External Debts and Current Account Adjustments

External Debts and Current Account Adjustments Levent Bulut November 21, 2008 Abstract This paper empirically investigates the effect of net external debt holdings on the size of medium-term current account balances. It utilizes an approach where net external debt holdings behave like a shadow interest rate in affecting the current account imbalances. This paper has four major findings: First, in a simple accounting framework, net external debt holdings have a significant dampening effect on mediumterm current account imbalances: an increase in the net external debt holdings by 10% of GDP improves medium-term current account balances by 1% of GDP. Second, net external debt holdings affect current account imbalances through their effect on domestic investment and government expenditures. Private consumption, on the other hand, isn t affected by net external debt holdings. Third, across the country groups, it is found that OECD countries differ from developing countries in current account adjustments as government expenditures deteriorate more in developing countries than in OECD countries across the sample period. And finally, net external debt holdings temper current account imbalances more during 1980s than 1990s. Keywords: External Debt Holding Costs; External Debts; Feldstein-Horioka Puzzle; Current Account Adjustments; Capital Mobility JEL Classification Numbers: E21, F32, F41. I am especially grateful to Bent E. Sørensen for his very useful comments. Thanks also to David Papell, Sebnem Kalemli- Ozcan, and Dietrich Vollrath, as well as seminar participants at the 76th Southern Economic Conference and University of Houston Workshop Series, for their comments. Visiting Assistant Professor, Emory University, Department of Economics, Atlanta, 30322-2240 GA, USA Telephone: 1-404-712-8688, FAX: 1-404-727-4639, e-mail: lbulut@emory.edu.

1 INTRODUCTION Domestic saving and investment are highly correlated both across time and within countries: in years, when and countries where, saving is high, so is investment. Feldstein & Horioka (1980 (henceforth FH interpret this correlation as evidence of low international capital mobility which is known as the Feldstein- Horioka puzzle (FH puzzle. This is an anomaly because under the assumption that capital is perfectly mobile between countries, domestic investments do not depend strongly on domestic savings. Rather than saving being contained within the country of origin, individual countries savings should be added to funds in the world capital market and distributed amongst countries according to the most favorable returns. The FH puzzle has been one of the most robust and intractable puzzles in international finance as the anomaly hasn t normalized with increased integration of world financial markets. The FH puzzle has been the subject of various attempts to reconcile the stylized facts of highly correlated investment-saving ratios with the increasing degree of capital mobility. 1 There remains however an ignored and unexplored side of the FH puzzle which has a direct bearing on the degree of capital mobility. Given the fact that a country s current account is the difference between national saving and domestic investment, as long as investment and saving ratios exhibit a strong correlation both across time and countries, the current account to GDP ratio will be fairly small in size relative to investment and saving ratios. Thus, it can be concluded that the observed dynamics of current account balances are not compatible with the increasing degree of capital mobility as countries couldn t take advantage of the globalized markets to finance their capital needs with foreign savings. Blanchard & Giavazzi (2002 theoretically document the channels through which financial integration affects current account balances. The cost of borrowing is expected to decrease with increasing financial integration: poor but developing countries with a lower level of capital, higher marginal productivity and high growth prospects are expected to run current account deficits by increasing their external borrowing to finance domestic investment. Accordingly, advanced economies with a higher level of capital, lower marginal productivity, and a relatively stable level of investment are expected to run current account surpluses. The dispersion of the current account positions, therefore, is expected to increase with the increasing integration of financial markets. Data show that for the last 30 years, the world average current account balance in absolute value is less than 3% of world GDP and this ratio is similar for both developing and OECD countries. 1 Sinn (1992 investigates the time-dependent relation between investment and saving, Obstfeld (1986 considers the country specific saving-investment relation, and Cardia (1991 shows that persistent shocks may generate saving and investment correlation even under the perfect capital mobility assumption. Summers (1989 and Bayoumi (1990 try to explain the anomaly by means of a current account targeting motive, while on the other hand, Harberger (1980 and Murphy (1984 argue that FH s results reflect a large country bias rather than low capital mobility. Coakley, Kulasi & Smith (1996 interpret the long-run strong correlation as a result of the external solvency constraint requiring the current account to be stationary in order for the external debts to be bounded. Obstfeld & Rogoff (2000 explain the puzzle with transport costs in goods trade. 1

In theory, it is expected that countries can achieve some welfare improvements by borrowing from or lending to the rest of the world. et it is a puzzle as to why countries, especially OECD countries, are not taking advantage of this and are running small-size current account balances on average even though the world financial markets have become more integrated over the last two decades. This paper suggests that net external debt holdings have a role in this stylized fact of small-size current account balances. Specifically, an approach is utilized where net external debt holdings behave like a shadow interest rate in affecting the behavior of the current account balances in the medium run. Shadow interest rate occurs when there is a discrepancy between the ongoing market interest rate in the economy and the interest rate relevant for decision making. It is suggested that running high current account deficits increases the cost of external borrowing for the domestic residents, thus debtor countries face a positive shadow interest rate on loans. On the other hand, running high current account surpluses decreases the benefit from lending abroad, thus creditor countries face a negative shadow interest rate on loans. As a result, the relevant real interest rate for decision making in the debtor countries is higher than the world real interest rate, while the real return to those lending abroad in creditor countries is below the world real interest rate. According to this construction, it will be harder for the debtor countries to run current account deficits and therefore there are current account adjustments in the medium-run. In other words, as countries keep borrowing from abroad, the accumulated external debt holdings start forcing the debtor countries to adjust or improve their current account imbalances. By the same logic, the creditor countries will be reluctant to run current account surpluses and in the medium run, they are also expected to adjust their current account imbalances. This method of current account adjustment mechanism working through the net external debt holdings is the contribution of this paper to the literature on the FH puzzle. In this paper, a simple model is introduced in which there are convex external debt holding costs through which external debts dampen the widening of the current account. This is a better structure by which to comprehend the role of the shadow interest rate on the dynamic of the current account imbalances. By assuming that there are some hidden costs in holding external debts, the model allows us to capture the shadow interest rate on loans as an unobserved part of the interest rate that a domestic borrower has to pay. The way external debts create a shadow interest rate on loans can be driven by some internal and external factors: internal factors are related to the capital market structure of the debtor country. In the domestic economy, the existence of capital market imperfections and underdeveloped financial systems can cause intermediation of external credit costly to the domestic residents due to the weak banking system in most of the developing and emerging economies. Also, high external debt ratio can force the government to reduce external borrowing to keep country creditworthiness at a reasonable level. External factors, on the other hand, summarize the behavior of the creditors in financing the debtor country. For instance, 2

