RDP 2005-09: The US Current Account Deficit: A Re-Examination of the Role of Private Saving 1. Introduction

The large United States (US) current account deficits of recent years have been the subject of an enormous amount of study in academia, among government and central bank economists, in business economics reports, and in the press.^[1] Many different explanations of the cause of the deficit have been offered, and to varying degrees we believe that all may have played a role in the evolution of the deficit: low private saving in the US generated by financial innovations or asset price inflation; large public-sector budget deficits; a ‘glut’ of savings in the rest of the world driven by demographic factors in the rich countries and capital market imperfections in the emerging markets; and, perhaps even misalignment of the nominal exchange rate leading to excessively cheap import prices for the US in the short run.

In this paper we explore the role of one other factor that also has been mentioned prominently: private saving in the US is low because income growth is expected to be strong in the US. We will discuss briefly some of the other factors that deserve consideration, but here our focus is on the question of how much of the current account deficit can be understood in the context of a model of optimal saving. We rework the standard neoclassical two-country model to show how a country will be a net borrower when its future share of world GDP (net of investment and government spending) is expected to increase above its current share.

Figure 1 is the starting point of our analysis. It shows what is perhaps one of the most striking economic developments of the last 25 years: the dramatic increase in US output relative to the rest of the high-income world. Specifically, the figure plots US GDP less government spending and investment as a share of the sum of G7 GDP less government spending and investment. We net out government spending and investment because our theory suggests that consumption spending ultimately depends on income that is available for household consumption, but a plot of unadjusted GDP shares looks very similar. The figure also eliminates the influence of exchange rate swings by comparing real GDP growth relative to GDP shares in 1990, in a way that will be explained in greater detail later.

The striking thing about this figure is that since some time in the 1980s, the US share appears as if it is on an upward linear trend. Its share has increased from a trough in 1982 of around 39.5 per cent to its level in 2004 of approximately 45.3 per cent. Prior to 1982, it appears that the US share fluctuated around a level of roughly 40.5 per cent.

GDP shares capture two factors that are typically considered separately in the neoclassical approach to the current account. First, if the country's income is expected to rise, it may borrow now and run a current account deficit. Second, if the world interest rate is low, the country's incentive may be to borrow more now.

One factor that may lead to lower interest rates is high saving in the rest of the world, which might in turn be generated by poor growth prospects in those countries. Typically, the two-country model (in which the world interest rate is determined endogenously) has been used when studying a ‘large’ country, such as the US.

Our models recast the neoclassical model to show how we can (under certain assumptions) express the optimal consumption path as a function of the current and expected future discounted sum of the country's share of world output (adjusted for government spending and investment). This representation is useful because it expresses the economic forces in a transparent and intuitive way; it also gives us a way of testing the model that does not require modelling of world real interest rates. The effects of real interest rates are implicitly captured by the ‘shares’ model. For example, if the rest of the world's income growth is expected to be slow, in the standard representation, their high saving will lower interest rates and these low rates will stimulate consumption at home. In the ‘shares’ approach, we see directly that home borrowing is encouraged by low growth in the rest of the world, because that low growth will lead to higher future output shares for the home country.

Another nice feature about the ‘shares’ representation of the two-country model is that it applies for countries of any size. There is no difference in the representation for countries that are too small to affect world real interest rates, and large countries that can.

Our research ultimately is motivated by the question of whether the US current account is ‘sustainable’. The definition of ‘sustainability’ is a bit slippery and differs from study to study, and we do not offer a definition here. We begin from the observation that while the US is building up debt obligations which may imply high debt service obligations in the future, it will almost certainly also be the case that US GDP will be much higher in the future. We need some yardstick to measure how burdensome those debt service obligations will be in the future. Will the US be worse off from its high current spending at the expense of having to devote some of future output to servicing its external debt?

It seems like the natural way to answer this question is in the context of an optimising model. This allows us to compare the trade-off between current and future consumption. The way we approach the question is to see whether the high level of US spending currently is compatible with an optimal path of borrowing. In particular, what assumptions about expected future growth of the US's share of world output could possibly justify its current account deficit?

It is obvious that the US cannot run primary deficits forever – although even here, a caveat is necessary. Figure 2 plots the decline in the US external net asset position since 1982. Although there are serious and well-known measurement issues, the chart reproduces the conventional wisdom that the US has emerged as a large debtor. But Figure 3 shows the US net investment income as a share of GDP. That also has declined since 1982, which is consonant with the decline in the US asset position. But the decline in investment income has not been nearly as sharp as the decline in the net asset position, and the chart shows that US net investment income in 2004 was still positive.

