7.8 Plotting the returns

We will now plot the returns for easy comparison. The best way to depict these returns will be by using a scatterplot superimposed on the error bars corresponding to the lower and upper prediction intervals.

We will first create a data set that is reshaped to be long. As the realized returns do not have the prediction intervals, we will drop these variables for the time being.

predictions_lg <- predictions %>% 
  select(ticker, ret, fit) %>% 
  reshape2::melt(id.vars = "ticker")

Take a look at this data using headTail() function from psych package, which will print 4 observations from the top and 4 observations from the bottom by default.

psych::headTail(predictions_lg) %>% 
  knitr::kable(caption = "Reshaped Predictions",
               booktabs = TRUE)
Table 3.5: Reshaped Predictions
ticker variable value
1 AMZN ret -0.05
2 BA ret 0
3 CBS ret 0
4 CCZ ret 0
… NA NA …
19 MKGAF fit 0
20 NYT fit 0
21 JWN fit 0
22 TM fit -0.01

Now make the plot.

ggplot(predictions_lg, aes(x = ticker)) +
  geom_errorbar(aes(ymin = lwr, ymax = upr), 
           data = predictions,
           color = "#03c3f6",
           width = 0.2) +
  geom_point(aes(y = value, 
                 fill = variable), 
             color = "#3b4252",
             shape = 21) +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_manual(values = c("#ef2e69", 
                               "#205aff"),
                    labels = c("Realized Returns",
                               "Estimated Returns")) +
  labs(x = "Ticker",
       y = "Daily Stock Returns", 
       fill = "") +
  theme_minimal()
Stock Reactions to Trump Attacks

Figure 7.2: Stock Reactions to Trump Attacks

Interestingly, except for Amazon, no other stock suffered from Trump attack! Nordstrom actually showed an unexpected increase in the stock price. Otherwise, rest 9 stocks have no effect of Trump Twitter attack.