Net Foreign Assets and Equilibrium Exchange Rates: Panel Evidence

ABSTRACT By exploiting the information in a panel data set, this paper is able to construct more powerful tests of various hypotheses on the determinants of real exchange rates than would be possible with single-country time-series data. Focusing on annual data for 20 industrial countries from 1973 through 1995, there are three major results. First, the evidence for a stationary real exchange rate is stronger when the exchange rate is defined in terms of wholesale prices than consumer prices, presumably because of the greater tradability of wholesale commodities. Second, the half-life of shocks to the real exchange rate is between two and three years. Third, there is a significant and robust relationship between real exchange rates and net foreign assets.

Net Foreign Assets and Equilibrium Exchange Rates: Panel Evidence Joseph E. Gagnon’ Introduction and Summary This paper exploits a panel of annual data for 20 countries from 1960 through 1995 to examine the relationships between real exchange rates and a number of variables that have been proposed in the literature on exchange rate determination. These variables are the level of net foreign assets, the ratio of consumer prices to wholesale prices, the level of real per capita income, and the share of government consumption in total output. The panel cointegration approach used in this paper estimates long-run equilibrium relationships while controlling for simultaneity and feedback from the exchange rate to the explanatory variables. Real per capita income does not have any statistically significant relationship with the real exchange rate either in the short run or in the long run. Government consumption also does not have any long-run relationship with the real exchange rate, but the future change in government spending does have a robust short-run effect. When government consumption is expected to increase in the following year, %enior Economist in the International Finance Division of the Board of Governors of the Federal Reserve System. This work was performed in part while I was a special advisor at the U.S. Department of the Treasury. I am deeply indebted to Andrew Rose, with whom I began this project, and to Andrew Levin for panel cointegration advice. I would also like to thank David Bowman, Menzie Chinn, Dale Henderson, Graciela Kaminsky, Mico Loretan, Catherine Mann, Jaime Marquez, and John Rogers for helpful comments. This paper represents the views of the author and should not be interpreted as reflecting those of the Board of Governors of the Federal Reserve System, the U.S. Department of the Treasury, or other members of their staffs.

2 the real exchange rate appreciates temporarily. The CPI-WPI ratio is strongly correlated with the real exchange rate defined in terms of CPIS, but it has only a weak long-run correlation with the WPI real exchange rate. This finding supports the view that CPIS contain a significant component of nontradable goods and services prices that are not arbitraged by international trade. The level of net foreign assets is significantly and robustly correlated with the real exchange rate, both in the short run and the long run. Focusing on the relationship between the WPI real exchange rate and net foreign assets, the estimates indicate that the half-life of real exchange rate disequilibria is about fivo and one-half years. For most countries in the sample, net foreign assets are essentially constant, and the real exchange fluctuates widely with no trend. For a few countries, net foreign assets have a pronounced trend. It is primarily for these countries that the real exchange rate appears to drift over time. Literature Review Following the econometric methodology of Levin and Lin (1992), a number of researchers recently have tested for stationarity of the real exchange rate using panel data, including Frankel and Rose (1996), Papell (forthcoming), and Wei and Parsley (1995). These studies typically claim to reject the hypothesis of a unit root in the real exchange rate using the greater power of panel data. However, O’Connell (1996) shows that the standard practice of defining all exchange rates relative to a single I country can distort ‘thesize of the test, and that correcting for this effect weakens the

3 evidence against a unit root. This paper extends the panel framework to control for the O’Connell critique and to consider cointegration between the real exchange rate and other variables. The literature on the determinants of the real exchange rate is too voluminous for an exhaustive review, especially as Froot and Rogoff (1995) and Rogoff (1996) provide up-to-date surveys. Dornbusch and Fischer (1980), Hooper and Morton (1982), and Gavin (1991) present theoretical treatments of the relationship between net foreign assets and the exchange rate that form the basis for this study. Hooper and Morton develop a model in which exogenous shocks to trade flows create a longrun positive correlation between net foreign assets and the real exchange rate (defined as the price of home goods relative to foreign goods). In a more complete theoretical model, Gavin shows that exogenous shocks to wealth create a positive correlation between net foreign assets and the real exchange rate when the Marshall- Lerner condition is satisfied.2 Hooper and Morton (1982), Faruqee (1995), and Obstfeld and Rogoff (1995) present empirical results confirming a positive correlation between net foreign assets and the real exchange rate. In a more general framework, both net foreign assets and the real exchange rate should be viewed as endogenous variables that influence each other and are determined simultaneously. The papers cited above focus on the effects of exogenous 2Thesame shocks create a negative correlation when the Marshall-Lerner condition does not hold. The Marshall-Lerner condition is that the price elasticity of demand for tradables should be sufficiently high that a real depreciation leads to an increase in the trade balance. Nearly all econometric estimates of trade elasticities satisfy this condition, at least in the long run. See Marquez (1990).

