Fans

Back in 1988 Richard Ashley wrote a paper for the International Journal of Forecasting called “On the relative worth of recent macroeconomic forecasts”. It is a little old now but it has some interesting findings which are often cited by those who have studied the accuracy of forecasts in the social sciences.

There are numerous ways that forecasters ply their art but one of the most favoured, particularly by real estate forecasters, is the use of “explanatory variables” to produce forecasts of another “dependent” variable (for example rental growth). Using this method means that you have to be able to forecast the explanatory variables to be able to forecast the dependent variable; to produce a forecast of rental growth for example you may need to be able to forecast a macroeconomic variable such as GDP growth. For Ashley, this is where the issues start.

He looked at the size of the forecast errors when experts forecast macroeconomic data such as GDP growth. He found that the forecast errors were bigger than the variability in the macroeconomic data being forecast. This means that you would be better off leaving out these explanatory variables from a forecasting model and purely relying on the history of the data itself to make a forecast. This is enormously inconvenient but if you can’t accurately forecast the long list of explanatory variables needed, (for total returns that might include inflation, real bond yields, a risk premium, an illiquidity premium, a demand variable such as GDP growth and a supply variable such as space completed) then there is little hope for accurately forecasting the dependent variable.

One common way around the issue of uncertainty is to:

• estimate all manner of models using different explanatory variables and relationships

• choose one (which is not a straightforward task)

• use it to produce a forecast

• add some expertise and write down the number you first thought of.

Another way around this issue, given the potential range of outcomes, is to produce a fan chart. These are commonly seen when central banks forecast inflation, which sets a range around the forecast. For example this may show “we forecast that inflation is likely to be 2% next year and we are 90% confident it will be between 0% and 4%”. As the forecast horizon is extended, the fan gets wider and wider and then settles down somewhere around the historic range of inflation; in the case of UK CPI inflation it takes about 18 months to settle down to show a range roughly between -0.5% and 5%. For UK GDP growth such a fan chart settles at around -1.5% to 5.5%.

There must be a strong case for property forecasters to produce fan charts given the uncertainty surrounding the inputs, and, the billions of dollars of other peoples’ money that may be invested based upon them.