def test_arg_validation():
with pytest.raises(ValueError):
TaxBrain(2018, 2020)
with pytest.raises(AssertionError):
TaxBrain("2018", "2020", use_cps=True)
with pytest.raises(AssertionError):
TaxBrain(TaxBrain.LAST_BUDGET_YEAR,
TaxBrain.FIRST_BUDGET_YEAR,
use_cps=True)
with pytest.raises(AssertionError):
TaxBrain(TaxBrain.FIRST_BUDGET_YEAR - 1, 2018, use_cps=True)
with pytest.raises(AssertionError):
TaxBrain(2018, TaxBrain.LAST_BUDGET_YEAR + 1, use_cps=True)
def cli_core(startyear, endyear, data, usecps, reform, behavior, assump,
baseline, outdir, name, make_report, author):
"""
Core logic for the CLI
"""
tb = TaxBrain(start_year=startyear,
end_year=endyear,
microdata=data,
use_cps=usecps,
reform=reform,
behavior=behavior,
assump=assump,
base_policy=baseline,
verbose=True)
tb.run()
# create outputs
dirname = name
if not dirname:
dirname = f"TaxBrain Analysis {datetime.today().date()}"
outputpath = Path(outdir, dirname)
outputpath.mkdir()
# create output tables
aggregate = tb.weighted_totals("combined")
aggregate.to_csv(Path(outputpath, "aggregate_tax_liability.csv"))
for year in range(startyear, endyear + 1):
yeardir = Path(outputpath, str(year))
yeardir.mkdir()
make_tables(tb, year, yeardir)
if make_report:
report(tb, name=name, outdir=outputpath, author=author)
def cli_core(startyear, endyear, data, usecps, reform, behavior, assump,
baseline, outdir, name, make_report, author):
"""
Core logic for the CLI
Parameters
----------
startyear: int
year to start analysis
endyear: int
last year for analysis
data: str or Pandas DataFrame
path to or DataFrame with data for Tax-Calculator
usecps: bool
whether to use the CPS or (if False) the PUF-based file
reform: dict
parameter changes for reform run in Tax-Calculator
behavior: dict
behavioral assumptions for Behavioral-Responses
assump: dict
consumption assumptions
base_policy: dict
parameter changes (relative to current law baseline) for baseline
policy
verbose: bool
indicator for printing of output
Returns
-------
None
reports saved to disk at path specified by outdir
"""
tb = TaxBrain(start_year=startyear,
end_year=endyear,
microdata=data,
use_cps=usecps,
reform=reform,
behavior=behavior,
assump=assump,
base_policy=baseline,
verbose=True)
tb.run()
# create outputs
dirname = name
if not dirname:
dirname = f"TaxBrain Analysis {datetime.today().date()}"
outputpath = Path(outdir, dirname)
outputpath.mkdir(exist_ok=True)
# create output tables
aggregate = tb.weighted_totals("combined")
aggregate.to_csv(Path(outputpath, "aggregate_tax_liability.csv"))
for year in range(startyear, endyear + 1):
yeardir = Path(outputpath, str(year))
yeardir.mkdir(exist_ok=True)
make_tables(tb, year, yeardir)
if make_report:
report(tb, name=name, outdir=outputpath, author=author)
def test_user_input(reform_json_str, assump_json_str):
valid_reform = {"II_rt7": {2019: 0.40}}
# Test valid reform dictionary with No assumption
TaxBrain(2018, 2020, use_cps=True, reform=valid_reform)
TaxBrain(2018, 2020, use_cps=True, reform=reform_json_str)
invalid_assump = {"consumption": {}}
# Test valid reform and assumptions dictionary
valid_assump = {
"consumption": {},
"growdiff_baseline": {},
"growdiff_response": {}
}
TaxBrain(2018, 2019, use_cps=True, assump=valid_assump)
TaxBrain(2018,
2019,
use_cps=True,
reform=reform_json_str,
assump=assump_json_str)
tb = TaxBrain(2018,
2019,
use_cps=True,
reform=valid_reform,
assump=valid_assump)
required_param_keys = {
"policy", "consumption", "growdiff_baseline", "growdiff_response",
"behavior", "base_policy"
}
assert set(tb.params.keys()) == required_param_keys
with pytest.raises(ValueError):
TaxBrain(2018, 2020, use_cps=True, assump=invalid_assump)
invalid_assump = {
"consumption": {},
"growdiff_baseline": {},
"growdiff_response": {},
"invalid": {}
}
with pytest.raises(ValueError):
TaxBrain(2018, 2020, use_cps=True, assump=invalid_assump)
with pytest.raises(TypeError):
TaxBrain(2018, 2020, use_cps=True, reform=True)
with pytest.raises(TypeError):
TaxBrain(2018, 2020, use_cps=True, assump=True)
def test_stacked_run():
# reforms to use
payroll_json = """{"SS_Earnings_thd": {"2021": 400000}}"""
CG_rate_json = """{
"CG_brk3": {"2021": [1000000, 1000000, 1000000, 1000000, 1000000]},
"CG_rt4": {"2021": 0.396}
}"""
reform_dict = {
"Payroll Threshold Increase": payroll_json,
"Capital Gains Tax Changes": CG_rate_json
}
tb = TaxBrain(2021, 2022, reform=reform_dict, stacked=True, use_cps=True)
tb.run()
# check that there is a stacked table now
assert isinstance(tb.stacked_table, pd.DataFrame)
def tb_dynamic(reform_json_str):
return TaxBrain(2018, 2019, use_cps=True, reform=reform_json_str,
behavior={"sub": 0.25})
def tb_static(reform_json_str):
return TaxBrain(2018, 2019, use_cps=True, reform=reform_json_str)
def test_baseline_policy():
base = {"II_em": {2019: 0}}
reform = {"II_em": {2025: 2000}}
tb = TaxBrain(2018, 2019, use_cps=True, reform=reform, base_policy=base)
tb.run()
def run_model(meta_params_dict, adjustment):
"""
Runs TaxBrain
"""
# update meta parameters
meta_params = MetaParameters()
meta_params.adjust(meta_params_dict)
# convert COMP user inputs to format accepted by tax-calculator
policy_mods = convert_adj(adjustment["policy"], meta_params.year.tolist())
behavior_mods = convert_behavior_adj(adjustment["behavior"])
user_mods = {"policy": policy_mods, "behavior": behavior_mods}
start_year = int(meta_params.year)
use_cps = meta_params.data_source == "CPS"
if meta_params.data_source == "PUF":
puf_df = retrieve_puf(AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)
if puf_df is not None:
if not isinstance(puf_df, pd.DataFrame):
raise TypeError("'puf_df' must be a Pandas DataFrame.")
