Are wolves driving sheep off French farms?

Find your issue, and check it is measured

An environmental issue: something happening, to something, because of something, with proof that somebody measures it.

Wolves came back to France across the Alps from Italy in 1992 and now number about 1,080. Sheep are what they kill most, France's sheep flock has been shrinking for years, and sheep farmers say the wolf is part of why.

Proof it is measured

The wolf presence records of the OFB, France's national wildlife agency, opened: every sighting, dropping, track and howl reported since 2013, each rated by an expert, by commune and department.

And the flocks

Sheep on French farms, opened: the agriculture ministry's count for every department at the end of every year since 2010.

In the news

France 3 Grand Est, 22 February 2025: a sheep farmer in the Vosges, after 42 years, selling his flock after a wolf attack. "Je n'élève pas des animaux pour les donner au loup": I don't raise animals to give them to the wolf.

The system

Store: the wolves living in France, about 1,082 in the winter of 2024 to 2025 by the OFB's estimate. Inputs: births, and young wolves dispersing into new departments. Outputs: deaths, including 190 shot legally in 2025. Food: deer, chamois and wild boar, and sheep where flocks are exposed: about 12,000 farm animals a year are compensated as wolf kills, most of them sheep. Feedback: negative. Attacks on flocks raise the number of wolves the state allows to be shot, which takes wolves out of the store.

Write your research question

One sentence naming the indicator and its units, the years and the places. No strategy in it.

A comparison of the percentage change in sheep kept on farms, 2013 to 2024, between the French departments where wolves were confirmed in every one of those years and the departments where they were never confirmed.

Independent variable

Wolf presence: whether a department had evidence confirmed as wolf in every year from 2013 to 2024, or in none of them. Two groups. A department is one of France's 96 mainland administrative areas, roughly a county. Natural.

Dependent variable

The percentage change in the number of sheep kept on farms, from the end of 2013 to the end of 2024. Human.

Why these two groups

They are the clearest contrast the files allow: twelve years of wolves against none. The 63 departments wolves reached part of the way through are left out, because their flocks had anything from one to eleven years of wolves, and there is no fair way to line those up.

Not obvious

Farmers expect flocks to have shrunk most where the wolves are. A geographer who studies the conflict expects the opposite (step 3). The groups also differ in other ways: the wolf departments are mostly mountains and southern hills, where sheep are often the main use of the land.

Find a real-world strategy and its tensions

One real-world strategy people are genuinely arguing about, two named positions, and what the tension does.

France's national action plan on the wolf and livestock farming, 2024 to 2029: money to protect flocks, and a ceiling on how many wolves may be shot, raised again in 2026.

The strategy

Run by the ministries of Ecological Transition and Agriculture. The 2024 to 2029 plan, published 23 February 2024, pays for shepherds, guard dogs and electric fencing (more than 40 million euros a year) and compensates farmers for animals killed. It also caps how many wolves may be shot: 19% of the estimated population in 2025. The Bern Convention then lowered the wolf from strictly protected to protected (in force 7 March 2025), the EU followed with Directive 2025/1237, and France raised the ceiling to 21% and, from 1 April 2026, let farmers in most areas shoot to defend a flock after declaring it rather than asking permission.

Position 1 · economic, cultural

Claude Font, a sheep farmer in the Haute-Loire and the Fédération nationale ovine's lead on predation, to Causeur, 18 June 2020: "Certains abandonnent le métier, directement en lien avec les prédations lupines": some are giving up farming as a direct result of wolf attacks. In December 2025 the federation said the 19% ceiling could not halt "la spirale des attaques", the spiral of attacks.

Position 2 · environmental, economic

Farid Benhammou, a geographer at the Université de Poitiers who studies the wolf, to Vivant, 25 April 2023: sheep farms "ont même tendance à décliner plus vite dans les zones où il n'y a pas de loups", even tend to decline faster where there are no wolves. The wolf, he says, is "un révélateur": it shows up the weaknesses of farming systems that were already struggling.

