RSF's 2013+ exports store score/dimension percentages (0-100, 2 implied decimal places) as bare digit strings with no decimal point (e.g. "9189" means 91.89). Values ending in one or two zeros have those trailing zeros silently dropped somewhere in RSF's own export pipeline (e.g. "844" means 84.40, not 8.44; "87" means 87.00, not 0.87), which is indistinguishable from a genuinely low sub-10 score for the worst-ranked countries (e.g. "46" for a rank-180 country can legitimately mean 0.46).
Arguments
- raw_chr
Character vector. Raw digit strings from the source CSV (no decimal point; may have a leading "-" for Period 1 legacy values).
- rank_chr
Character vector, same length as `raw_chr`. The corresponding rank column, used to disambiguate short values by comparing against neighboring, unambiguous (4+ digit) values at nearby ranks (scores are approximately monotonic in rank).
Details
Values with 4 or more digits are unambiguous ("confirmed"): a 4-digit value is divided by 100 (2 implied decimals), and a small number of 2025 rows have 5-digit values (RSF apparently breaks near-tied ranks with a third decimal place, e.g. "65487" means 65.487, not 654.87), so n-digit confirmed values are divided by `10^(n_digits - 2)` generally. For shorter (2-3 digit) values, two candidates are computed: right-padding with zeros to 4 digits before dividing by 100 (the "dropped trailing zero" interpretation), and dividing the raw digits by 100 directly (the "already complete, genuinely low score" interpretation). The candidate closer to a rank-based linear interpolation of neighboring confirmed values is selected. Missing rank or missing value inputs fall back to the zero-padded interpretation.