WEBVTT 00:00:00.000 --> 00:00:03.444 align:middle line:90% [SQUEAKING][RUSTLING][CLICKING] 00:00:03.444 --> 00:00:21.533 align:middle line:90% 00:00:21.533 --> 00:00:22.700 align:middle line:90% RICHARD DE NEUFVILLE: Right. 00:00:22.700 --> 00:00:24.420 align:middle line:90% Richard de Neufville here again. 00:00:24.420 --> 00:00:27.390 align:middle line:84% Pleasure to talk with you, and for this session, 00:00:27.390 --> 00:00:30.230 align:middle line:84% I'm going to talk about porcupine graphs. 00:00:30.230 --> 00:00:34.160 align:middle line:84% It's all part of this topic of the forecast being always 00:00:34.160 --> 00:00:38.360 align:middle line:84% wrong, not every single time, but in general, we 00:00:38.360 --> 00:00:40.040 align:middle line:90% can't count on it being right. 00:00:40.040 --> 00:00:42.740 align:middle line:84% And we just expect it, the reality 00:00:42.740 --> 00:00:45.620 align:middle line:90% that we face during the-- 00:00:45.620 --> 00:00:49.860 align:middle line:84% our project lifetime is not what we anticipated. 00:00:49.860 --> 00:00:52.550 align:middle line:90% Now, what's a porcupine graph? 00:00:52.550 --> 00:00:55.880 align:middle line:84% Well, let me first describe what a porcupine animal is in case 00:00:55.880 --> 00:00:58.190 align:middle line:84% you don't know it or haven't seen it. 00:00:58.190 --> 00:01:01.220 align:middle line:84% The porcupine is a North American animal. 00:01:01.220 --> 00:01:02.750 align:middle line:90% It maybe exists elsewhere also. 00:01:02.750 --> 00:01:05.390 align:middle line:84% I think it does, but I'm not an expert in the area. 00:01:05.390 --> 00:01:07.670 align:middle line:84% But the important thing about a porcupine 00:01:07.670 --> 00:01:11.510 align:middle line:84% is that it is an animal, four-footed, small, 00:01:11.510 --> 00:01:14.690 align:middle line:84% about the size of a small dog, and it 00:01:14.690 --> 00:01:18.030 align:middle line:84% has a lot of spikes on it that go up from its back. 00:01:18.030 --> 00:01:20.180 align:middle line:84% And as you can see in this example here, 00:01:20.180 --> 00:01:24.920 align:middle line:84% they go all back in a variety of ways 00:01:24.920 --> 00:01:27.120 align:middle line:90% in many different directions. 00:01:27.120 --> 00:01:31.970 align:middle line:84% And what I'm going to show you is some forecasts which 00:01:31.970 --> 00:01:34.940 align:middle line:84% contrast with the-- as with a porcupine, that 00:01:34.940 --> 00:01:38.690 align:middle line:84% contrast with the actual back of the porcupine which 00:01:38.690 --> 00:01:42.800 align:middle line:84% in this case goes down from its head down towards its tail, 00:01:42.800 --> 00:01:46.010 align:middle line:84% but the forecasts go up like this, 00:01:46.010 --> 00:01:49.610 align:middle line:84% like my fingers on my hand, but are 00:01:49.610 --> 00:01:52.010 align:middle line:90% like the spines on a porcupine. 00:01:52.010 --> 00:01:54.740 align:middle line:84% Let's take a look at what those might be. 00:01:54.740 --> 00:01:56.420 align:middle line:90% Here's the first one. 00:01:56.420 --> 00:01:58.630 align:middle line:90% This is an old one. 00:01:58.630 --> 00:02:03.320 align:middle line:84% It's has a lot of history around MIT. 00:02:03.320 --> 00:02:06.890 align:middle line:84% It's called the NERF fan, the New England Regional 00:02:06.890 --> 00:02:08.870 align:middle line:84% Forecast for the use of electricity 00:02:08.870 --> 00:02:13.670 align:middle line:84% over a long period from about 1960 to about 1990. 00:02:13.670 --> 00:02:17.390 align:middle line:84% And what you see is that the actual growth 00:02:17.390 --> 00:02:20.150 align:middle line:84% was exponential at a reasonable rate. 00:02:20.150 --> 00:02:25.320 align:middle line:84% And then later on, it fell off and leveled off. 00:02:25.320 --> 00:02:27.650 align:middle line:84% Meanwhile, during all this period, 00:02:27.650 --> 00:02:31.010 align:middle line:84% the forecast kept going up, as you can see. 00:02:31.010 --> 00:02:33.890 align:middle line:84% This is the porcupine, the spines going up, 00:02:33.890 --> 00:02:35.840 align:middle line:90% but the actual use was down. 00:02:35.840 --> 00:02:39.590 align:middle line:84% Now, the first one to go up is an extension 00:02:39.590 --> 00:02:42.270 align:middle line:90% of the original use. 00:02:42.270 --> 00:02:45.300 align:middle line:84% So you can just imagine the exponential going on. 00:02:45.300 --> 00:02:48.920 align:middle line:84% But what's interesting is that while over the 10 years 00:02:48.920 --> 00:02:53.840 align:middle line:84% or so while the graph flattened out, as you can see, 00:02:53.840 --> 00:02:56.750 align:middle line:84% the big dark line flattened out, people 00:02:56.750 --> 00:02:59.840 align:middle line:84% kept projecting as if it went up again. 