The tension

The future of sheep farming against the recovery of a protected predator. Both sides want flocks on the hills. They disagree about what is emptying them.

What it does

Dilution: the wolf's protection has been weakened step by step, from strictly protected to protected, and from shooting by permission to shooting by declaration. Resistance: conservation groups take shooting orders to court; two in the Doubs were annulled by the Besançon court on 18 June 2024. Uneven costs: farmers bear the attacks and the work of guarding, and the wolves bear the shooting.

For the environment

Shooting breeding adults can break up a pack, which can change where the survivors hunt. And hill pasture that flocks leave is taken over by scrub, and loses the open grassland that grazing kept.

Link to the question

The two positions predict opposite results for these two groups of departments. The farmers predict that flocks shrank more where wolves have lived longest; Benhammou predicts the reverse. The test can say which the data fits better. It cannot say why.

Select and document your data

A dataset you can defend: counted, documented, and cut to a rule fixed before you looked.

Source 1 · the wolves

OFB Réseau Loup-Lynx (the agency's wolf and lynx monitoring network), Les indices de présence du Loup: one CSV of about 14 MB, which this map-service link downloads directly. The data.gouv.fr page describes it. 67,212 records for all of France.

Source 2 · the sheep

Agreste (the agriculture ministry's statistics service), Statistique agricole annuelle, departmental series 2010 to 2025: a zip of two workbooks. Use the departmental one, sheet EFA, row 35 - Ensemble espèce ovine (all sheep) for each department, columns EFF_2013 and EFF_2024. The headings are on row 6. If the link is refused, open the series page and choose the departmental ZIP there.

Filters

Wolf verdict (f) is Retenu, meaning confirmed as wolf; duplicates removed on the reference code (r); dates 2013 to 2024, since records from 2025 are still being assessed. Mainland departments only: Agreste's five overseas departments have no wolves.

Inclusion rule

A department is established if it has at least one confirmed record in each of the twelve years, and never reached if it has none. Departments with fewer than 1,000 sheep at the end of 2013 are left out, because one farm closing would move them by tens of percent: that removes four around Paris, one of which had a single sheep. Written down before any change was worked out.

The join

Both files carry the department as a code and a name in one cell: 04 - ALPES-DE-HAUTE-PROVENCE in the wolf file's column d, 004 - Alpes-de-Haute-Provence in Agreste's LIB_DEP. Take the code before the dash, drop Agreste's leading zero, and match. Corsica is 2A and 2B in both.

Derived quantities

For each department: sheep in 2024 minus sheep in 2013, divided by sheep in 2013, times 100. And the number of years, out of twelve, with at least one confirmed wolf record.

How the numbers were made

Wolf evidence is reported by OFB staff, parks, hunters, farmers and the public, then rated by OFB experts, some by genetic test; places are blurred to the commune on purpose. Sheep are not counted animal by animal. Each region's statistics service builds the figure from surveys, farm-aid claims and local contacts, and where a department has few sheep it may move last year's figure in line with the national average. Every figure carries a quality mark from 1 (reasonably reliable) to 3 (to be used only in totals): Val-d'Oise and Pas-de-Calais, both never reached, are marked 3.

Controls

Normalise: a percentage change, so Aveyron with a million sheep and Val-d'Oise with 1,400 sit in one test. Restrict: the departments wolves reached part of the way through are left out, and so are those with fewer than 1,000 sheep.

Credit

OFB, Réseau Loup-Lynx, and Agreste, Statistique agricole annuelle, both under Licence Ouverte (France's open-data licence), with the dates downloaded.

Treat your data

Run a statistical test, leading to figures you built and a statistical result.

Mann-Whitney U, two-tailed, at p = 0.05.

Why this test

A difference between two independent groups of departments, 16 and 12. Groups that small cannot show that they are bell-shaped, so a rank test rather than a t-test.

Decided before running it

Critical value of U for groups of 16 and 12: 53 (check it in the table you cite). A U at or below it is significant. Report both medians as well as p, because a test says whether there is a difference, not how big it is. Free calculator: Social Science Statistics.