Venezuela Price Index (VPI) methodology

The VPI is a daily index of posted online retail prices for consumer goods in Venezuela. It follows the Billion Prices Project playbook: scrape a fixed panel of SKUs every day, compute a chained index from price relatives, and aggregate using the official consumer price index's category weights.

It is a goods index, not headline inflation. It covers 40–48% of the official INPC basket by weight: food, personal care, household consumables and OTC pharmacy goods sold by formal online retail. It excludes services, rent, transport, utilities, education, and everything sold in informal markets.

That caveat is published everywhere the number appears. Credibility is the product.

The official reference is the BCV/INE INPC: a fixed-basket Laspeyres index, base December 2007 = 100, covering 362 items across 13 groups, weighted by the third National Household Budget Survey.

The panel's four categories map to INPC groups. The weights used are the full-group weights renormalised over the covered set: food and beverages 67.2%, household and cleaning 11.7%, OTC pharmacy 9.0%, personal care 12.1%.

Weight vintage caveat, stated plainly: those weights come from a survey on a 2007 base, and the BCV has never published a re-based structure. After two decades of economic collapse, dollarisation and migration, actual expenditure shares have moved. We use them anyway, because they are the only citable official structure and because they make the monthly aggregate directly comparable to BCV and OVF prints. A food-heavier alternative weighting is published as a sensitivity series, never as the headline.

Within each category the index is a daily chained Jevons: the geometric mean of price relatives across the SKUs that have a price on both days. Above the category, aggregation is arithmetic and weighted by the renormalised weights.

Every SKU with a price on both days is in that mean, whether or not its price moved. An unchanged price enters at a relative of exactly 1, which is zero inflation, and is never dropped. On a typical day most of the panel holds its price, and the index reflects that.

Daily chaining carries a known chain-drift risk. It is declared as a risk and cross-checked against a fixed-basket construction over the same panel.

A SKU with no usable price on a given day (not returned, or marked out of stock) is held at its last observed price for up to seven days, at a relative of 1. After seven days it leaves the calculation; when it comes back it re-enters from its new price, with no single catch-up move spanning the gap. It is never interpolated and never chained from its base price.

Outliers are logged, not silently dropped. Every exclusion is recorded with its reason.

Source eligibility (§3.4) has been corrected twice, and the second correction reverses the first. From 10 August 2026 a chain was dropped from the index when its prices had barely moved in thirty days. That test selects sources on the very thing the index measures, and the bias runs one way: a chain holding its prices is posting zero inflation, and removing it leaves only the chains that moved. Withdrawn on 17 September 2026 and replaced by a liveness test, which drops a catalogue only when nothing in it moves at all: not its prices, not its stock, not its discounts. Every chain in the panel now passes, the index history has been restated accordingly, and flat prices count as the zeros they are. Price rigidity at a chain is still measured and still reported. It is a finding about Venezuelan retail, not grounds for removal.

There are three series: the bolívar index, a dollar index built from native USD prices the retailers themselves post, and a dollar index derived from the official rate. The USD headline is the native one; the derived one is published labelled as such.

No series is ever simulated. Where a category has prices in only one currency, it is published in that currency only, and we say so.

Collected daily by deterministic scrapers from each chain's own public storefront. Collector implementation detail is deliberately not published.

The identities of the panel's chains are withheld in product-level publications. This is not methodological opacity: it is the condition on which the panel continues to exist.

Two counts describe the panel, and they are two different sets. Products priced on a given day is the figure in the site footer. Products in the index that day is smaller: the ones in stock, in one of the four headline categories, with a price on both that day and the previous one, in the currency of the series. Only the second is the set the index is computed over; today's figures are in the opening paragraph, and a single public size is never quoted without saying which set it is.

Success criteria were pre-registered on 3 August 2026, before the BCV's July print, and the July figures were locked on 10 August. A pre-registered number that later lands inside its band is the only evidence that counts.

The comparison is against the BCV and the OVF, at monthly frequency, across two prints. The first gate, panel stability, passed on 2 August 2026.

The panel skews to Caracas and to formal online retail. These are posted shelf prices, not transaction prices: they miss checkout discounts and negotiation. They miss informal markets, which in Venezuela are a large share of real consumption.

Substitution bias: a fixed panel cannot see consumers moving to cheaper pack sizes. It is declared rather than silently corrected.

Every restatement of published history is announced with its date, its reason, and its numerical effect on the headline. Nothing is deleted and no collected price row is ever modified.

The 10 August 2026 restatement moved the VES headline for 3 July → 2 August from +3.31% to +13.39%, and the USD headline as of 10 August from −8.38% to −1.44%. The change is in which sources are chained, not in what is collected.