00:02:59.840 --> 00:03:03.080 align:middle line:84% That is, they knew somehow that it 00:03:03.080 --> 00:03:06.350 align:middle line:84% was going to go to where it had been even though the world has 00:03:06.350 --> 00:03:07.960 align:middle line:90% changed. 00:03:07.960 --> 00:03:11.440 align:middle line:84% This is remarkable, but it is not the unique case. 00:03:11.440 --> 00:03:16.180 align:middle line:84% This is an old one, but here is another one, a very recent one 00:03:16.180 --> 00:03:19.630 align:middle line:84% for the last few years of the teens 00:03:19.630 --> 00:03:26.200 align:middle line:84% of the 2010 to about 2017, The US consumption of AV fuel, 00:03:26.200 --> 00:03:27.740 align:middle line:90% aviation fuel. 00:03:27.740 --> 00:03:31.000 align:middle line:84% So the actual consumption was this black line which 00:03:31.000 --> 00:03:33.400 align:middle line:84% goes-- bounces up and down at the bottom, 00:03:33.400 --> 00:03:38.860 align:middle line:84% and all those things going up, they are the forecast. 00:03:38.860 --> 00:03:41.230 align:middle line:84% Now, you'd say, well, my goodness. 00:03:41.230 --> 00:03:42.430 align:middle line:90% How could this be? 00:03:42.430 --> 00:03:47.290 align:middle line:84% We know it's possible to know that the number of people 00:03:47.290 --> 00:03:50.260 align:middle line:84% flying in the United States has in fact increased remarkably 00:03:50.260 --> 00:03:51.010 align:middle line:90% over this period. 00:03:51.010 --> 00:03:55.300 align:middle line:84% How is it possible that consumption of AV fuel 00:03:55.300 --> 00:03:57.500 align:middle line:90% didn't grow along with it? 00:03:57.500 --> 00:04:00.130 align:middle line:84% Well, the answer is that there have 00:04:00.130 --> 00:04:03.070 align:middle line:84% been developments of newer aircraft that 00:04:03.070 --> 00:04:04.810 align:middle line:90% are more fuel efficient. 00:04:04.810 --> 00:04:07.630 align:middle line:84% People go in larger planes rather than smaller planes, 00:04:07.630 --> 00:04:11.500 align:middle line:84% which are less efficient in terms of consumption 00:04:11.500 --> 00:04:16.089 align:middle line:84% of fuel per seat mile, and so that there's 00:04:16.089 --> 00:04:18.440 align:middle line:90% been a technological change. 00:04:18.440 --> 00:04:21.519 align:middle line:84% So you can explain that in terms of, well, yeah, now 00:04:21.519 --> 00:04:23.920 align:middle line:84% we can see that would happen this way. 00:04:23.920 --> 00:04:30.070 align:middle line:84% But what's remarkable is that the experts in the area making 00:04:30.070 --> 00:04:32.510 align:middle line:90% these forecasts got it wrong. 00:04:32.510 --> 00:04:34.030 align:middle line:84% They got it wrong the first time, 00:04:34.030 --> 00:04:37.390 align:middle line:84% OK, but why did they not correct it later on? 00:04:37.390 --> 00:04:39.850 align:middle line:84% It went down, but it will go up again. 00:04:39.850 --> 00:04:42.530 align:middle line:84% We'll go up again, but it doesn't happen that way, 00:04:42.530 --> 00:04:43.730 align:middle line:90% in fact. 00:04:43.730 --> 00:04:47.690 align:middle line:84% And it wasn't as if that the switch 00:04:47.690 --> 00:04:52.070 align:middle line:84% from less efficient aircraft to more efficient aircraft 00:04:52.070 --> 00:04:55.700 align:middle line:84% wasn't obvious because it takes a long time 00:04:55.700 --> 00:04:58.580 align:middle line:84% to manufacture aircraft, and people 00:04:58.580 --> 00:04:59.960 align:middle line:84% who are experts in the area could 00:04:59.960 --> 00:05:03.200 align:middle line:84% see the aircraft coming down the production line knowing down 00:05:03.200 --> 00:05:04.250 align:middle line:90% to the fleets. 00:05:04.250 --> 00:05:06.470 align:middle line:84% So they know what will be happening. 00:05:06.470 --> 00:05:09.410 align:middle line:84% This does not come as a surprise-- oops, 00:05:09.410 --> 00:05:12.990 align:middle line:84% United Airlines has gotten a lot of new aircraft, for example. 00:05:12.990 --> 00:05:14.510 align:middle line:90% This is well known in advance. 