Utilities for costing and evaluating Australian tax policy, including high-performance tax and transfer calculators, a fast method of projecting tax collections from ATO sample files, and an interface to common indices from the Australian Bureau of Statistics. Written to support Grattan Institute’s Australian Perspectives program.
income_taxCalculates the income tax for a given taxable income and financial year:
## [1] 8547income_tax is designed to work well with the ATO’s sample files. You can obtain the sample files from my repo:
# install.packages("taxstats", repos = "https://hughparsonage.github.io/tax-drat")
library(taxstats)
library(hutils)
library(data.table)
library(magrittr)
library(ggplot2)Simply pass the sample file to .dots.ATO and the complexities of things like Medicare levy and the Seniors and Pensioners Tax Offset are handled for you. For example:
s1314 <- as.data.table(sample_file_1314)
s1314 %>%
.[, tax := income_tax(Taxable_Income, "2013-14", .dots.ATO = s1314)] %>%
.[, .(Taxable_Income, tax)]## Taxable_Income tax
## 1: 4800 0.000
## 2: 126122 36503.970
## 3: 39742 4655.410
## 4: 108123 29574.355
## 5: 85957 21040.445
## ---
## 258770: 24462 1111.710
## 258771: 37055 3701.525
## 258772: 45024 6530.520
## 258773: 5134 0.000
## 258774: 46368 7007.640model_income_tax: modelling changes to personal income taxWhile income_tax is designed to inflexibly return the tax payable as legislated, model_income_tax is designed to calculate income tax when changes are made. For example,
s1314 %>%
# reduce top threshold from 180,000 to 150,000
model_income_tax(ordinary_tax_thresholds = c(0, 18200, 37000, 80000,
150e3),
baseline_fy = "2013-14") %>%
.[, .(Taxable_Income, baseline_tax, new_tax)]## Taxable_Income baseline_tax new_tax
## 1: 4800 0.000 0.000
## 2: 126122 36503.970 36503.970
## 3: 39742 4655.410 4655.410
## 4: 108123 29574.355 29574.355
## 5: 85957 21040.445 21040.445
## ---
## 258770: 24462 1111.710 1111.710
## 258771: 37055 3701.525 3701.525
## 258772: 45024 6530.520 6530.520
## 258773: 5134 0.000 0.000
## 258774: 46368 7007.640 7007.640projectGiven a sample file, we can project forward a number of years
or to a particular financial year
Together with model_income_tax, this allows us to make point-predictions of future years. The function revenue_foregone prettily prints the resultant revenue:
sample_file_1314 %>%
project_to("2018-19") %>%
model_income_tax(baseline_fy = "2017-18",
ordinary_tax_thresholds = c(0, 18200, 37000, 87000,
150e3)) %>%
revenue_foregone## [1] "$1.7 billion"compare_avg_tax_rates:Create comparison of average tax rates:
lapply(list("30k" = 30e3,
"36k" = 36e3,
"42k" = 42e3),
function(T2) {
model_income_tax(s1718,
baseline_fy = "2017-18",
ordinary_tax_thresholds = c(0,
18200,
T2,
87000,
180e3))
}) %>%
rbindlist(idcol = "id",
use.names = TRUE,
fill = TRUE) %>%
compare_avg_tax_rates(baseDT = .[id %ein% "36k"]) %>%
ggplot(aes(x = Taxable_Income_percentile,
y = delta_avgTaxRate,
color = id,
group = id)) +
geom_hline(yintercept = 0) +
geom_line()## Warning: Removed 6 row(s) containing missing values (geom_path).
_inflator functions have been moved to grattanInflator, breaking (for example) cpi_inflator_general_date which are now cpi_inflator.income_tax no longer accepts family_status, n_dependants, allow.forecasts and .debug
model_income_tax Arguments prefixed Budget_ and lito_ are deprecated and now offsets = set_offsets are preferred. However, using System with income_tax is likely to be just as convenient.
lito and lmito no longer accept varying parameters, but lmito is now exported
sapto_rcpp, income_tax_sapto, new_income_tax, new_medicare_levy, and new_sapto have now been removed. Use System.
The transfer functions have been removed as they became unreliable / difficult to maintain. See repository hughparsonage/grattanTransfers for possible future development
age_pensioncarer_paymentchild_care_subsidyenergy_supplementfamily_tax_benefitmodel_child_care_subsidymodel_rent_assistancenewstart_allowancepension_supplementrent_assistancestudent_repaymentunemployment_benefityouth_allowanceyouth_unemploymentThe grattan v2 package will be more focused on income tax with other elements being spun off into other packages. In particular, _inflator functions and related forecasting methods, functions essentially accessing ABS data, and functions relating to tranfers will be put in other packages, and imported later. Functions that are made available in the grattan NAMESPACE may have changes to their API when this happens.
