Juan Arismendi Zambrano, D.Phil.
This is my personal webpage. Here, you will find a mix of different topics between my academic, industry, and entrepreneur careers and my personal life.
​
I'm Visiting Research Fellow from ICMA Centre since 2013 and from Catolica Lisbon since 2024..
​
PROJECTS - TOPICS OF RESEARCH
EQUITY RISK PREMIUM PREDICTABILITY
We illustrate the role of left tail dependence variables – left exceedance correlation (LEC) and left tail mean (LTM ) – in equity risk premium (ERP) predictability. LEC and LTM measure the average of pairwise left tail dependency among major equity sectors incorporating shocks that are imperceptible at the aggregate level. LEC and LTM, as well as the variance risk premium, significantly predict the ERP in- and out-of-sample, which is not the case with commonly used predictors. We find this predictability is the result of pro-cyclical shocks in a stable business cycle. This paper contributes to the ongoing debate on ERP predictability.
SPARSE FACTORS PRICING
The results on approximate factor pricing models from Chamberlain and Rothschild (1983) triggered a machinery of theoretical and empirical results based on the natural rate assumption (Connor and Korajczyk ,1986; Stock and Watson, 2002; Bai, 2003). Nevertheless, the natural rate assumption is not required in the original model of Chamberlain and Rothschild (1983), and seems instead too restrictive. In this project we develop statistical tests to estimate the diffuse/sparse vs. strong/semi-strong properties of known factors with an alpha that is allowed to variate below the natural rate.
CENTRAL BANK'S COMMUNICATIONS - NLP
We develop a market model for monetary policy stability of decisions taken by the executive board of a Central Bank (Federal Reserve, the European Central Bank and Banco Do Brasil). The Central Bank institution is considered as an agent that maximises a bivariate function of stability: i) price stability and ii) financial system stability. The results show that the increased frequency, reduced sentiment and clear stance of the communications' have a significant impact int the stability of the monetary policy by improving the effects of communication.
BIG DATA - POLITICAL STABILITY: FINANCIAL MARKET APPROACH
Decisions in politics regarding the economy are held by people, decision takers–ministers, based on knowledge, research, reports and a particular set of facts. Those kind of decisions are then implemented as a set of economic policies, following previous experiences. Developed nations have advanced in their democratic systems for the decision-taking process, adopting the wave of the technology for having as an input some sort of the feedback of citizen’s impressions about economic policies. Nevertheless, two aspects of actual democratic systems are particularly relevant, i) the purely expert-based management system with no citizen
feedback has not changed in the majority of non-developed countries, and ii) financial markets technology have advanced in a very fast way, where most decisions are assisted and at sometimes conducted by machine intelligent systems, avoiding all the decision-taking ethical problems related to humans, such as corruption.
​
Research of social media interaction in economic decision-taking problems using financial markets technology may assist the evolution of the less developed political decision systems. Transparency on a real-time basis of the citizen’s perception of economic decisions by the executive may introduce a policy watcher and prevent from extreme economic measures that depart from the public will from which the executive power is inherited.
​
​
We propose an index for government stability and political risk that includes three components that measure i) sudden and unexpected change of executive government, ii) increase of the government economic decisions uncertainty, and iii) increase in the foreign risk perception of government ability to fulfill citizens expectations and foreign expectations. An index of the President’s popularity is used for measuring i), the Economic Political Uncertainty Index (EPU) of Baker et al. (2015) is used for measuring ii), and an index of the country macroeconomic data and foreign debt riskiness is used for measuring iii). All these measures are components of a vector representing the multivariate measure of government stability and political risk.
Our model assume that decisions are taken in the best effort of the agent (executive) to improve its popularity and its foreign risk perception. As a consequence, popularity is a direct measure that links executive decisions with citizen’s well-being. Pastor and Veronesi (2013) and Pastor and Veronesi (2012) have modelled the effects of political decisions in stock prices. Pastor and Veronesi i) proposed and tested in a theoretical model of general equilibrium for stock prices response to future political decisions–signaling, ii) analysed an equilibrium model where stock price reacts to effective political decisions. We adopt Pastor and Veronesi results as it incorporates a endogenous learning process where the executive government is feed from price impact of its own decisions.
QUANTITATIVE FINANCE - RISK MANAGEMENT
Our research on moments methods and applications to finance is motivated on skewness and kurtosis importance in asset returns. Financial theory of risk and return has been developed over first two moments of distributions: mean return and standard deviation (portfolio optimisation, CAPM, risk management). But several studies as Kraus and Litzenberger (1976) show the importance of skewness to asset valuation. The research of Ang and Chen (2002) and Longin and Solnik (2001) tests asymmetries of equity market returns. They found the markets are asymmetric during bullish and bearish regimes. However, correlations are complex measures in a multivariate setting. Higher-order moments like skewness and kurtosis are also complex measures important for option pricing and risk management. It is our intention to develop parametric statistical models for empirical researchers. Additional research is developed in credit risk and derivatives pricing theory.
LEARNING THE BEHAVIOUR OF TRADERS - ALGORITHMIC TRADING
Technical analysis is regarded as one of the most used trading techniques in the industry. In the high-frequency level, programmed rules similar to technical analysis are used by super computers to trade the markets, based on optimal dynamic programming theory and Bell's equation. There exist advanced statistical methodologies to asses the performance when using technical analysis for trading, in terms of information's aggregation and predicting abilities.
Nevertheless, practitioners rarely asses performance by such statistical methodologies, generating the disagreement of the results by the academics. Friesen et al. (2009) provides a review of the literature on technical analysis `patterns', and test a behavioural model that analyse and describe the success of technical analysis trading rules based on certain patterns. The results of Friesen et al. (2009) explain that common patterns applied by traders, like `head-and-shoulders', generate autocorrelation that could be exploited by trading systems. The traditional approach of `buy-and-hold' is constantly tested against an active strategy of trading the assets. Recent advances in the literature has been done by Zhu and Zhou (2009), where technical analysis is used into a sub-optimal dynamic portfolio strategy, where the allocation occurs between a fixed-income and a equity instrument on a partial basis, i.e., the allocation of the wealth is less than certain percentage (80%), when the technical analysis indicator signals a buying transaction.
This strategy is tested against the traditional total allocation (100%) in each asset (fixed or variable) of an optimal dynamic strategy with certain information about the market. The results show that under uncertainty of future information, the sub-optimal strategy of partial allocation using technical analysis over-performs the optimal dynamic strategy.
We use statistical findings of the micro-structure markets with machine learning tools to develop optimal investment strategies. The optimal strategies will be used by a computer as in the technical analysis literature, assuming that markets are not Efficient as in Fama (1970) but Adaptive Efficient as in Lo (2004).
HIGH FREQUENCY TRADING
Recent empirical findings suggest that events like the `Flash-Crash' of May, 2010, delivered levels of multivariate kurtosis never observed in the past. This finding attempt against the Efficient Market Theory of Fama (1970) and in favor of the Adaptive Market Hypothesis of Lo (2004), having important implications in the understanding of the markets at the micro-structure level.