What to Make of Revenue Surprises
A Democratic-economist friend of mine emailed me last night irate at Republicans. Yesterday, when the Administration reported that tax revenues were coming in stronger than expected, many Republicans were quick to argue that this showed that their fiscal policies were working.
My friend pointed out that forecasting tax revenue entails a lot of inherent uncertainty. A surprise even as large as $100 billion is not all that surprising, and we shouldn't make too much of it. It is absurd, he said, to suggest that this revenue surprise tells us much about current policies. My friend thinks it's just a stroke of luck.
We saw a similar phenomenon in the late 1990s, when positive revenue surprises drove the federal budget into surplus. Democrats were then quick to claim credit for Clinton policies. They said, "See, raising taxes did not have all the negative effects that Republicans predicted." Meanwhile, Republicans thought Clinton just got lucky.
Both cases are examples of confirmation bias--the tendency to interpret evidence in favor of one's preconceptions. When there is good news, the party in charge overinterprets the evidence as establishing the rectitude of their policies. The party out of power is too dismissive of the evidence.
A good Bayesian updates priors in response to news. The evidence from the Clinton expansion should have reduced one's estimates of the adverse effects of tax hikes, and yesterday's evidence from the Bush expansion should move one's view in the opposite direction. But remember that we have a lot of data, and each year adds only one more data point.
My friend pointed out that forecasting tax revenue entails a lot of inherent uncertainty. A surprise even as large as $100 billion is not all that surprising, and we shouldn't make too much of it. It is absurd, he said, to suggest that this revenue surprise tells us much about current policies. My friend thinks it's just a stroke of luck.
We saw a similar phenomenon in the late 1990s, when positive revenue surprises drove the federal budget into surplus. Democrats were then quick to claim credit for Clinton policies. They said, "See, raising taxes did not have all the negative effects that Republicans predicted." Meanwhile, Republicans thought Clinton just got lucky.
Both cases are examples of confirmation bias--the tendency to interpret evidence in favor of one's preconceptions. When there is good news, the party in charge overinterprets the evidence as establishing the rectitude of their policies. The party out of power is too dismissive of the evidence.
A good Bayesian updates priors in response to news. The evidence from the Clinton expansion should have reduced one's estimates of the adverse effects of tax hikes, and yesterday's evidence from the Bush expansion should move one's view in the opposite direction. But remember that we have a lot of data, and each year adds only one more data point.
<< Home