Financial Exchange Rates and International Currency Exposures

That while the valuation channel of exchange rates was well established in theory and for a few countries, “remarkably little is known about the currency composition of the foreign assets and liabilities of most countries” – so the paper’s major contribution is to build that empirical profile for a large country set. The authors situate the work between two existing strands: one on emerging markets with large stocks of foreign-currency debt, for whom depreciation is balance-sheet-damaging; and one on major advanced economies, which are “typically short in domestic currency and long in foreign currencies,” so that unanticipated depreciation raises the net international investment position (Section 2, pp. 3-4). Relative to earlier work – Eichengreen, Hausmann and Panizza (2003) and Goldstein and Turner (2004) on the currency composition of developing-country external debt, Tille (2003) on the United States, Lane and Milesi-Ferretti (2007c) on European countries plus Japan and China – the paper claims two advances: coverage of a far larger number of countries, and estimation of the full currency composition of the international balance sheet rather than only the debt component (p. 6). The earlier debt-focused contributions, they note, “do not take into account the portfolio equity and FDI components of the international balance sheet.”

By a two-step procedure – currency composition within each asset class, then weighting asset classes by their shares in the international balance sheet – with inference procedures filling gaps. The asset side is divided into five classes: portfolio equity, direct investment, portfolio debt, other (generally bank-related) debt, and reserves, each with its own sources and methodology (Appendix A.1). Where bilateral positions are missing, the authors “rely on recent advances in the modeling of the geographical distribution of international financial portfolios to generate predictions for asset holdings that allow us to fill in missing observations,” for instance a gravity-based model of bilateral equity holdings built on CPIS coverage of 68 reporter countries across 220 host countries (Section 3, p. 7; Appendix A.1.1). Asset-class weights come from the External Wealth of Nations database, which reports foreign assets and liabilities for 145 countries over 1970-2004 split between portfolio equity, direct investment, reserves and debt (Section 4, p. 8). The full sample is 117 countries with full data; hyperinflation episodes are excluded as outliers and a country’s data begin after a hyperinflation ends, with countries whose hyperinflations fall late in the sample dropped entirely. Results on the 1994-2004 change use a smaller 102-country sample with full data over that window (Section 3, p. 7).

Because the authors want an index that behaves like the value of a portfolio holding those currency positions, and a geometric index does not. The index averages percentage changes in bilateral exchange rates using the relevant weights and multiplies the previous period’s index, rather than taking a geometric average of levels (equations 9-10, Section 4, p. 11). Two design choices follow. First, the exchange rate is defined as the home price of foreign currency, so a partner’s hyperinflationary collapse sends its rate toward zero rather than toward infinity, and the index falls by at most that currency’s weight – “the equivalent of some portion of a portfolio becoming worthless.” A geometric index, by contrast, is driven toward zero by any near-zero entry regardless of its weight: the authors’ example is a rate falling from 100 to 0.1 with a 10 percent weight, which moves a summation index from 100 to 90 but a geometric index from 100 to 50 (fn. 13, p. 12). Second, because the index combines percentage changes, a change in weights with no change in exchange rates produces no change in the index, so “more complex chain weighting is not necessary; we can simply employ new weights whenever they are available” (p. 12). Where weights are unavailable the authors average across gaps, and 1997 weights are extended back to 1990 for asset classes whose earliest currency-composition data are from 1997.

Different enough that the trade index “does a poor job of summarizing the net financial impact on a country when the exchange rate changes.” The asset and liability indices are highly correlated with each other and each is correlated with the trade index, though more weakly for liabilities, reflecting the importance of domestic-currency liabilities. But the net financial index is strongly negatively correlated with the trade index on average in the full and developing samples (Table 1, Section 5.1.1). The authors are careful about the mechanism: this is not because asset and liability indices move in opposite directions, but because “it is the net positions and also the size of the movements of assets- and liability indices that generate the diverging pattern.” The cross-sectional distribution shows a cluster of countries correlated near minus one – typically countries that had very large depreciations while holding negative foreign-currency positions – and weak correlations for many others simply because trade partners and financial partners differ (Figure 1, p. 13). For industrial countries, which on average hold net positive foreign-currency positions, the mean correlation is 0.41 and the median 0.70, against above 0.85 for every other pair of indices.

