Forecasting Growth

When the Australian Government announced (Jan 1996) investigation of Holsworthy Military Training Reserve as a second site for Sydney's second airport, there was a sudden jump in the forecast load on the airport from 25 million to 30 million passengers per year (in barely six months). Why the jump ?

It's possible that this inflated figure was to be used to rule out Badgery's Creek, the site planned since 1984 with a forecast of 25 million passengers per year underpinning it. So let's see how good the new number is.

January 1996 Air Traffic Forecasts by the Department of Transport claimed that the number of passenger's using KSA each year has grown at 6.9% for the past 30 years, and predict Sydney will have 60 million passengers arriving at its airports in 2025 (originally published at "http://www.dot.gov.au/programs/avpol/avstats/sydbasin.htm")

The 60 million figure comes from a simple model of exponential growth. Exponential growth works just like interest growing on a loan with no payments. The graph here shows how exponential growth fits the historical data (see the line labeled Best Exp Fit 1997). This actually has a growth rate of 5.9% pa - lower than the DOT's original proposal 7% forecast; the 5.9% figure comes from a more statistically valid regression analysis. It's a fair fit to the historical record. But it's hardly a perfect fit to the bumpy actual data - there's plenty of gaps there.

Also shown in the graph are some alternative models. The Draft EIS prediction from the DOT (late 1997) has been included, as has been the Final EIS revised predictions from July 1999.

The magenta line labelled 1997 Logistic shows a mathematical model of logistic growth, using just the data points available in early 1997.

Market Saturation

While there's not much difference in the historical fit, the picture soon changes a little when you project into the future.

The DOT's model is simple in that it assumes the population of travelers will grow without end - if you let it run on (this is shown in the next graph below). This is like expecting VCR, car sales or mobile phone sales to grow without end. Or for a population of rabbits to grow without end.

In reality, this never happens. Every market or environment will eventually "saturate" (and interestingly, for many technologies, this occurs over about 30 years - or one generation - just about the time we've had cheap 747 based air travel technology). For natural populations, competition for scarce food increase the death rate as the population grows. For technologies, competing products are developed at an increasing rate as the market grows.

NASA's SATS 1999 research supports this (presented two years after BEAR advocated logistic curve modelling; click on NASA graph to learn more about SATS).

Even if air travel becomes cheaper, there's a limit to how many times you'll want to go to Paris in one year (Been there, done that!). If you're going to Sydney for an Olympic holiday in year 2000, you might not take your usual visit in 1999 or 2001.

In the logistic growth model, both the initial exponential growth and its slow-down due to limits being reached are modelled This is clearer when you look at the next 30 years in the Projected Movements graph.

This graph also shows the trend from the DOT's forecast for the next 30 years. Interestingly, although they've stuck by their exponential growth model, they've not chosen to use their historical growth rate (6.9%) for the next 30 years. Why ?

Could it be that it gave the wrong answer ? If you accepted it's prediction of 110 million passengers per year, you'd really have to start talking of lots more than a second airport - at a time when talk of a 2^nd airport scares the hell out of most of us.

Cooking the Books

So what did they do ? They didn't throw away their simple model - even when they suspect it's wrong. Rather, they've gathered a few of their mates around a table, and cooked up 3 lower exponential growth estimates - 6.1% to 2000, 4.6% to 2010, and 3% to 2025 (this was the 1997 Draft EIS forecast). Why 2000, 2010 & 2025 ? Is something special due to happen then ? Not !

This looks dreadfully like they worked back from a target of 60 million passengers per year. Such a figure (if true) would "prove" that KSA remains viable at its present capacity, and a 2^nd airport with double KSA's capacity would nicely handle the rest. Or alternatively, grow KSA by 50%, and have two 30 million passenger airports. This might be the sort of evidence you need to maximise your return on the sale of KSA, but is clearly not objective value-free science

For the Final Supplement to the EIS (see Section 4.3.3, page 4-5), the DOT revised it's optimistic forecast further to better match what had come to pass over the 1997-1998 year. They fudged the figures down a little to 4.0% growth to 2000 (down 2%), 4.2% until 2010 (down by 0.4%), and 2.8% (down by 0.2%) to 2020 (not 2025 as previously). Once again there was no public scrutiny of the forecast process. And there is no explanation of why the forecast period has been truncated to 2020. Are they losing confidence in their forecasting ability ?

