assignment 3

assignments/el3220_3.R

@@ -0,0 +1,119 @@
+# In-class assignment 3:
+
+# Submit to NYU Classes as: [NetID]_3.R.
+
+# Load the libraries.
+library(quantmod)
+library(stargazer)
+library(fredr)
+library(tidyr)
+library(PASWR2)
+library(MASS)
+library(repmis)
+library(latex2exp)
+library(dplyr)
+library(ggplot2)
+library(tidyverse)
+library(RCurl)
+library(haven)
+fredr_set_key('30e6ecb242a73869e11cb35f6aa3afc3') # My key, please don't abuse.
+
+
+getwd() 
+setwd("C:/Users/erikl/OneDrive/Documents/NYU_CUSP/Summer2020/REDA")
+
+
+# 1. Re-run all code above to ensure it works.
+# 2. Do FF1F, FF3F and FF5F for IBM and Microsoft stock.
+
+runFF <- function(a, b, stock) {
+  data = merge(a, b, join='right')  ## Merge into single time-series dataset
+  names = c("date", "mktrf", "smb", "hml", "rmw", "cma", "rf", stock) # factors (per Ken French) and AAPL returns
+  colnames(data) = names
+  data = data[, colnames(data) != "date"]  # Narrow dataframe 
+
+  ff1f.ols = lm(paste(stock,'- rf ~ mktrf'), data = data)
+  stargazer(ff1f.ols, type="text", title=paste("Baseline Results:", stock), single.row=TRUE, 
+            ci=TRUE, ci.level=0.95)
+
+  ff3f.ols = lm(paste(stock,'- rf ~ mktrf + smb + hml'), data = data)
+  stargazer(ff3f.ols, type="text", title=paste("FF3F Results:", stock), single.row=TRUE, 
+            ci=TRUE, ci.level=0.95)
+
+  ff5f.ols = lm(paste(stock,' - rf ~ mktrf + smb + hml + rmw + cma'), data = data)
+  stargazer(ff5f.ols, type="text", title=paste("FF5F Results:", stock), single.row=TRUE, 
+            ci=TRUE, ci.level=0.95)
+}
+
+ff5f = read_dta("FF5F.dta") # Data are stored in multiple formats.  
+ff5f$mktrf = ff5f$mktrf / 100
+ff5f$smb = ff5f$smb / 100
+ff5f$hml = ff5f$hml / 100
+ff5f$rmw = ff5f$rmw / 100
+ff5f$cma = ff5f$cma / 100
+ff5f$rf = ff5f$rf / 100
+
+a = xts(x=ff5f, order.by = ff5f$date)
+
+
+#IBM
+getSymbols(c('IBM'), from="2000-01-01", to="2019-12-31") # Read in AAPL data using quantmod.
+b = IBM$IBM.Adjusted
+b = diff(log(b), lag=1)  ## Log returns
+b = na.omit(b)
+
+runFF(a, b, 'ibm')
+
+
+#MSFT
+getSymbols(c('MSFT'), from="2000-01-01", to="2019-12-31") # Read in AAPL data using quantmod.
+b = IBM$IBM.Adjusted
+b = diff(log(b), lag=1)  ## Log returns
+b = na.omit(b)
+
+runFF(a, b, 'msft')
+
+# 3. Replicate the omitted variable bias example inducing negative correlation.  
+# (Pay attention to Z1.) <-- X1??
+
+rm(list=ls())
+set.seed(1993)
+z = rnorm(10000, mean=0, sd=1)  
+v = rnorm(10000, mean=0, sd=1)  
+w = rnorm(10000, mean=0, sd=1) 
+e = rnorm(10000, mean=0, sd=1) 
+x1 = z + v
+x2 = -1*(z + w)
+y = 1 + 1* x1 + 1* x2 + e
+cor(x1, x2)
+lm = lm(y~x1)
+stargazer(lm, title="Regression Results When We Omit Correlated Relevant Variable", 
+          single.row=TRUE, type="text", ci.level=0.95, ci=TRUE)
+
+#lm = lm(y~x1+x2)
+#stargazer(lm, title="Regression Results When We Include (Negatively) Correlated Relevant Variable", 
+#          single.row=TRUE, type="text", ci.level=0.95, ci=TRUE)
+
+# Now that they're negatively correlated, X1 is moved to the left to compensate for the missing X2
+# In our specific case, If both are included X1 and X2 are 1
+# If X2 is missing and positively Correlated --> X1 is 1.5
+# If X2 is missing and negatively correlated --> X1 is .5
+# ^^^Assuming X1, X2 were generated our same way
+
+
+# 4. In most circumstances, we seek to draw inferences about the effects of some type of policy or action.
+# Considering the ideas discussed in class regarding causality:
+# Do opportunity zones actually result in benefitial development, 
+# and how could we evaluate potential benefits using data? 
+
+# If an OZ actually benefits the community depends on what you choose to measure as an indicator of success
+# Lets suppose an investor brings a tech office to an OZ (and that it qualifies as an OZ investment)
+# If you use real estate prices as an indicator of success
+#, you may think a rise in housing prices shows success when it could actually be indicating a displacement
+#  of the original population as white-collar workers buy-out OZ real estate.
+# However if you instead decide to track employment outcomes for "original" OZ residents (which sounds harder to do)
+# , you could have a better idea of the impact of OZ investments. 
+# Determining OZ benefits depends on who is receiving the benefits we are measuring.   
+
+
+