The usual explanation for this disconnect – that the US is a net debtor, but its net investment income is positive – is that the US earns more from its foreign investments than foreigners earn from their investments in the US. This in turn is explained by the claim that the US external portfolio is dominated by direct investment and equities, while foreigners' portfolio of US assets is dominated by holdings of Treasury securities. Whatever the reason, if the US can always earn more on its foreign investments than foreigners earn on their investments in the US, then indeed even a primary current account deficit could be sustained forever. The US could maintain a position of zero net investment income in this situation if the ratio of foreign assets that it acquires to US assets acquired by foreigners were to equal the rate of return on these US assets relative to the rate of return on foreign assets acquired by the US. If the latter ratio is less than one, then the former ratio can also be less than one, which implies a primary current account deficit.

We will abstract from this possibility in our discussions. Primarily we will take an approach that sets the current US net investment income to zero, and assumes that on future transactions the returns on US borrowing and lending are equal. We acknowledge that in some circumstances this may underestimate the future debt service obligations of the US if markets have properly priced long-term asset positions. But under the conventional wisdom, this assumption actually is conservative. That is, we try to uncover what assumptions about US growth relative to the rest of the world would justify the high US current account deficit. High US growth combined with low growth in the rest of the world is required. The most likely deviation from this situation is one in which total world GDP growth turns out to be lower than the markets currently expect. But if that were to happen, future short-term interest rates would be lower than those implied by our model. The way in which the US could get into trouble with its current debt obligations (which we are, in essence, ignoring on the grounds that its current net investment returns are not negative) is if interest rates are higher when the current debt is rolled over at maturity. But that would require future interest rates to be higher than the rates that are implicit in our model, not lower.

Returning to Figure 1, it is obvious that it is difficult to forecast the US's future share of adjusted G7 GDP. As we have noted, it appears that the US share has been on an upward linear trend since some time in the 1980s. But mathematically it is impossible for the share to continue upward on a linear trend. If nothing else, the trend must stop when the US share reaches one, but obviously it will stop long before then. But it is hard to read the tea leaves from Figure 1. Will the US share continue to rise, and then level out at a higher share? If so, what will that share be? Or will the trend reverse, and the US share return to somewhere around 40.5 per cent? We consider different scenarios based on alternative econometric approaches, but ultimately there is no answer in which we can be very confident. Indeed, that is the core conclusion we reach: there is a great deal of uncertainty about the future path of the US share of world output. There are certainly plausible scenarios in which modest continued growth in the US share could justify high US external borrowing today.

Our study is an exercise in generating intuition about the magnitude of the effect of growth in future output shares, rather than a definitive study of the US current account. As we noted from the start, there are many factors that contribute to the deficit, and we consider only one of them. We perform a delicate balancing act in ignoring many of these factors.

One of the first questions that might be raised is why Figure 1 considers the US share of adjusted G7 GDP? If we included China, India, and the east Asian tigers, for example, would not the figure show a falling share? Certainly, the increase in the US share would not appear as dramatic if it were to include these countries, but we leave them out for good reason. If we were to treat these countries symmetrically with the US, the model implies that they should be running large current account deficits. Indeed, it is a puzzle that this set of countries, which must be expecting high future income growth, have current account surpluses in a world of low real interest rates. Bernanke (2005) offers the ‘saving glut’ explanation for their behaviour. He argues that this set of countries saves a lot now because they do not have full access to world capital markets. The aftermath of the 1997 crisis has led outside lenders to be cautious, and required the emerging-market countries as a whole to finance investment internally. In addition, some countries (such as China) may have been building up a ‘war chest’ as protection against the event of a future crisis, even if private capital markets have been willing to lend. And, Bernanke notes, the oil-producing countries have had high saving rates recently that have been associated with the recent increase in the price of oil.

From the perspective of the US, the relevant point is that these countries are not borrowers on international capital markets. Since they are running current account surpluses, they are contributing to the pool of world saving. This increases the incentive for any other country to borrow today. It is a conservative assumption on our part to ignore this pool of saving in assessing what sort of income growth in the US could justify its current low level of saving.

