Detecting changes for several companies - r

Question

I wonder if anyone can provide me with some tools/packages/codes to detect changes in the peer groups that are used for relative performance evaluation.
I have a dataframe with all peers that are used for a certain firm (CIK) over the years. An example of this data is given below:
CIK <- c("21344","21344", "21344", "21344", "21344", "21344", "21344", "21344", "21344")
FiscalYear <- c("2013", "2014", "2015", "2016", "2017", "2014", "2015", "2016", "2017")
PeerCIK <- c("1800","1800","1800","1800","1800","21456","21456","21456","21456")
dataframe <- data.frame(CIK, FiscalYear, PeerCIK)
This results in the following table:
CIK FiscalYear PeerCIK
1 21344 2013 1800
2 21344 2014 1800
3 21344 2015 1800
4 21344 2016 1800
5 21344 2017 1800
6 21344 2014 21456
7 21344 2015 21456
8 21344 2016 21456
9 21344 2017 21456
Now, I want to identify whether the peers (PeerCIK) are present for the whole period that is covered by the firm (CIK). I thus first need to identify the first and last year per CIK (in this example it is clear (2013-2017), but I need to do this for many firms). A code I used for this is:
df2 <- dataframe %>%
group_by(CIK) %>%
summarise(
start = min(FiscalYear),
end = max(FiscalYear)
)
> df2
CIK start end
1 21344 2013 2017
and following I need to identify whether all different peers are present for that period.
If this is not true, then a change must have taken place in the peer group (the peer is added to or deleted from the group). This is where I have trouble with how to continue. The outcome that I ultimately want, is a dataframe with for every firm (CIK), per fiscalyear whether a change has taken place in the peer group compared to last year (where change is a dummy variable with value 1 if change takes place). Such a change is thus when a peer is added (after the starting date) or when a peer is no longer included while the end date is not yet reached for that particular CIK.
Expected outcome
For the example above, I would have the following outcome, as company 21456 is added from 2014 onwards and thus a change has taken place compared to 2013:
CIK FiscalYear change
1 21344 2013 0
2 21344 2014 1
3 21344 2015 0
4 21344 2016 0
5 21344 2017 0
I really hope someone can help me, please let me know

Answer 1

A slightly different approach via expand(), full_join, and some helper variables which should cover most of your edge cases:
library(tidyverse)
dataframe %>%
# Add helper variable to indicate present relationships.
mutate(
present = 1
) %>%
# Generate all possible variations of CIK, FiscalYear, and PeerCik
# and join with our data.
full_join(
dataframe %>% expand(CIK, FiscalYear, PeerCIK),
by = c("CIK", "FiscalYear", "PeerCIK")
) %>%
# Set the helper variable to 0 wherever it is missing,
# which is the case in your newly joined empty data from `expand(...)`.
mutate(
present = ifelse(is.na(present), 0, present)
) %>%
# Sort the data because now the order will be important.
arrange(CIK, PeerCIK, FiscalYear) %>%
# Group by CIK-PeerCIK relationship...
group_by(
CIK, PeerCIK
) %>%
# ...and compare each FiscalYear to the previous FiscalYear.
mutate(
# Check if a relationship was added compared to the year before.
added = case_when(
row_number() == 1 ~ 0,
lag(present) == 0 & present == 1 ~ 1,
TRUE ~ 0
),
# Check if a relationship was removed compared to the year before.
removed = case_when(
row_number() == 1 ~ 0,
lag(present) == 1 & present == 0 ~ 1,
TRUE ~ 0
),
# Combine those two into one variable.
change = ifelse(abs(added) + abs(removed) > 0, 1, 0)
) %>%
ungroup() %>%
# Now to the summary: Group by CIK and FiscalYear...
group_by(
CIK, FiscalYear
) %>%
# ...and calculate all sums for each CIK and FiscalYear.
summarize(
# Total number of present relationships in this year.
num_present = sum(present),
# Number of added relationships in this year.
num_added = sum(added),
# Number of removed relationships in this year.
num_removed = sum(removed),
# Was there any change in this year?
# An alternative would be `sum(change)` to
# indicate the number of changed relationships.
change = max(change)
) %>%
ungroup()
Result:
# A tibble: 5 × 6
CIK FiscalYear num_present num_added num_removed change
<chr> <chr> <dbl> <dbl> <dbl> <dbl>
1 21344 2013 1 0 0 0
2 21344 2014 2 1 0 1
3 21344 2015 2 0 0 0
4 21344 2016 2 0 0 0
5 21344 2017 2 0 0 0