Tuesday, January 24, 2006

Fed Economist: Current Account Deficit near Optimal Levels

by Calculated Risk on 1/24/2006 08:54:00 PM

Federal Reserve economist John Rogers (Chief, Trade and Financial Studies Section) and University of Wisconsin Professor Charles Engel, in a new paper "The U.S. Current Account Deficit and the Expected Share of World Output", Journal of Monetary Economics suggest the US Current Account Deficit may be near optimal levels.

From their conclusions:

We have asked whether the U.S. current account deficit could be consistent with expectations that the U.S. share of world GDP will increase. Under assumptions about the growth in the net GDP share that are not wildly implausible, the level of the deficit can be consistent with optimal saving behavior. But, in making this assessment, we emphasize that there are many difficult issues to deal with, and the conclusion is sensitive to how one handles these questions.

First, our findings are sensitive to how we treat two problems: the high saving rate in East Asian emerging economies, and the "exorbitant privilege" (the term used by Gourinchas and Rey (2005)) that allows the U.S. to receive a much higher return on its foreign investments than foreigners earn on their U.S. investments.

On the first point, most forecasters predict that the emerging market's share of world GDP will be increasing over time. Our empirical work does not include these countries, and if it did, the forecast path of the U.S. share of world GDP would not be as rosy. But, according to themodel, these countries ought to be borrowers in international capital markets. They are not -- they are large net lenders. It is puzzling that they are net lenders. Bernanke (2004) refers to this as a "savings glut", and hypothesizes that these countries are in essence building up a nest egg in order to protect them against a possible future international financial crisis such as the one that beset East Asia in 1997-1998.

We are not sure how to handle this in our model. It may be that these countries will continue to be high savers, in which case their saving will hold down world interest rates and the U.S. deficits will be more justifiable. On the other hand, their saving rate may fall and real interest rates may rise, which works toward the U.S. optimally having a smaller deficit.

We make the "heroic" assumption in our work that these countries are not contributing to net world saving at all. On the one hand, this is a conservative assumption (if one is trying to explain the large U.S. deficits), because the countries are in fact large net savers. On the other hand, if their net saving is reversed, the assumption is too optimistic.

It does seem like markets favor the position that these countries will maintain their positions as large savers, because long term real interest rates are very low. However, much of the recent scholarly and policy-oriented research on the U.S. current account deficit has taken the position that the markets may not be correctly foreseeing events.

Finally, it is possible that the saving rate is high in East Asian countries because of demographic factors. It has been noted that because of the one-child policy, the ratio of old to young is increasing rapidly in China. There are other countries for which demographic factors may be very important as well, and this deserves further study.

We take a similar neutral position on the exorbitant privilege. One possibility is that the U.S. will continue to receive higher returns on its foreign investments than it pays out on its foreign borrowing. On the other hand, that privilege may disappear, and worse, it may disappear not only for future borrowing but also for our outstanding debt when it is refinanced. Our work takes a somewhat neutral position by assuming future borrowing and lending takes place at the same rate of return, but that there is no additional burden to be encountered from refinancing existing debt at less favorable rates of return.

There really are a variety of scenarios that could play out. As Gourinchas and Rey (2005) demonstrate, it is not only that the return on U.S. assets within each asset class is lower than on foreign assets (implying the market views U.S. assets as less risky), but also that the mix of U.S. investments abroad favors riskier classes of assets. It is possible that the U.S. net return will fall in the future both because the risk premium on U.S. assets rises (as in Edwards (2005) or Blanchard, Giavazzi and Sa (2004)), and because foreigners shift toward investing in more risky U.S. assets. But, again, it is notable that markets do not reflect any increasing riskiness of U.S. assets.

With these major caveats in mind, we find that the size of the U.S. current account deficit may be justifiable if markets expect further growth in the U.S. share of advanced-country GDP. The growth that is needed does not appear to be implausible.
But, what the model cannot explain is why the U.S. current account deficit continues to grow. If households expect the U.S. share of world GDP to grow, they should frontload consumption. The deficits should appear immediately, not gradually.

We have allowed in our Markov-switching model for the possibility that there was a shift in regime that U.S. households only gradually learned about. But that turned out not to be able to explain the rising U.S. current account deficits. However, our simulations and estimation assumed that households understood that if a regime shift took place, the U.S. share of world GDP in the long term would be much higher than it was in the early 1980s. In practice, it may be that markets only gradually learned the U.S. long-term share. Examination of the model when there is only gradual learning about the parameters of the model will be left for future work. It is possible that because U.S. households only gradually came to the realization that their share of advanced country GDP was going to be much higher in the long run, they only gradually increased their borrowing on world markets.

