Horseshoes and Hand Grenades

The year: 1991. We are in Bordeaux, in the heart of French wine country. Still working through the after-effects of a long dinner the night before at L’Entrecôte, we find ourselves en-route to Merignac Airport; to the home of Dassault Aviation. The boss is with us. We are picking up his new, shiny, Falcon 900B today.


“Hey, what will my plane be worth in 10 years? How about 20 or 25 years?”


(Blank stares from the back of the van) … finally, somebody remembers a number they overheard at a cocktail party: “apparently they lose, like, 3% or 4% a year, on average…”


So, now it’s 2016. Lot’s of things have changed in the world since 1991. And our Falcon 900B is 25 years old. Just for kicks, let’s see how our back-of-the-van (“BOV”) prediction has held up against history. To do this, we are going to use historical data from Aircraft Bluebook, for each Q1 edition since 1991. Let’s average our BOV prediction at minus (-) 3.5% per year.


Screen Shot 2016-05-18 at 2.18.44 PM


So how did we do? Well, at first glance, not very well … the actual values appear to be mostly nowhere near the estimate. In fact, we seem to have done a really good job predicting what the value of the aircraft won’t be.


This illustrates the problem of thinking in terms of averages when it comes to residual value forecasts. Whatever “average” number you pick, most years are not going to be “average.” Moreover, history shows a tendency to throw multiple “non-average” years in a row at us. The business cycle. Global Warming. Locusts. Whatever the root cause, residual values are undeniably volatile and can stray from the theoretical average by large magnitudes, for considerable periods of time. And if we are interested in what the average should be, we should also be interested in what a best-case or worst-case scenario might look like.


Here’s the thing. In retrospect, we began with only half of a prediction. We were ignoring the most important part of any estimate: margin of error. Anybody paying remote attention to the zillions of Presidential primary-election forecasts will be familiar with this little term, usually written in tiny print at the bottom of your favorite cable news graphic. Simply put, we need some metric to account for how volatile the historical data is compared to the average.


If we had done a little math on the year-over-year volatility in residual values for the decade preceding 1991, we could have concluded that our margin of error (with 95% confidence) was about +/- 18%!  That’s right – the movement in value for any given year can dwarf the average by several orders of magnitude. So for our forecast to “get close,” we are going to have to expand our definition of what close means.


Knowing the margin of error allows us to create a different kind of chart – one that takes into account the probabilities of future value based not only on the average – but also incorporates the compounding impact of volatility around the average:


Screen Shot 2016-05-18 at 2.18.55 PM

The yellow line represents our minus (-) 3.5% average guestimate. The red area represents +/- one standard deviation from the average (about 68% of probable outcomes). The green area represents a 95% probability low and high. Finally, the blue line represents our 1991 Falcon 900B historical bluebook value.


Although the history of this aircraft has stressed the boundaries of the green area – it has, for the most part, fallen within a range that could have been predictable in 1991.


Obviously, the range of forecast values is really broad. In the year 2001, the low is $9.4M and the high $24M. This causes many people to discount the value of such probabilistic forecasts- after all – “anybody could get it right with that range.” But that is exactly the wrong conclusion to draw.


This level of uncertainty and volatility (at least) does exist with aircraft values. It is merely a mathematical calculation based on history. The correct conclusion should be – 1) Awareness, and; 2) What does this residual value uncertainty mean when making aircraft purchase and ownership decisions?


The point here is not to debate if 3.5% was a fair estimate of average annual residual loss in 1991 (or ever for that matter), but rather to illustrate that looking at residual values as a function of uncertainty and probability will result in a much more useful forecast – even if the assumptions are drawn crudely as they are here.


Any estimate of yearly depreciation should be much higher now than in 1991. That was a different era, and three or four percent is not nearly indicative of the past few years. However, the popular belief is that current asset values are “off the chart” compared to historical precedent. Even though today’s forecast may be more pessimistic than what’s depicted above – we can’t exactly argue ignorance when it comes to the effects of volatility. We had that data long ago.


Finally, there is 2008 and the radical loss in value which has since transpired – “averaging” 17.3% per year in the case of our 1991 Falcon 900B. How the heck would we have been able to predict that?


More on that subject in a future edition of theJetWatch.