Noelle Natale: Ind Assignment 2 - Data Analytics in Accounting

1. Bar Chart by Frequency

## 1. Bar Chart by Frequency
# group by year, quarter, digit and count occurrences 

#df2 <- df1 %>% group_by(fyearq,fqtr,digit_diluted) %>%
  #summarize(count_filings = n())

#df2 %>% 
  #ggplot(aes(x=digit_diluted, y=count_filings)) + 
  #geom_col(aes(fill=digit_diluted)) + theme_bw()
#plot by quarter

df2 %>% 
  ggplot(aes(x=digit_diluted, y=count_filings)) + 
  geom_col(aes(fill=digit_diluted)) + facet_wrap(vars(as_factor(fqtr))) +
theme_bw()

1b: The smallest difference between integer 4 and 5 looks to be Q4, much less variation observed in that quarter, interestingly enough.

2a - 1998 Only

2b - 2019 Only

2C: Differences observed between 2019 and 1998: In 1998, there are much larger discrepancies between the 4 integer and the surrounding integers, particularly the difference between 4 and 5. This issue is less prevalent in 2019 as more audit tools and internal controls are implemented in organizations.

3 Top 10 Worst Offenders

#3 
#aggregate results for company by 1 row each

df_3 <- df1 %>%
  mutate (
    digit_4 = if_else(digit_diluted == 4,1,0), 
    digit_5 = if_else(digit_diluted == 5,1,0)
    ) %>% group_by (tic,conm) %>%
  summarize(count_4 = sum(digit_4),count_5 = sum(digit_5),count_filings = n()) %>% mutate(
  percent_4 = count_4/ count_filings,
  percent_5 = count_5/ count_filings) %>% filter(count_filings >56,percent_4< 0.0118, percent_5 > 0.11)

3b - These companies have almost no 4 digit integers in their EPS, but much higher % of the 5 integer. We would expect to see close to 10% of EPS have a 4 in the tenths place. This is definitely a cause for an investigation due to a high likelihood of “strategic rounding”.

3c - Yes, Fulton, Interface and HCSG. That’s crazy!