creditor countries will be reluctant to extend new loans or roll-over existing loans because of the increasing burden of external borrowing on the debtor country s economy through the exchange rate channel. In the exchange rate channel, increase in external debts increases the need for foreign currency in order to service the debts. Herrero, Berganza & Chang (2004 find a positive relationship between a country s risk premium and increases in the external debt service of the country. In addition, monitoring and repudiation costs may cause creditors to be reluctant to keep financing deficit countries. When the creditor allows for the funds to go to the debtor country, then the performance of the borrower is to be monitored in terms of default risk. This creates a cost that is relatively higher than the monitoring of any domestic borrower. Repudiation costs occur when the debtor country either fails to repay its debt obligations or prefers not to repay when external debt level reaches a threshold level. Therefore, creditors show their unwillingness to finance highly debtor countries by asking extra premiums. There may be some other observed or unobserved internal and external factors which may lead to the external debt holding costs for which convex debt holding costs are assumed to capture them all. A convex debt holding costs assumption is one of several ways to modify an incomplete small open economy model in order to have stationary equilibrium-policy function at the steady state. 2 The convex debt holding costs assumption was offered first by Turnovsky (1985. 3 It was later applied in Heathcote & Perri (2002, Schmitt-Grohe & Uribe (2003, Kim, Kim & Kollmann (2005, and Neumeyer & Perri (2005. In fact, what has been introduced by the recent literature is not directly external debt holding costs but portfolio adjustment costs. It is assumed that there is a long run steady-state level of net external debt holdings, and when a country deviates from this steady state level, a cost is incurred in adjusting the existing portfolio composition. Thus, an a priori assumption of non-zero long run level of net external debt holdings changes the structure of the cost in the following way: any amount of external debt holdings away from the long-run steady state level is punished in proportion to the increased size of the deviation from the long-run level. If a country is very close to its long-run level, the costs associated with external debt holdings (e.g. monitoring costs, repudiation costs, transaction costs, etc. are muted. But the problem is to identify the long run steady state level of external debt holdings. To overcome this, in this paper a zero level of net external debt holdings is assumed at the steady state, and it is argued that no matter what the level of external debt holdings, a cost is associated with the holding of external debt. 2 In incomplete small open economy models, since there is no complete risk sharing, macroeconomic variability affects the mean net foreign asset position which results in a non-stationary equilibrium relation as it depends on the initial level of net foreign asset positions. In order to overcome this technical difficulty, convex debt holding costs assumption is added to the model where there is a cost to adjust the existing portfolio composition of external debt holdings from its long run steady state level. See Schmitt-Grohe & Uribe (2003 for a recent survey. 3 Turnovsky (1985 rationalize the convex debt holding costs by addressing the fact that there is imperfect substitutability between domestic and foreign bonds within a certainty equivalent structure, and there is a cost differential in the acquisition of foreign bonds and domestic bonds. Information and transaction cost along with brokerage fees for obtaining foreign exchange to purchase foreign assets or bonds are some possible sources of that cost differential. 3

The remainder of this paper is organized as follows: section (2 provides a theoretical background for external debt holdings as a shadow interest rate for the borrower country. Section (3 tests the empirical validity of the external debt holding costs phenomena within a simple accounting framework: in the first stage, the effect of net external debt holdings on medium-run domestic investment, private consumption, and government expenditure are separately estimated. In the second stage, the overall effect on the current account balances is obtained. Section (5 provides concluding remarks. 2 THEORETICAL MODEL There is a small open economy populated by a large number of infinitely-lived households with preferences described by the following utility function: Max U t = β t s U(c t, (1 t=s where c t denotes consumption in period t and β denotes the subjective discount factor. The representative agent will get income from production and holdings of foreign bond, and she will invest her wealth net of consumption into domestic capital and/or foreign bond. 4 When she holds her wealth in foreign bonds, there will be convex external debt holding costs. For a debtor country, income from production (y t will constitute the only source of total current income. Expenditures on consumption goods (c t, on investment in gross domestic capital (i t, interest payments on external debt holdings (rb t 1, and payments due to convex external debt holding costs (b 2 t will be financed with total current income, and any excess ( ψ 2 expenditures over current period income will be financed from abroad by increasing the existing stock of external debt holdings. The following is the inter-period budget constraint of a representative agent who borrows internationally: b t b t 1 = c t + i t + ( ψ (b 2 t + rb t 1 y t. (2 2 There is an aggregate production function that is homogenous to degree one with given labor and productivity parameters. Production function, y t = AF (k t, represents constant returns to scale technology with the standard capital accumulation equation of k t+1 = k t + i t. 5 It is also assumed that goods are reversible such that one unit of consumption can be transformed into one unit of capital at no cost. After inserting the production function and capital accumulation equation to the budget constraint and solving it 4 The absence of state contingent assets makes the model have incomplete markets, such that the insurance against country specific shocks is incomplete. 5 It is assumed that there is no depreciation. 4

