Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 22 views

    New Assign

    Uploaded by

    Qianyu Gu
    This document presents calculations and regressions relating to a log-linearized model of stock returns. It begins by log-linearizing an identity relating stock returns, prices, and dividends. Taking a Taylor series expansion leads to an approximation of returns as a function of prices, dividends, and other terms. Regressions are then used to estimate coefficients in the model and compare an implied coefficient to one estimated directly, finding a very small RMSE, supporting the theoretical relationship.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    New Assign

    Uploaded by

    Qianyu Gu
    22 views4 pages
    This document presents calculations and regressions relating to a log-linearized model of stock returns. It begins by log-linearizing an identity relating stock returns, prices, and dividends. Taking a Taylor series expansion leads to an approximation of returns as a function of prices, dividends, and other terms. Regressions are then used to estimate coefficients in the model and compare an implied coefficient to one estimated directly, finding a very small RMSE, supporting the theoretical relationship.

    Original Title

    New-Assign

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document presents calculations and regressions relating to a log-linearized model of stock returns. It begins by log-linearizing an identity relating stock returns, prices, and dividends. Taking a Taylor series expansion leads to an approximation of returns as a function of prices, dividends, and other terms. Regressions are then used to estimate coefficients in the model and compare an implied coefficient to one estimated directly, finding a very small RMSE, supporting the theoretical relationship.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    22 views4 pages

    New Assign

    Uploaded by

    Qianyu Gu
    This document presents calculations and regressions relating to a log-linearized model of stock returns. It begins by log-linearizing an identity relating stock returns, prices, and dividends. Taking a Taylor series expansion leads to an approximation of returns as a function of prices, dividends, and other terms. Regressions are then used to estimate coefficients in the model and compare an implied coefficient to one estimated directly, finding a very small RMSE, supporting the theoretical relationship.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 4

    Question 3.

    1
    Pt +1
    +1
    Pt +1 + Dt +1 Dt +1 Dt +1
    Rt +1= =
    Pt +1 Pt Dt
    Dt

    Log-linearlizing, define:
    r t =log ( Rt )

    d t =log ( Dt )

    pt =log ( Pt )

    r t +1=log
    ( Pt +1
    Dt +1 )
    +1 − ( pt − dt ) + Δ d t+1

    Thus:

    r t +1=log ( 1+ e p t +1 − dt+ 1
    ) − ( pt − d t ) + Δ d t +1
    Taking a Taylor series expansion of the term about a point P/D = exp(p-d):

    log ( 1+ e p t+ 1 −d t+ 1
    )
    So:
    p −d
    log ( 1+ e
    pt+ 1 −d t+ 1
    )=log ( 1+ e p − d ) + e p − d ( p t +1 − d t +1 − ( p − d ) )
    1+ e
    Setting:
    p− d
    e
    ρ= p−d
    1+e

    k =log 1+ ( P
    D )
    − ρ ( p − d )=log ( 1+e p −d ) − ρ ( p − d )

    So we can substitute rho and k to get:

    r t +1=log ( 1+ e p t +1 − dt+ 1
    ) − ( pt − d t ) + Δ d t +1
    r t +1=k + ρ ( p − d ) + ρ ( pt +1 − d t +1 − ( p − d ) ) − ( pt − d t ) + Δ d t +1

    Hence,using the identity can derive the following approximation


    r t +1=k + ρ ( pt +1 − d t +1 )+ Δ d t +1 − ( pt − d t )
    Question 3.2
    By rearranging identity 4 we can get:
    d t − p t=− k +r t +1 − Δ d t +1+ ρ ( p t+1 − dt +1 )

    Then to find analytical relationship between coefficients b_r,b_d,b_y, we need to solve the
    variance of d_t-p_t
    var ( d t − pt )=cov ( d t − p t , d t − pt )

    var ( d t − pt )=cov ( d t − p t , r t +1 ) − cov ( d t − pt , Δ d t +1 ) + ρcov ( d t − pt , pt +1 − d t +1 )

    divide
    cov ( d t − p t , r t +1 ) − cov ( d t − pt , Δ d t +1 )+ ρcov ( d t − pt , pt +1 − d t +1 )
    1=
    var ( d t − pt )

    Then from these three regressions, we can get regression coefficients:


    cov ( d t − p t ,r t+ 1)
    b r=
    var ( d t − pt )

    cov ( d t − p t , Δ d t +1 )
    b d=
    var ( d t − pt )

    cov ( d t − pt , pt +1 − d t +1 )
    b y=
    var ( d t − pt )

    The following results can be obtained by dividing each term in the relationship above by
    the denominator:
    1=br −b d + ρ b y

    Question 3.3
    library(readxl)
    library(dplyr)

    ##
    ## Attaching package: 'dplyr'

    ## The following objects are masked from 'package:stats':


    ##
    ## filter, lag

    ## The following objects are masked from 'package:base':


    ##
    ## intersect, setdiff, setequal, union
    data_3<-read_xls("ps2_data.xls")
    colnames(data_3)[4]<-"Return_inc_dividends"
    colnames(data_3)[5]<-"Return_exc_dividends"
    colnames(data_3)[6]<-"Rf"

    ##Compute ratios (the logarithm of the excess returns, the Dividend Growth,
    D/P)
    data_3 <- data_3 %>%
    filter(year>1946 & year<2010) %>%
    mutate(DP=(1+Return_inc_dividends)/(1+Return_exc_dividends)-1) %>%
    mutate(Excess_r = log(Return_inc_dividends+1)-log(Rf +1)) %>%
    mutate(Div_Growth=(DP/lag(DP))*(Return_exc_dividends+1))
    data_3$Div_Growth[is.na(data_3$Div_Growth)]<-1
    data_3<-data_3 %>%
    mutate(Acc_Div_Growth=cumprod(Div_Growth))

    ##Regression relationships from part ii


    ##'tail' means at r_t+1, 'head' means at r_t
    Regress_r<-lm(tail((data_3$Excess_r),-1) ~ log(head(data_3$DP,-1)))
    Regress_d<-lm(log(tail(data_3$Acc_Div_Growth,-1)/head(data_3$Acc_Div_Growth,-
    1))
    ~ log(head(data_3$DP,-1)))
    Regress_y<-lm(log(tail(data_3$DP,-1))~log(head(data_3$DP,-1)))

    ##Beta Estimates by using direct regression


    sum_r<-summary(Regress_r)
    beta_r<-sum_r$coefficients[2,1]
    sum_d<-summary(Regress_d)
    beta_d<-sum_d$coefficients[2,1]
    sum_y<-summary(Regress_y)
    beta_y<-sum_y$coefficients[2,1]

    ##Implied Beta r calculation using part ii regression


    ##First, we need to calculate rho
    data_3<-data_3 %>%
    summarise(PD=mean(1/DP))
    rho<-data_3$PD/(data_3$PD+1)

    ##Then, we can get b_r by the relationship of part ii


    implied_b_r<-1+beta_d-rho*beta_y

    ##Comparison of the implied return beta and the regression estimated beta
    ##by RMSE, implied_beta is predicted value, regress_beta is actual value
    rmse<-sqrt((implied_b_r-beta_r)^2)
    print(rmse)

    ## [1] 0.01043867
    To sum up, after comparing the implied beta and the estimated regression beta, we
    discovered RMSE is 0.01, which is extremely close to 0 and gives a strong evidence to
    support the relationship of part ii. Hence, this comparison strengthens the blief in stock
    market predictability.

    You might also like