Code for preliminary Yesteryear Bias results.

1 prog--yesteryear-bias--14apr2017.R

## #########################################################################
## #########################################################################
## @purpose: plot rate at which WSJ articles about the S&P 500 each month
##           mention horizons ranging from 1 day to 1 year over the period
##           from JAN 1995 to DEC 2016.
## -------------------------------------------------------------------------
## @author:  alex chinco
## @date:    14-APR-2017
## #########################################################################
## #########################################################################





## #########################################################################
## #########################################################################
## @section: prep workspace
## #########################################################################
## #########################################################################

rm(list=ls())

options(width=200, digits=6, digits.secs=6)

library(foreign)
library(lubridate)
library(grid)
library(plyr)
library(ggplot2)
library(RColorBrewer)
library(scales)   
library(reshape)

library(foreach)

library(tikzDevice)
options(tikzLatexPackages =
        c("\\usepackage{tikz}\n",
          "\\usepackage[active,tightpage,psfixbb]{preview}\n",
          "\\PreviewEnvironment{pgfpicture}\n",
          "\\setlength\\PreviewBorder{0pt}\n",
          "\\usepackage{amsmath}\n",
          "\\usepackage{xfrac}\n"
          )
        )
setTikzDefaults(overwrite = FALSE)

scl.str.DAT_DIR  <- "~/Dropbox/research/technical_notes/"
scl.str.FIG_DIR  <- "~/Dropbox/research/technical_notes/"





## #########################################################################
## #########################################################################
## @section: load data on S&P 500 return
## @source:  WRDS
## #########################################################################
## #########################################################################

mat.df.RET        <- read.csv(paste(scl.str.DAT_DIR, "data--sp500-return.csv", sep = ""),
                              stringsAsFactors = FALSE,
                              header           = TRUE
                              )
names(mat.df.RET) <- c("t", "r")
mat.df.RET$t      <- as.Date(as.character(mat.df.RET$t), format = "%Y%m%d")
mat.df.RET$t      <- format(mat.df.RET$t, "%Y-%m")
mat.df.RET$t      <- as.Date(paste(as.character(mat.df.RET$t), "01", sep="-"), format = "%Y-%m-%d")


scl.int.T       <- dim(mat.df.RET)[1]
mat.df.RET$rLag <- NA
mat.df.RET$rFwd <- NA
for (t in 13:(scl.int.T-12)) {
    mat.df.RET$rLag[t] <- prod(1 + mat.df.RET$r[(t-1):(t-11)])^(12/11) - 1
    mat.df.RET$rFwd[t] <- prod(1 + mat.df.RET$r[(t+1):(t+12)]) - 1
}
mat.df.RET$rL     <- (mat.df.RET$r < 0)
mat.df.RET$rLagH  <- (mat.df.RET$rLag > 0)





## #########################################################################
## #########################################################################
## @section: plot S&P 500 return, 1997-99
## #########################################################################
## #########################################################################

if (TRUE == FALSE) {
    mat.df.PLOT      <- mat.df.RET[, c("t", "r")]
    mat.df.PLOT      <- mat.df.PLOT[mat.df.PLOT$t <= as.Date("1999-09-01"), ]
    mat.df.PLOT      <- mat.df.PLOT[mat.df.PLOT$t >= as.Date("1997-09-01"), ]
    mat.df.PLOT$rCum <- cumprod(1 + mat.df.PLOT$r)/(1 + mat.df.PLOT$r[1]) - 1
    
    
    theme_set(theme_bw())
    
    
    scl.str.RAW_FILE <- paste("plot--sp500-return--14apr2017", sep = "")
    scl.str.TEX_FILE <- paste(scl.str.RAW_FILE,'.tex',sep='')
    scl.str.PDF_FILE <- paste(scl.str.RAW_FILE,'.pdf',sep='')
    scl.str.AUX_FILE <- paste(scl.str.RAW_FILE,'.aux',sep='')
    scl.str.LOG_FILE <- paste(scl.str.RAW_FILE,'.log',sep='')
    
    
    tikz(file = scl.str.TEX_FILE, height = 2, width = 4, standAlone=TRUE)    
    
    
    obj.gg2.PLOT <- ggplot(data = mat.df.PLOT, aes(x = t, y = rCum))
    obj.gg2.PLOT <- obj.gg2.PLOT + geom_path(size = 3)
    obj.gg2.PLOT <- obj.gg2.PLOT + ggtitle("S\\&P 500 Cumulative Return", subtitle = "September `97 to September `99")
    obj.gg2.PLOT <- obj.gg2.PLOT + coord_cartesian(ylim = c(-0.04, 0.46), expand = TRUE)
    obj.gg2.PLOT <- obj.gg2.PLOT + scale_y_continuous(breaks = round(mat.df.PLOT$rCum[c(1, 13, 25)], 2))
    vec.dt.BREAKS <- as.Date(c("1997-09-01", "1998-09-01", "1999-09-01"))
    vec.dt.LABELS <- c("S`97", "S`98", "S`99")
    obj.gg2.PLOT <- obj.gg2.PLOT + scale_x_date(breaks = vec.dt.BREAKS, labels = vec.dt.LABELS)
    
