Wildlife data analysis · Alberta, Canada

Alberta banned bear baiting in half the province. Here is what the harvest data says it changed.

Baiting black bears is legal in 42 of Alberta's wildlife management units, prohibited in most of the rest, and permitted over part of seven more. That makes the province an accidental controlled experiment — five seasons, 12,939 harvested bears, and a built-in intermediate group to test the gradient.

Headline findings

Four results, one of which is unusually clean for observational data.

Natural experiment

Baiting units run roughly double the success

In all five seasons, units where baiting is legal returned 26–36% hunter success against 13–19% where it is banned. The ratio stayed between 1.72× and 2.02× — a narrow band, while total harvest swung 58%.

Dose-response

Partial-baiting units land in between — every year

Seven units allow baiting in part of their area. They sat between the two groups in all five seasons without exception. That ordering is what separates a real effect from a regional coincidence.

Null result

Baiting does not shift the sex ratio

A common objection is that baiting leads to sows being shot. In 2025 the female share of harvest was 20.6% in baiting units and 21.3% where it is banned — essentially identical, and marginally lower where baiting is legal.

Structural zero

Not one cub in five seasons

The young column is zero in every unit in every year. It is illegal to take a bear under one year or a female with a cub, and the harvest reflects that absolutely — unlike deer, elk and moose where young run 2–6%.

The baiting experiment

Alberta's regulations split the province into three groups. Nobody designed this as a study, but it behaves like one.

Hunter success by baiting status, five seasons
Harvest-weighted success: total harvest divided by total estimated hunters in each group, rather than an average of unit percentages. The three lines never cross.
baiting permitted permitted in part of the unit baiting prohibited
Which units you can bait in
Every unit with recorded bear harvest, grouped by baiting status. Hover any unit for its 2025 numbers; click to open it in the dashboard. This is a reading of the 2026 regulations — the partial units allow bait over only part of their area, and additional restrictions apply near dwellings, parks and industrial sites.

Why the middle group matters

Baiting-legal units are concentrated in the boreal north, so the obvious objection is that this is geography rather than bait — northern units might simply hold more bears. The partial-baiting group is what makes that hard to sustain. Those seven units are scattered across the foothills and boreal, are governed by the same regulations as their neighbours in every other respect, and allow bait over only part of their area. They land between the two groups in all five seasons, every time, without a single inversion.

That is a dose-response relationship: more permitted bait area, more success. It does not prove causation on its own, but a confounder would have to track partial permissions unit by unit to reproduce it.

The size of the effect is also stable in a way that regional differences usually are not. Across five seasons the ratio between the two extreme groups stayed inside a narrow band — 1.72× at the low end and 2.02× at the high end — while total provincial harvest swung by 58%. Whatever baiting is worth to a hunter, it is worth about the same amount every year.

What the harvest is made of

Bear composition is shaped almost entirely by what hunters choose, because bear regulations set very few limits on sex.

Male share of harvest, five seasons
Males outnumber females roughly three to four times over, every year, with no trend. Compare deer and elk, where the sex mix moves more than 20 percentage points as tag policy changes.

Why the young column is empty

It is unlawful to hunt a black bear under one year old, or a female accompanied by a cub. Five seasons and 12,939 bears later, the young column is still zero in every unit. That is not rounding — it is a regulation being followed essentially perfectly, which is worth noting given how often compliance is assumed rather than measured.

It also means the sex split here is a clean two-way comparison, unlike the other species where a sexless "young" column absorbs part of the harvest.

The sow-avoidance question

Because a female with cubs is protected but a lone female is not, the female share reflects hunter judgment in the field: identifying sex on a bear at distance is hard, and many hunters pass anything they cannot confirm. That figure sat between 20.4% and 25.4% across all five seasons, which suggests a settled behavioural norm rather than a response to any regulatory change.

Which is why the baiting comparison is interesting in the other direction. Bait gives a hunter a stationary animal at known range and time, exactly the conditions that should improve sex identification. The data shows no penalty and, if anything, a marginally lower female share.

Five seasons

One year in this series does not belong with the others.

Total black bear harvest by season
2021 stands 58% above the average of the four seasons that follow it, then the series settles into a narrow band.

The 2021 outlier, and what I will not claim about it

2021 recorded 3,670 bears against 2,172 to 2,431 in every following season. Both sexes moved together and the baiting gradient held, so this is not a composition change — it is more bears taken, everywhere.