4 shocks that affect net foreign assets directly. The transmission to the real exchange rate is based on the conclusion that in equilibrium, a country with negative net foreign assets must have a trade surplus to finance the stream of interest and dividend payments. The mechanism to generate this trade surplus is a real exchange rate depreciation. Similarly, countries with positive net foreign assets must have trade deficits in equilibrium. Thus, a shock to net foreign assets has a long-run effect on the real exchange rate as long as goods produced in different countries are not perfect substitutes. While shocks to the real exchange rate have a well-defined short-run effect on net foreign assets, their long-run effect is ambiguous. Such shocks have no effect on net foreign assets unless they affect a country’s saving rate permanently. In some simple models of trade and asset accumulation, exchange rate shocks have no longrun effect on net foreign assets, and net foreign asset shocks do have a long-run effect on the real exchange rate. It is important to recognize that exchange rate dynamics may be the mechanism by which shocks to desired net foreign assets are equilibrated. Thus, an increase in desired net foreign assets may cause an immediate depreciation of the exchange rate in order to generate a trade surplus, followed by a long-run real appreciation of the exchange rate above its initial level once the desired net foreign asset stock is achieved. Balassa (1964) and Samuelson (1964) argued that technological progress is biased toward the p~oduction of tradables, and that countries experiencing rapid (

5 technological progress should exhibit a rising real exchange rate in terms of price indexes that include nontradables, such as the CPI. A direct test of the Balassa- Samuelson hypothesis would require data on productivity over time, which are difficult to obtain for many countries. As a proxy, this paper uses real per capita GNP. Chinn and Johnston (1996) and Mark and Choi (1995) find evidence for a positive correlation between productivity or per capita income and the CPI real exchange rate using a subset of the countries examined in this paper. A partial test of the Balassa-Samuelson hypothesis is to examine the impact of the CPI-WPI ratio on the CPI real exchange rate to confirm whether movements in the tradables-nontradables price ratio plays a significant role in real exchange rate movements. Rogoff (1996) shows that the WPI real exchange rate between Japan and the United States has exhibited a much smaller drift over time than the CPI real exchange rate. Froot and Rogoff (1991) and DeGregorio, Giovannini, and Wolf (1994) find evidence of a temporary effect of government spending on the CPI real exchange rate. They argue that higher government spending tends to appreciate the real exchange rate because it falls more heavily on nontraded goods. Rogoff (1992) argues that the long-run effects of government spending are much smaller than the short-run effects. Nevertheless, Chinn and Johnston (1996) claim that the ratio of real government consumption to real GDP does have a long-run effect on the real CPI exchange rate.

6 The Data The dataset is an annual panel with observations for 20 industrial countries from 1960 through 1995.3 The basic regressions were run over the floating-rate period beginning in 1973. The series were obtained from the International Financial Statistics database maintained by the International Monetary Fund.4 For the basic regressions, all exchange rates were defined vis-a-vis Germany, and the German data were dropped from the regressions.s The data mnemonics and definitions are presented in Table 1. Net foreign assets are scaled by trade flows because equilibrium is achieved through the operation of the real exchange rate on the trade sector of the economy. Lhivariate Analysis Figures 1-6 plot the data starting in 1970. Most of the real exchange rate series exhibit large swings around a slowly drifting or constant mean. For the WPI real exchange rate, the primary exceptions are those of Germany’s immediate neighbors-- 3Australia,Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Due to missing observations for certain series the sample begins in 1980 for Germany, and it ends in 1992 for Finland and 1994 for Belgium, Germany, Italy, and Greece. 4Finnishnet factor income was obtained from OECD Annual National Accounts. Belgian and French wholesale prices were obtained from national sources. 5Frenkel (1981) was the first to note that the evidence for purchasing power parity is weaker when us” g dollar exchange rates, a finding confirmed by Edison, Gagnon, and Melick (forthco%‘ing). As discussed below, use of the United States as the partner country has only a smaHeffect on the results of this paper.

