The devil's in the detail of Treasury data

When Treasury and the Department of Finance issued their pre-election economic and fiscal outlook statement this0week it had something written on the cover in invisible ink: Why don't you all grow up!

Although the figures in the PEFO ("pee-fo") for the forecast and projected growth in the economy and the change in the budget balance over the four years to 2016-17 were virtually identical to those in the Labor government's economic statement 11 days earlier - no surprise to anyone except conspiracy theorists - the words were quite different.

What Treasury issued was a kind of adults only version of the government's document, a rebuke to people who think knowing what the future holds is easy peasy and anyone who gets their forecasts wrong must be either incompetent or corrupt.

The Labor government was so unsophisticated in its understanding of the limitations of forecasting it took a Treasury projection of the budget balance in four years' time and raised it to the status of a solemn promise. No one working in Parliament House thought this a foolish thing to do.

The first thing Treasury does in the PEFO is stress that, while all forecasts are uncertain, the economy's transition to new sources of growth make these forecasts particularly so. It said the transition "may not occur as smoothly as forecast" twice on the first page.

Cop this for a product warning: "This uncertainty surrounding global growth prospects poses a risk to the terms of trade and nominal gross domestic product forecasts. There is also a risk that the anticipated fall in resources investment following its peak could be sharper than expected, especially around the middle of the decade. In addition, the transition to new sources of growth may not occur as smoothly as anticipated. Unexpected global or domestic developments could also generate further sharp movements in the exchange rate."

It's long been the convention to express forecasts as a "point estimate" - a single figure rather than a range. But quoting single figures gives the forecasting exercise an air of false precision which can mislead the uninitiated.

So Treasury has joined the Reserve Bank in showing the "confidence interval" surrounding its key point-estimate forecasts. It has examined the (lack of) accuracy of its forecasts over the past 13 years and used this to show its latest forecasts over a symmetrical range, with its point estimate the central forecast within that range.

Its central forecast is that real GDP will grow at an average annual rate of 2.75 per cent over the two years to 2013-14. So if you assume its forecast errors are similar to those in the past, and also assume its forecasts are just as likely to prove too high as too low, there is a 70 per cent probability that actual real growth will average somewhere between 2 per cent and 3.5 per cent (that is, the central forecast plus or minus 0.75 percentage points).

Its central forecast is that nominal GDP will grow at an average annual rate of 3.125 per cent over the two years. So there's a 70 per cent chance the actual rate of growth will average between 1.75 per cent and 4.5 per cent (central forecast plus or minus 1.375 percentage points).

Why is the confidence interval for nominal GDP so much wider than for real GDP? Because, to get to nominal, you also have to forecast the GDP inflation rate (strictly, the GDP deflator) and it's much more uncertain because it's heavily affected by the change in the terms of trade (export prices divided by import prices) and thus the prospects for world commodity prices.

Why is the GDP inflation rate forecast to be so small, just an average rate of 0.375 a year? Mainly because export prices are expected to fall a fair bit further.

Why does the growth in nominal GDP matter much? Because, as Wayne Swan never tired of pointing out, we live in - and pay tax in - the nominal economy; the real economy is just a (useful) concept.

It was because Treasury kept under-forecasting the rise in export prices that it kept underestimating the improvement in the budget balance in the early years of the resources boom. It's because it's been under-forecasting the fall in export prices that it's been overestimating the improvement in the budget balance in recent years.

Another aspect of the politicians' and media's incomprehension of the budget figuring is their failure to understand the difference between forecasts and projections. By government decision, the figures for the budget year and the first year of the forward estimates are forecasts - that is, Treasury's best guess on what will happen. But the last two years of the forward estimates are merely projections - that is, you assume it will be an average year and mechanically plug in the figures accordingly.

You assume "trend" real growth of 3 per cent, trend employment growth of 1.5 per cent, inflation at the middle of the Reserve Bank's target range - 2.5 per cent - and unemployment at what the econocrats consider to be its lowest sustainable rate (aka full employment), 5 per cent. This makes it all the more foolish for the government to turn a mere projection of the budget balance in four years' time into a solemn promise, and for the rest of us to take it seriously.

It also means you can get some literally incredible jumps between the last forecast year and the first projection year.

For instance, between 2014-15 and 2015-16, the unemployment rate is supposed to drop from 6.25 per cent to 5 per cent, even though real growth stays unchanged at just 3 per cent.

Treasury uses the PEFO to show what it would have forecast for the last two years of the forward estimates had it not been required to use projections, and drops a big hint it will ask the "future government" to change the rules to four years of forecasts.