This possibility is supported by our examination of the consensus long-term forecasts of U.S. GDP relative to G-7 GDP since 1993. These forecasts have consistently underestimated U.S. GDP growth relative to other countries, by wide margins. The current forecasts for the future, however, show that the markets expect a large increase in the U.S. share of GDP – almost precisely the amount that we calculate would make the current level of the deficit optimal.

There are at least two other possible explanations to explain this gradual emergence of the current account deficit. One possibility is that it takes time for consumption to adjust. This could be modeled either with adjustment costs, or, as is popular in many calibrated macro models, with habit persistence in consumption.
Another possibility is that there has been a steady relaxation of credit constraints for many U.S. households, as well as increased access to U.S. capital markets for foreign lenders. The relaxation of credit constraints was one of the possibilities that Parker (1999) explored in his study of the decline in U.S. saving. He found that it could explain at most 30% of the increase in consumption from 1959 to 1998.

The starting point of Parker's back-of-the-envelope calculation is the observation that the consumption boom is the equivalent of three-quarters of one year's GDP in present value terms. The rise in debt, as measured by the difference in ratios of household total assets to income and net worth to income, was about 20 percent over the period. Therefore, debt can explain at most .20/.75 < 30 percent of the increase in consumption. Since the time Parker wrote his paper, debt has continued to rise, by another 25% through 2005Q2 when the ratio of total assets to income exceeded the ratio of net worth to income by 1.24.

Of course, the other obvious candidate for the increasing U.S. current account deficit is through the effect of U.S government budget deficits. It is useful to note that what we are really talking about is the effects of tax cuts. In the first place, government spending as a share of GDP has not changed dramatically, so could not account for the large current account deficit. Moreover, our analysis allows for the effects of increases in government spending. An increase in current spending above the long-run spending levels would lower the U.S. share of GDP net of government spending and investment relative to future shares, thus inducing a greater consumption to net GDP ratio.

But our model assumes that the timing of taxes does not matter for household consumption -- that Ricardian equivalence holds. Obviously that might not be correct. Recent empirical studies do not show much support for Ricardian equivalence, though the point is debated.8 We note that to the extent that credit constraints have been relaxed in recent years, Ricardian equivalence becomes a more credible possibility. It may be that in more recent years, lower taxes do not boost consumption as much, and instead allow households to pay off some of their credit card debt or prepay some of their mortgage. It may be interesting to pursue empirically the hypothesis that the effects on national saving of tax varies change with the degree of credit constraints in the economy.

Another argument that needs to be explored is the distributional effects of the recent tax cuts. It has been argued that the tax cuts were less stimulative than previous cuts because they accrued mostly to wealthy individuals, who simply saved the additional after-tax income. (That is, the rich act more like Ricardian consumers.) But if that is the case, then it is more difficult to make the case that the tax cuts are responsible for the decline in U.S. national saving.

Finally, we cannot reach firm conclusions about the future path of U.S. real exchange rates. We have calibrated a model that is essentially identical to the one examined by Obstfeld and Rogoff (2004), but one in which the consumption path is determined endogenously as a function of current and expected discounted real income in each country. We found that under one set of baseline assumptions, there should not be much change in the equilibrium real exchange rate as the U.S. current account adjusts. Our model assumes the U.S. will experience higher growth in productivity in both traded and non-traded sectors, and that there is factor mobility between the traded and non-traded sector. On the one hand, if traded/non-traded productivity growth in the U.S. is slightly higher than in the rest of the world, the price of non-traded goods will rise in the U.S. from the Balassa-Samuelson effect. On the other hand, the U.S. terms of trade should fall as the supply of its exports increases. If there is home bias in consumption of tradables, that would work toward causing a U.S. real depreciation. In our baseline calibration, these two effects approximately cancel.

But as we have noted, the conclusions about the real exchange rate depend on assumptions about parameters of the model. Particularly, if the elasticity of substitution between imports and exports in consumption is much lower than our baseline simulation assumed, the U.S. could experience a substantial real depreciation over the next 25 years.

The basic message of our paper is that there are many aspects of the current account adjustment that are just not possible to predict. Under some scenarios that we do not regard as entirely unreasonable, we find that the U.S. current account deficit can be explained as the equilibrium outcome of optimal consumption decisions. But some of our modeling simplifications and assumptions might be wrong in important ways, and so it may turn out, as many have been warning, that the deficits have put the U.S. on the path to ruin.