Argentina Food & Beverage Price Index — Methodology

What does this index measure?

This index measures the daily change in online food and beverage prices in Argentina, set to 100 on the first day of the series. A reading of 103 means tracked prices are 3% above the base day. It is a high-frequency, transparent complement to official monthly statistics — daily, with a three-day lag rather than weeks.

Where does the data come from?

Prices are collected daily from publicly listed product pages at major Argentine online retailers (initially Buenos Aires; ~50–100 SKUs across dairy, bakery, produce, meat, pantry, and beverages). Only public prices are used — no personal data, no logged-in sessions. Every data response reports its sources, coverage, and sample size so the number can be independently checked.

How is the index calculated? (three estimators, one dataset)

Every estimator is computed from the same underlying observations: for each tracked item, today's price is divided by its base-day price to form a price relative, then aggregated across items:

• Jevons (primary) — the geometric mean of price relatives. The internationally standard elementary index (used by national statistics agencies); it avoids the upward bias of the arithmetic mean.
• Carli — the unweighted arithmetic mean of price relatives. Simple, but it has a well-documented upward bias (it fails the time-reversal test), so it is offered for comparison, not as the headline.
• Trimmed mean — drops the top and bottom 10% of relatives before averaging, a robust "core" reading less sensitive to individual outliers.

The headline uses the primary estimator (Jevons); all three are published free for comparison on the index page and via the API. Out-of-stock items are excluded that day; promotional prices are included. Every response records which method produced it. A planned v1 adds category (Laspeyres) weighting.

How is data quality kept stable?

Each basket item is pinned to a fixed product identifier and fetched by that ID every day, so the index tracks the same products over time rather than whatever a search happens to return — keeping composition stable. A daily quality gate accepts a day only if coverage and the day-over-day move fall within expected bounds; a failed day is quarantined (the index holds its last good value) and flagged for review. Individual readings that move implausibly in a single day are dropped as outliers. Any corrections or rebaselining are recorded in the changelog.

How often is it updated?

Daily. Each page and API/MCP response carries a machine-readable last_updated timestamp and the methodology version. Historic values are not silently revised; corrections appear in the changelog.

Who maintains it, and what are the limits?

The index is built and maintained by Oscar E. López, Ph.D., whose field is information systems, data administration, and business intelligence — building clean, well-governed data pipelines. The economic index methodology is co-signed by an economist advisor [advisor name/credentials]. Known limits: coverage is a curated basket (not the full CPI basket), it reflects online prices (not in-store), and single-day readings are noisy — interpret trends over weeks, not days.

How should this be cited?

Cite as: López, O. E. (2026). Argentina Food & Beverage Online Price Index. data.oscar-lopez.com/methodology. A citable methodology note with a DOI is available via [SSRN/Zenodo link]. The dataset is licensed CC BY 4.0; attribution required.