Didier Darcet (of Gavekal Intelligence Solutions, abbreviated to Gavekal IS) explains how, from being a client of Gavekal who he was first, he came to work with Charles Gave and his son Louis, after attracting the attention of the latter and their collaborators on the interest represented by the scientific approach in the analysis of financial risks. The aim was to test, in a rational manner, the validity of the hypotheses developed by Gave, over many years, based on his experience and his understanding of financial markets, concerning for example the influence of growth or inflation on stocks.
Databases have grown considerably over the last decades, as has processing capacity, which allows hypotheses to be scientifically tested, which must be kept in mind. The required state of mind consists of approaching economic and financial questions logically, so as to understand their causal relationships, by having an intellectual approach to learning and research methods. In artificial intelligence, it involves building neural networks capable of learning by themselves from input variables, responses and verifications.
Rather than simply using artificial intelligence techniques, it was necessary to control the process internally to understand, for example, the weight given to a given economic variable, and to verify its consistency in relation to economic logic. The experiment first consisted of observing the network's behavior in relation to the same parameters in three different countries (the United States, Japan and Germany), before extending the observation to forty different countries. The right decision-making method is therefore not necessarily the one that feeds the network with as much data as possible in order to make the formulas more complex, but rather the one that focuses on finding the determining parameters.
There is, therefore, a work of elimination of information. However, this work is not enough to overcome uncertainty. When we talk about a rising market, we must first define the terms, and ask ourselves if it is significant or if it is just noise. We then base ourselves on a signal, and we study the relationships of this signal with the noise which surrounds it. For example, the S&P 500, for 70 years, has gained three basis points per day (0.03%), with a volatility of 1% (or thirty times more). In this case, trading the market on a daily basis means looking for a signal that is 97% noisy. Even if we extend the time range, the uncertainty remains high and it is honest to recognize this. It is better, in fact, to say that we do not know because, when we know, the decision-making is all the clearer.
The experts have developed what they call, in professional jargon, the Quant pole. The idea is to try to take the data not to make predictions but to see, for example, whether logical expectations can arise from repetitive constellations of events. We cannot predict anything, but it is possible to measure a state, as in physics, and we know, from a good description, that certain paths are impassable. The inflationary depression that hit England in the 1970s was, as its name suggests, marked by both inflation and depression, resulting in falling profits. Buying shares, in such a context, amounts to wanting to get out of money: it is an escape. In the financial professions, hundreds of billions of dollars are spent on research which, surprisingly, concerns 20% of expected results (80% remaining in a gray area).
The aim of economic verification and fundamental research is to make money, with a significant probability of error: it is a school of modesty. Statistical finance began at the beginning of the twentieth century with Louis Bachelier (1870-1946). In a market, we observe trends and, around these trends, random agitation, which is similar to Brownian movement in physics. The method adopted is similar to a simulation of dice rolls. Rolling a die a million times increasingly confirms, according to the Gaussian curve, the probability of the 6 coming out one time in six. Knowing that, in a given portfolio, there can be, let's say, 1 chance in 10 of losing more than 10%, we become able to calculate the price of an option or that of the risk (which is also useful in insurance).
The Bachelier equations being approximately correct at the first order, they are not correct at the second order. The importance of the second order can be understood in light of the difference between Newton's physics and Einstein's physics, which appears when we take into account the fact that certain objects have a high speed, which can have practical consequences (such as geolocation of mobile phones). This observation leads to the specific characteristics of financial assets, and to the notions of fragility and anti-fragility as developed by Nassim Taleb. Apple is a fragile company in this sense, because its income depends on adaptation to a defined environment, which can always experience upheavals likely to call this income into question.
Ricardian growth is based on the availability of raw materials for a densely populated area, leading to better use of time, labor and energy: this is the fragile example of the corridor between Saint Petersburg and India, which is not based on any invention, only copying what was already done in China, Japan, the United States and Europe in the nineteenth century. According to the law of comparative advantages of David Ricardo (1772-1823), everyone continues to do what they do best. Schumpeterian growth is based on invention: this is the anti-fragile example of Elon Musk. The Malthusian model is based on the idea that there will not be enough resources for everyone. Today, by mathematically taking up Nassim Taleb's model, we can precisely measure the degree of fragility (decreasing in the event of agitation) or anti-fragility (increasing in the event of agitation) of any asset.
