Model income tax doc

[HughParsonage]

@MattCowgill let me know what you think of the most recent commit (i.e. the vignette)

[MattCowgill]

Thanks @HughParsonage. I have to try and finalise the dental report today and will try to review this over the weekend.

[MattCowgill]

Thanks @HughParsonage. Sorry for the delay. @k-griffiths is having a look to see if she can think of anything else that would be useful to add.

PS Github won't let me merge the PR while Travis is failing

[HughParsonage]

• Under changing ordinary tax parameters – “Note that the top marginal tax rate must include the temporary budget repair levy. This causes a slight difference between the two taxes because the small business tax offset does not offset levies.” – it’s not clear to me what ‘the two taxes’ refers to or what effect the ‘slight difference’ has? Bit more explanation here would help clarify.
• Under changing Medicare levy parameters – unclear how the default changes to the thresholds are calculated, a bit more explanation here would help
  • Can you change other things e.g. the medicare levy phase in rate? (medicare_levy_taper =?)
  • Looks like you can change medicare_levy_rate OR medicare_levy_taper OR medicare_levy_upper_threshold and the other two variables will adjust accordingly?
  • The default setting for each of these is NULL so presumably that means no change from the level specified in medicare_tbl.tsv?
• Change to the LITO so that the maximum offset is $1000 – I think this is replacing the current $445 limit with $1000 and then the amount reduces by 1.5 cents for each dollar over $37,000? But the current description implies a simple max offset of $1000. Bit more explanation here would help clarify.
• The project function uses “the internal forecast methods” – can you elaborate on this a bit more?
  • What’s the default for wage.series and lf.series?
  • Is this function dependent on the wage_inflator function?
  • What’s happening here: multiply_by(s1819[["WEIGHT"]][1L])
  • You seem to be multiplying total tax by the projected weight, but I don’t get the 1L bit
  • What is Sw_amt? excl_vars is documented but I can’t find anything on Sw_amt

[MattCowgill]

Thanks @HughParsonage this is all really helpful. As I said yesterday, I'm happy to merge the PR once @k-griffiths is happy, although Github won't let me merge while Travis is failing. From Tuesday I'll have some time to devote to this, so I hope we can talk next week.

[k-griffiths]

I'm trying to construct a simple income tax change (increasing the top rate to 50%) and look at the budget impact in different years. I've then compared the outputs to STINMOD+ and they’re wildly different:

• Under model_income_tax I get extra revenue of $5.9 billion in 2018-19 and $7.5 billion in 2022-23
• Under STINMOD+ only $1.8 billion in 2018-19 and $5.4 billion in 2022-23 Any ideas?

## Get latest year of ATO data (2015-16)
s1516 <- sample_file_1516_synth %>%
         mutate(fy.year = "2015-16",
                WEIGHT = 50, # Years with a 1% sample file have a weight of 100; years with a 2% sample file have a weight of 50
                id = rownames(sample_file_1516_synth)) %>% 
         as.data.table

## Project data forward to current financial year (2018-19)
s1819 <- project(s1516, h = 3L) %>%
         mutate(fy.year = "2018-19") %>% 
         as.data.table

## Increase highest income tax rate to 50% in 2018-19
s1819_alt_tax_rate <- model_income_tax(s1819, baseline_fy = "2018-19",
                                   ordinary_tax_thresholds = c(0, 18200, 37000, 80000, 180000),
                                   ordinary_tax_rates = c(0, 0.19, 0.325, 0.37, 0.50),
                                   return. = "sample_file.int")

revenue_foregone(s1819_alt_tax_rate) #"$5.9 billion" extra revenue in 2018-19
revenue_foregone(s1819_alt_tax_rate, revenue_positive = TRUE) #"$5.9 billion" (same result with or without revenue_positive = TRUE)

