of pleasing Spreitzer that all professional judgment was thrown away. Which raises still another question, how can we trust any research carried out in an environment where prospects of political and personal gain outweigh the quest for truth?
Alan Horowitz, Whitefish Bay, October 13, 2019.
Technical note on Chang’s speed calculation: Chang’s speed data came from a heavily instrumented automobile that had been driven around the city. This is precisely the same instrumentation used to create the speed profile underlying the original EPA urban driving fuel economy test. The instruments gave us the average speed calculated over exactly 1 second of time. To find the trip speed, Chang simply averaged all those one second speeds in the trip.
The controversy between “space mean speed” and “time mean speed” arises when finding the average speed from spot, instantaneous, speeds of many vehicles in a traffic stream. The consensus of traffic experts is we should be using “space mean speed”, which can be approximated from the harmonic mean of all spot speeds. “Time mean speed”, which is not correct, is found from the simple average of spot speeds. Chang’s data were not spot speeds of many vehicles. They were speeds from a single vehicle at short, regular time intervals. This distinction is critical.
The best way to explain Chang’s logic is to give you a famous riddle about a car being driven over a two mile course. The driver is testing a race car by traveling 2 miles. For the first mile she drives at 30 mph. At what speed does she need to drive the second mile in order to average 60 miles per hour? The correct answer is that she cannot ever average 60 mph, since she has used up her allotted 2 minutes of driving time in the first mile. Most people will give the wrong answer of 90 miles per hour. Let’s work with this wrong answer.
The time mean speed, which we know to be wrong is:
S = (30 + 90)/2 = 60 mph
The space mean speed, which we know to be correct is:
S = 2/(1/30 + 1/90) = 45 mph
Chang’s mean speed comes from averaging speeds by seconds. It takes exactly 120 seconds to travel 1 mile at 30 mph and it takes exactly 40 seconds to travel 1 mile at 90 miles per hour. So the mean speed over the 160 seconds for the whole trip is:
S = (120*30 + 40*90)/160 = 45 mph.
Chang’s method indeed gives us the same answer as space mean speed, exactly and always.