A start-up needs initial funding to try to create its disruptive products, which will cause chaos. During this phase, it has no turnover. Then, it remains a loser if nothing ultimately happens, or becomes a winner if it manages to break the other's income. Being able to measure these fragility and anti-fragility ratios precisely, we can say for example that the fragility of the S&P 500 is 3, that of Google -0.5 at the first order. At the second order, we know that a fragile asset will earn money under stable conditions. This market asymmetry was unknown in Bachelier’s time. It is comparable to changes of state in thermodynamics (like liquid water transforming into vapor at 100 degrees).
There are therefore two states in a portfolio: the usual one, where nothing happens, and the one, on the sides, where we must be extra wary, because that is where the possible losses will be the greatest. Gavekal IS has successfully conceptualized and measured these two states. A key notion is that of correlation. When the crowd increases on the platforms during rush hours, single lines are naturally created to regulate passage. The dispersion of results in a system according to agitation led experts to take up the idea of Harry Browne's (1933-2006) permanent portfolio, the one that would get through all situations.
The question is whether you want your portfolio to be comparable to a Jeep (to stay rich) or a Ferrari (to get rich). In summary, there are four main asset classes to balance in a portfolio, depending on the risks one is willing to take: stocks, government bonds, cash and gold (the latter two corresponding to a present value, while stocks and bonds are oriented towards the future). Cash has no volatility, bonds have 10% volatility, stocks have 15% and gold has 20%. We noticed that, in a portfolio oriented towards sustainability, the result of the balancing could give a volatility of only 6.5% instead of the expected 12%, which confirms the interest of this model, which can bring in 3% to 4% per year.
Balancing must precede a more tactical phase consisting for example, depending on economic or monetary circumstances, either of replacing bonds (according to profitability criteria) or of reducing shares (according to security criteria). The economic dial needs to be monitored, with the two main drivers of stock valuation being inflation and growth. In a Ricardian growth zone like India in 2023, redistribution is phenomenal, however the Americans still master information processing (including, precisely, artificial intelligence), which remains another determining aspect of the creation of value.
Databases have grown considerably over the last decades, as has processing capacity, which allows hypotheses to be scientifically tested, which must be kept in mind. The required state of mind consists of approaching economic and financial questions logically, so as to understand their causal relationships, by having an intellectual approach to learning and research methods. In artificial intelligence, it involves building neural networks capable of learning by themselves from input variables, responses and verifications.
Rather than simply using artificial intelligence techniques, it was necessary to control the process internally to understand, for example, the weight given to a given economic variable, and to verify its consistency in relation to economic logic. The experiment first consisted of observing the network's behavior in relation to the same parameters in three different countries (the United States, Japan and Germany), before extending the observation to forty different countries. The right decision-making method is therefore not necessarily the one that feeds the network with as much data as possible in order to make the formulas more complex, but rather the one that focuses on finding the determining parameters.
There is, therefore, a work of elimination of information. However, this work is not enough to overcome uncertainty. When we talk about a rising market, we must first define the terms, and ask ourselves if it is significant or if it is just noise. We then base ourselves on a signal, and we study the relationships of this signal with the noise which surrounds it. For example, the S&P 500, for 70 years, has gained three basis points per day (0.03%), with a volatility of 1% (or thirty times more). In this case, trading the market on a daily basis means looking for a signal that is 97% noisy. Even if we extend the time range, the uncertainty remains high and it is honest to recognize this. It is better, in fact, to say that we do not know because, when we know, the decision-making is all the clearer.
The experts have developed what they call, in professional jargon, the Quant pole. The idea is to try to take the data not to make predictions but to see, for example, whether logical expectations can arise from repetitive constellations of events. We cannot predict anything, but it is possible to measure a state, as in physics, and we know, from a good description, that certain paths are impassable. The inflationary depression that hit England in the 1970s was, as its name suggests, marked by both inflation and depression, resulting in falling profits. Buying shares, in such a context, amounts to wanting to get out of money: it is an escape. In the financial professions, hundreds of billions of dollars are spent on research which, surprisingly, concerns 20% of expected results (80% remaining in a gray area).