The DOT's downward revision of the exponential growth rate seems to acknowledge that there are limits to growth. Can we do better with a more realistic logistic model that accounts for competition and market saturation effects ?

Looking at the lower (1997) Logistic Growth curve in the second graph above, you can see it begins to flatten out. It suggests that we'll need airports to handle only 37 million passengers per year in 2025 - not quite double what KSA does now.

The 1997 logistic growth model matches very closely with the historical record. Note that if you were less careful about the historical fit, you could show lower future growth scenarios (some even as low at 30 mpy for 2025).

Extraordinarily, SACL's 2003 Masterplan for KSA has stuck with the 64 million passenger movements plus target for 2025. That's in spite of the very heavy down-turn in aviation due to SARS, Sept 11, the Bali bombing etc.,. etc.,. Since they've passed the DOT's breakpoints in the curve, SACL's is a simple 4.2% pa growth exponential curve. Wishful thinking !! Maybe they hope the share-market isn't smart enough to doubt it.

In case you're getting the idea that numbers can be produced to fit any argument, you're learning -let's take a reality check. The real value of the numbers is only known when the passage of time provides new data to check the model with.

Verification of Logistic Model

The proof of any forecasting model is in how well it predicts the future. So how does the logistic model, based on data available up til 1997, do at predicting the following two years of data (1998 and 1999) ? Take a close look at the top graph above.

The Final EIS provides an additional data point for 1997-98 passenger movements of 21.3 million. This matches the logistic curve prediction of 21.23 million for 1998 well - and better than both the Draft EIS DOT forecast of 22.72 million and it's earlier 23.1 million forecast. It provides strong substantiation of the arguments constructed here in 1997 (before the 1997/98 data point was available).

The SACL 1999 Annual Report gives another data point - 21.88 million passenger movements for 1999. Again, the logistic curve prediction of 22.5million is closer than the Draft EIS forecast of 24.11 million and the earlier 24.7 million forecast.

So for the first two years of its "future", the logistic model (based on data to 1996/97) produced better results than the exponential models used by the airport proponents. When 99/00 data is included, the score is 3-0 to the logistic model (see Updated Airport Growth for Sydney Airports for more details, including data points for 2001/2002).

The logistic equation prediction is better science whereas the DOT's methods appeal to guesswork.

It is sensible to ask whether their guesswork could be wrong. The logistic estimate shows that if the past is like the future, the EIS forecasts may be too low. On the other hand, there are new factors like future oil prices and lower population growth rates than in the past that might mean even the DOT's forecasts are too high.

But this doesn't mean the DOT numbers are meaningless - they do show the dangers we are in.

Risks of a Second Airport

1. Firstly, there's a possibility the Second Airport will be one huge white elephant. At the extreme, airport traffic may fail to grow and no new airport would be necessary (e.g. if oil supplies peter out - see The End of Cheap Oil or Scientific American, March 1998 or for Jean Laherrere's analysis of oil supplies at http://www.oilcrisis.com/laherrere or look for the International Energy Agency's paper for the March 98 G8 Energy Minister's meeting, http://www.iea.org/g8/world/oilsup.htm ).

2. While this will be good for the victims of the airports flight paths and vehicular traffic, it won't be good for whoever paid for the airport - and in all likelihood this will be the taxpayer or travelers who underwrite or fund the private airport developer..

3. On the other hand, there is a possibility that the planned airport will be grossly inadequate - and Sydney should be planning BOTH Holsworthy and Badgery's Creek Airports plus a trebling of KSA now.

There's a serious risk that the private airport owner will sue the government when he can't make enough profit because the forecasts were wrong (too high or low). Guess who'll foot the bill then ?

Naturally, the DOT's traffic forecasts make no mention of the risks of these errors. They've buried them in a morass of pseudo-scientific fudgery and econometric mumbo-jumbo. And they're hoping we'll be bamboozled into accepting this.

Write and tell them they've failed. Demand that they properly identify the risk's that the forecasts are wrong and that they consider more appropriate models. Better still, tell them to get out of the business and let the private sector do its own dirty work of justifying a new airport - EIS & all !