for consumption, the maximization problem becomes: Max U t = β t s U t=s [ b t + AF (k t (k t+1 k t (1 + rb t 1 ( ] ψ (b 2 t. (3 2 It is assumed that there is no uncertainty in the economy. The representative agent has two decision variables; the holdings of domestic physical capital (k t and next period holdings of external debt (b t. The first order conditions with respect to those variables are: [1 ψb t 1 ] U (c t 1 = β(1 + ru (c t, U (c t 1 = β [AF (k t + 1] U (c t. (4 The introduction of debt holding costs changes the traditional Euler equation for consumption by imposing a shadow interest rate in the domestic country. According to the Permanent Income Hypothesis (PIH, at the optimum, a temporary income shock wouldn t affect consumption; thus, consumption growth would be zero as households would always try to smooth consumption. In contrast, with the debt holding costs, consumption will be sensitive to net external debt holdings. As long as the country stays in the debtor position, marginal utility of current consumption would be higher than the marginal utility of future consumption. This would induce a reduction in consumption growth, because current consumption would be more expensive in terms of foregone future consumption and would induce substitution effect toward future consumption with elasticity ψ. On the other hand, if the country is in a creditor position, then the marginal utility of present consumption would be smaller than the marginal utility of future consumption which would lead to consumption growth in the future. A simple banking model can be constructed to explain the debt holding costs in external borrowing. 6 Assume that there is a small open economy with imperfectly established capital markets. The economy compensates for the shortage of savings by external borrowing. In addition, households and firms can only access bank financing where a domestic capital market possesses systems of atomistic banks in the market. 7 Furthermore, banks in this structure function as financial intermediaries between domestic residents and foreign residents. Banks help intertemporal smoothing by saving buffer stocks in good times and dissaving in recession periods. It is assumed that banks face some intermediation costs which is convex in the volume of intermediation. The convex intermediation cost is consistent with the current literature of U-shape 6 This is based on Uribe & ue (2006. 7 Bank financing is mostly relevant in developing countries where most of the intermediation is dominated by banks. The majority of private savings are in the form of bank deposits and the major source of finance for firms are in the form of bank loans. 5

operation costs for the banks in their operations. If ψ(b t represents a convex cost function, in a perfectly competitive environment, banks would take the borrowing and lending rates as given and try to maximize their profits by optimizing the amount of borrowing from abroad. In a debtor country, any atomistic bank would borrow b t from abroad at the fixed rate of r and lend it to the domestic residents at a given rate of r by facing the convex intermediation cost of ψ(b t. They will then maximize the following profit function, (1 + r [b t ψ(b t ] (1 + r b t, by optimally choosing b t, the amount of external borrowing. The first order condition gives the following interest rate differential between the domestic country and ( the rest of the world: (1 + r =, where r > r for any b t 0. Then, in debtor countries, due 1+r 1 ψ (b t to the intermediation costs, domestic residents would face a higher interest rate than the world interest rate. The amount of the interest rate differential is determined by the amount of external borrowing and the convexity of external debt holding costs. If the cost was a linear function of external debt holdings, the interest rate differential would be constant regardless of the level of external debt. But, as suggested in this paper, when the cost function is convex in net external debt, then the interest rate differential will be increasing in net external debt, hence the depressing effect of net external debts on the components of GDP arises. 2.1 Euler Equation for Consumption Assuming that there is a logarithmic utility function such that U(c t = ln(c t, the Euler equation for ( ( consumption would be in the following form: ct c t 1 = β(1+r 1 ψb t 1. After taking the natural logarithm of both sides and assuming that β(1 + r = 1, one can get: 8 ln(c t ln(c t 1 c t = ln[β(1 + r] ln[1 ψb t 1 ], (5 and by the linear approximation that ln(x = x 1 for any x > 0 but close to 1, then the Euler equation for a debtor country becomes: c t = ln[1 ψb t 1 ] = [1 ψb t 1 1] = ψb t 1. (6 In contrast to the standard approach, consumption growth rate is sensitive to the net external debt holdings when the economy is off the equilibrium. Even though this paper assumes a constant discount factor and 8 This is the standard assumption to prevent any perpetual growth in the economy. 6

