Weather forecasting has improved significantly and a five day forecast is reasonably accurate. This has been so because a lot of data points made on atmospheric pressure, temperature, wind speed, wind direction, humidity, precipitation are collected routinely from trained observers, automatic weather stations or buoys, which are then used in conjunction with a numerical model’s most recent forecast for the time that observations were made to produce the meteorological analysis.
Given the information in a forecast, nothing changes in the atmosphere to make the actual outcome change.
But it is not so with price forecast for example. Given a price forecast, the actors will act to respond to the signal positively, negatively or remain neutral which will change what the data points had originally considered; reactions to a price signal cannot be modeled by the algorithm that originally modeled the signal.
This is known as level two complexity, where the reactions to an output will change the output.
Most economic forecasts have this inherent challenge and interested actors act to change the outcome and the outcome deviates therefore from the original forecast.
Think of oil forecasts, which states a range for oil prices far into the future based on some modeling, like the weather forecasts. But once the forecast is known by the actors, they would act and not necessarily in concert; there would be consumers who might defer or pre-pone their buying decisions, producers who could raise output or curtail production and intermediaries who could speculate further. All this cannot be modeled by any algorithm.
The particularly unpredictable response is that of the speculator, who could sway the outcome in any given direction. This is precisely the reason most forecasts on oil in the current times have been wrong. This is where synthetic instruments like hedges of all kind which the financial markets and exchanges provide is inadequate to provide a risk cover completely.
If we look at economic forecasts like growth rates in income, it also has to face the same challenge; economic actors respond to such signals in various different ways. Here we deal with multiple complexities and monetary policy acts as a balancing force as well.
Take the example of a rising expectation of growth as given in a forecast, it would influence the commercial banks to a far more relaxed lending standards than in a situation where the growth forecast is tempered down. This would lead to a cascade of events where a larger capacity will be built and in no time this would have a reverse effect of lower capacity utilization thus impacting prices and income levels.
The central bank therefore in such a situation would act as a balancing force to bring in some sense of parity by modulating the flow of money so that the situation does not go out of hand.
To keep the results and outcome to follow the forecast, the government will also initiate a slew of measures, which either aid or come in the way of price performance in the economy.
Then you have the cyclical nature of output, where government spending remains cyclical, which means as the growth increases government spending also increases. What this does is that in down cycle, when the actual spending of the private sector slows down the government spending also follows suit, which is just the reverse of what the need of the hour could be.
When government actions become unpredictable and if too many actions are on the anvil, it becomes even more complex for respondents to take a definitive stand.
The forecasters therefore have to keep up their ante by continuing to factor in the effect of new information as the actors respond to the forecasts positively or negatively. The forecasts by themselves are a signal to act, inaccuracy is inherent in them.
This has wide spread implications in other walks of life as well. Think of a forecast on how many years a person would live; forecast of a shorter life span induces serious changes in the way a person would like to lead his balance days. He could be far more objective or productive given his balance days, whereas a longer life would make him space out his actions in many different ways.
It is good to know a forecast, but it is another matter how we respond, individually or collectively. This response is far more unpredictable and less deterministic. This is where the probabilistic domain takes over and models become more complex.
Second order complexities like economic forecasting or price forecasting better be seen through a probabilistic model. But human brains are far more comfortable with a deterministic outlook; the markets are similar, they want to keep matters simple. This is where the friction generates peculiar results, some of which are so baffling that it better be left to simple explanations.
I am reminded of Jerome K Jerome’s example, when he wrote, “reading that there will be no rain today, I took my umbrella with me.”