The tempting explanation is pandemic-era licence sales, and the timing fits. But this dataset contains no participation figures, no licence sales, and no earlier seasons for a baseline, so I cannot separate a surge in hunters from a change in survey methodology or a genuinely good bear year. Flagging it as unexplained is more useful than guessing.

Harvest by region
2025 season. Boreal and foothills carry almost the entire provincial harvest; the mountain units produce almost nothing despite being bear country.

The mountain 400-series result is the same pattern the elk data shows: 29 units, 73 bears, 6% success. Bears are certainly there. Access, terrain and the absence of legal bait are what keep the harvest near zero, and it is a useful reminder that a low harvest number is a statement about hunters at least as much as about animals.

Scout the units (WMUs)

Every unit and season, filterable. Click any row or bar for that unit's five-season history. Deep-link here with #dashboard.

beta Ask in plain English — a question gets parsed into the filters below and answered from the data

Prototype: a rules-based parser, running entirely in your browser with no server. In production the parser swaps for a language model behind a small serverless endpoint — the data layer stays exactly as it is here.

Season
Baiting
Class
Region
Top units
Coloured by baiting status. Click a bar for unit detail.
permitted partial prohibited
Harvest vs success
Bubble area is estimated hunters. The baiting groups separate vertically — green sits above red at almost every harvest level.
permitted partial prohibited
Male / female split of selection
Totals across every unit currently matching.
Click a column to sort · a row for detail

2026 outlook

Which units look strongest for the coming spring and fall — from a model you can check by hand.

The model, in four steps

No 2026 bear data exists yet: the spring season opens in April 2026 and the survey publishes around March 2027. So this is a forecast — but built on five seasons per unit, which is enough for a real trend rather than a line drawn through three points.

  1. Recency-weighted baseline — the five seasons weighted 1, 2, 3, 4, 5 from 2021 to 2025, so the most recent year carries a third of the weight.
  2. Fitted slope — least squares across all five seasons.
  3. Volatility — mean absolute year-over-year change as a share of the mean.
  4. Volatility-weighted shrinkage — the projection blends trend with the most recent actual season, weighted by 1 − volatility, so noisy units are pulled back toward what actually happened rather than following a line through scattered points.

2021 is the complication. It sits 58% above every following season, so a naive five-year mean is dragged upward for every unit. The recency weighting handles most of that, and the reported range is the projection ± half the volatility, so units still carrying 2021's spike show visibly wide intervals.

Baiting
Confidence
Rank by
Projected 2026 harvest — top 20 units
Hollow outline is the actual 2025 harvest for reference; the solid bar is the projection, coloured by confidence. Click any bar for that unit's five-season history.
actual 2025 projected — high confidence medium low — treat as noise

What the model says, and what would break it

The projection assumes baiting rules, season dates and supplemental-licence availability hold. Those are the levers that would move a unit far more than any trend line: if a unit gains or loses bait permission, the five-season evidence above suggests its success rate moves by roughly a factor of two, which swamps everything this model estimates. Treat the high-confidence rows as the useful output and the low-confidence rows as noise.

Method, limits and integrity

Derived figures

hunters ≈ harvest ÷ (success% ÷ 100). Group success is harvest-weighted — total harvest over total estimated hunters — rather than a mean of unit percentages, so a unit with two bears at 100% cannot distort a group.

Baiting classification

Taken from the 2026 regulations, which permit baiting in WMUs 322, 330–338, 348, 358–360, 500–506, 509, 510, 512–520, 522, 523, 529–536 and 539–544, and in portions of 320, 324, 357, 507, 521, 526 and 528. The partial group is those seven. Baiting is further restricted near dwellings, parks and industrial sites, and is limited to the open season plus the preceding two weeks — none of which this classification captures.

Limits

  • This is observational. Units were not randomly assigned to baiting. The partial-unit gradient strengthens the case considerably but does not make it an experiment.
  • Baiting status is current, harvest is historical. Classification comes from the 2026 regulations applied to 2021–2025 seasons. If a unit's baiting status changed in that window, it is misclassified here.
  • Spring and fall are pooled. Alberta runs two bear seasons a year and the reports do not separate them, so seasonal effects are invisible.
  • Success rates are survey estimates. Several small units report rates of 100% on two animals. They are shown as published but should not be ranked.
  • Supplemental licences are not modelled. A second bear licence is available across much of the province, so "hunters" derived from success rate understates participation there.
Data integrity check
Each season's unit rows are summed and compared against the "estimated total" line printed in the source report, live in the browser.