ignores the consumption tilting motive for current account behavior, the introduction of external debt holding costs creates an additional source of consumption tilting channel other than that of impatience. Based on the first order conditions, creditor countries tend to tilt their consumption upward while debtor countries tend to tilt downward. Ghironi, Iscan & Rebucci (2005 find the same kind of behavior with different sets of assumptions in a 2 country-olg setting with a time-variant discount factor. They look for the effect of the long run level of non-zero net foreign asset (NFA on consumption dynamics, and find that more patient countries with positive long run NFA positions tilt their consumption upward after a worldwide productivity shock, whereas less patient countries display downward consumption tilting. Therefore, the NFA position would not be independent of the consumption or saving decision. Even though this approach has the same implication as this paper, the basic differences are that in this study a time-invariant discount factor is taken along with the assumption that the long run steady state level of NFA holdings would be zero. 2.2 Euler Equation for Investment What is the implication of the convex holding costs of external debt on the domestic economy s investment demand? As shown in relation to consumption, in the case of external debt holding costs, domestic residents face a shadow interest rate which is higher than the ongoing domestic market rate, and this would have some consequences for the aggregate investment demand in the economy. Based on the model introduced above, the first order condition for domestic physical capital of a debtor country is: ( 1 + r AF (k t + 1 =. (7 1 ψb t 1 The introduction of external debt holding costs to the model introduces a non-standard investment decision. In a standard small open economy with frictionless trade in goods and financial assets, domestic investors actual gross return, if they decide to invest in the domestic physical capital, would be (1 + r, where r stands for the exogenous world interest rate. When there is friction in the financial markets in the form of external debt holding costs, the optimality condition would be equation (7. After defining the marginal product of domestic physical capital in the following way: AF (k t = r d, where r d refers to real domestic interest rate, ( domestic residents would face the following gross return on domestic physical capital; (1 + r d 1+r = 1 ψb t 1 which this paper assumes as a shadow rate of return where the actual return is not independent of their net external debt holdings. Depending on the current position of net holdings of external debt, countries would hold domestic physical capital different from that which would be held if there was no external debt holding costs in the economy. This is the imposition of home bias to the model. In the absence of external debt holding costs, ψ = 0, then (1 + r d = (1 + r, the economy would decide investment behavior and holdings 7

of domestic physical capital based on the exogenous world interest rate. When economy is off- the steady state, the creditor and debtor countries would exhibit different preferences on domestic physical capital at the margin: a creditor country with negative external debts would face a shadow domestic interest rate that ( 1 is lower than the world interest rate such that 1 ψb t 1 < 1 for b t 1 < 0. From the optimality condition ( (1 + r d 1+r =, then, it is easy to see that (1 + r > (1 + r d. This straightforward result comes from 1 ψb t 1 the following partial derivation: [AF (k t ] ψ(1 + r = < 0, (8 b t 1 (1 + ψb t 1 2 where a creditor country increases its holdings of external debt and gives loans to the rest of the world, productivity of the domestic physical capital decreases. Because they want to diversify their country portfolio by investing abroad, this would crowd out domestic investment. Debtor countries, on the other hand, would exhibit foreign bias at the margin as they have a shadow interest rate in the domestic market which is ( 1 higher than the world interest rate; given that 1 ψb t 1 > 1, from equation (7, (1 + r < (1 + r d. These findings are crucial because when there are no external debt holding costs, ψ = 0, external debt holdings have nothing to do with the optimum investment decision; the only factors would be the world interest rate along with any productivity shocks. Then, external borrowing would only function to smooth consumption. When external debt holding costs are introduced to the model, however, external debt holdings matter. The creditor countries would temper their lending to the rest of the world as they increase their holdings of foreign bonds, due to the existence of external debt holding costs, the domestic economy would face a decrease in the marginal productivity of domestic capital. For the debtor country, the external debt holdings create a shadow interest rate that increases the cost of capital which decreases the investment incentives in the economy. The implications of the model are tested in the empirical part to see the role of external debt holdings on consumption and investment behaviors. 3 EMPIRICAL FINDINGS Empirical methodology is based on the investment-saving approach to the current account. Because this approach defines the current account balance as the difference between national saving and domestic investment, it allows determinants of saving and investment to explain the current account adjustments, specifically the average small size of current account balances. This paper, as a key contribution, looks at the role of net external debt holdings on the current account adjustments channeled through the GDP components. This paper does not aim to extrapolate all possible factors which explain the current account 8

adjustment, and attention is focused on the effect of net external debt holdings in the adjustment process of the current account. In contrast to the increasing integration of financial markets, the countries experience a small average size of the current account balances. Table (1 summarizes its evolution through 5 year time intervals across the OECD and developing countries as well as for all countries in the sample. The data for the current account balances are from the World Bank s World Development Indicators (WDI. Detailed information about the data is provided in the appendix. As shown in Panel A of Table (1, all countries in the sample exhibit a steady increase in the size of the current account balance (measured as the mean-absolute current account balances. The 5 year average (absolute current account to GDP ratio almost tripled from the 1975 1989 period to the 2000 2003 period. However, most of the increase was observed in the OECD countries as the developing countries exhibited a very small size increase. The average size of current account balances is around 3% of the world GDP which is far below the size one would expect to observe in free international capital mobility. In Panel B, cross country standard deviation of current account balances to GDP ratios are shown for 5 year time intervals. 9 Again, even though there is an increase in the standard deviation measurements, it is mostly driven by the OECD countries. Sections (3.1, (3.2 and (3.3 look for the role of external debt holdings on gross domestic investment, private and government expenditures, respectively. Finally, in section (3.4, the overall effect of net external debt holdings on the current account adjustment is analyzed within the context of the investment-saving approach to the current account. 3.1 External Debt Holdings and Investment For the empirical testing of the model, the partial adjustment extension of the neo-classical theory of investment is followed, Jorgenson (1971. According to this theory, optimal long run capital stock is related to the production technology and factor prices through a representative firm s maximization of present value of lifetime net revenue for a given production technology. The theory predicts the following long run co-integration relationship between optimal capital stock (K, level of output (, cost of capital (CC and elasticity of substitution (σ in the form of K = ( CC σ. It is assumed that there are capital adjustment costs in production. After adding them to the model in an ad hoc partial adjustment approach, these adjustment costs create sluggishness in capital accumulation. 10 9 The standard deviation is another way of examining the widening of the current account balances. The bigger the standard deviation, the higher the size of the current account imbalances. 10 Aggregate investment spending exhibits some persistency which is difficult to explain. It may be because of uncertainty of the source of demand shock, irreversibility in production or the cost of adjusting the existing stock of capital. 9