## STINMOD+ comparison $1.8 billion in 2018-19 
```r

[HughParsonage]

What's happening here: multiply_by(s1819[["WEIGHT"]][1L]). You seem to be multiplying total tax by the projected weight, but I don’t get the 1L bit

[1L] just means the first element. Since all weights are the same, we only need to multiply by the first.

What is Sw_amt? excl_vars is documented but I can’t find anything on Sw_amt

Sw_amt is a column in the sample file, salary/wages amount.

[HughParsonage]

Re the comparison with STINMOD, the reasons for the difference from most important to least:

• The second-highest marginal rate starts at $87,000 not $80,000. (In general, it's best to supply the minimal arguments that represents the change.)
• You need to use the real sample file, not the synthetic version
• Likely there are differences in the way we project from 2015-16 to 2018-19

[k-griffiths]

Thanks Hugh. I've changed the second-highest rate now (it's $90k for 2018-19) and the results are still wildly different ($4.5b vs. $1.8b). The synthetic file doesn't make any difference to the total revenue gained. It must be something to do with the baseline comparison point, I'll keep working on it for now, but this would be a good example to add to the vignette.

[k-griffiths]

I've now tried aligning model_income_tax outputs with both STINMOD+ and the Roach model. The big difference is for 2018-19 and STINMOD+ is the odd one out. All three fairly close for 2021-22.

# Get latest ATO sample file (2015-16)
data2016<- read.csv("2016 sample file/2016_sample_file.csv", stringsAsFactors=FALSE)

d1516 <- data2016 %>%
  mutate(fy.year = "2015-16",
         WEIGHT = 50, # Years with a 1% sample file have a weight of 100; years with a 2% sample file have a weight of 50
         id = rownames(data2016)) %>% 
  as.data.table

# Project data forward to current financial year (2018-19)

d1819 <- project(d1516, h = 3L) %>%
  mutate(fy.year = "2018-19") %>% 
  as.data.table

# Increase highest income tax rate to 50%

d1819_alt_tax_rate <- model_income_tax(d1819, baseline_fy = "2018-19",
                                       ordinary_tax_thresholds = c(0, 18200, 37000, 90000, 180000),
                                       ordinary_tax_rates = c(0, 0.19, 0.325, 0.37, 0.50))

revenue_foregone(d1819_alt_tax_rate) # "$4.1 billion" in extra revenue in 2018-19
# STINMOD+ comparison $1.8 billion in 2018-19 
# Roach comparison $4.3 billion in 2018-19

# Model income tax in 2021-22 under the higher income tax rate of 50%

d2122 <- project(d1516, h = 6L) %>%
  mutate(fy.year = "2021-22") %>% 
  as.data.table

d2122_alt_tax_rate <- model_income_tax(d2122, baseline_fy = "2018-19",
                                       ordinary_tax_thresholds = c(0, 18200, 37000, 90000, 180000),
                                       ordinary_tax_rates = c(0, 0.19, 0.325, 0.37, 0.50))

revenue_foregone(d2122_alt_tax_rate) # "$5.1 billion" in extra revenue in 2021-22
# STINMOD+ comparison $5.5 billion in 2021-22
# Roach comparison $5.6 billion in 2021-22

# Model income tax in 2022-23 under the higher income tax rate of 50%

d2223 <- project(d1516, h = 7L) %>%
  mutate(fy.year = "2022-23") %>% 
  as.data.table

# Assume thresholds stay the same as 2018-19, only the top tax rate changes
d2223_alt_tax_rate_only <- model_income_tax(d2223, baseline_fy = "2018-19",
                                            ordinary_tax_thresholds = c(0, 18200, 37000, 90000, 180000),
                                            ordinary_tax_rates = c(0, 0.19, 0.325, 0.37, 0.50),
                                            return. = "sample_file.int")

revenue_foregone(d2223_alt_tax_rate_only) # "$5.4 billion" in extra revenue compared to 2018-19 if thresholds are kept as per 2018-19 and top tax rate increased)
# STINMOD+ comparison $5.4 billion in 2022-23
# No Roach comparison available