policyengine-canada

Name: policyengine-canada
Author: PolicyEngine

31stars5forks

How to install policyengine-canada?

policyengine/policyengine-claude/policyengine-canada

$npx -y skills add PolicyEngine/policyengine-claude --skill policyengine-canada

Installs into the current project.

Summary

How to compute Canadian household tax/benefit outcomes with PolicyEngine, and the hard limits. Load when a task involves Canadian taxes or benefits (CCB, GST/HST credit, CWB, OAS, provincial tax) at the household level. Key facts: pe.ca does NOT exist — Canada is not in the polic

Files · 1

View on GitHub

SKILL.md

1	# PolicyEngine Canada domain knowledge
2
3	Canada is the exception to the standard PolicyEngine stack. Read this before writing any Canadian
4	analysis — the entry points and the limits are different from the US/UK.
5
6	Verified against `policyengine-canada` 0.99.0 (2026-07), installed standalone. Do not hardcode
7	benefit amounts; look them up live (below), because provincial and federal parameters re-index
8	every year.
9
10	## Two hard constraints
11
12	1. `pe.ca` does not exist. The `policyengine` wrapper ships US and UK only; there is no `pe.ca`.
13	Use the `policyengine_canada` package directly.
14
15	<!-- verify -->
16	```python
17	import policyengine as pe
18	assert not hasattr(pe, "ca") # Canada is not in the policyengine wrapper
19	```
20
21	2. Canada is household-only — there is no representative microdata. The package ships a small
22	synthetic template, not a survey-weighted population sample. So you cannot run population
23	microsimulation, and you cannot produce national program costs, revenue estimates, caseloads,
24	or poverty rates for Canada. If asked "what would this cost nationally" or "how many families
25	benefit," say that population estimates are not available for Canada and offer a household example.
26	What you can do: compute taxes/benefits for a specific family, compare baseline vs. reform for
27	that family, sweep it across an income range, and compare provinces.
28
29	## One household calculation
30
31	Install the package on its own (it is not in the shared analysis venv), then build a `situation`
32	and use `policyengine_canada.Simulation`. Values below verified against 0.99.0.
33
34	```bash
35	uv pip install policyengine-canada
36	```
37
38	```python
39	from policyengine_canada import Simulation
40
41	sim = Simulation(situation={
42	"people": {
43	"parent1": {"age": {"2026": 35}, "employment_income": {"2026": 45_000}},
44	"parent2": {"age": {"2026": 33}, "employment_income": {"2026": 20_000}},
45	# full_custody defaults to False, which HALVES the CCB (see gotcha below):
46	"child1": {"age": {"2026": 4}, "full_custody": {"2026": True}},
47	"child2": {"age": {"2026": 9}, "full_custody": {"2026": True}},
48	},
49	"households": {"household": {
50	"members": ["parent1", "parent2", "child1", "child2"],
51	"province_code": {"2026": "ONT"}, # note: "ONT" for Ontario, "QC" for Quebec
52	}},
53	})
54	assert round(float(sim.calculate("child_benefit", 2026)[0]), 2) == 11_030.75 # federal CCB
55	assert float(sim.calculate("adjusted_family_net_income", 2026)[0]) == 65_000 # AFNI
56	```
57
58	Entities are just `household` and `person` (far simpler than the US six). `province_code`
59	is an enum on the household: `AB, BC, MB, NB, NL, NS, NT, NU, ONT, PE, QC, SK, YT` — Ontario is
60	`"ONT"`, not `"ON"`.
61
62	Key variable names (they are not the acronyms): federal Canada Child Benefit = `child_benefit`,
63	GST/HST credit = `gst_credit`, Canada Workers Benefit = `canada_workers_benefit`, Old Age
64	Security = `oas_net`. Browse the full list with
65	`from policyengine_canada import CountryTaxBenefitSystem; CountryTaxBenefitSystem().variables`.
66
67	### Gotcha: `full_custody` defaults to False and halves the CCB
68
69	Each child's `full_custody` defaults to False, which the model treats as 50/50 shared custody
70	and halves that child's benefit base. A sole-custody two-parent family gets the full amount only
71	if you set `full_custody: {"YEAR": True}` on each child — otherwise every CCB (and child-benefit-
72	derived) figure comes out at half. Verified: the family above yields $3,658.25 with the default
73	and $11,030.75 with `full_custody=True`.
74
75	### Gotcha: prefer the specific benefit variable over `household_net_income`
76
77	In 0.99.0, computing `household_net_income` (or `gst_credit`, which depends on it) can raise a
78	`ClimateActionIncentiveCategory` enum error via the climate-incentive chain. Calculate the specific
79	variable you need (`child_benefit`, `adjusted_family_net_income`, `income_tax`) rather than the full
80	net-income roll-up.
81
82	## Look parameters up live — n