7 Table1 Data Definitions REc Log real exchange rate using CPIS (increase = appreciation) REW Log real exchange rate using WITS (increase = appreciation) NFA Net foreign assets (calculated by cumulating exports plus net factor income minus imports since 1960) divided by exports — plus imports of goods and services CWR Log ratio of CPI to WPI YPc Log real GNP per capita GOV I Ratio of government consumption to nominal GDP Austria, Denmark, the Netherlands, and Switzerland--which are characterized by . much smaller movements and very little drift. A downward drift is noticeable for Australia, Canada, and the United States. An upward drift is apparent for Japan. NFA drifts upward in Japan and Switzerland, and downward in Australia, New Zealand, and the United States, with little trend in the remaining countries. .. CWR drifts upward in most countries. YPC exhibits a pronounced upward trend in every country. Finally, GOV has either a small upward trend or no trend in every country. Table 2 presents the results of panel unit root tests for each variable. The tests were conducted by regressing the first difference of the variable on its lagged level, lagged first differences of the variable, and a full set of country and year dummy variables, as shown in equation (l). Four lagged differences were included in the initial regressions and the lag length was truncated sequentially by dropping

8 coefficients that were not significant at the 10 percent level.6 h most cases only one lagged difference was retained, but in all cases the adjustment coefficient was completely insensitive to the choice of lag length and the test statistic was only moderately affected. AWit = ~~it_l + ~=1 ~kA-it_k + ~i + .t (1) i = 19 industrialcountries t = 1973, ..., 7995 Levin and Lin (1992) show that estimation of country-specific intercepts biases Lb~adj’.::t~eiltcoefficient, p, and the associated test statistic downward. They pi< v;d:’ ~tatle of adjusted critical values from Monte Carlo simulations of different sample sizes which is the basis for the significance levels in Table 2. O’Connell (1996) points out that the definition of all exchange rates in terms of a common country cai~ lead to a cross-country correlation in the residual that also tends to bias the test stat~sticdownward. By estimating a complete set of time effects, the regression is ale te c~ntrol fsr the common influence of the omitted country, Germany. Both real exchange rates appear to be stationary.7 The real exchange rate using WPISexhibits stronger mean reversion than that using CPIS, which may reflect the greater tradability of the components of the WPI. There is weak evidence of mean reversion of the CPI-WPI ratio and the government consumption ratio. There is no 61also tested for the significance of a full set of country-specific dynamic coefficients. I found them to be significant only for YPC and GOV, and their inclusion had no material effect on the unit root test statistics for YPC and GOV. 7Usingexchange rates vis-a-vis the United States leads to nearly identical results.

evidence of mean reversion in net Table2 Panel Unit Root Tests foreia assets and real per ca~ita u A J. (Equation (l)) incomes. Variable Adjustment Test Further evidence on the time- Coefficient Statistic REc -.25 .SOO*** series properties of these data can be Ww -.32 -8.8*** derived by analyzing the estimated time NFA -.01 -1.3 effects from the panel data regressions. CWR -.14 -6.6* WC -.08 -4.0 These time effects capture the GOV -.19 -6.4 movements common to all countries ***,**,and*denotesignificancaetthe1, 5,and10percentlevels,respectively. over time, and they may themselves contain unit roots. To perform the analysis, I combined the time-effect coefficients sequentially to create a new data vector for each variable listed in Table 2. I then conducted augmented Dickey-Fuller tests with one lagged difference on these data vectors. The tests revealed that the estimated time effects are stationary at the 5 percent level for all variables except Yrc. Multivariate Analysis In light of the evidence of the previous section that real exchange rates are stationary, it may seem unlikely that there exist significant correlations between real exchange rates and the remaining variables in the dataset, which appear to be nonstationary, or, at most, borderline stationary. However, it is important to

I 10 recognize that the maintained assumption of the panel unit root test is that either all series are stationary or all series have unit roots. Recall that the plots of the data revealed that real exchange rates exhibit drifts in certain countries, but not in other countries. The pattern of drifts over time also varied across countries for the other variables, raising the possibility that some of the other variables may be able to explain the apparent long-run movements of the real exchange rates of certain countries. The approach taken here is purely empirical. The goal is to search for a longrun correlation between the real exchange rate and any or all of the explanatory variables. No attempt is made to identify a theoretical structural relationship. However, significant evidence of a long-run correlation, or cointegrating vector, may be taken as evidence on the types of shocks and transmission charnels that are important in the data. In a truly long sample, one could simply regress the real exchange rate on the explanatory variables. However, in the relatively short sample available, dynamic adjustments to shocks, and feedback from the exchange rate to some of the explanatory variables, may lead to biased estimates of the long-run correlations. Moreover, the adjustment dynamics and feedback effects may be of interest in themselves. Phillips and Loretan (1991) provide a methodology for dealing with these issues in short samples. As shown in equation (2), the regressions of the previous section are augmented to include the other four variables. The term in parentheses is the long-