Firms will react to this change in the economic condition with some lags. Assuming an arbitrary fraction (λ is the adjustment parameter, the partial adjustment function will be as follows: k t = k t λ(k t k t 1. (9 The above equation can be used to define the current level of capital stock holdings in terms of lag values of optimal capital stock. Based on the co-integration relationship, it can be defined as a function of past values of output growth and cost of capital. The empirical tendency is to go beyond this general specification and define current capital stock holdings in terms of gross investment to GDP ratio. The reason for defining gross investment as a ratio of GDP is to reduce measurement error. From the capital accumulation formula of K t = I t + (1 δk t 1 where δ shows the constant depreciation rate of capital from one period to another, dividing the capital accumulation function by K t 1, one gets k t Kt Kt 1 K t 1 = It K t 1 δ. Gross investment can then be defined as a ratio of GDP based on the long run capital stock optimality condition. This paper adds external debt holdings to the existing neo-classical theory of investment as a deteriorating factor. As a result of external borrowing, external debt holdings raise the shadow interest rate in the domestic economy. 11 The country fixed effect panel regression equation is estimated as follows: ( I i,t ( F Debt = η i + β 1 + β 2 r i,t 3 + β 3 y i,t 3 + ε i,t. (10 i,t 3 In equation (10, ( I shows the 3 year average data of gross investment to GDP ratio from time (t 2 i,t to (t. 12 Unobserved and time-invariant determinants of the gross investment ratio are captured by the country specific intercept, η i. Adding the ratio of stock of external debt to GDP, ( F Debt, to the regression equation originates from the model where there are convex holding costs of external debt. Theoretically, the incurred shadow interest rate in the domestic economy would affect the investment decision of domestic residents causing the expected sign of the coefficient to be negative for the debtor countries. Here, β 1 represents the role of external debt holdings as an adjustment mechanism for the current account to revert to its long run average. The expected sign of this coefficient would be negative for the current account to be mean reverting. 13 At the annual frequency, the strong negative correlation between the current account and investment may bias the coefficient estimate of β 1 if contemporaneous values of net external debt are 11 The mechanism is shown in the Euler equation for investment in the earlier theoretical discussion. 12 Because this paper analyzes the current account from the perspective of longer-run saving-investment balances, taking the 3 year average data absorbs the factors that affect investment and saving at the business cycle frequency. In this way, the interpretation of the results is less influenced by issues related to the business cycle. 13 Of course, there are other variables which affect the current account besides investment, but, keeping all other variables constant, a country cannot finance domestic investment with continuous external borrowing if it is to achieve solvency constraint. 10

used in the regression. Thus, non-overlapping values of net external debt holdings, ( F Debt, is used as i,t 3 a regressor. Due to the absence of a perfect measure of cost of capital, country-specific real interest rate is used as an imperfect proxy with the coefficient of β 2. Three periods of lagged value of real interest rates is used as contemporaneous values of real interest rate will be an endogenous variable determined by the interaction of investment and saving in the economy. 14 The last parameter, β 3, comes from the neo-classical approach which relates the optimal stock of capital to lag levels of output, thereby tying investment with output growth. y refers to the logarithm of real GDP and y i,t 3 corresponds to the 3 periods lagged real GDP growth. Table (2 shows the results for all countries in the sample at different time periods. 15 In column (1 of Table (2, only 3 year lagged net external debt liabilities to the GDP ratio is regressed on the following non-overlapping 3 year-averaged gross investment to the GDP ratio. The result is significant at 1% for the full time span as well as for the 1975 1988 period. However, for the last decade, even though the coefficient estimate is negative it is not significant. In column (2, the lagged value of real interest rate is used as the only regressor for the same regression equation. The estimates are not significant for any time period. In column (3 both external debt and the real interest rate are included in the regression equation to see if the results in column (1 are capturing the real interest rate effect. The coefficient estimate on the external debt is robust to the inclusion of the real interest rate. In column (4, 3 year lagged value of real GDP growth is included in the regression equation as a control variable similar to the accelerator model. Again, the coefficient estimate on external debt is very robust. Panel regressions show that, on an average base, when a country increases its external debt liabilities by 1% of its GDP, then for the following 3 year averages of investment to GDP ratio decrease by 0.22 of GDP for the 1975 2003 period. The depressing effect of external debt is more pronounced for the 1975 1988 period with coefficient estimate of 0.7% of GDP. Therefore, the implicit interest rate caused by the external debt holdings is so high that it causes a significant decrease in the investment, and it is not captured by the real interest rate. As shown in the appendix, the same results occur when the sample splits into OECD and developing countries. In fact, the less precise relationship for the last decade is not surprising. Increasing financial integration creates extra financing opportunities for the debtor countries and they keep financing their domestic investment by borrowing more without defaulting and the trade-off is the long-lasting current account deficits. 14 β 2 can also be defined as a measure of the crowding out effect of the real interest rate on investment. 15 There are 21 OECD countries and 36 developing countries in the sample. The list of countries are provided in the appendix. 11

