A few more points regarding the analysis and example earlier (warning – reading this will give you a headache):
Corrections: In the caveats section, strike the work “substantially”, since a smaller retirement tax bracket will return more (I meant that a substanially smaller tax bracket will probably be required to offset increases in the income tax, reduced flexibility of a std. 401k, large distributions & RMDs), also regarding the ROTH it should read “… the ability to retire a few years earlier” – see #2 below.
1) For clarification, I should have stated an important assumption – that the person ONLY had funds for the maximum allowable 401k contribution (ROTH or standard) *pretax*.
Which means that the std. 401k contributions were at the maximums, while the ROTH 401 contributions were NOT (max *less* the income tax rate). For example, 15k pretax savings would be fully allocated to the std. 401k, while only ~11k went into the ROTH 401.
E.g. if the employee *had* the ability to fully fund the ROTH 401k with pretax dollars, then:
~21k pretax contribution -> 15k into ROTH, 6k paid in taxes (28% tax rate)
vs.
~21k pretax contribution -> 15k into a std. 401k, wih 4.2k into a taxable account (invested buy & hold, withdrawn @ 15% cap gains)
Under this fully-funded scenario, the ROTH 401 will distribute more than a standard 401k (including taxable account) at equal input/output tax rates (with the larger the input tax rate, the greater the benefit of the ROTH, as counter-intuitive as that sounds).
In the 24 year example posted earlier:
Before distribution taxes, the 845k ROTH would’ve been 1175k fully funded, while the std. no match 401 would’ve also been 1175k (since the inputs were the same), in addition, the std. 401k must take into account the portion of taxes not paid to fully fund the ROTH, which would’ve amassed to 410k in the taxable account, making for a 1585k total. After distribution taxes on each portion (28% & 15%) the std. 401 would’ve dist. 1125k (less if the capital gains rate were to increase in time, in the taxable account) – which is less then the ROTH even though the same input & output tax was used…
Expressed another way, at a 28% contribtion tax rate, both accounts would be equal if withdrawing the std. 401 at a lower ~24% distribution tax, meaning lower tax rates during retirement don’t necessarily mean a std. 401k is better.
Since a ROTH 401 (not to be confused with the ROTH IRA) automatically includes the std. 401k employer match (if offered), there is no need to worry about the matching factor. If you were comparing a ROTH IRA to a MATCHED 401k, then fully funding the ROTH IRA is only advantageous if the match is *extremely* small, so its much better to fund the std. 401k to receive the full match, using the remains for the ROTH IRA.
With all that mess said, fully funding (if possible) the ROTH 401/IRA is more advantageous, even if you will be in a somewhat lower tax bracket at retirement (up to the difference in input/output tax rates).
Tax laws are so convoluted…
2) Early retirement:
If one is thinking about early retirement, which would entail taking non-qualified (early) disbursements from their std. 401k or ROTH, then the scenarios are much different.
Retiring a few years early may favor the ROTH, since you would most likely be withdrawing contributions which are tax and penalty free, but withdrawing the earnings early will be horrendous, due to paying income tax plus a 10% penalty, making it far worse then a taxable accounts if done withdrawing well before retirement.
Whether a taxable account is better then dipping into a standard 401k (and taking the 10% penalty) depends on how long the account has a chance to grow, the rate of return, and how early distributions begin. It appears that given at least 15 years, and at an avg. 10% return, distributions at the tax input rate would be roughly the same for each.
So using a ROTH 401 exclusively may not be the best for a substantially early retirement. The early retirement scenarios are far more complex, with many indeterminate variables interacting, so I will leave that analysis to one of our budding MC simulation-ists.