11 run cointegrating relationship between the real exchange rate and the explanatory variables. The error-correction coefficient, p, captures the long-run adjustment speed to deviations from equilibrium. In order to control for short-run dynamics and feedback effects from the real exchange rates to the independent variables, I included backward and forward differences of the independent variables as well as lagged differences of the real exchange rate. Additional lags of the explanatory variables were not significant, mesumablv reflectimzthe fact that exchange rates respond L J u quickly to economic news. Finally, the regressions include a full set of country and time effects. ARExti= p(RExti_-l Y1N..Ati_l-51c~-l - e, YPCu_l- AlGO~t_l) + Yz‘NFAit+l + Y3ANFAti+ 52Ac~”t+l + /33AC~.t + 02AYPCit~1+ 03AYPC&+ L2AGOVit+1+L3AGOVit (2) hAR%t-+k + Z=l (x i + .t i = 19 industn”alcountn”es t = 1973, ..., 1994 In order to obtain a parsimonious and robust specification, I performed sequential tests on the estimates of equation (2). First, I tested the lagged differences of the real exchange rates. The second, third, and fourth lags were never individually or jointly significant at the 10 percent level, so they were dropped from the regression. Next I tested the significance of the independent variables.g For each variable I tested the joint significance of the lagged level, forward difference, and backward difference. YPC was never significant at the 10 percent level. GOV was ‘According to Levin and Lin (1992), if one accepts the hypothesis of cointegration (based on the estimated error-correction coefficien~ p, and its standard error) the coefficients on the levels of the explanatory variables have standard distributions.

12 significant at the 5 percent level for REC and the 10 percent level for REW. NFA and CWR were always significant at the 1 percent level. After dropping YPC from the regressions, further tests revealed that only the forward difference of GOV was significant. The lagged level and backward difference of GOV were dropped. The results are presented in Table 3. According to the first column of Table 3, a future increase in government spending of 1 percent of GDP causes an immediate increase in the real exchange rate of 0.87 percent. While it is possible that this correlation arises because shocks to the exchange rate affect future government spending, a more plausible interpretation is that the political process makes fiscal policy highly forecastable. Thus, forwardlooking exchange markets can price in the effect of future government spending. As discussed above, government spending has no significant long-run effect on either REC or REW. Changes in the CPI-WPI ratio show up one-for-one in the CPI real exchange rate (REC) immediately (from the coefficient on ACWRJ, but in the 1ongrun only about two-thirds of these changes are reflected in the CPI real exchange rate (from the coefficient on CW&.J. These estimates are consistent with those from the REW equation: changes in the CPI-WPI ratio have no immediate effect on the WPI real exchange rate, but they tend to reduce the WPI real exchange rate in the long run. If international arbitrage equates the prices of the components of WPIS and not the components of CPIS, one would expect to observe a long-run coefficient on CWR of 1 in the REC equation and Oin the REW equation. On the other hand, if CPIS are

13 arbitraged and WPIS are not, one would Table3 Multivariate Results expect to observe a coefficient on CWR (Equation (2)) of Oin the REC equation and -1 in the ~Ci, ~Wi~ REW equation. These estimates imply E.C.(P) ..34*** -.34*** (.04) (.04) that arbitrage works better for the .11*** .1O*** NFAi~.l (.03) (.03) components of WPIS,but that a .66*** cwlq.~ -.26* weighted average of WPIS and CPISis (.16) (.16) ~FAi~+l -.15*** -.14*** more stable than either index alone. (.05) (.05) .24*** .24*** The positive coefficient on future ~FAi~ (.05) (.05) changes in the CPI-WPI ratio reflects the .34*** .33*** Acwt+l (.11) (.11) feedback of the exchange rate onto ACW& 1.14*”” .16 (.11) (.10) domestic prices. Exchange rate AGOVi~+l .87** .84** (.35) (.35) appreciations tend to reduce WPIS more .13*** .19*** ~Xit-~ than CPIS,thereby increasing CWR in (.04) (.05) R* .66 .59 the following period. std.error .046 .045 Turning to the effect of NFA, No. Obs. 414 414 ***,**,and*denotesignificancaetthe1, Table 3 displays a significant positive 5,and10percentlevels,respectively. correlation between the real exchange rate and the level of net foreign assets. When net foreign assets increase by an amount equal to total exports plus imports (ANFA=l) the real exchange rate appreciates by 24 percent in the short run and 10-11 percent in the long run. The short-run effect is much larger because the exchange

