Menzie Chinn on the Budget and Current Account Deficits

Following up on this post on the "twin deficits", here's Menzie Chinn of econbrowser:

The Debate over the Impact of the Budget Deficit on the Current Account Deficit, by Menzie Chinn: What does the empirical literature say? In a recent Washington Post article, Treasury Secretary Snow was quoted as follows:

Speaking to the International Monetary Fund, Snow repeated the U.S. position that ratcheting down big imbalances in trade and capital flows "cannot be anything other than a shared responsibility" because no one country caused them.

Interestingly, this assertion is much less nuanced than the Treasury Occasional Paper on the subject... That study cited point estimates for the response of the current account to GDP ratio to a a budget balance to GDP ratio ranging from 0.1 (Erceg et al.) to 0.44 (IMF).

Since the study also cited the 0.13 coefficient Chinn and Prasad (2003), it seems only natural to discuss what Hiro Ito and I have discovered in our updating of these estimates ... in a forthcoming revision of our paper here. The updated point estimate is 0.20 -- for a pooled panel cross section regression... Our fixed effects regression estimates for different sets of countries is presented below:

Estimated coefficients on the government budget balance (% of GDP) in the fixed effects regressions. Source: Chinn and Ito, forthcoming revision of Current Account Balances, Financial Development and Institutions: Assaying the World "Savings Glut"

The critical estimate for this discussion is the industrial country point estimate of 0.40. The estimate is significantly different from zero at the 10% ... level. Why is this point estimate so different from the pooled panel-time series estimate? The biggest impact likely arises from allowing each individual country to have a different constant, something ruled out in the pooled OLS procedure...

Is the pooled estimator appropriate? It depends upon the question one is asking. If one wants to know ... the average response for an industrial country's current account balance to changes in the budget balance, assuming a high degree of homogeneity across the countries, then this is the correct number. If, on the other hand, one is interested primarily in how the United States behaves, ... then the fixed effects estimate is more relevant. ... I would assert the empirical evidence -- as opposed to calibration-based results -- is on the side of effects greater than 0.2. ...

As I have noted elsewhere, when the coefficient is 0.4 (remarkably close to the OECD's Interlink model, and similar to that in other estimated macroeconometric models), 0.4 means that a 6% percentage point swing in the budget balance -- like the one that took place after 2001 -- would result in a 2.4 percentage point swing in the current account balance. Not enough to eliminate the deficit, but certainly enough to make a substantial impact.

Update: Brad DeLong adds his comments and notes:

Menzie Chinn critiques the Bush administration's latest Snow job ... Menzie, however, is kind, relative to what the Washington Post did to John Snow. It got medieval on him, by putting this headline atop his argument: "Don't Blame Just Us."