income_tax now uses System to define the tax systemmodel_income_tax does not throw an error for incorrect specification of lito_ arguments (which are deprecated)age_grouper now works with long vectors containing missing valuesrent_assistance() example no longer works because of outdated data.income_tax and medicare_levy no longer accept family_status as an input, since it was giving misleading results. Use .dots.ATO with appropriate variables to define the spouse’s income or the number of children.income_tax gives a slightly different error message when an invalid financial year is passed. Previously “not in correct form”, now “not a valid financial year”.useABSConnection = TRUE is no longer supported because of ABS server issuesprojectproject now has r_super_balance to project super balances by a user-supplied factor, rather than a hard-coded 1.05.package:rsdmx is now Suggested, since a dependency has been orphaned.grattan’s tax modelling functions now work with the 2016-17 ATO sample filegrattan now depends on R 3.5.0 due to serialization format version 3 becoming the default in R 3.6.0.cpi_inflator fails more gracefully when the ABS’s website is not availableincome_tax now gives consistent results modulo the existence of completely empty columns that are inputs for sapto (#158)income_tax will work if .dots.ATO is a non-data.table data.frame.model_income_tax:
sapto_rcpp is now more careful about passing length-one vectors to vectorized C++ functions.project now correctly prioritizes excl_vars over variables with otherwise predefined uprating mechanisms (such as Sw_amt).awote for weekly earningsage_grouper can now have a custom first label prefix, and is much faster when length(age) is large.income_tax now emits a warning when both age and .dots.ATO are provided, indicating that age will be ignored.mutate_ntile and weighted_ntile now use the hutils equivalents. This broke 3 unit tests because of the specific phrasing of some error messages.small_business_tax_offset() is now always positive, fixing the original misinterpretation of the legislation whereby negative business income resulted in a negative offset.*_inflator functions now return correct results for non-standard but supported financial years.inflator no longer fails when to_fy is length > 1 and unordered.mutate_ntile and mutate_weighted_ntile for adding quantile columnsage_pension,carer_paymentcarers_allowanceenergy_supplementfamily_tax_benefitnewstart_allowancepension_supplementrent_assistance the Commonwealth Rent Assistancemodel_rent_assistance as experimental function for modelling changes to rent assistance.youth_allowance() now available, though limitedcompare_avg_tax_rates: create a difference in average tax rates between multiple models and a baseline tax, by percentile.install_taxstats() as a convenient means to install the non-CRAN taxstats dependency.prohibit_vector_recycling() and friends return more informative error messages.income_tax, income_tax_sapto: the default value for fy.year is the current financial yearcpi_inflator, lf_inflator_fy, wage_inflator: if both from_fy and to_fy are missing, the default values become the previous and current financial years respectively. If only one of the two are missing, an error appears.income_tax is about twice as fast since 1.6.0.0: 1.5-2.0s down from 3.0-3.7s on the 100% population (13 million)inflator and cpi_inflator, lf_inflator_fy, and wage_inflator are now much faster when either from_fy or to_fy have more than 100,000 elements:set.seed(19952010)
from_fys <- sample(yr2fy(1995:2010), size = 1e6, replace = TRUE)
microbenchmark(cpi_inflator(from_fy = from_fys, to_fy = "2015-16"))
# Old
Unit: seconds
expr min lq mean median uq
cpi_inflator(from_fy = from_fys, to_fy = "2015-16") 1.519483 1.54438 1.550628 1.549735 1.554507
max neval
1.661502 100
# New
Unit: milliseconds
expr min lq mean median uq
cpi_inflator(from_fy = from_fys, to_fy = "2015-16") 40.71753 41.94061 47.93162 42.93946 48.08461
max neval
191.3497 100yr2fy(x) no longer works for x = 1900L, despite a unit test, for the sake of performance. #> Last change: NAMESPACE at 2018-08-19 14:47:14 (4 mins ago).
#> Unit: milliseconds
#> expr min lq mean median uq max neval cld
#> yr2fy(z) 75 88 98 90 101 161 100 a
#> .yr2fy(z) 274 286 298 297 302 359 100 bUse yr2fy(x, assume1901_2100 = FALSE) if you need the old behaviour.
taxstats1516 is now a suggested dependency.model_income_tax() no longer coerces WEIGHT to integer.lito_multi Permits multiple pieces to the linear offset.Budget2018_lamington The Low And Middle Income Tax Offset proposed in the Budget 2018 budget.Budget2018_lito_202223 The proposed change to LITO from 2022-23.Budget2018_watr The offset proposed by the Opposition the Budget Reply.sbto_discount Allows modification of the small business tax offset.clear_tax_cols By default, old tax columns are deleted.warn_upper_thresholds If changed to FALSE allows the automatic changes to be applied without warning.progressivity() for Gini-based measure of the progressivity of income taxrevenue_foregone() as a convenenience for returning the revenue foregone from a modelled sample file.model_income_tax which attempts to provide every lever of the income tax system that is visible from the tax office’s sample files. Users can model the sample file by changing single parameters to observe the effect on tax collections.small_business_tax_offset: Include the small business tax offset as a standalone function and within income_tax.project and project_to no longer require fy.year.of.sample.file. However, they expect the supplied data.frame to be compatible with the sample file provided. Failling to provide a sample file with the expected number of rows or not providing a sample file with a valid number of rows is a warning, which can be silenced by check_fy_sample_file = FALSE.mgcv was used but not declared in Suggests: Thanks to BDR for reporting.prohibit_vector_recycling to return the maximum permissible length of a list of vectors.lf_trend internal data table used to report the labour force in thousands of persons, as the ABS does. This seemed a bit strange, so now obsValue uses integers (i.e. just the labour force).taxstats to a temporary directory if not already installed, rather than the user or system’s library.income_tax, and related functionsage in income_tax now NULL rather than 42..dots.ATO. However, if .dots.ATO is supplied (and the age variable has not been removed from it), the individuals’ SAPTO eligibility is determined by the age variable in .dots.ATO, rather than setting each individual’s SAPTO to 0.NEWS.md file to track changes to the package.This is a package update
0 ERRORS | 0 WARNINGS | 1-2 NOTEs
NOTES:
Notes for taxstats are routine