14 rate is a forward-looking variable and current account balances are very persistent. Thus, a small increase in the growth rate of NFA tends to lead to a large permanent increase in the level of NFA. The negative coefficient on future changes in NFA reflects the equilibrating nature of the system, as an appreciating exchange rate puts downward pressure on the future growth rate of NFA. To put these estimates in perspective, the United States’ current account balance in 1994 was equal to about 10 percent of exports plus imports of goods and services. A permanent decrease in NFA of this magnitude would be expected to depreciate the real exchange rate by I percent in the long run. As in the previous section, an augmented Dickey-Fuller test strongly rejects the hypothesis of a unit root in the estimated time effects of both regressions presented in Table 3. Figure 7 plots the time effects from the REW regression. The horizontal line represents the mean value of the estimated time effects.9 Robustness Table 4 compares the basic results for the WPI real exchange rate with five alternative specifications to provide evidence on the robustness of the estimated relationship between net foreign assets and the real exchange rate. The first column reprints the estimates from the last column of Table 2. The short-run dynamic coefficients on CWR and GOV are omitted. The second column presents estimates of 9Theinclusion of a full set of country dummies necessitates a zero restriction on one of the time effects, in this case the 1994 dummy was dropped.

15 Table4 Alternative Specifications Basic Basic NFIproxy country- country- U.S.as Model Model forNFA Specific Specific Reference 1962-94 Trends Dynamics country E.C.(~) -.34*** -.23*** -.35*** -.54*** -.40*** -.35*”* (.04) (.03) (.04) (.05) (.04) (.04) NFAit.l .1O** .14*** 1.22*** .11** .09*** .08** (.03) (.04) (.45) (.05) (.03) (.03) Cwq.l -.26* -.17 -.16 .25 -.23 -.27* (.16) (.11) (.16) (.17) (.14) (.15) ~FAit+l -.14*** -.12*** .11 -.15*** n.a. -.15*** (.05) (.04) (.31) (.06) (.05) ~FAit .24*** ,19*** -.35 .17** n.a. .25*** (.05) (.04) (.32) (.05) (.05) ~Wj~~ .19*** .18*** .20*** .25*** n.a. .l&+** (.05) (.04) (.05) (.05) (.05) R* .59 .50 .57 .65 .69 .80 No. Obs. 414 593 411 414 414 406 ***,**,and*denotesignificancaetthe1,5,and10percentlevels,respectively. the basic specification over the entire sample from 1962 through 1993. Not surprisingly--given the switch from fixed to floating exchange rates--there is evidence of a structural break in the adjustment speed coefficient, but one can still reject noncointegration at the 1 percent level. (It is because of this structural break that most of the regressions focus on the floating-rate period only.) Using the entire sample, the long-run estimated effect of NFA on REW is even greater and more significant than in the floating rate subsample.

16 The second alternative examines the effect of measurement error in net foreign assets. NFA is calculated by cumulating the own-currency current account balances over time.l” The lack of an initial benchmark value has no effect on the estimated coefficient since there is a complete set of country-specific intercepts in the regression. However, the calculation of NFA ignores valuation effects on the gross stocks of foreign assets and liabilities due to movements in exchange rates, equity prices, and land prices. In addition, there are large errors and omissions in the international current accounts data. The third column presents estimates using the ratio of net factor income to total trade (NFI) as an alternative measure of NFA. In theory, NFI should be equal to a market rate of return times the level of NFA. If the rate of return were around 5 to 10 percent, one would expect the coefficient on NFI to be around 10 to 20 times larger than that on NFA. As shown in Table 4, the estimated coefficient on NFI is 12 times larger than that on NFA, implying an average rate of return of 8 percent. The short-run impact of changes in NFI appear quite different from those of NFA. This difference probably reflects the tendency for foreign assets to be denominated in foreign currencies and foreign liabilities to be denominated in domestic currency. Periods of exchange-rate appreciation are thus associated with declining net factor income, as the domestic-currency value of earnings on foreign assets declines relative to payments on foreign liabilities. 1°Asthe sum of net factor income and exports minus imports of goods and services, NFA differs from a cumulated current account balance slightly due to the omission of unilateral transfers.

-.17 The next two alternatives consider different econometric specifications. The fourth column displays the effect of including country-specific time trends in the regression. The specification with time trends yields a somewhat faster estimated adjustment speed and a positive, but insignificant, long-run effect of CWR on REW. The estimated long-run effect of NFA is essentially unaffected. The fifth column displays the effect of estimating separate coefficients for each country on the lagged difference of the exchange rate and the forward and backward differences of net foreign assets. Despite the profligate parameterization of this specification, the coefficient on the level of NFA is nearly identical to its value in the basic specification. Finally, the last column presents the estimates of the basic specification when all exchange rates are defined relative to the United States. In this regression the U.S. data are dropped and the German data are included. The results are very similar to those of the first column. Interpretation of Results The statistical results of the previous three sections suggest a paradox: Real exchange rates are stationary; net foreign assets are nonstationary; yet there is a longrun correlation between real exchange rates and net foreign assets. The solution to this paradox is apparent in Figure 8. The solid lines are log WPI real exchange rates (REW), demeaned by country. The cross symbols represent the long-run effect of the ratio of net foreign assets to