Liability indices are much more stable than asset indices, so a generalized currency movement shifts net foreign wealth even when the net foreign asset position is zero. The average standard deviation of the percentage change in the liabilities index is 3.5 percent for industrial countries against 5.9 percent for assets, reflecting the greater share of domestic currency on the liability side; the United States is the extreme case, with over 90 percent of liabilities in dollars and a liability index volatility of less than 1 percent a year (Table 2, Section 5.1.2, p. 14). Regressing the change in the liability index on the change in the asset index yields an estimated coefficient of 0.80 for the full sample and the developing group, with very high R-squared, and 0.66 for advanced countries; “since the estimated coefficient is below 1 in all cases, a generalized movement in the value of the home currency against other currencies will induce a shift in the value of the net foreign asset position, even for a country with an initially-balanced international investment position” (Table 3, Section 5.1.3, p. 15). Net financial indices are far more stable than any other index for all country groups, reflecting the offsetting effects on foreign-currency assets and liabilities – but with a fair degree of volatility remaining for developing countries.

The net financial index is more stable than the trade index but moves in the opposite direction, and the associated losses are large. In sudden stops and in cases where a country depreciates 50 percent or more against its base currency, “the net index is both strikingly more stable and moves in an opposite direction of the trade index,” with the net index moving about minus 8 percent for sudden stops and minus 30 percent for large depreciations. The corresponding wealth losses are that “the sudden stop countries lost 6 percent of GDP on average and the large depreciation countries 29 percent of GDP” (Panel B of Table 2, Section 5.1.2, pp. 14-15). These calculations rest on 17 sudden stops – taken from the list in Durdu et al (2007) – and 52 large depreciations, in both cases excluding hyperinflations (fn. 16, p. 15).

A clear majority of countries were short foreign currency in 1994, and by 2004 the distribution had shifted substantially toward balance and beyond. In 1994, 70 percent of countries had a net negative foreign-currency position with an average weight of minus 27 percent, and over 20 percent were below minus 50 percent. Industrial countries were close to balance, with mean and median weights between zero and 10 percent and 60 percent of them positive. Emerging countries were negative on balance but much closer to zero than poorer developing countries (Figure 3 and Table 4, Section 5.2.1, pp. 17-18). By 2004, 17 percentage points more of the sample had a positive position, the mean and median were minus 7 percent, roughly 10 percent were at minus 50 percent or worse, industrial countries’ mean and median were close to plus 10 percent with 86 percent positive, and emerging countries were on average positive. The authors add the obvious but easily missed caveat: “shifting to a positive net position does not eliminate exchange rate based valuation effects: it simply means that the sign will be positive when the country depreciates against the rest of the world” (p. 18). They also stress that a zero aggregate exposure does not mean no exposure – all but 10 countries were short some individual currency in 2004, 50 percent had a negative weight of 11 percent or more against some currency (half of those largest short positions against the dollar, the rest roughly split between euro and yen), and all but one country was long another currency, with long positions spread across the dollar (33 percent of cases), the euro (28) and the pound (20) plus 16 further currencies (fn. 18, p. 18).

Large: the paper’s own worked example is a 10 percent of GDP loss from a 10 percent depreciation for a typical country. “A negative foreign-currency exposure of 50 percent against the rest of the world means that a 10 percent depreciation would generate a valuation loss of 10 percent times 50 percent times total assets and liabilities divided by GDP… Thus, a country at the average gross position of 200 percent of GDP would experience a 10 percent of GDP loss from such a depreciation” (Section 5.2.1, p. 18). The paper insists the scale term matters independently of the exposure weight: for many industrial countries the exposure weight barely moved over 1994-2004 but “their scale of financial globalization (IFI) has increased considerably, so their overall net long exposure against foreign currencies has increased as a share of the economy” (p. 19). The same logic explains why the euro-area group’s NET FX rose despite a falling exposure weight: “the growth in gross cross-border holdings was sufficiently large to dominate the declining share of foreign currencies in these positions” (Table 7, Section 5.2.2, p. 21).

