Economic forecasting is a highly uncertain activity that aims to identify the future direction of key variables in the economic and financial system to help make decisions on policy setting, investments and business planning.
Forecasting is often wrong, there is no doubt about that. So the aim is not to get the actual numbers spot on, but to find the general trajectories or scenarios and to assess the size of any forecast errors and the reasons behind those errors.
Under normal economic conditions when the economy is growing broadly in line with its underlying, long-term trend, forecasting is relatively easy. The trend is your friend, as we say.
Good economic forecasting becomes particularly necessary when there is an economic shock and a degree of uncertainty such as we have seen — and are seeing — in the current scenario.
In particular, if the shock to the economy is large, it can cause changes in how businesses and consumers are behaving and this structural uncertainty makes good forecasting even more essential to help decision-makers formulate feasible policies and strategies. When they give a “reasonably good” and “reasonably precise” description of the evolution of economic activity, they are useful for making decisions and projections ahead.
Timing and risk play a decisive role in forecasting. When we perform a one year ahead forecast or even longer, the uncertainty is high and the forecaster has to take an active risk in formulating his numbers. As we approach the target date of the forecast, more and more data become available, the main features start to show in the numbers and the range of the possible values of the unknown variables gets narrower.
For example, when we forecast the annual growth of the economy six months ahead, the final outcome is an arithmetic average of two numbers, the first-half outcome and the second-half forecast. The first one is already known.
For forecasts to be useful, we need to distinguish the fine-tuning search of the final value as we are almost there from the more important and longer-term forecast, at least a year ahead. In the second case, we need to consider different scenarios and run different models and analytics in order to make the forecast.
For short-term forecasting, it is possible to anticipate where the final figure will be because most of its value is already “in the numbers” from earlier periods. For example, it takes a huge outlier, either positive or negative, in the fourth quarter of a year to change the average of the preceding three quarters.
When forecasting is based on scenario analysis, we get quite important variability among the different forecasts, depending on the subjective evaluation of how a very complex and dynamic economy will evolve. Where we move closer to the target, we see that the different forecasts tend to converge progressively towards the final value.
From an informative point of view, longer-term scenario-based forecasting adds much more value, since the decisions to be made and the policy measures to be implemented take time to produce their effects.
In advanced economies, a large pool of private forecasters produces a wide range of these scenario-based forecasts, putting together many different assumptions and evaluations coming from the market. These forecasts can come from financial institutions, policy institutions, academic forecasters or government and semi-government agencies. The range and the central tendency of the forecasts constitute an important source of information for the government and the central bank, as well as the wider domestic and international markets.
Furthermore, private forecasters do not have the constraints and risk aversion in taking unconventional positions. They can quickly change or correct their views, sometimes in quite a decisive way, something that government institutions try to avoid. So government institutions have a different timing and risk attitude in formulating forecasts.
To illustrate this, take a look at the table which shows the recent forecasts of major institutional forecasters for Malaysia.
The important difference between the forecast and the final growth outcomes tells us two things. First, there is a big discrepancy between the forecasts and the outcomes, especially 12 months in advance. Second, the process of updating the forecasts to a completely different set of final results is systematically and inefficiently slow.
Compared to the 1990s, for example, it appears that we are living in a world that is increasingly subject to structural shocks, each of a different nature. For example, the 2001 dotcom crisis; the 2008 sub-prime crisis; the 2020 Covid-19 crisis and now, the 2022 Russian-Ukraine crisis.
In this environment, economists and, especially, economic institutions have to adopt a different approach to figure out possible scenarios that are more flexible and rely on a wider set of models and tools.
The central point is that traditional approaches to economic forecasting often assume that there is an underlying process in the economy, a trend or an “initial equilibrium”, and try to estimate how long short-term deviations from that trend will last before the economy “gets back to normal”.
The idea of the easy “back to normal” process is misleading and does not allow us to understand how the changes and reactions in the behaviour of businesses and consumers are affecting the new situation. It also may overlook the changing context in which economic and financial decisions are being made if the economic structure has changed.
The actual adjustment of the economy is path dependent, that is to say, where it will go depends on where it came from and how it got to where it came from. It is not an easy convergence towards the previous situation in which the only uncertainty is the time it takes to get “back to normal”.
The forecasts that are based on this type of approach rely on looking back to how things were and assuming that we will adapt back to that point in the future. In this respect, they are backward-, not forward-looking, and so, are always late and always wrong.
When looked at from this perspective, we can see that it is important for Malaysia to have an ample set of independent forecasters specialised in scenario analysis who can produce their forecasts transparently, creating timely information and explanations of their analysis and the way the different models or tools they use work to give their projections.
Bank Negara Malaysia and the government also have to do their independent work in assessing and comparing market forecasts to their institutional forecasts, having a lag in the timing but higher precision in the fundamental recommendation of policy and to inform the economic agents
The current situation in which the central bank or government lead the forecasting and market forecasters follow is not an efficient one. The government should incentivise the activities of independent forecasters to create value for the whole system so that everyone can do his own job, with different perspectives and responsibilities.
Professor Paolo Casadio is an economist at HELP University. Professor Geoffrey Williams is an economist and provost for research and innovation at Malaysia University of Science and Technology.