STA457 数值分析

STA457 Time Series Analysis Assignment 1 (Winter 2019)
Jen-Wen Lin, PhD, CFA
Date: February 07, 2019
Please check in Quercus regularly for the update of the assignment.
Background reading:

  1. Assignment and solution (Fall 2018)
  2. Moskowitz et al. (2012), “Time series momentum”, Journal of Financial Economics
    General instruction
    § Download daily data of 30 constituents in the Dow Jones (DJ) index from 1999 December to
  3. December. Please see for the list of DJ
    § Calculate the performance based on a 60-month rolling window and rebalance the portfolio
    monthly but calibrate/estimate parameters () at the end of each year.
    § Performance: Annualized expected return, annualized volatility (standard deviation), and
    Annualized Sharpe ratio. Annualization is done using the squared root of time. Use Sharpe
    ratio as example
    where assume that annual risk free rate , = 0.02 and ) is the sample mean of monthly
    strategy returns and ./ is the monthly volatility.
    A. Technical trading rule
    1) Find the optimal double moving average (MA) trading rules for all 30 DJ constituents
    (stocks) using monthly data.
    Hint: see Assignment (Fall 2018) for more details.
    Copyright Jen-Wen Lin 2019
    2) Construct the equally weighted (EW) and risk-parity (RP) weighted portfolio using all
  4. DJ constituents. Summarize the performances of EW and RP portfolios (trading
    Hint: For simplicity, assume the correlations among stocks are zero when
    constructing the risk-parity portfolio.

    BCD #D3B#E

    /G = ∑ H IJKL
    ∑ IM NO KL M; is defined in Equation (1) (see question B)
    B. Time Series Momentum
    1) Calculate the ex-ante volatility estimate 3 for all 30 DJ constituents using the
    following formula:
    R = 261 T(1)X(2)
    where the weights X
    (1) add up to one, and
    ;,3 is the exponentially weighted
    average return computed similarly.
    Hint: Solve using
    T(1XR\8XF8= 1and;,3 = T(1)XR\8XF8
    Copyright Jen-Wen Lin 201932) Consider the predictive regression that regresses the (excess) return in month on
    its return lagged months, i.e.
    where ;,3 denotes the -th stock in the DJ constituents and in the prediction
    regression, returns are scaled by their ex-ante volatilities ;,3=6. Determine the
    optimal for both predictive regressions for all 30 DJ constituents.
    Remark: For simplicity, students only need to consider Equation (4) in this question
    and use R-squared to evaluate the predictive regression.
    3) Consider a time series momentum trading strategy by constructing the following