Blame it on bad weathermen

As a public figure, I'm quite accustomed to seeing my name and face in the media. But it was still quite a surprise to load The Sydney Morning Herald's site on Monday last week and see my craggy visage staring back at me from the front page:

The reason was Peter Martin's article "A Keen eye on the global economy: how our forecasters fared”, which reported the results of the Melbourne newspaper The Age's annual survey of economic forecasters.

On Peter's evaluation, I topped the pool. Even on the numbers that I clearly got wrong – the rate of economic growth of Australia in 2012, and hence also the unemployment rate – Peter was willing to attribute this to the government deficit being $23 billion larger than I had guessed it would be. That's equivalent to 1.6 per cent of GDP, and neatly accounts for the 1.4 per cent error I made in guessing what economic growth would be:

"This time last year, only one of our forecasting panel was bold enough to say the Reserve Bank would cut its cash rate to 3 per cent by December 31," Peter wrote.

"It had just cut the rate twice, from 4.75 per cent to 4.25 per cent. The equivalent of five more cuts was unthinkable, except for Steve Keen…

"Professor Keen is also the only member of our panel to come close to forecasting the inflation rate.

"He picked a mere 2 per cent by December (which is exactly the most recent published figure). The other forecasts were about 2.8 per cent – not shabby but not Keen."

So three cheers for me. But now, let's get real. I used the word "guess” above because that's what my numbers were: educated guesses of course, but no more than guesses. Some of the economists in this poll do have models they use for their predictions, but I do not (though I may one day if my Minsky Project reaches maturity).

So how come my model-free predictions were close to the money (and for the second time in the last four years) while the others – even those with formal econometric models – got it wrong?

Figure 2: Tim Colebatch's article on forecasts in 2009

It's because of the nature of economic modelling. Conventional economic modelling tools can extrapolate forward existing trends fairly well – if those trends continue. But they are as hopeless at forecasting a changing economic world as weather forecasts would be, if weather forecasters assumed that, because yesterday's temperature was 29 degrees Celsius and today's was 30, tomorrow's will be 31 – and in a year it will be 395 degrees.

Of course, weather forecasters don't do that. When the Bureau of Meteorology forecasts that the maximum temperature in Sydney on January 16 to January 19 will be respectively 29, 30, 35 and 25 degrees, it is reporting the results of a family of computer models that generate a forecast of future weather patterns that is, by and large, accurate over the time horizon the models attempt to predict – which is about a week.

Figure 3: The BOM forecast for Sydney from January 16-18 2012, made on January 12

Weather forecasts have also improved dramatically over the last 40 years – so much so that even an enormous event like Hurricane Sandy was predicted accurately almost a week in advance, which gave people plenty of time to prepare for the devastation when it arrived:

Almost five days prior to landfall, the National Hurricane Center pegged the prediction for Hurricane Sandy, correctly placing southern New Jersey near the centre of its track forecast. This long lead time was critical for preparation efforts from the Mid-Atlantic to the Northeast and no doubt saved lives.

Hurricane forecasting has come a long way in the last few decades. In 1970, the average error in track forecasts three days into the future was 518 miles. That error shrunk to 345 miles in 1990. From 2007-2011, it dropped to 138 miles. Yet for Sandy, it was a remarkably low 71 miles, according to preliminary numbers from the National Hurricane Center.

Within 48 hours, the forecast came into even sharper focus, with a forecast error of just 48 miles, compared to an average error of 96 miles over the last five years.

Meteorological model predictions are regularly attenuated by experienced meteorologists, who nudge numbers that experience tells them are probably wrong. But they start with a model of the weather than is fundamentally accurate, because it is founded on the proposition that the weather is unstable.

Conventional economic models, on the other hand, assume that the economy is stable, and will return to an 'equilibrium growth path' after it has been dislodged from it by some 'exogenous shock'. So most so-called predictions are instead just assumptions that the economy will converge back to its long-term growth average very rapidly (if your economist is a Freshwater type) or somewhat slowly (if he's a Saltwater croc).

Weather forecasters used to be as bad as this, because they too used statistical models that assumed the weather was in or near equilibrium, and their forecasts were basically linearly extrapolations of current trends. Meteorologists were dissatisfied with this state of affairs, but couldn't see an alternative.

The mathematician turned meteorologist Edward Lorenz could: he developed a system of three time-varying equations, starting from models of fluid dynamics, that, though simple (just three equations and three parameters), were nonlinear. Their behaviour was dramatically different to that of much larger linear models, and established to meteorologists that forecasting had to start from the fact – not the assumption – that the weather was unstable, nonlinear, and far from equilibrium.