18 total trade (NFA) and the CPI-WPI ratio (CWR) on REW from the basic regression, also demeaned by country. The cross symbols are estimates of the long-run equilibrium real exchange rate. In every country, the variance of the real exchange rate exceeds the variance of the equilibrium exchange rate. In many countries, the equilibrium real exchange rate is essentially a constant, reflecting the stability of NFA and the very small effect of CWR. For these countries, the real exchange rate is stationary. Since these countries dominate the sample, the panel unit root test is able to reject the hypothesis that all real exchange rates contain a unit root. However, for several countries--notably Australia, Japan, New Zealand, Switzerland, and the United States--NFA does have a visible trend. These countries provide the evidence in favor of a long-run effect of NFA on REW.11 To take two important examples: Between 1973 and 1994 NFA fell from .53 to -.45 in the United States, implying a permanent real exchange rate depreciation of nearly 10 percent. During the same period in Japan, NFA rose from .29 to 2.0, implying a permanent real appreciation of roughly 17 percent. Thus, movements in net foreign assets can explain a cumulative drift of 27 percent in the U.S.-Japanese real exchange rate over the past 20 years, which is nearly two-thirds of the total movement in the real bilateral exchange rate over this period. 1lWhenthese countries are excluded from the regression, the long-run coefficient on NFA is greatly reduced but the new estimate is only about one standard deviation lower than the original estimate. The remaining coefficients are affected only slightly.

-19 References Alesina, Alberto and Roberto Perotti (1995) “Taxationand Redistribution in an Open Economy,” European Economic Review 39,961-79. Balassa, Bela (1964) “ThePurchasing Power Parity Doctrine: A Reappraisal,” Journal of Political Economy 72,584-96. Chin-n,Menzie, and Louis Johnston (1996) “RealExchange Rate Levels, Productivity and Demand Shocks: Evidence from a Panel of 14 Countries,” NBER Working Paper No. 5709, August. De Gregorio, Jose, Alberto Giovannini, and Holger Wolf (1994) “International Evidence on Tradables and Nontradables Inflation,” European Economic Review 38, 1225-44. Dornbusch, Rudiger, and Stanley Fischer (1980) “ExchangeRates and the Current Account,”American Economic Review 70, 960-71. Edison, Hali J., Joseph E. Gagnon, and William R. Melick (forthcoming) “Understandingthe Empirical Literature on Purchasing Power Parity: The Bretton Woods Era,”Journal of International Money and Finance. Faruqee, Hamid (1995) “Long-RunDeterminants of the Real Exchange Rate: A Stock- Flow Perspective,” IMF Staff Papers 42, 80-107. Frankel, Jeffrey A., and Andrew K. Rose (1996) “MeanReversion Within and Between Countries: A Panel Study of Purchasing Power Parity,”Journal of International Economics 40, 209-24. Frenkel, Jacob A. (1981) “Flexible Exchange Rates, Prices, and the Role of News: Lessons from the 1970s,”Journal of Political Economy 89, 665-705. Froot, Kenneth A., and Kenneth Rogoff (1991) “TheEMS, the EMU, and the Transition to a Common Currency,“in Blanchard and Fischer, eds., IW?ER Macroeconomics Annual 1991 (Cambridge: The MIT Press). (1995) “Perspectiveson PPP and Long-Run Real Exchange Rates,” in Grossman and Rogoff, eds., Handbook of International Economics (Amsterdam: North Holland Press). Gavin, Michael (1991) “Termsof Trade, The Trade Balance, and Stability: The Role of Savings Behavior,” International Finance Discussion Papers No. 397, Board of Governors of the Federal Reserve System.