00:05:14.510 --> 00:05:18.440 align:middle line:84% But nonetheless, the experts create what 00:05:18.440 --> 00:05:20.990 align:middle line:90% amounts to a porcupine graph. 00:05:20.990 --> 00:05:22.310 align:middle line:90% I say they create it. 00:05:22.310 --> 00:05:23.850 align:middle line:90% They don't do it on purpose. 00:05:23.850 --> 00:05:25.760 align:middle line:84% It's just that when we look backwards at it 00:05:25.760 --> 00:05:30.500 align:middle line:84% and say what was the forecast and what actually happened 00:05:30.500 --> 00:05:36.850 align:middle line:84% that we observe here, as in so many cases, a porcupine graph. 00:05:36.850 --> 00:05:38.120 align:middle line:90% Here is another one. 00:05:38.120 --> 00:05:42.190 align:middle line:84% This is also coming back to my experience at airports. 00:05:42.190 --> 00:05:46.360 align:middle line:84% This is now a notorious one in the field. 00:05:46.360 --> 00:05:50.440 align:middle line:84% This is the forecast for terminal areas. 00:05:50.440 --> 00:05:55.960 align:middle line:84% That means this is the number of operations, operations being 00:05:55.960 --> 00:05:58.690 align:middle line:84% flights, being landings and takeoff, that's 00:05:58.690 --> 00:06:00.850 align:middle line:84% an operation, the number of operations 00:06:00.850 --> 00:06:02.530 align:middle line:90% within metropolitan areas-- 00:06:02.530 --> 00:06:06.430 align:middle line:84% New York, Washington, San Francisco, Chicago, Texas, 00:06:06.430 --> 00:06:08.270 align:middle line:90% Houston, and so forth. 00:06:08.270 --> 00:06:10.460 align:middle line:90% So these are the forecasts. 00:06:10.460 --> 00:06:13.900 align:middle line:84% So they-- the line consistently going 00:06:13.900 --> 00:06:18.790 align:middle line:84% down here like this is what actually happened, 00:06:18.790 --> 00:06:22.060 align:middle line:84% and all during this period, the agency 00:06:22.060 --> 00:06:24.220 align:middle line:84% responsible for these forecasts, which 00:06:24.220 --> 00:06:27.970 align:middle line:84% I will spare them the notice of which they were, 00:06:27.970 --> 00:06:32.200 align:middle line:84% but the agency, while this was happening, they projected this. 00:06:32.200 --> 00:06:39.040 align:middle line:84% This is now a classic version of the porcupine graph 00:06:39.040 --> 00:06:40.420 align:middle line:90% that happens. 00:06:40.420 --> 00:06:42.460 align:middle line:90% Now, how is this possible? 00:06:42.460 --> 00:06:46.915 align:middle line:84% I mean, if you're observing the drop of what's happening 00:06:46.915 --> 00:06:49.540 align:middle line:84% and you know why it's happening because you know how 00:06:49.540 --> 00:06:52.040 align:middle line:84% the airlines are scheduling their fleets and everything, 00:06:52.040 --> 00:06:58.010 align:middle line:84% why is it that the forecast are, whoops, in the wrong direction? 00:06:58.010 --> 00:07:00.190 align:middle line:90% What is it that's happening? 00:07:00.190 --> 00:07:02.240 align:middle line:84% Because this is not just this case. 00:07:02.240 --> 00:07:05.920 align:middle line:84% This is a case that happens again and again. 00:07:05.920 --> 00:07:07.070 align:middle line:90% Why is this? 00:07:07.070 --> 00:07:13.600 align:middle line:84% Well, the fact is that very often experts don't learn. 00:07:13.600 --> 00:07:16.570 align:middle line:84% They know better or they think they do. 00:07:16.570 --> 00:07:20.710 align:middle line:84% Whatever it is that somehow it happens again and again as I've 00:07:20.710 --> 00:07:23.350 align:middle line:84% shown through three examples, but there are plenty more, 00:07:23.350 --> 00:07:26.650 align:middle line:84% and people send me more examples all the time 00:07:26.650 --> 00:07:30.220 align:middle line:84% because it's such a strange phenomenon. 00:07:30.220 --> 00:07:34.240 align:middle line:84% But the experts in the field don't acknowledge, 00:07:34.240 --> 00:07:36.820 align:middle line:84% don't digest, don't take on board 00:07:36.820 --> 00:07:39.310 align:middle line:84% the fact that the world has changed 00:07:39.310 --> 00:07:43.120 align:middle line:84% and they don't predict accurately. 00:07:43.120 --> 00:07:46.660 align:middle line:84% There's mental biases going on in the profession which 00:07:46.660 --> 00:07:50.650 align:middle line:84% lead to this thing called porcupine graphs which 00:07:50.650 --> 00:07:56.700 align:middle line:84% is another example of why the forecasts are always wrong. 00:07:56.700 --> 00:08:08.316 align:middle line:90%