The aim of economic verification and fundamental research is to make money, with a significant probability of error: it is a school of modesty. Statistical finance began at the beginning of the twentieth century with Louis Bachelier (1870-1946). In a market, we observe trends and, around these trends, random agitation, which is similar to Brownian movement in physics. The method adopted is similar to a simulation of dice rolls. Rolling a die a million times increasingly confirms, according to the Gaussian curve, the probability of the 6 coming out one time in six. Knowing that, in a given portfolio, there can be, let's say, 1 chance in 10 of losing more than 10%, we become able to calculate the price of an option or that of the risk (which is also useful in insurance).
The Bachelier equations being approximately correct at the first order, they are not correct at the second order. The importance of the second order can be understood in light of the difference between Newton's physics and Einstein's physics, which appears when we take into account the fact that certain objects have a high speed, which can have practical consequences (such as geolocation of mobile phones). This observation leads to the specific characteristics of financial assets, and to the notions of fragility and anti-fragility as developed by Nassim Taleb. Apple is a fragile company in this sense, because its income depends on adaptation to a defined environment, which can always experience upheavals likely to call this income into question.
Ricardian growth is based on the availability of raw materials for a densely populated area, leading to better use of time, labor and energy: this is the fragile example of the corridor between Saint Petersburg and India, which is not based on any invention, only copying what was already done in China, Japan, the United States and Europe in the nineteenth century. According to the law of comparative advantages of David Ricardo (1772-1823), everyone continues to do what they do best. Schumpeterian growth is based on invention: this is the anti-fragile example of Elon Musk. The Malthusian model is based on the idea that there will not be enough resources for everyone. Today, by mathematically taking up Nassim Taleb's model, we can precisely measure the degree of fragility (decreasing in the event of agitation) or anti-fragility (increasing in the event of agitation) of any asset.
A start-up needs initial funding to try to create its disruptive products, which will cause chaos. During this phase, it has no turnover. Then, it remains a loser if nothing ultimately happens, or becomes a winner if it manages to break the other's income. Being able to measure these fragility and anti-fragility ratios precisely, we can say for example that the fragility of the S&P 500 is 3, that of Google -0.5 at the first order. At the second order, we know that a fragile asset will earn money under stable conditions. This market asymmetry was unknown in Bachelier’s time. It is comparable to changes of state in thermodynamics (like liquid water transforming into vapor at 100 degrees).
There are therefore two states in a portfolio: the usual one, where nothing happens, and the one, on the sides, where we must be extra wary, because that is where the possible losses will be the greatest. Gavekal IS has successfully conceptualized and measured these two states. A key notion is that of correlation. When the crowd increases on the platforms during rush hours, single lines are naturally created to regulate passage. The dispersion of results in a system according to agitation led experts to take up the idea of Harry Browne's (1933-2006) permanent portfolio, the one that would get through all situations.
The question is whether you want your portfolio to be comparable to a Jeep (to stay rich) or a Ferrari (to get rich). In summary, there are four main asset classes to balance in a portfolio, depending on the risks one is willing to take: stocks, government bonds, cash and gold (the latter two corresponding to a present value, while stocks and bonds are oriented towards the future). Cash has no volatility, bonds have 10% volatility, stocks have 15% and gold has 20%. We noticed that, in a portfolio oriented towards sustainability, the result of the balancing could give a volatility of only 6.5% instead of the expected 12%, which confirms the interest of this model, which can bring in 3% to 4% per year.
Balancing must precede a more tactical phase consisting for example, depending on economic or monetary circumstances, either of replacing bonds (according to profitability criteria) or of reducing shares (according to security criteria). The economic dial needs to be monitored, with the two main drivers of stock valuation being inflation and growth. In a Ricardian growth zone like India in 2023, redistribution is phenomenal, however the Americans still master information processing (including, precisely, artificial intelligence), which remains another determining aspect of the creation of value.