Of course our neoclassical model could directly incorporate an explanation for the glut of saving in the emerging markets, and the extended model could be tested. We do not take this approach for three reasons. First, we do not believe the exercise of determining optimal behaviour for the US necessarily depends on the explanation for the saving glut. Whatever the reason, it will lead to lower interest rates than those implied by our model, and imply an even greater incentive for the US to borrow. Second, we are intent on examining a simple and intuitive version of the model. Third, there is a sense in which the ‘shares’ model can incorporate a high desired saving level in the rest of the world in a straightforward way. We can simply make the ad hoc assumption that the rest of the world puts a higher weight on future utility of consumption than does the US. We show that in this context, the optimal deficit for the US is still a weighted average of current and future expected output shares. But the model implies, naturally, that the US consumption level will be higher the higher the discount factor (the lower the discount rate) in the rest of the world.

We also do not directly consider demographic explanations for different saving rates around the world. There is a sense, however, in which our model does capture these effects. We model behaviour in each country as determined by an infinitely-lived representative household. We do not even consider changes in the number of people in each household. Maintaining this fiction, one might suppose that in some households, future production will fall (or grow more slowly) as members of the household retire and leave the workforce. In other households, income will grow as the fraction of working members of the households increases and their output rises. In the ‘young’ countries, output is expected to increase more rapidly than in the ‘old’ countries. From this vantage, demographics can be seen as an underlying determinant of income growth, and therefore a driving variable in the determination of future output shares. For example, the OECD recently has predicted very low GDP growth in the eurozone economies in the decade of the 2020s.^[2] The primary reason for this low growth projection is demographic – the OECD forecasts a shrinking workforce in these economies.

One of the most controversial issues that we are sidestepping is the role of the US government budget deficits in contributing to the current account deficit. Figure 1 nets out government spending. To the extent that budget deficits are caused by increased public expenditure, we have already factored them into our analysis. That is, higher public expenditure implies a lower share of GDP left over for private consumption. So we are only trying to explain that part of the US current account deficit that cannot be explained by higher government spending.

On the other hand, our model is based on an infinitely-lived representative household, and so the timing of taxes does not matter for consumption decisions. That is, Ricardian equivalence holds. In practice, however, it may be that a reduction in taxes does not lead to a dollar for dollar increase in private saving. The empirical evidence is ambiguous on this point, but our feeling is that Ricardian equivalence does not hold. So when we ask whether the saving behaviour of the US could be consistent with an optimising framework, we are to some extent assuming that the government is using its taxing powers to determine optimal saving. That is, it lowers taxes in those cases in which it believes that the outcome of decisions by private households and businesses leads to too much saving. Put another way, the question is whether the tax policy decisions of the government are irresponsible, or whether they could be consistent with a responsible government that has higher expectations of growth, or possibly a lower discount factor (higher discount rate), than the public. In particular, we want to judge whether the expected growth or the discount factor that is needed to reconcile the US current account with plausible estimates of its future income share is ‘responsible’ or ‘irresponsible’.

We examine the current account question entirely from the perspective of a neoclassical model. We take this approach because we view the current account deficits of the US, and the eventual adjustment of its primary deficit, as a long-run phenomenon. This does not mean that short-run considerations are unimportant. They certainly might matter for the short run, and Keynesians believe that short-run considerations matter for policy. We do not believe, however, that short-run fluctuations in exchange rates or income will account for much of the long-run adjustment process, so we ignore such factors in order to simplify the analysis.

In Section 2 below, we lay out the basic ‘shares’ model and consider a couple of simple generalisations. Section 3 takes up a related issue – does the adjustment of the US current account require a large real depreciation? We show that if indeed the deficit can be explained by higher future income shares, then the size of the required real depreciation may be quite small indeed. Section 4 then provides some simple numerical exercises with the model to assess the impact of the role of growth in future output shares on the current account. Readers not interested in the technical details may want to skip Sections 2 and 3, and head straight for the bottom line in Section 4. Conclusions are drawn in Section 5.

Footnotes

See, for example, Backus et al (2005), Bernanke (2005), Blanchard, Giavazzi and Sa (2005), Chinn (2005), Clarida, Goretti and Taylor (2005), Edwards (2005), Gourinchas and Rey (2005), Kouparitsas (2005), Kraay and Ventura (2005), Obstfeld and Rogoff (2000, 2004, 2005), and Truman (2005). The US current account deficit stood at almost 6 per cent of GDP in 2004, the largest deficit on record over the last 40 years. [1]

See, for example, Martins et al (2005). [2]