    
    obj.gg2.PLOT <- obj.gg2.PLOT + theme(plot.margin       = unit(c(0.15,0.15,0.15,0.15), "lines"),
                                         axis.text         = element_text(size = 12),
                                         axis.title        = element_blank(),
                                         plot.title        = element_text(size = 18),
                                         panel.grid.minor  = element_blank(),
                                         legend.position   = "none"
                                         )
    
    
    print(obj.gg2.PLOT)
    dev.off()
    
    
    system(paste('lualatex', file.path(scl.str.TEX_FILE)), ignore.stdout = TRUE)
    system(paste('rm ', scl.str.TEX_FILE, sep = ''))
    system(paste('mv ', scl.str.PDF_FILE, ' ', scl.str.FIG_DIR, sep = ''))
    system(paste('rm ', scl.str.AUX_FILE, sep = ''))
    system(paste('rm ', scl.str.LOG_FILE, sep = ''))
}





## #########################################################################
## #########################################################################
## @section: load data on news horizon
## @source:  ProQuest
## #########################################################################
## #########################################################################

mat.df.WSJ        <- read.csv(paste(scl.str.DAT_DIR, "data--sp500-horizon-mentions.csv", sep = ""),
                              stringsAsFactors = FALSE,
                              header           = FALSE
                              )
mat.df.WSJ        <- mat.df.WSJ[, 1:3]
names(mat.df.WSJ) <- c("horizon", "t", "newsCount")
mat.df.WSJ        <- mat.df.WSJ[mat.df.WSJ$horizon %in% c("day", "week", "quarter", "year", "any"), ]


mat.df.WSJ$t         <- as.Date(paste(mat.df.WSJ$t, "01", sep = ""), format = "%Y%m%d")
mat.df.WSJ$newsCount <- as.integer(mat.df.WSJ$newsCount)


mat.df.TOTAL        <- mat.df.WSJ[mat.df.WSJ$horizon == "any", c("t", "newsCount")]
names(mat.df.TOTAL) <- c("t", "newsCountTot")


mat.df.WSJ                         <- mat.df.WSJ[mat.df.WSJ$horizon != "any", ]
vec.str.LE_W                       <- c("day", "week")
vec.str.GE_Q                       <- c("quarter", "year")
vec.int.ROWS                       <- which(mat.df.WSJ$horizon %in% vec.str.LE_W)
mat.df.WSJ[vec.int.ROWS, ]$horizon <- "< Week"
vec.int.ROWS                       <- which(mat.df.WSJ$horizon %in% vec.str.GE_Q)
mat.df.WSJ[vec.int.ROWS, ]$horizon <- "> Quarter"
mat.df.WSJ                         <- ddply(mat.df.WSJ, c("t", "horizon"), function(X)sum(X$newsCount, na.rm = TRUE))
names(mat.df.WSJ)                  <- c("t", "horizon", "newsCount")
vec.str.HORIZON                    <- c("< Week", "> Quarter")
mat.df.WSJ                         <- mat.df.WSJ[mat.df.WSJ$horizon %in% vec.str.HORIZON, ]


mat.df.WEEK        <- mat.df.WSJ[mat.df.WSJ$horizon == "< Week", c("t", "newsCount")]
names(mat.df.WEEK) <- c("t", "newsCountW")

mat.df.QTR         <- mat.df.WSJ[mat.df.WSJ$horizon == "> Quarter", c("t", "newsCount")]
names(mat.df.QTR)  <- c("t", "newsCountQ")





## #########################################################################
## #########################################################################
## @section: merge data and standardize variables
## #########################################################################
## #########################################################################

mat.df.DATA                <- merge(mat.df.WEEK, mat.df.QTR, by = c("t"), all.x = TRUE)
mat.df.DATA                <- merge(mat.df.DATA, mat.df.TOTAL, by = c("t"), all.x = TRUE)
mat.df.DATA$yesteryearBias <- (mat.df.DATA$newsCountQ - mat.df.DATA$newsCountW)/mat.df.DATA$newsCountTot


mat.df.DATA <- merge(mat.df.DATA, mat.df.RET,  by = c("t"), all.x = TRUE)
mat.df.DATA <- mat.df.DATA[, c("t", "yesteryearBias", "r", "rLag", "rFwd", "rL", "rLagH")]


obj.fun.STANDARDIZE <- function(x) return((x - mean(x))/sd(x))

mat.df.DATA <- mat.df.DATA[is.na(mat.df.DATA$yesteryearBias) == FALSE, ]
mat.df.DATA <- mat.df.DATA[is.na(mat.df.DATA$r) == FALSE, ]
mat.df.DATA <- mat.df.DATA[is.na(mat.df.DATA$rLag) == FALSE, ]
mat.df.DATA <- mat.df.DATA[is.na(mat.df.DATA$rFwd) == FALSE, ]


mat.df.DATA$yesteryearBiasPos <- (mat.df.DATA$yesteryearBias > 0)
mat.df.DATA$yesteryearBias    <- obj.fun.STANDARDIZE(mat.df.DATA$yesteryearBias)        


mat.df.DATA$r       <- obj.fun.STANDARDIZE(mat.df.DATA$r)
mat.df.DATA$rLag    <- obj.fun.STANDARDIZE(mat.df.DATA$rLag)





