PSLmodels
Extending taxdata projection beyond 2029
Originally in Tax-Calculator issue 2257 @codykallen said:
Pull request #2255 updates Tax-Calculator to run through 2029. However, I think it might be useful to extend this even further. The CBO Long-term budget projections from June 2018 includes several applicable forecasts through 2048, including:
- population growth
- fertility (children per woman)
- GDP growth (real and nominal)
- growth of hours worked
- earning share of compensation
- fraction of earnings below the FICA taxable maximum
- inflation
- Social Security spending
These are presumably not enough to directly forecast each of the series for the growth factors. In the absence of better long-term forecasts, we could perhaps assume that these series grow at a rate tied to the growth of GDP per capita, and move all of imperfectly specified growth factors in tandem in the long-term forecast.
@andersonfrailey, How much taxdata work would be required to implement @codykallen's suggestion in taxdata issue #310?
@MattHJensen @feenberg @donboyd5
Just to be sure I understand precisely, I am assuming @codykallen means that dollar-valued items on individual records (income components, deduction components) would grow at the rate of nominal GDP per capita? (Presumably record weights, handled separately, would grow at the rate of population or something similar to overall population.)
Seems like a reasonable and efficient-use-of-time assumption to me, particularly given the level of confidence we should have in any forecasts for those years. There doesn't look to be anything surprising going in the drivers of years 2030-2048 in https://www.cbo.gov/system/files/2018-06/51119-2018-06-ltbo.xlsx, Tab 2, that would suggest we could predict deviations from per-capita-gdp-growth for components of income and deductions with any confidence. (For example, by the time we get out there, labor force participation has stabilized, and productivity growth and interest rates are all stable.)
@donboyd5 asked:
Just to be sure I understand precisely, I am assuming @codykallen means that dollar-valued items on individual records (income components, deduction components) would grow at the rate of nominal GDP per capita?
Yes, that's generally correct, and I think weights should grow at the population growth rate. However, it may make sense to age some variables separate from GDP. For example, perhaps Social Security income should be aged based on Social Security expenditures per capita (or something similar).
That does seem like the most-likely candidate for special treatment. But if that is treated specially, several other large related items come to mind: (1) pension income (which I believe includes not just payments from DB plans, but also employer-sponsored DC plans) also might grow at a population-aging-related rate, and these are quite large, (2) IRA distributions probably should, also, (3) we might expect medical expense deductions to grow in a compatible manner given that they, presumably, are also linked partially to population aging, (4) deductions for homeowner property taxes might slow given that property taxes paid by the elderly tend to be lower due to explicit policies and also housing choices, and (5) the rapidly growing older population is coupled with a relatively slower-growing working-age population, which could suggest relatively slower wage growth, not just in relation to income of the older population but also relative to history.
On the other hand, given that the most rapid aging will occur over the next 10 years, after which it will slow, all 5 of these items might be a bigger concern in the 10-year forecast than in the 30-year forecast.
Anyway, it might make sense to either (a) put few resources into this and just make the gdp-per-capita-growth assumption for everything, or (b) put enough resources into it to decide carefully which items are large enough and certain enough to be treated specially, and which can be ignored. I'd probably vote for "a".