Vehicle fleet and registrations — Algeria (1996-2019)

The Algerian vehicle fleet and its registrations, 1996 to 2019. 29,199 observations and sixteen crossings: vehicle type against wilaya, make, country of origin, fuel, year first registered, horsepower, payload and age bracket.

The national fleet grows from 2,680,819 vehicles in 1997 to 6,418,212 in 2018. And diesel's share of the passenger-car fleet rises from 6.9% in 1997 to 22.3% in 2019 — 110,665 diesel cars out of 1.6 million, then 945,685 out of 4.2 million.

Two natures coexist and are labelled plainly: the fleet at 31 December, and the registrations of the period — yearly or half-yearly depending on the edition.

These sheets' structure is discovered, not declared. Their headers wrap over three lines, and one sheet in three carries none that can be used. The total column is therefore the one equal to the sum of the others on every row, and finding it is verifying it. Two hundred and thirty-seven sheets out of two hundred and ninety-one yield at least one verified identity; those that yield none do not ship.

The percentage does not mean the same thing from one sheet to the next, and that is this series' trap. In the 2006 fleet, "84.67" is the share of the row: 1,729,823 petrol vehicles out of the 2,042,986 passenger cars. In the 2010 registrations, "53.73" is the share of the whole table: 52,941 out of 98,538 vehicles registered. Nothing in the header separates them — both are written "%". Publishing one for the other would read as a fleet 54% petrol where it is 74%.

We tested both conventions on every sheet and kept the one that holds. The percentages are nonetheless not published: they recompute exactly from the counts, and one more ambiguous column would serve nobody. The convention served the check; the file serves counts.

Orientation is not readable from the title either. One title puts vehicle type in rows, another puts it in columns. The title names both dimensions; it does not say which one carries the rows. We recognise it from the labels themselves: a column of wilayas is recognisable by its wilayas. Both dimensions are named in the file, so the buyer never has to guess what they are reading.

What we do not publish. Fifty-four sheets yield no checkable identity and are refused. 9,232 values are set aside because their column did not recover its name — better an absent column than one called "column 3". And years are read from the title, never from the body: an earlier reading took the document's largest four-digit number, and on a 1998 sheet the value "2059" became the year.

Unit and source. Vehicle counts, as published. This file joins our other wilaya-level products on the official code, and completes our "motorisation rate", which related the 2010 fleet to population.