Research

     Paper Publications
    
Abstract
The paper addresses an inefficiency in the traditional approach of modeling the tail risk, particularly the 1-day-ahead forecast of Value-at-Risk (VaR), using Extreme Value Theory (EVT) and a GARCH model. Specifically, I apply both models to the daily losses in the stock markets of major countries, including the U.S., U.K., China and Hong Kong, between 2006 and 2015, and compare the relative forecasting performance. The paper differs from other studies in two important ways. First, it incorporates an asymmetric shock to volatility in the financial time series. Second, it applies a skewed fat-tailed return distribution using the Generalized Error Distribution (GED). The back-testing result shows that, on the one hand, the conditional EVT performs equally well relative to a GARCH model under the Generalized Error Distribution; but on the other hand, the Exponential GARCH based model is the best performing one in Value-at-Risk forecasting, because it not only correctly identifies the future extreme loss, but more importantly, its occurrence is independent.

Keywords: Value-at-Risk, Extreme Value Theory, Conditional EVT and Backtesting

JEL Classifications: C53, G32

Work In Progress 

  •     "Optimal Portfolio Selection Strategy using Monte-Carlo Simulation--Evidence from Asian Markets" (Under Review) [Link]
        Abstract: Recently the portfolio optimization has become widely popular in risk management, and the common practice is to use mean-variance or Value-at-Risk(VaR), despite the VaR an incoherent risk measure because of the lack of sub-additivity. This has led to the emergence of the conditional value-at-risk (CVaR), consequently, the gradual development of mean-CVaR portfolio optimization. To seek an optimal portfolio selection strategy and increase the robustness of the result, the paper studies the performance of portfolio optimization in Asian markets using a Monte-Carlo simulation tool, creates a variety of randomly selected portfolios that consists of Asian ADRs listed in NYSE from 2010 to 2016, and applies both optimization frameworks with different skewed fat-tailed distribution function, namely the Generalized Hyperbolic(GH) and skewed-T distribution. The main result shows that Generalized Hyperbolic distribution produces the lowest risk under a given rate of return, while the skewed-T distribution creates a diversification allocation outcome similar to that of historical simulation.

Keywords:  Portfolio Optimization, Mean-CVaR, Monte Carlo, Hyperbolic Distribution, Skewed Fat-tailed Distribution

JEL Classification: C53, C61, F37, G11

  •   "Is Conditional Value-at-Risk A Better Measure than Variance In Portfolio Optimization?  "
       Abstract: One of the most popular tools in portfolio optimization is the mean-variance, that is, to minimize the volatility of the portfolio. But the main drawback of this method is that the variance does not differentiate positive and negative shocks. Because portfolio optimization is about minimizing the risk under target return over a certain time horizon, the risk is measured as a potential loss due to negative shocks. The paper proposes taking a different approach, using mean-CVaR and examines its performance against mean-variance. In this paper, we construct a representative portfolio that is a good mix of the top 24 companies in the Asian market in terms of market cap from 2006 to 2015, and we create a portfolio allocation strategy that achieves the smallest risk under the constraint. The purpose is to show how allocation strategy differs between mean-variance and mean-CVaR. Also, the paper investigates whether CVaR is a robust measure of the financial crisis by analyzing the portfolio performance pre-and post-crisis. Lastly, the paper evaluates the comparative advantage of mean-CVaR versus mean-variance.

   Keywords: Conditional Value-at-Risk, Portfolio Optimization, Variance, Backtesting

   JEL Classification: G11, G32, G17

  •    "The Determination of Chinese Business Cycle Synchronization at Province Level" 
Abstract: The business cycle synchronization has been an important tool in studying the macroeconomic co-movement, while most of the existing literature studies the Chinese business cycle at a national level using aggregated national data, primarily the real GDP. This paper aims to understand the business cycle at a provincial level, that is, how the economic co-movement behaves across different regions and provinces. Using a gravity model with the CEIC data between 1993 and 2013, the main contribution of the paper could be summarized in two aspects. First, it describes how the synchronization of the business cycle changes over time, and how the synchronization differs across different regions. Second, it identifies what the driving force is responsible for the business cycle synchronization. Also, it answers how the main determinants vary over time, and how they differ across each province. The results show that bilateral gap in GDP, labor force, public infrastructure and freight traffic determines the synchronization of the business cycle at the national level. However, at a regional level, the synchronization is largely affected by its own region’s factor. In other words, there is a clear disparity in explaining the regional business cycle synchronization.

Keywords: Business Cycle, Synchronization, Regional and Provincial Level and Disparity
            
JEL Classifications:  E32, E60
         
  •    "The Comparison between GARCH, EVT, Regime Switching  in Modeling Expected Shortfall "

Comments