20 Hooper, Peter and John Morton (1982) “Fluctuationsin the Dollar: A Model of Nominal and Real Exchange Rate Determination,” Journal of International Money and Finance 1, 39-56. Levin, Andrew, and Chien-Fu Lin (1992) “UnitRoot Tests in Panel Data: Asymptotic and Finite-Sample Properties,” University of California at San Diego Discussion Paper 92-23. Mark, Nelson and Doo-Yu1lChoi (1995) “RealExchange-Rate Prediction over Long Horizons,” Ohio State University manuscript, November. Marquez, Jaime (1990) “Bilateral Trade Elasticities,” Review of Economics and Statistics 72, 70-77. Obstfeld, Maurice (1982) “Aggregate Spending and the Terms of Trade: Is There a Laursen-Metzler Effect?” Quarterly Journal of Economics 97, 251-70. Obstfeld, Maurice and Kenneth Rogoff (1995) “TheIntertemporal Approach to the Current Account,” in Grossman and Rogoff, eds., Handbook of International Economics (Amsterdam: North Holland Press). O’Comell, Paul G. (1996) “TheOvervaluation of Purchasing Power Parity,” Harvard University manuscript, January. Papell, David H. (forthcoming) “Searching for Stationarity: Purchasing Power Parity under the Current Float,” Journal of International Economics. Phillips, Peter and Mico Loretan (1991) “EstimatingLong-Run Economic Equilibria,” Review of Economic Studies 58,407-36. Rogoff, Kemeth (1996) “ThePurchasing Power Parity Puzzle,”JournaZof Economic Literature 34, 647-68. Samuelson, Paul A. (1964) “Theoretical Notes on Trade Problems,” Review of Economics and Statistics 46, 145-54. Wei, Shang-Jin and David Parsley (1995) “PurchasingPower Disparity during the Floating Rate Period: Exchange Rate Volatility, Trade Barriers, and Other Culprits,” NBER Working Paper No. 5032, February.

‘) o 0 m 07 r-! a) -P (73 CE x m al CE I z d H n (-l m o IQoul mom Irloul lnoln- 1“ 1“ i“ . 1“

22 0 w H 0 (n m a) -+J m o m m o m m - 0 h UI m a) m u w N IA rl =- mr, --l m.om mom. lqotq q o q“ 1“ - I I I

-.23 .E m a) w 1- u on 1- ’11- 1- m c- 1- (U -w a u N / 0 “n- H k =n—n— T- I 1 1 1 t- 0 x IJJ \ mm =?—1—rr =l-t- n c m -1-1 c1) c-. 0 Lu i U ‘n—r—n z LLJ -r-r-. N ~Tll\l— -1 I 1 I ? I-L -v CD z l-l

E= AT BE DE .2 0 -.2 -.4 FI F13 GR .2 0 -.2 -.4 JA NE NO NZ so .2 0 -.2 -.4 & Sz UK .2 L— 0 -.2 -.4 -1 J- 19010 19 1 90 1970 laiii___ 1990 1970 19s0 1990 Figure 4 Log CPI-WPI Ratio (1990=0)

25 . 0 w o ,~, \ 0 H l“- 07 .Z m w-l CD UI \ .: 0E =nl—l— m c-) c q-+ .: m c-) ccl-l n (u al CE

L BE CA DE .3 1 -.\ .2 .1 1 +, 0 b- 1 1 k FI Gi7 IR .3 .2 .1 -1 1 0 i, i -4 1 1 I L JA NZ .3L .2 -1 .1 h) m 0 SP !& e UK us 1970 19180 19910 .3 4– — :; ~ 1, o 1L9 0 19B0 1990 1970 1980 1990 1970 1980 1990 1970 19 1 B0 19 I 90 Figure 6 Government Share of GDP

27 \ > < o .0-) > (n m -P (-) CD Lny .Co @-+JJ c1) LCD < u o CD .Co -w 07 m E -1-1 -v m w u-) -b m < a u-) Y-1 4 . Ln 0 0 . 1- 1“

28 u a) c m a) E CD n u) (D -P m m CE H a) x -d 0 Ll- W a c1 J m cl) Ct w 1- u z < U_ w H IL (nor! mom. moq m.orq I I I I r

29 International Finance Discussion Papers IFDP Number Titles Author(s) 1996 574 NetForeignAssetsandEquilibriumExchangeRates: Joseph E. Gagnon Panel Evidence 573 TimingofChildbearingF,amilySizeandEconomicGrowth Murat F. Iyigun 572 Money,politicsandthepost-Warbusinesscycle Jon Faust John Irons 571 ImplicationosfEconomicInterdependenacned Jay H. BVson Exchange Rate Policy on Endogenous Wage Chih-huan Chen indexation Decisions David D. VanHoose 570 The Reaction of Exchange Rates and Intemt Ha.liJ. Edison Rates to News Releases 569 The Role of C~dit in Post-Stabilization Martina Copelman Consumption Booms 568 Hazards in Implementing a Monetary Conditions Kari H. Eika Index Neil R. Ericsson Ragnar Nymoen 567 Financial Innovation And The Speed of Adjustment Martina Copelman of Money Demand: Evidence From Bolivia, Israel, And Venezuela 566 Long-Term Evidence on the Tobin and Fisher Shaghil Ahmed Effects: A New Approach John H. Rogers 565 Some Evidence on the Efficacy of the UK Inflation Chan Huh Targeting Regime: An Out-of-Sample Fo~cast Approach 564 The Use of the Parallel Market Rate as a Guide Nita Ghei to Setting the Official Exchange Rate Steven B. Kamin 563 Country Fund Discounts and the Mexican Crisis of Jeff~y A. Frankel December 1994: Did Local Residents Turn Sergio L. Schmukler Pessimistic Befo~ InternationalInvestors? Please address ~quests for copies to InternationalFinance Discussion Papers, Division of InternationalFinance, Stop 24, Board of Governors of the Federal Rese~e System, Washington, DC 20551. Email: graym@frb.gov. Fax: (202)736-5638.