3.2 External Debt Holdings and Private Consumption The cost of holding external debts channels through the implicit interest rate in the domestic economy and has implications for national savings. The implicitly higher domestic interest rate increases savings which necessitates a decrease in total consumption. In this and the following sections, it is searched whether the private or public sector is most affected by external debt holding costs. As no country can continue to be financed by the rest of the world, both the private and public sectors should start paying off the accumulated external debts so as to achieve solvency constraint. Since the aim is to look at the role of components of GDP on the mean-reverting current account dynamics, this paper focuses on the role of external debt holdings on the variation of aggregate private consumption to GDP ratio, not the real per capita consumption growth. Aggregate private consumption is the biggest component of GDP with an average ratio of more than 50% in both developed and OECD countries. et, to date there exists no fully satisfactory explanation for the behavior of the aggregate consumption which is valid at different time periods or sample selection. Since there are various unobservable factors such as cultural, physiological, institutional and economical factors, a fixed effect panel regression is used in the following regression equation: ( C i,t ( F Debt = ϕ i + θ 1 + θ 2 r t 3 + ε i,t, (11 i,t 3 where ( C refers to the 3 year average aggregate private consumption as a ratio of GDP from time (t 2 i,t through (t; ϕ i captures the unobserved and time-invariant country specific factors which may affect the dependent variable; θ 1 captures the effect of non-overlapping external debt holdings at time (t 3 on the following 3 year average aggregate private consumption and θ 2 will detect the sensitivity of aggregate consumption to country specific real interest rate. Table (3 shows the fixed effect panel regression results. Regression results give insignificant coefficient estimates, regardless of the time period. This result is surprising because, from the regression results, there is no clear evidence that private consumption cares about solvency constraint though most of the countries have exhibited persistent current account deficits and increasing external debt ratio in the last decade. As shown in the appendix, when the sample is divided into OECD and developing countries, the results predict a positive effect of external debt holdings on the private consumption of OECD countries. 12

3.3 External Debt Holdings and Government Expenditures In both OECD and developing countries, external borrowing constitutes a big portion of the financing source of governments. Hence, as in the case of the private sector, the public sector also will be subject to external debt holding costs. Existing literature on the growth and size of government considers their tax smoothing motives. In the short run, stabilization policies shape government expenditure. Besides the elderly dependency ratio, the population growth rate, openness in trade of goods and services, and capital account restrictions may also affect government expenditure. 16 Therefore, the following parsimonious regression equation is estimated by taking the 3 year averages to isolate the effect of short run government policies and some other business cycle effects on government expenditures: ( G i,t ( Debt = α i + δ 1 + δ 2 r i,t 3 + δ 3(OLDi, t + δ4(op ENi,t + ε i,t. (12 i,t 3 In equation (12, ( G i,t refers to 3 year average government expenditure to GDP ratio; ( Debt is the i,t 3 net external debt position at the end of time (t 3 and δ 1 shows the effect of existing external debt holdings on government expenditure; r i,t 3 is the real interest rate, (OLD is the 3 year average elderly dependency ratio and (OP EN is the 3 year average openness ratio. 17 Table (4 shows the regression results. Estimates of δ 1 are significantly negative in the full sample. When external debt liabilities increase by 1% of GDP above its sample mean, then the following 3 year average government expenditure to GDP ratio decreases by 0.26% of GDP which is in accordance with the mean reverting current account dynamics. 3.4 External Debt Holdings and Current Account Adjustments What is the overall effect of net external debt holdings on the mean reverting current account dynamics? According to the inter-temporal approach to the current account, based on the assumptions of Fisherian separability and PIH in consumption, the current account is used as a shock absorber in the case of a temporary shock to domestic income. This helps to achieve consumption smoothing through lending to or borrowing from the international markets. From the national income identity, the current account is the difference between national saving and domestic investment. Since consumers would change their consumption behav- 16 Growth literature approaches to government expenditures as like investment and assumes that government expenditures are subjects to diminishing return. Therefore, a government with an initially low level of government expenditures will have a higher growth rate of expenditures than countries initially having a high level of government expenditures. Another approach on the government expenditures is the Wagner s law which states that government spending increases with GDP, but later on diverges from GDP due to the growth of the welfare state. 17 Elderly population variable is the fraction of 65-years-old or older population to the total population. It is expected that as elderly dependency ratio increases, the government expenditures will decrease which is consistent with the life-cycle approach as people will decrease their saving at older ages. 13

ior only if they see an increase in their permanent income, any extra income generated from a temporary domestic positive income shock would not be consumed but saved internationally. Investment would not be affected from this temporary shock as it is determined by the exogenous world interest rate such that firms will invest in domestic physical capital to the point that the marginal product of capital would be equal to the world interest rate. Therefore, extra savings would go to foreign markets without affecting any real variable. However, when external debt holding costs is introduced, it will create a shadow interest rate in the domestic economy which will alter the investment decision. 18 As shown in the investment part, when the external debt liabilities increase above its sample average by 1% of GDP, the investment ratio will deteriorate relative to its sample average by almost 0.22% of GDP. In addition, the Euler equation for consumption suggests that the consumption decision, thus saving, is not independent from the level of external borrowing. This is the current account adjustment channel that goes through private and government expenditure and investment. When a country increases its holdings of external debt above the long-run average, private and government expenditure tends to grow lower than the average. From an accounting framework, this decrease goes to saving as the shadow interest rate implies a higher domestic interest rate than the exogenous world interest rate. In Table (5, the overall effect of external debt holdings on mean-reversion of the current account is calculated through the channels of investment, private consumption and government expenditure. The current account is affected by external debt holdings in the following way; the investment channel through the coefficient of β 1 from equation (10, private consumption channel through the coefficient of θ 1 of equation (11, and government expenditure channel through the coefficient of δ 1 of equation (12. The overall effect of external debt holdings to GDP ratio on the current account will be (β 1 + θ 1 + δ 1. The adjustment role of external debt holdings on the current account balances is supported by the data as the accumulated external debts depress domestic investment and increase the domestic saving by decreasing total expenditure. Between 1975 and 2003, after an increase of net external debts by 1% of GDP, the following 3 year average investment and government expenditure ratios decrease in total by almost 0.05% of GDP. Even though the 0.015% of that extra borrowing is used to finance private consumption, the overall recovery effect on the current account surplus is estimated to be 0.033% of GDP. This finding is higher for the 1975 to 1988 period when the overall effect on the current account is 0.10% of GDP. But in the last period it decreased to 0.045%. In fact this result is in accordance with the changing degree of integration in global capital markets over time. Through the first decade of the sample period, countries were experiencing a rigid current account deficit as it is harder for them to have a prolonged current account deficit due to the relatively low degree 18 A debtor country would face a shadow interest rate which is higher than the world interest rate. This would encourage saving and discourage investment. 14

