Tuesday, October 15, 2013

Deficits and Climate Change

Larry Summers makes the link and it is not the one you might expect; instead he is looking at the innate uncertainty in policy projections:

These figures lie well within any reasonable confidence interval for deficit forecasts. The most recent comprehensive CBO evaluation found that, leaving aside any errors due to policy changes, the expected error in projections out only five years is 3.5 per cent of GDP. Put another way, given the magnitude of forecast uncertainties there is a chance of close to 40 per cent that with no new policy actions the ratio of debt-to-GDP will decline over 25 or 75 years.

Of course, debt problems could also be much worse than is now forecast.

But in most areas policy makers avoid taking strong actions unless there is statistically compelling evidence to support them. Few would favor action to curb greenhouse gas emissions without evidence establishing that substantial climate change is overwhelmingly likely. Yet it is conventional wisdom that urgent action must be taken to cut the deficit, even as prevailing short-run deficit forecasts suggest no problems and long-run forecasts are within margins of error.
When the variance in the projection models are explicitly discussed, it makes the whole matter seem more prosaic.  After all, if we were frequentists we would reject the hypothesis of an increase in debt to GDP deficit at 10 years as not statistically significant (versus the null hypothesis that this change = 0). 

Much of this uncertainty is on the growth side of the equation, which may actually be the easiest policy lever to move in the short term.  It seems to me, as a student of history, that it is rarely the case that a country has been ruined by a burst of prosperity, growth, and development of human capital.

But the second part of Dr. Summers' argument is really the best -- Climate change is at least as likely as an increase in the debt to GDP ratio in the next 75 years.  I think the balance is much higher and the possible side effects of climate change are pretty dire for some really key places in America (think Miami).  Yet it is not the flashpoint for this discussion and that, I believe, is what gives the game away. 

Whatever people are fighting about, it is not the long term state of the debt.  Sure, it would be nice to reduce it but obvious tactics (raising taxes) are completely off of the table and there is a constant return to trying to cut specific government programs.  This has all of the hallmarks of a policy duel and not a real worry about the long term future.

But one way or the other, the discussion has been way too focused on the point estimates of the statistical models and not the confidence intervals.  Epidemiologists have already learned, to their regret, how dangerous that can be. 

P.S. And no, I am not suggesting a p-value based approach either, but rather measures of certainty that are explainable in ordinary language and make clear that these are projections, so that the level of sacrifice to improve the point estimate can be properly evaluated. 

No comments:

Post a Comment