Five Countries Missing From the Average — Python Bug Hunt

Briefing

Modelled on the Reinhart–Rogoff spreadsheet error. Their 2010 paper "Growth in a Time of Debt" was widely cited in austerity debates. In 2013 Thomas Herndon, Michael Ash and Robert Pollin re-ran the authors' own spreadsheet and found, among other issues, that an AVERAGE formula's cell range stopped short and left out five countries — Australia, Austria, Belgium, Canada and Denmark — from the headline growth figures. Reinhart and Rogoff acknowledged the coding error.

This reconstruction's summary.py averages a worksheet's growth column through a spreadsheet-style range helper, and the range it asks for starts several rows too late.

Fix average_growth so every country row is counted.

Bug report

BUG-RR2013 · Priority: High · Reported by: replication study

average_growth(rows) — rows[0] is the header, every later row is one country [name, bucket, growth]:

• average the growth column over EVERY country row (sheet row 2 through the last row)
• blank cells (None) are skipped, not counted as zero
• no values at all -> None
• round the result to 2 decimals

Observed: the first five countries in the sheet never reach the average, so the figure changes when the sheet is re-sorted.

Logs

[replicate] AVERAGE over L30:L44 — sheet data runs L30:L49
[replicate] 5 of 20 countries excluded from the 90%+ bucket

The code as shipped

src/sheet/summary.py (editable)

sheet = bug_require("./sheet.py")

GROWTH_COLUMN = 2


def average_growth(rows):
    cells = sheet.cell_range(rows, 7, len(rows), GROWTH_COLUMN)
    values = [c for c in cells if c is not None]
    if not values:
        return None
    return round(sum(values) / len(values), 2)

Read-only context: src/sheet/sheet.py.