No. The improvement came from better net foreign asset positions, a shift toward equity and FDI financing of liabilities, and reserve accumulation, not from denominating international debt in domestic currency. Splitting the sample into quartiles by the change in aggregate exposure, the top quartile shows a 34 to 92 percentage point increase in the exposure index, driven by a 16 percentage point rise in the asset share of gross assets and liabilities and a 29 percentage point fall in the foreign-currency share of liabilities without a shift in the asset-side share (Table 5, Section 5.2.2, p. 19). The mechanism is compositional: “FDI and equity are denominated in local currency, so increasing their share of liabilities will lower the foreign currency component of liabilities,” and the top two quartiles saw substantial shifts toward equity-oriented financing, most strongly among emerging and developing countries (Table 6, p. 20). On the asset side, “all quartiles increased the reserves share of total assets. For the top quartile, over 50 percent of the increase in total assets came from an increase in reserves,” and only the non-advanced countries were “truly stockpiling reserves”; for emerging countries it was reserve accumulation that drove the asset-share shift, since their non-reserve net external position was on average negative. The authors’ conclusion is stated flatly: “the shift away from negative foreign currency positions is not coming from borrowing in domestic currency but from the shift towards equity finance and improvements in the net foreign asset position” (p. 20). The euro area is the distinct case, having cut the foreign-currency shares of assets and liabilities by 52 and 42 percentage points respectively – which, combined with an essentially average net foreign asset position, is why euro-area countries saw little improvement in aggregate exposure (p. 19).

Yes for developing and emerging countries, roughly one-for-one; much less so for advanced countries. Regressing the aggregate valuation term on the currency valuation term, both scaled by GDP, the authors note that full offset by local-currency returns would imply a zero coefficient. Instead, “for developing or emerging countries, the ‘pass through’ is approximately one-to-one: a currency gain of 1 percentage point of GDP (according to our measure) is associated with a 1 percentage point aggregate net capital gain,” with R-squared between 0.4 and 0.6. For advanced countries the term remains significant but the coefficient is roughly 0.6 and the R-squared only 0.06 to 0.09, “which suggests that there is some degree of offset by which capital gains via currency movements are partially cancelled out by lower foreign-currency returns” – a difference the authors attribute to advanced countries’ larger equity positions making price valuation shocks more important (Table 9, Section 5.3, pp. 22-23). They flag a measurement complication: some valuation shocks cannot be tied to either exchange rates or market prices, including data revisions, debt reduction schemes and capital transfers, and there may be correlation between depreciations and debt reduction schemes (fn. 22, p. 22).

The paper’s claim is that the shocks are not undone by quick exchange rate reversals – but its own two statements of the evidence differ, and a reader should note the discrepancy. The abstract and introduction say “exchange rate valuation shocks are sizable, not quickly reversed” and that “the autocorrelation of exchange rate valuation shocks is in fact positive” (Abstract; Introduction, p. 2). Section 5.3 reports the underlying regressions differently: “In regressions of VAL on lagged VAL, we find that all three types of valuation effects are essentially stationary. They all have autocorrelation coefficients of nearly zero. Individual country coefficients are quite noisy, but only a handful have point estimates lower than -0.2 (suggesting some reversals) for the exchange rate valuation shocks” (p. 23). Both statements support the paper’s load-bearing conclusion – that there is no systematic reversal, so the wealth change is not a paper gain that unwinds – but they characterise the sign of the autocorrelation differently, and the stronger “positive autocorrelation” phrasing is not the one the results section supports. On magnitudes the paper is consistent: the 75th percentile of absolute currency valuation effects is 2.8 percent of GDP for advanced, 3.8 for emerging and 5.3 for developing countries, “meaning that one in four observations has a shock of these magnitudes,” effects “sizable enough to dominate current account flows in some years” and, depending on market capitalisation, to rival stock-market wealth effects (Table 10, p. 23).

