bug: filter financial data unusable in ownership_weight methodology

[jdhoffa]

current_shares_outstanding_all_classes–a value that is synonymous with the shares_all_classes value– is used downstream to calculate ownership_weight.
This value is unusable by the ownership_weight calculation if it is 0, as it will cause divide by 0 problems and yield Inf results: (https://github.com/RMI-PACTA/pacta.portfolio.allocate/blob/4c96adb9856788d97293e41c0cd68b8951c3c56b/R/calculate_ownership_weight.R#L5).

There is already a filter in prepare_financial_data to remove this case, however it filters a wrong (but similar) column.

An edge case was found by @Antoine-Lalechere in which ADR holdings with no similar EQ holdings pass through and yield a 0 value for shares_all_classes, which identified this bug.

This PR adjusts the filter to ensure that the correct value is filtered and usable in the downstream analysis.

See the reprex below for more information:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(pacta.data.preparation)

# consider the case there an entity has an equity holding and an ADR holding
# consider further, that the equity holding that reflects the ADR holding has no price or share information
# note: this was a real case in the data
data <- tibble::tribble(
  ~fsym_id,          ~isin, ~factset_entity_id, ~adj_price, ~adj_shares_outstanding, ~issue_type, ~one_adr_eq,
  "ABCDE-F", "US123456789",         "12ABCD-E",      5.5,               12345678,        "AD",          NA,
  "ABCDE-G", "JP123456789",         "12ABCD-E",         NA,                      NA,        "EQ",          NA
)

issue_code_bridge <- tibble::tribble(
  ~issue_type_code, ~asset_type,
  "AD",    "Equity",
  "EQ",    "Equity"
) |> 
  pacta.data.preparation::standardize_asset_type_names() 

# our current methodology does NOT consider ADR holdings in the calculation of ownership_weight denominator
# what this means is that, in the case above, it would be impossible to calculate an appropriate ownership_weight
# since the denominator would be 0

# however, as can be seen below, the ADR holding does appear in the output, with 
# a value of 0 for current_shares_outstanding_all_classes. This is unusable downstream by the ownership_weight calculation
# 
# See: https://rmi-pacta.github.io/pacta.data.preparation/articles/share_ownership_methodology.html for more information
out <- pacta.data.preparation::prepare_financial_data(data, issue_code_bridge)

print("US123456789" %in% out$isin)
#> [1] TRUE
print(filter(out, isin == "US123456789")["current_shares_outstanding_all_classes"])
#> # A tibble: 1 × 1
#>   current_shares_outstanding_all_classes
#>                                    <dbl>
#> 1                                      0

^{Created on 2024-07-17 with reprex v2.1.1}

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 1 line in your changes missing coverage. Please review.

Project coverage is 18.82%. Comparing base (0b3d197) to head (0540029).

Files	Patch %	Lines
R/prepare_financial_data.R	0.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #33   +/-   ##
=======================================
  Coverage   18.82%   18.82%           
=======================================
  Files          35       35           
  Lines        1126     1126           
=======================================
  Hits          212      212           
  Misses        914      914           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.