People who liked this also viewed

6 picks

Type

policyengine-canada

31stars5forks

View on GitHub

How to install policyengine-canada?

policyengine/policyengine-claude/policyengine-canada

$npx -y skills add PolicyEngine/policyengine-claude --skill policyengine-canada

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# PolicyEngine Canada domain knowledge
2
3	Canada is the exception to the standard PolicyEngine stack. Read this before writing any Canadian
4	analysis — the entry points and the limits are different from the US/UK.
5
6	Verified against `policyengine-canada` 0.99.0 (2026-07), installed standalone. Do not hardcode
7	benefit amounts; look them up live (below), because provincial and federal parameters re-index
8	every year.
9
10	## Two hard constraints
11
12	1. `pe.ca` does not exist. The `policyengine` wrapper ships US and UK only; there is no `pe.ca`.
13	Use the `policyengine_canada` package directly.
14
15	<!-- verify -->
16	```python
17	import policyengine as pe
18	assert not hasattr(pe, "ca") # Canada is not in the policyengine wrapper
19	```
20
21	2. Canada is household-only — there is no representative microdata. The package ships a small
22	synthetic template, not a survey-weighted population sample. So you cannot run population
23	microsimulation, and you cannot produce national program costs, revenue estimates, caseloads,
24	or poverty rates for Canada. If asked "what would this cost nationally" or "how many families
25	benefit," say that population estimates are not available for Canada and offer a household example.
26	What you can do: compute taxes/benefits for a specific family, compare baseline vs. reform for
27	that family, sweep it across an income range, and compare provinces.
28
29	## One household calculation
30
31	Install the package on its own (it is not in the shared analysis venv), then build a `situation`
32	and use `policyengine_canada.Simulation`. Values below verified against 0.99.0.
33
34	```bash
35	uv pip install policyengine-canada
36	```
37
38	```python
39	from policyengine_canada import Simulation
40
41	sim = Simulation(situation={
42	"people": {
43	"parent1": {"age": {"2026": 35}, "employment_income": {"2026": 45_000}},
44	"parent2": {"age": {"2026": 33}, "employment_income": {"2026": 20_000}},
45	# full_custody defaults to False, which HALVES the CCB (see gotcha below):
46	"child1": {"age": {"2026": 4}, "full_custody": {"2026": True}},
47	"child2": {"age": {"2026": 9}, "full_custody": {"2026": True}},
48	},
49	"households": {"household": {
50	"members": ["parent1", "parent2", "child1", "child2"],
51	"province_code": {"2026": "ONT"}, # note: "ONT" for Ontario, "QC" for Quebec
52	}},
53	})
54	assert round(float(sim.calculate("child_benefit", 2026)[0]), 2) == 11_030.75 # federal CCB
55	assert float(sim.calculate("adjusted_family_net_income", 2026)[0]) == 65_000 # AFNI
56	```
57
58	Entities are just `household` and `person` (far simpler than the US six). `province_code`
59	is an enum on the household: `AB, BC, MB, NB, NL, NS, NT, NU, ONT, PE, QC, SK, YT` — Ontario is
60	`"ONT"`, not `"ON"`.
61
62	Key variable names (they are not the acronyms): federal Canada Child Benefit = `child_benefit`,
63	GST/HST credit = `gst_credit`, Canada Workers Benefit = `canada_workers_benefit`, Old Age
64	Security = `oas_net`. Browse the full list with
65	`from policyengine_canada import CountryTaxBenefitSystem; CountryTaxBenefitSystem().variables`.
66
67	### Gotcha: `full_custody` defaults to False and halves the CCB
68
69	Each child's `full_custody` defaults to False, which the model treats as 50/50 shared custody
70	and halves that child's benefit base. A sole-custody two-parent family gets the full amount only
71	if you set `full_custody: {"YEAR": True}` on each child — otherwise every CCB (and child-benefit-
72	derived) figure comes out at half. Verified: the family above yields $3,658.25 with the default
73	and $11,030.75 with `full_custody=True`.
74
75	### Gotcha: prefer the specific benefit variable over `household_net_income`
76
77	In 0.99.0, computing `household_net_income` (or `gst_credit`, which depends on it) can raise a
78	`ClimateActionIncentiveCategory` enum error via the climate-incentive chain. Calculate the specific
79	variable you need (`child_benefit`, `adjusted_family_net_income`, `income_tax`) rather than the full
80	net-income roll-up.
81
82	## Look parameters up live — n

People who liked this also viewed

6 picks

Type