That a common dollar shock produces sharply different outcomes across country groups, and that some developing countries gain from it. Simulating a 20 percent dollar depreciation against all currencies, all countries face trade-weighted appreciations, with emerging markets seeing the largest trade-side shift because of their tight trade relationship with the US. But “it is non-EMU advanced countries that face the largest net financial index change, a greater than 1 percent change in the index and almost 5 percent of GDP loss from valuation.” Non-emerging developing countries “in fact benefit from a dollar depreciation on average,” because their negative dollar positions are large enough that dollar depreciation lifts their net index, producing net financial gains “on the order of 3 percent of GDP.” The authors are careful not to net this against the trade effect: “whether this sufficiently offsets the effects of an appreciating trade-weighted exchange rate is unclear, but it certainly dampens the effect when compared to emerging market countries that lose on both trade and financial dimensions” (Table 11, Section 5.4, pp. 23-24).

Three, stated directly: the analysis is partial equilibrium, the currency positions are estimates built on many assumptions, and derivative hedging is unobserved. On the first: “Our analysis is partial equilibrium in nature, since we effectively treat exchange rate movements as exogenous,” though the authors argue that “understanding why the exchange rate changes does not change the positive aspects of our work – the examination of the wealth effects – but it does have implications in terms of the optimal composition of international portfolios” (Section 1, pp. 2-3). On the second: “While our work represents a dramatic improvement relative to the status quo, it is important to be clear about its limitations… Obviously, estimated data will not be perfectly accurate, nor will every assumption made fit every country perfectly,” with cross-checks made where possible and choices defended in the appendix (p. 6). On the third, the authors address two ways currency weights might overstate true risk exposure. Local-currency asset prices might offset exchange rate movements, but they argue against a full offset: the failure of uncovered interest parity and the profitability of carry trades show bond returns reinforce rather than counter currency movements; equity and FDI returns face conflicting forces (depreciation may boost exporters but also accompanies a slowing economy); and bank loans, deposits, reserves and other items not marked to market “do not have price valuation effects, only exchange rate based valuation effects, so there is no offset for these asset classes” (pp. 5-6). On derivatives, “lack of data means that the extent of cross-border currency hedging is difficult to assess,” but much derivative trade is between domestic residents and so does not change aggregate exposure, Hau and Rey (2006) estimate only 10 percent of foreign equity positions are hedged, and hedging achieved by balancing asset and liability exposures is already captured in the paper’s weights (pp. 6-7).

That “new portfolio balance” models need to model the exchange rate’s dual role, and that the empirical exposures the paper measures should discipline that modelling. The paper connects to the DSGE literature computing optimal international portfolios, noting the general pattern there that a positive domestic productivity shock raises welfare and depreciates the currency – so a negative foreign-currency position is the good hedge – while a positive demand shock raises welfare and appreciates the currency, favouring a positive position (Section 5.2, pp. 16-17). It also notes work relating portfolio configurations to structural differences across economies, such as Mendoza et al (2007), in which differences in financial development make the advanced economy a net debtor holding a long equity position in the developing economy. Its own exploratory cross-section regresses aggregate exposure on GDP per capita, trade openness, institutional quality, country size and an EMU dummy, finding a clear positive relation between GDP per capita and exposure across all samples, and that larger and more open countries have more positive positions – which the authors read through “the ability of larger and more open developing countries to issue domestic-currency liabilities via portfolio equity and FDI channels” (Table 8, Section 5.2.3, p. 22). They present this specification with a caveat about its status: “we lack strong theoretical guidance in formulating this specification” (fn. 21, p. 22). The discussion closes with three modelling implications – the dual role of trade- and finance-weighted exchange rates; the interaction of external wealth effects with domestic sectoral balance sheets, where it may be useful “to establish the conditions under which such valuation movements may have a stabilising influence versus scenarios under which the impact is pro-cyclical”; and the dependence of optimal open-economy monetary and fiscal policy on structural characteristics such as financial development and the contracting environment (Section 5.5, p. 24).