30 International Finance Discussion Papers IFDP Number Titles Author(s). 1996 562 Eastern European Export Performance during Nathan Sheets the Transition Simona Boata 561 Inflation-Adjusted Potential Output Jane T. Ha]tmaier 560 The Management of Financial Risks at Gen-nan Allen B. Frankel Nonfinancial Firms: The Case of Metallgesellschaft David E. Palmer 559 Broad Money Demand and Financial Liberalization Neil R. Ericsson in Greece Sunil Sharma 558 Stockholding Behavior of U.S. Households: Evidence Carol C. Bertaut from the 1983-89Survey of Consumer Finances 557 Fim Size and the Impact of Profit-Margin Uncertainty Vivek Ghosal on Investment: Do Financing Constraints Play a Role? Prakash Loungani 556 Regulation and the Cost of Capital in Japan: A Case John Ammer Study Michael S. Gibson 555 The Sovereignty Option: The Quebec Referendum and Michael P. Leahy Market Views on the Canadian Dollar Charles P. Thomas 554 Real Exchange Rates and Inflation in Exchange-Rate Steven B. Kamin Based Stabilizations: An Empirical Examination 553 Macroeconomic State Variables as Determinants John Ammer of Asset Price Cova.riances 552 The Tequila Effect: Theory and Evidence from Martin Uribe Argentina 551 The Accumulation of Human Capital: Alternative Murat F. Iyigun Methods and Why They Matter Ann L. Owen 550 Alternatives in Human Capital Accumulation: Murat F. Iyigun Implications for Economic Growth Ann L. Owen

31 International Finance Discussion Papers IFDP Number Titles Author(s) 1996 549 More Evidence on the Link between Bank Michael S. Gibson Health and Investment in Japan 548 The Syndrome of Exchange-Rate-Based Enrique G. Mendoza Stabilization and the Uncertain Duration of Martin Uribe Currency Pegs 547 German Unification: What Have We Learned Joseph E. Gagnon from Multi-Country Models? Paul R. Masson Warwick J. McKibbin 546 Returns to Scale in U.S. Production: Estimates Susanto Basu and Implications John G. Femald 545 Mexico’sBalance-of-Payments Crisis: A Chronicle Guillen-noA. Calvo of Death Foretold 544 The Twin Crises: The Causes of Banking and Graciela L. Kaminsky Balance-of-Payments Problems Carmen M. Reinhart 543 High Real Interest Rates in the Aftermath of Graciela L. Karninsky Disinflation: Is it a Lack of Credibility? Leonardo Leiderman 542 Precautionary Portfolio Behavior from a Life-Cycle Carol C. Bertaut Perspective Michael Haliassos 541 Using Options Prices to Infer PDF’sfor Asset Prices: William R. Melick An Application to Oil Prices During the Gulf Crisis Charles P. Thomas 540 Monetary Policy in the End-Game to Exchange-Rate Steven B. Karnin Based Stabilizations: The Case of Mexico John H. Rogers 539 Comparing the Welfare Costs and the Initial Dynamics Martin Uribe of Alternative Temporary Stabilization Policies 538 Long Memory in Inflation Expectations: Evidence Joseph E. Gagnon from International Financial Markets

32 International Finance Discussion Papers IFDP Number Titles Author(s) 1996 537 Using Measures of Expectations to Identify the AlIan D. Bmnner Effects of a Monetary Policy Shock 536 Regime Switching in the Dynamic Relationship Chan Huh between the Federal Funds Rate and Innovations in Nonborrowed Reserves 535 The Risks and Implications of External Financial Edwin M. Truman Shocks: Lessons fmm Mexico 534 Currency Crashes in Emerging Markets: An Jeffrey A. Frankel Empirical Treatment Andrew K. Rose 533 Regional Patterns in the Law of One Price: The Charles Engel Roles of Geography Vs. Currencies John H. Rogers

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 574 December 1996 NET FOREIGN ASSETS AND EQUILIBRIUM EXCHANGE RATES: PANEL EVIDENCE Joseph E. Gagnon NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International FinanceDiscussionPapers (otherthan anacknowledgment thatthewriter hashadaccess to unpublished material) should be cleared with the author or authors.