## #########################################################################
## #########################################################################
## @section: plot yesteryear bias, 1997-99
## #########################################################################
## #########################################################################

if (TRUE == FALSE) {
    mat.df.PLOT <- mat.df.DATA[, c("t", "yesteryearBias")]
    mat.df.PLOT <- mat.df.PLOT[mat.df.PLOT$t <= as.Date("1999-09-01"), ]
    mat.df.PLOT <- mat.df.PLOT[mat.df.PLOT$t >= as.Date("1997-09-01"), ]

    
    theme_set(theme_bw())
    
    
    scl.str.RAW_FILE <- paste("plot--yesteryear-bias--14apr2017", sep = "")
    scl.str.TEX_FILE <- paste(scl.str.RAW_FILE,'.tex',sep='')
    scl.str.PDF_FILE <- paste(scl.str.RAW_FILE,'.pdf',sep='')
    scl.str.AUX_FILE <- paste(scl.str.RAW_FILE,'.aux',sep='')
    scl.str.LOG_FILE <- paste(scl.str.RAW_FILE,'.log',sep='')
    
    
    tikz(file = scl.str.TEX_FILE, height = 2, width = 4, standAlone=TRUE)    
    
    
    obj.gg2.PLOT <- ggplot(data = mat.df.PLOT, aes(x = t, y = yesteryearBias))
    obj.gg2.PLOT <- obj.gg2.PLOT + geom_path(size = 3)
    obj.gg2.PLOT <- obj.gg2.PLOT + ggtitle("WSJ Yesteryear Bias", subtitle = "September `97 to September `99")
    obj.gg2.PLOT <- obj.gg2.PLOT + coord_cartesian(ylim = c(-2.29, -0.12), expand = TRUE)
    obj.gg2.PLOT <- obj.gg2.PLOT + scale_y_continuous(breaks = round(mat.df.PLOT$yesteryearBias[c(1,13,25)],2), labels = c("", "-2.11", "-0.90"))
    vec.dt.BREAKS <- as.Date(c("1997-09-01", "1998-09-01", "1999-09-01"))
    vec.dt.LABELS <- c("S`97", "S`98", "S`99")
    obj.gg2.PLOT <- obj.gg2.PLOT + scale_x_date(breaks = vec.dt.BREAKS, labels = vec.dt.LABELS)
    
    
    obj.gg2.PLOT <- obj.gg2.PLOT + theme(plot.margin       = unit(c(0.15,0.15,0.15,0.15), "lines"),
                                         axis.text         = element_text(size = 12),
                                         axis.title        = element_blank(),
                                         plot.title        = element_text(size = 18),
                                         panel.grid.minor  = element_blank()
                                         )
    
    
    print(obj.gg2.PLOT)
    dev.off()
    
    
    system(paste('lualatex', file.path(scl.str.TEX_FILE)), ignore.stdout = TRUE)
    system(paste('rm ', scl.str.TEX_FILE, sep = ''))
    system(paste('mv ', scl.str.PDF_FILE, ' ', scl.str.FIG_DIR, sep = ''))
    system(paste('rm ', scl.str.AUX_FILE, sep = ''))
    system(paste('rm ', scl.str.LOG_FILE, sep = ''))
}






## #########################################################################
## #########################################################################
## @section: estimate regressions
## #########################################################################
## #########################################################################

if (TRUE == TRUE) {
    print("                                                  ")
    print("                                                  ")
    print("    yesteryear = a + b1*r")
    print(round(summary(lm(yesteryearBias ~ r, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    yesteryearBias = a + b1*rLag")
    print(round(summary(lm(yesteryearBias ~ rLag, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    yesteryearBias = a + b1*r + b2*rLag")
    print(round(summary(lm(yesteryearBias ~ r + rLag, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    rFwd = a + b1*yesteryearBias")
    print(round(summary(lm(rFwd ~ yesteryearBias, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    rFwd = a + b1*rLag")
    print(round(summary(lm(rFwd ~ rLag, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    rFwd = a + b1*yesteryearBias + b2*rLag")
    print(round(summary(lm(rFwd ~ rLag + yesteryearBias, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    rFwd = a + b1*(yesteryearBias - E[yesteryearBias|r,rLag])")
    mat.df.DATA$residuals <- lm(yesteryearBias ~ r + rLag, data = mat.df.DATA)$residuals
    print(round(summary(lm(rFwd ~ residuals, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    rFwd = a + b1*{(yesteryearBias > 0) x (r < 0) x (rLag > 0)}")
    print(round(summary(lm(rFwd ~ yesteryearBiasPos * rL * rLagH, data = mat.df.DATA))$coef[,c(1,3)], 2))
    print("                                                  ")
    print("                                                  ")
    print("    sum(coef|YBPos = TRUE) - sum(coef|YBPos = FALSE) = 135% per year")
}

2 data--sp500-horizon-mentions.csv

3 data--sp500-return.csv