Editing policy_current_law.json to incorporate ARPA.json values

[amshoulders]

This pull request edits the policy_current_law.json file to incorporate the ARPA.json parameter values.

cc @jdebacker

[jdebacker]

This year should be 2022. See error message from tests about CTC_new_c:

The Value objects for CTC_new_c do not span the specified parameter space. Duplicate combinations:
14085
E           	[((2013,), 2)]

[jdebacker]

@amshoulders After fixing that year index in the JSON file, you will need to do a couple things to make sure tests pass:

1. You'll need to update the csv files against which output are tested (since these cached files need to match the results produced from the model under current law). If you run the test suite locally, you'll see a message displayed with the names of the files to replace. Note that you will only be able to do this with the CPS-based files. I'll open a PR to your branch to update the PUF-based results.
2. You'll need to update the the tests/test_reforms.test_round_trip_tcja_reform() function. This function starts with 2017 law and layers on tax law changes since to get to current law. If you look at some in-line comments there you can see how the TCJA and the CARES Act were added. You'll want to do similar with the ARPA law change. For the ARPA changes, you can either paste in the dictionary of parameter updates from the ARPA.json file (this is what is done for the CARES Act) OR you can point to the url for ARPA.json file in the PSL/Examples repository (the TCJA changes point to the JSON file within the Tax-Calculator repo - you'd do similar, but passing a URL rather than file path).

[jdebacker]

I'm don't understand the failure with test_parameters.py::test_json_file_contents[policy_current_law.json]. Here's the error message:

>           raise ValueError(failures)
247
E           ValueError: param:<ALD_BusinessLosses_c>; len(value_yrs)=9; known_years={2016, 2017, 2018, 2019, 2026, 2013, 2014, 2015}
248
E           param:<EITC_c>; len(value_yrs)=9; known_years={2016, 2017, 2018, 2019, 2013, 2014, 2015}
249
E           param:<EITC_ps>; len(value_yrs)=9; known_years={2016, 2017, 2018, 2019, 2013, 2014, 2015}
250
E           param:<EITC_InvestIncome_c>; len(value_yrs)=8; known_years={2016, 2017, 2018, 2019, 2013, 2014, 2015}
251
E           param:<CTC_new_ps>; len(value_yrs)=9; known_years={2016, 2017, 2018, 2019, 2013, 2014, 2015}

But the ALD_BusinessLosses_c looks like it is defined correctly in my view. I thought gaps in years would be filled in by the extrapolation in ParamTools. Can anyone let me and @amshoulders know what the issue is here? Thanks!

cc @hdoupe

[jdebacker]

@hdoupe I the issue here with the ALD_BusinessLosses_c parameter (highlighted in comment above) an instance where we need to use the clobber function you demonstrated in the PSL Demo Day today?

[hdoupe]

@jdebacker sorry I just saw this. It looks like this test only loads the JSON file and does not feed it through ParamTools for extrapolation:

Tax-Calculator/taxcalc/tests/test_parameters.py

Lines 204 to 252 in 2f94e6e

	with open(os.path.join(tests_path, "..", fname)) as f:
	allparams = json.loads(f.read())
	for pname in allparams:
	if pname == "schema":
	continue
	# check that param contains required keys
	param = allparams[pname]
	# check that indexable and indexed are False in many files
	if fname != 'policy_current_law.json':
	assert param.get('indexable', False) is False
	assert param.get('indexed', False) is False
	# check that indexable is True when indexed is True
	if param.get('indexed', False) and not param.get('indexable', False):
	msg = 'param:<{}>; indexed={}; indexable={}'
	fail = msg.format(pname,
	param.get('indexed', False),
	param.get('indexable', False))
	failures += fail + '\n'
	# check that indexable param has value_type float
	if param.get('indexable', False) and param['type'] != 'float':
	msg = 'param:<{}>; type={}; indexable={}'
	fail = msg.format(pname, param['type'],
	param.get('indexable', False))
	failures += fail + '\n'
	# ensure that indexable is False when value_type is not real
	if param.get('indexable', False) and param['type'] != 'float':
	msg = 'param:<{}>; indexable={}; type={}'
	fail = msg.format(pname,
	param.get('indexable', False),
	param['value_type'])
	failures += fail + '\n'
	# check that indexed parameters have all known years in value_yrs list
	# (form_parameters are those whose value is available only on IRS form)
	form_parameters = []
	if param.get('indexed', False):
	defined_years = set(
	vo["year"] for vo in param["value"]
	)
	error = False
	if pname in long_params:
	exp_years = long_known_years
	else:
	exp_years = known_years
	if pname in form_parameters:
	if defined_years != exp_years:
	error = True
	else:
	if defined_years != exp_years:
	error = True

Maybe a better test would be:

• check the min / max years of the JSON contents
• load JSON with appropriate class (tc.Policy, etc.)
• make sure years are filled in correctly

[jdebacker]

Ah - thanks @hdoupe! I'll take a stab at rewriting this test.

[codecov]

Codecov Report

Merging #2588 (5133def) into master (cb0297d) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #2588   +/-   ##
=======================================
  Coverage   98.46%   98.46%           
=======================================
  Files          14       14           
  Lines        2611     2611           
=======================================
  Hits         2571     2571           
  Misses         40       40           

Flag	Coverage Δ
unittests	98.46% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

[jdebacker]

@amshoulders This PR looks good to me -- thanks for your work! You can go ahead and remove the [WIP] from the title.

@MattHJensen I have one question confirming that you are ok with a change to a unit test, but after that, this is ready for your final review.

[MattHJensen]

Thanks very much @amshoulders for the PR, @jdebacker for your review and contributions, and @hdoupe for your review and suggestions. Merging.