The resulting paper, "Deterministic Nonperiodic Flow”, revolutionised not just meteorology but science as well, because it introduced the concepts of chaos, complexity, and "sensitive dependence on intial conditions” into science. In fact, about the only area of human study that has not been largely transformed by Lorenz's work is – you guessed it – economics.

Simulations of Lorenz's model also strikingly illustrated how different a "linear, equilibrium” world is from a "nonlinear, far from equilibrium” one. If the model began in its equilibrium, it stayed there – as in this simulation in my modelling program Minsky (the black-letter symbols are variables; the red ones are constants). The top plot shows the location of the x variable against the y in Lorenz's model; the lower shows the value of the z variable over time:

Figure 4: Lorenz's fluid flow model at equilibrium

Exciting, eh? Compare this to what happens if one of the variables is a tiny, tiny distance from its equilibrium: in this case, 0.01 (when the plot range above is for x & y values from -15 to 15):

Figure 5: Lorenz's model starting with one variable 0.033 per cent away from equilibrium

Which figure looks more like the weather to you – and which one looks more like the economy too? Yet economists – or rather the neoclassical school that dominates the discipline – continue to use models which innately assume that the economy, in the absence of "exogenous shocks”, would look like figure 4. Is it any wonder that they continually get the future of the economy wrong?

The other factors in my favour are that I have a credit-based view of aggregate demand, and that I know we're in a debt-deflationary environment which is unprecedented in history, and likely to persist for more than another decade.

Figure 6: Debt-deflation's driver, the Debt to GDP ratio, Australia

I've expounded on the first issue many times here, and proven it with the help of the mathematician Professor Matheus Grasselli of the Fields Institute too, but most economists (not just neoclassical ones) think I'm making an obvious mistake of double-counting. I'm not, but as long as they keep believing that, I'm going to have another advantage in the guessing game.

Figure 7: Debt-deflation's driver, the Debt to GDP ratio, USA

The impact of the second issue, debt-deflation, can be attenuated by sensible government policies (like running deficits now rather than surpluses), as clearly happened in 2010-2011 in the US and Australia – which is why my pessimistic forecasts for those years, predicated on wimpish government action, were wrong. But we're going to remain in a debt-deflationary fix for many years because we can only get out of it by abolishing large slabs of the private debt, and governments won't do that – or at least won't seriously consider doing so for another half decade.

In the meantime, governments are going to fall for the "government debt is the problem” myth and make things worse.

Other economists think that too – the surplus fetish is really restricted to politicians themselves, who seem incapable of thinking of an economy as anything other than a blown-up household (well, there are some economists well to the right of Paul Krugman who still believe surpluses are needed, but the empirical data is forcing them to change tune – as happened with the IMF's Olivier Blanchard recently). So that per se doesn't give me an advantage.

But they also don't understand the role of credit in aggregate demand, so they can make forecasts – for rising house prices, for example – without considering what that would require in terms of growth of private debt from its already ridiculous levels.

Figure 8: How much higher to make house prices rise 10 per cent in 2013?

So how do I actually forecast, if I don't have a model? Well firstly I use a linear (yes, linear) extrapolation routine in Mathcad to take the past data and extrapolate the cyclical aspects of it forward (see figure 6).

Figure 9: Using Burg's method to forecast cyclical data

If the number fits within my overall credit-driven, debt-deflationary expectations, then I write it down and skip to the next question.

But if it's wildly at variance with my expectations – as it was for example with next year's inflation forecast – then I scratch it out and write my own down.

Figure 10: Berg's method surely got this one wrong

That method works well enough to get close enough to the actual outcome except in years where the government "flicks the switch to Vaudeville” on fiscal policy. Of course, the fact that it does is more an indictment of the state of economics today than praise for me. I hope that, one day, economists will embrace the realities that the economy is dynamic, out of equilibrium (but still stock-flow consistent), and fundamentally monetary. My Minsky modelling platform accepts all of that, and makes it possible to build simple models that demonstrate the advantages of such as approach, as Lorenz's three simple equations did in meteorology half a century ago.

Until then, my slapdash method is likely to work better than the precise, linear, equilibrium models that economists continue to use despite the evident non-linearity of the world around them.

Steve Keen is Associate Professor of Economics & Finance at the University of Western Sydney and author of Debunking Economics and the blog Debtwatch.