fuzz = True
use_cps = False
sampling_frac = 0.05
sampling_seed = 2222
full_sample = puf_df
else:
# Access keys are not available. Default to the CPS.
print("Defaulting to the CPS")
meta_params.adjust({"data_source": "CPS"})
if meta_params.data_source == "CPS":
fuzz = False
use_cps = True
input_path = os.path.join(TCDIR, "cps.csv.gz")
# full_sample = read_egg_csv(cpspath) # pragma: no cover
sampling_frac = 0.03
sampling_seed = 180
full_sample = pd.read_csv(input_path)
if meta_params.use_full_sample:
sample = full_sample
end_year = min(start_year + 10, TaxBrain.LAST_BUDGET_YEAR)
else:
sample = full_sample.sample(frac=sampling_frac,
random_state=sampling_seed)
end_year = start_year
tb = TaxBrain(start_year,
end_year,
microdata=sample,
use_cps=use_cps,
reform=policy_mods,
behavior=behavior_mods)
tb.run()
# Collect results for each year
delayed_list = []
for year in range(start_year, end_year + 1):
print('delaying for', year)
delay = delayed(nth_year_results)(tb, year, user_mods, fuzz)
delayed_list.append(delay)
results = compute(*delayed_list)
# process results to get them ready for display
# create aggregate plot
agg_plot = aggregate_plot(tb)
all_to_process = defaultdict(list)
for result in results:
for key, value in result.items():
all_to_process[key] += value
results, downloadable = postprocess(all_to_process)
agg_output, table_output = create_layout(results, start_year, end_year)
model_versions_str = ""
for model, version in TaxBrain.VERSIONS.items():
model_versions_str += f"{model}: {version}\n"
comp_outputs = {
"renderable": [agg_plot, agg_output, table_output],
"model_version": model_versions_str,
"downloadable": downloadable
}
return comp_outputs
from taxbrain import TaxBrain
reform_url = "https://raw.githubusercontent.com/PSLmodels/Tax-Calculator/master/taxcalc/reforms/Larson2019.json"
# run static analysis
tb_static = TaxBrain(2019, 2028, use_cps=True, reform=reform_url)
tb_static.run()
static_table = tb_static.weighted_totals("c00100")
print("Tax Liability by Year\n")
print("Static Results")
print(static_table)
# run dynamic analysis
tb_dynamic = TaxBrain(2019,
2028,
use_cps=True,
reform=reform_url,
behavior={"sub": 0.25})
tb_dynamic.run()
dynamic_table = tb_dynamic.weighted_totals("c00100")
print("Dynamic Results")
print(dynamic_table)
# produce a differences table
diff = tb_static.differences_table(2019, "weighted_deciles", "combined")
print("\nDifferences Table for 2019")
print(diff)
def run_tbi_model(start_year,
data_source,
use_full_sample,
user_mods,
puf_df=None):
"""
Run TBI using the taxbrain API
"""
tbi_path = os.path.abspath(os.path.dirname(__file__))
tcpath = inspect.getfile(Records)
tcdir = os.path.dirname(tcpath)
# use taxbrain
if data_source == "PUF":
if not isinstance(puf_df, pd.DataFrame):
raise TypeError("'puf_df' must be a Pandas DataFrame.")
fuzz = True
use_cps = False
sampling_frac = 0.05
sampling_seed = 2222
full_sample = puf_df
else:
fuzz = False
use_cps = True
input_path = os.path.join(tbi_path, '..', 'cps.csv.gz')
if not os.path.isfile(input_path):
# otherwise read from taxcalc package "egg"
input_path = os.path.join(tcdir, "cps.csv.gz")
# full_sample = read_egg_csv(cpspath) # pragma: no cover
sampling_frac = 0.03
sampling_seed = 180
full_sample = pd.read_csv(input_path)
if use_full_sample:
sample = full_sample
end_year = min(start_year + 10, TaxBrain.LAST_BUDGET_YEAR)
else:
sample = full_sample.sample(frac=sampling_frac,
random_state=sampling_seed)
end_year = start_year
tb = TaxBrain(start_year,
end_year,
microdata=sample,
use_cps=use_cps,
reform=user_mods["policy"],
behavior=user_mods["behavior"])
tb.run()
# Collect results for each year
delayed_list = []
for year in range(start_year, end_year + 1):
print('delaying for', year)
delay = delayed(nth_year_results)(tb, year, user_mods, fuzz)
delayed_list.append(delay)
results = compute(*delayed_list)
all_to_process = defaultdict(list)
for result in results:
for key, value in result.items():
all_to_process[key] += value
results = postprocess(all_to_process)
return results