Bias In CBO Estimates

The Congressional Budget Office (CBO) produces economic reports used by Congress to justify its positions. It has mostly permanent staff, and is supposed to be non-partisan. But, it is a creature of D.C. and Congress, so I think it leans left most of the time, more left when the Democrats control both houses.

Hanno Lustig, a macroeconomist at Stanford posted a Twitter thread this week showing that it's projections about the financing of fiscal policy have been too optimistic for the last 25 years or so. This takes the form of forecasts that are always on one side of the truth: in this case on the side of "oh no government spending isn't as out of control as everyone says".

This first one shows the debt/GDP ratio (not my favorite statistic, but the choice is theirs not mine) in black. Every year they make forecasts going several years into the future. Before 2000 they were forecasting it to get bigger and it didn't. Since then they've forecast it to get bigger, and its's outpaced their forecasts.

The story is the same for deficit/GDP ratio (a more useful statistic). Here surpluses are above 0 and deficits are below it. They had a good mix of forecasts before 2000. But since then, they've mostly forecast deficits to be smaller than they actually were.

In defense of the CBO, they are tasked with analyzing bills as written. Members of Congress have gotten very good at writing into law social policies with expiration dates ... that everyone knows will be renewed rather than allowed to go away. But because they do expire formally, the CBO isn't allowed to take into account what everyone knows is going to happen. That happens with tax cuts sometimes too: reductions in FICO taxes late in the Bush II administration persisted for longer than intended because Obama didn't find it politically feasible to have them cancelled.

 

***

One other thing to note about the colored forecasts. They all have a lot more variation towards their left than towards their right. This is mentioned further down in the thread, and is evidence that the forecasts are mean-reverting. This means they kind of behave like a guitar string being plucked (in one direction), and then vibrating back and forth a bit, before settling down to where it was before.

 

Mean reversion is a property of many time series whose behavior is tied to some fundamental value like its mean. Note that the black lines, showing actual political outcomes appear unhinged rather than tied to some fundamental.