The General Accounting Office takes a shot at forecasting future federal debt levels, and the results are unfortunately and unsurprisingly scary, as the above chart shows. The Alternative baseline — the non-magic unicorn baseline — is the one you should pay attention to. The real difference between Baseline Extended and Alternative is that the latter assumes an unreformed entitlement system.
And the longer we wait to deal with the entitlement debt problem, the bigger the problem becomes and the more dramatic action will be needed to wrangle it.
For example, to keep debt held by the public as a share of GDP in 2086 from exceeding its level at the beginning of 2012 (roughly 68% of GDP) in our Alternative simulation, the fiscal gap is 8.3% of GDP. This means that revenue would have to increase by 46% or noninterest spending would have to be reduced by about 32% (or some combination of the two) on average over the 75-year period. Even more significant changes would be needed to reduce debt to lower levels. …
However, the longer action is delayed the greater the risk that the eventual changes will be disruptive and destabilizing. Under our Alternative simulation, waiting 10 years would increase the fiscal gap to nearly 10% of GDP— meaning a revenue increase of more than 54% or a noninterest spending cut of about 37% or some combination of the two would be required to bring debt held by the public back to its level in 2012 by 2086.
Actually, even the Alternative baseline numbers are overly optimistic because they assume — despite the massive increase in debt — that GDP growth, inflation, and interest rate levels stay constant at 2.1%, 2.0%, and 5.2%. They assume the massive increase in debt has no economic impact.
Of course, that is unlikely. Indeed, those impacts would surely make the debt situation even worse even faster, adding an additional 50 percentage points to debt/GDP levels by the 2030s, as this CBO analysis shows:
Also note that these charts assume historical levels of federal tax revenue. The problem is spending.