of capital mobility. At low level of capital mobility, the discretionary effect of external debt on investment as well as on total consumption was higher. However in the last decade, with the increasing integration and high level of capital mobility, this discretionary effect is reduced, hence countries were able to achieve a high and persistent current account deficit without crowding out investment. 19 This finding is consistent with Taylor (2002 who studied the current account dynamics in an AR(1 setting and showed that the adjustment speed of the current account to GDP ratio towards its equilibrium is a measure of how a country can externally smooth shocks to saving and investment. 20 4 The Size of the Current Account Adjustments Through Net External Debts The size of the effect of external debt holdings on current account adjustments is very much correlated with the size of the net external debt liabilities as a percentage of GDP. Between 1975 and 2003, in the developing countries, un-weighted sample average net external debt liabilities was 42% of GDP, while this average ratio was just 16% of GDP in the OECD countries. Country average net external debt liabilities fluctuate within the range of 8% to 104% of GDP in developing countries while in OECD countries, it is at the range of 78% to 58% of GDP. 21 Considering the fact that there are some large and small countries with high external debt holdings in each group of countries, the GDP-weighted net external debt holdings will give a better understanding of the size of the external borrowing. A time-varying weighting scheme is constructed in such a way that each country gets a time-varying weight in proportion to its GDP in the overall GDP of that group of countries. More specifically, at time (t, county (i has a weight of: ( GDP w i,t = i,t. Then, for each country group, the GDP-weighted net external debt holdings at time N i=1 GDP i,t (t will be: ( F Debt GDP t = N i=1 w i,t F Debt i,t. Figure (1 gives the GDP-weighted net external debt holdings for OECD and developing countries for the period 1975 2003. According to the figure (1, OECD countries follow a narrow band at the range of 2% to 15%, but it is basically driven by the US. 22 Developing countries exhibit larger external borrowing values: the average 19 Assume that at the beginning, the country is in a zero level of current account and external debt positions. A temporary shock hits the economy, say an earthquake or a hurricane, and the country borrows from abroad by 5% of GDP to absorb the shock. Initially, the economy would start with a current account deficit of 5% of GDP. Then, given that the initial debt would improve the following 3 year-averaged current account by.10 of GDP through private and government expenditures and investment channels, it would take around 6.5 years for the current account to decrease by half. If the ratio was 0.045%, as in the last decade, then the half life would be 12.2 years. 20 He proposed that flexibility of the current account is a result of high capital mobility as it would allow a country to run persistent deficits or surpluses. On the other hand, rigidity of the current account means a lower half life because of the difficulty of achieving prolonged deficits which is considered as a measure of low capital mobility. 21 Recall that positive values corresponds to net debtor positions while negative values show net credit positions in external borrowing. 22 The US increases the OECD averages to a significant extend beginning in the second half of the 1980s. After excluding the US from the sample, average GDP-weighted net debt ratio drops from 6% to 4% of GDP and the maximum debt ratio occurs 15

GDP-weighted net external debt ratio reaches 28% of GDP and countries show a trending increase in external borrowing until the second half of the 1980s. After hitting the peak value at the 41% of GDP in 1985, they experience a trending decrease in external borrowing almost to the level of the 1970s. In light of these actual observations, the adjustment role of external borrowing on current account balances is calibrated within a simple debt accumulation construction. 4.1 An Illustrative Example An illustrative example is constructed to capture the effect of the external debt holdings on current account adjustments. A series of net external debt holdings is simulated, as a benchmark, based on the assumption that country will have a series of constant current account deficits at a ratio determined by the average ex post current account deficits. Then, another series of net external debt ratios will be simulated by using the estimated coefficients measured in the empirical part to compare it with the benchmark series. The benchmark external debt ratios will evolve through time by the following equation where external debt accumulates due to debt servicing and new borrowing: 23 ( ( F Debt F Debt = (1 + r t t 1 ( CA. (13 For both developing and OECD countries, a benchmark series of external debt ratio is simulated based on the actual observations derived from the GDP-weighted current account and net external debt positions. Thus, sample average GDP-weighted current account balances will be used as the average current account balance, and GDP-weighted net external debt ratio in 1975 will be used as the initial net external debt ratio. These benchmark series of net external debt ratios will be compared to another simulated series which extrapolate the adjustment channel from net external debts to current account balances. For this purpose, it will be assumed that each GDP component flows at a fixed fraction of GDP, but, as suggested by this paper, they will deviate from this steady state path if a country holds some non-zero external debt. Accordingly, as shown below, private consumption, gross investment and government expenditures will flow at fixed fractions of GDP at a rate of (c, (i and (g, respectively, and net external debt holdings affect each component by at 7% of GDP instead of 15%. 23 It is assumed that it is a small open economy which can borrow from abroad at the fixed rate of world real interest rate. Current account balances are used as a proxy for new borrowing. Another assumption made here is the absence of valuation effect as the focus of interest is the external debt holdings which are less influenced by the change in equity market prices and foreign currency rates. 16

