In lachlandeer/hayashir: R Companion to "Econometrics" by Fumio Hayashi

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Load Libraries

library(hayashir)
library(ggplot2)
library(dplyr)
library(zoo)

Inspecting the Data

head(mishkin)

Figure 2.3

First we construct a date variable and then calculate CPI inflation The library zoo contains the function as.yearmon() to construct dates containing only year and months.

mishkin2 <- mishkin %>%
                mutate(time_period = as.yearmon(paste(year, month, sep = "-")),
                       cpi_inflation = (cpi/lag(cpi) - 1) * 100 * 12
                       )

Now we plot the data:

ggplot(mishkin2) +
    geom_line(aes(time_period, tbill_1, linetype = "tbill_1")) +
    geom_line(aes(time_period, cpi_inflation, linetype = "cpi_inflation")) +
    xlab("Date") +
    ylab("percent") +
    theme_bw() + 
    theme(legend.position="bottom") +
    scale_linetype_manual(name = element_blank(), 
                          values = c(tbill_1 = "solid", cpi_inflation = "dotted"),
                          labels = c("inflation rate", "1 month T-bill rate")
                          )

Figure 2.4

Compute real interest rate:

mishkin2 <- mishkin2 %>%
                mutate(real_rate = tbill_1 - cpi_inflation)

And now plot:

ggplot(mishkin2) +
    geom_line(aes(time_period, real_rate, linetype = "tbill_1")) +
    geom_hline(yintercept = 0, linetype = "dashed", color = "red") +
    xlab("Date") +
    ylab("percent") +
    theme_bw() + 
    theme(legend.position="none")

Table 2.1

First we subset the data. The relevant Date range is January 1953 to July 1971.

fama <- mishkin2 %>%
            filter(between(as.yearmon(time_period), 
                           as.yearmon("Jan 1953"), 
                           as.yearmon("Jul 1971")
                           )
                   )

library(skimr)
skim(fama$real_rate)

Then the ACF

acf(fama$real_rate, lag.max = 12, plot = FALSE)

Then the Ljung Box Q tests (use the portes package for more flexibility)

library(portes)

LjungBox(fama$real_rate, lags = seq(1, 12))

Is the Nominal Interest Rate the Optimal Predictor?

model1 <- lm(cpi_inflation ~ tbill_1, data = fama)
summary(model1)

How to get robust standard errors?

library(sandwich)
library(lmtest)

coeftest(model1, vcov = vcovHC(model1, "HC0")) # white's

library(car)

linearHypothesis(model1, "tbill_1 = 1", white.adjust = "hc0")

lachlandeer/hayashir documentation built on Feb. 9, 2023, 2:01 p.m.