The Geneva Nautical Association is full of commotion during every Bol d’Or Mirabaud, when it opens its doors to the public as a one-off. The only time you could hear a pin drop is when Pierre Eckert took the mic late on Friday afternoon. The expert from MétéoSuisse's regional centre gave the teams a weather forecast for the race. Whether it be trong wind or a gentle breeze, dead calm or a risk of storms, with winds blowing towards Switzerland or France – this information shapes their strategy for the regatta.
In both meteorology and economics, various digital models have been developed to try to forecast the future. To compare practice in these two fields where forecasting is key, Valentin Bissat and Pierre Eckert met at the MétéoSuisse regional centre, based at the headquarters of the World Meteorological Organization in Geneva. This is where meteorologists prepare their weather forecasts for French-speaking Switzerland.
"Throughout time, people have tried to predict the future," the weatherman says. "In ancient Greece, a priestess called the Pythia was Apollo’s Oracle at Delphi. Today, prophesising is done through differential equations and supercalculators." Where does this need to understand the future come from? "People want to know what tomorrow will be like," answers Valentin Bissat. "Forecasts are reassuring and support decision-making."
On a day-to-day basis, weather forecasting shapes people's choices – about what to wear before going out, a weekend barbecue, racing in the mountains or sailing on the lake. "There may also be a direct economic impact, such as in farming, given that you need at least three days of good weather for hay to dry," continues Pierre Eckert.
Economists also use weather forecasts: "For example, we look carefully at the length of winter," explains Valentin Bissat. "If it's particularly long, the construction sector will slow and that will affect economic growth."
In both meteorology and economics, forecasts are basically made using digital models. MétéoSuisse has just launched a new forecast model for the alpine arc, Cosmo 1. Its estimates are accurate to within just one kilometre, compared with two kilometres previously.
"A weatherman uses several models to make a forecast," explains the meteorologist. It's the combination of results from these different models and the professional's experience that make weather forecasting possible."
The weather model works in two stages. "You first need to know the initial state of the atmosphere, that is, the temperature, pressure, wind (strength and direction), humidity (concentration and state) over the whole area covered by the model," notes Pierre Eckert. Observations are made continuously thanks to ground-level sensors, satellites, radar, aircraft and balloons.
"Once we know the initial state at every altitude, we make calculations using physics equations relating to friction, atmospheric forces, thermodynamics and radiation as well as the force of the Earth's rotation, called the Coriolis force," the meteorologist explains.
In contrast to weather forecasting models, which are based solely on the current situation and physics equations, economic models rely on historical data. "In econometrics, we mainly use time series. We look at historical developments and replicate them. We analyse previous behaviour to identify trends. If the conditions at the start are the same, we use this pattern to forecast the future, using probability theory," states Valentin Bissat.
Whereas, meteorology is an exact science using physics equations, economics is a social science based on simplified theories of reality. "We also use differential equations to model the behaviour of complex systems," Mirabaud's economic strategist continues.
"Economic forecasts support investment decisions, in both finance and industry," notes Valentin Bissat. "Positive economic growth forecasts could encourage companies to increase production and therefore to hire or even expand."
Mirabaud has developed digital forecasting models to determine the inflation rate. "These are error correction models. The variables may diverge in the short term, but converge over the long term," explains Valentin Bissat. "We favour a qualitative approach. The digital models we use support decision-making. Members of the Investment Committee analyse the figures and take investment decisions based on their experience."
Economic models can handle large amounts of data. Central banks use the most sophisticated models, called DGSE (dynamic stochastic general equilibrium). These model behaviour at the microeconomic level and aggregate it in a dynamic global model. The results obtained relate to growth, inflation and unemployment. Other models forecast market trends and developments in specific business sectors.
As economics is a social science, human factors can affect results. "The way the results are presented is sometimes more important than the values themselves," warns Mirabaud's economist. "When central banks present their growth forecasts, they try to give positive messages so individual behaviour is not affected by negative interpretations. Weaker than expected growth rates could actually provoke a wave of pessimism, resulting in falling consumption and the economy slowing." This is a parameter that is hard to model.
In economics, forecasts generally cover periods from one week to several years, but accuracy diminishes with the time horizon. In meteorology, the global model gives a forecast of 10 to 15 days and is usually updated every 12 hours. More precise models can give one to fiveday forecasts and are refreshed every few hours. "We can also make six-month forecasts," explains Pierre Eckert. "This works relatively well for the Tropics: 'El Niño' was forecast last winter. These forecasts are embargoed, to prevent speculative moves. Whoever has the information could buy high or low."
Although the models are effective, the reliability of forecasting may evolve further. "Sometimes a one-degree difference in temperature can have repercussions that will change the whole weather system. You get rain when you expected sunshine," the MétéoSuisse expert continues. "We extrapolate from the effect of a small difference in the initial state on the forecast. This is what enables us to give a confidence index. If on day five, for example, the chances the weather will change are very high, this index will be low."
Valentin Bissat explains that such scenarios also occur in economics: "We carry out stress tests. We simulate a rise in unemployment, changes in interest rates and analyse the resilience of a financial institution or a financial tool under these conditions. This is really useful for risk management."
In both economics and meteorology, the models could be improved. "This requires significant investment," emphasises Pierre Eckert, "not only in supercalculators but also to take into account new parameters such as humidity levels in the soil or the volume of foliage." The potential for improvement is also a delicate subject in economics, particularly because it is hard to model economic mechanisms. "In forecasting, having the experience and knowledge needed to interpret the results is what makes the difference," says Mirabaud's economist. For now, no supercalculator can do this.
Valentin Bissat has been an economic strategist within the Strategy Team at Mirabaud Asset Management since 2013 and is a member of the Investment Committee at Mirabaud & Cie SA. He works alongside the Chief Economist, who is in charge of macroeconomic studies and analyses. He is responsible in particular for analysing and drawing up financial indicators relating to the investment strategy. Valentin Bissat has a BSc in Economics from the University of Geneva, a Master's degree in Banking and Finance from the University of St. Gallen (HSG) and is a qualified Chartered Financial Analyst.