the fraction of the coefficients estimated in the empirical part which are ˆθ, ˆβ, and ˆδ, respectively. Such that: ( C t ( I t ( G t = c ˆθ 3 = i ˆβ 3 = g ˆδ 3 ( F debt ( F debt ( F debt, t 3, t 3. (14 t 3 In construction of the simulated series of net external debt holdings, the above equations will be embedded to the following national income accounting definition of current account to grasp the role played by the external debt holdings on the current account adjustments: ( CA t ( CA t ( ( ( C I G = 1, t t t ( (F ˆθ + ˆβ + ˆδ debt = 1 (c + i + g +. (15 3 t 3 In contrast to the benchmark debt accumulation equation, after including the depressing effect of external debts on GDP components, net external debt ratios will evolve in the following way: ( ( F Debt F Debt = (1 + r t t 1 ( (F ˆθ + ˆβ + ˆδ debt 3 t 3 (1 (c + i + g. (16 Figure (2 compares the benchmark external debt ratios with those constructed after including the role of external debt holdings on the adjustment of the current account balance. Since the idea is to quantify how external debt holdings behave as a shadow interest rate in the current account adjustments, this simulation analysis does not claim to proxy the observed path of the external debt holdings but rather focuses on the potential quantitative effect of shadow interest rate on the evolution of the external debt holdings in a hypothetical country that runs a constant current account deficit for a given initial external debt holdings. It is allowed that countries will run the exactly same current account deficits as in the benchmark case. The depressing effect of external debts on GDP components will improve the current account balances, which in turn, pin down the long run external debt holdings. It is estimated in Table (6 that for the developing countries this effect, (ˆθ + ˆβ + ˆδ, was 0.094 for the 1975 to 1988 period and 0.052 in the 1989 to 2003 period. For the OECD countries, it was 0.111 in the first half of the sample period and 0.029 in the second half of the sample period. 24 The simulation results in Figure (2 show that, for a typical developing Country 24 Since those values show the long run effect of external debt holdings on each GDP component, in the simulation, one third 17

with a 3% average current account deficit and 0% initial external debt holdings, the difference between the benchmark external debt ratio and the one which takes into consideration the shadow interest rate effect would be 6.5% of GDP in 13 years. This gap will continue to increase and reaches 41% of GDP by the end of 2003. 25 On the other hand, a typical OECD country that starts with a 0% initial external debt to GDP ratio and runs current account deficits by 3% of GDP every year, the percentage difference in the external debt positions between the simulated benchmark series and the one which takes into consideration the shadow interest rate effect would be 7% of GDP and this ratio would increase to 31% of GDP by the end of 2003. of the estimated coefficients are used to simulate the series of net external debt ratios. Also, in all simulations the world average real interest rate is taken as the sample mean of the GPD weighted real interest rates of USA, Germany, and Japan which is around 5%. 25 If the 1975 1988 effect had prevailed throughout the sample period, the gap would be 61% of GDP. 18

5 CONCLUSION Contrary to the prediction of a frictionless open economy model, long-term averages of savings rates and investment rates are highly correlated across countries. This is known as the savings-investment puzzle which was first identified by Feldstein and Horioka in 1980. This puzzling finding has stimulated a large empirical literature that attempts to refute it by studying different data samples and periods and by controlling for some common factors that may effect both investment and savings. Across empirical studies, however, the strong correlation has remained large and significant, though it has tended to decline in recent years. Another way to examine the Feldstein-Horioka finding is by looking at differences between savings and investment rates, that is, current account balance to GDP ratios. Feldstein and Horioka argue that a frictionless international financial market should allow countries domestic investment rates to diverge widely from their savings rates. However, in the data, differences between savings and investment rates have not been large for most of the countries. Thus, the observed (absolute current account balances remain at small levels in size for both OECD and developing countries. This paper concentrates on small current account balances and focuses on the role of external debt holdings in this outcome. It utilizes an approach where net external debt holdings behave like a shadow interest rate in affecting the current account imbalances. A friction in the financial markets is introduced by a simple model where net external debts are subject to convex debt holding costs. In this way, the borrowing country faces a higher effective real interest rate than the creditor country, and it is convex in net external debt holdings. Four basic outcomes arise: One, in the medium-run, net external debt holdings have a role in this stylized fact of small size current account balances. Second, the way net external debt holdings affect the size of current account balances is channeled through investment and government expenditure in the medium-run. Private consumption, on the other hand, isn t affected by net external debt holdings. Third, as shown in the appendix, it is found that OECD countries differ from developing countries in current account adjustments. Finally, net external debt holdings tempered current account imbalances more during the 1980s than the 1990s. The decrease in the dampening effect of external debt holdings on the current account balances is in accordance with the stylized fact of the historical literature concerning capital mobility. In the beginning of the financial integration, a low degree of capital mobility prevents countries from sustaining prolonged current account deficits. This causes a rigid current account behavior with the help of feeding mechanism of external debts. As capital mobility increases, it becomes easier to access external financing as the simultaneous improvement in the financial intermediation decreases the role of external debt holdings in tempering current account balances. 19