Home Bias in Open Economy Financial Macroeconomics

Home bias is the empirical fact that investors hold “a disproportionate share of domestic equities in their portfolio,” a pattern the authors trace to French and Poterba (1991) and measure via EHBi = 1 - (share of foreign equities in country i’s holdings) / (share of foreign equities in the world market portfolio) (p. 2). By this metric, EHBi = 1 is full home bias and EHBi = 0 matches the basic International CAPM prediction that a country’s investors should hold the world market portfolio. Using IFS/FIBV data, the authors report that “despite better financial integration, the home bias has not decreased sizably: in 2007, US investors still hold more than 80 percent of domestic equities” (p. 1), and that across developed countries the average degree of home bias in 2008 was 0.63 – meaning investors hold roughly a third of the foreign equities the CAPM benchmark implies they should – versus 0.9 in emerging markets, where investors hold only about a tenth of the benchmark amount (p. 4). Home bias has declined gradually with “financial globalization” but “remains high in most countries” (p. 4) and is lower within the Euro Area, consistent with a monetary-union effect (p. 4).

The authors organize the literature into “(i) hedging motives in frictionless financial markets … (ii) asset trade costs in international financial markets … (iii) informational frictions and behavioural biases” (p. 1), all of which are departures from the benchmark of Lucas (1982), in which homogeneous investors across the world hold an identical, fully diversified “market portfolio” of claims to global output. The authors state plainly that they “do not intend to provide a definite answer nor choose among alternative explanations, as they probably all contribute to part of the gap” (p. 1) – the survey’s stance throughout is pluralist rather than adjudicating a single winning theory.

Perturbation methods developed by Devereux and Sutherland (2008a) and, independently, Tille and van Wincoop (2008) made it possible to pin down non-trivial gross portfolio holdings in standard DSGE models, which earlier techniques could not do because “to a first order approximation, assets are perfect substitutes … so portfolio choice is not pinned down” (p. 6). The key insight is that the steady-state (“zero-order”) portfolio can be recovered from a second-order expansion of the portfolio (Euler) equations while only first-order dynamics are needed for the rest of the model, and that first-order model dynamics in turn depend only on that steady-state portfolio (p. 6). The authors flag real limitations: these are local methods valid only near the point of approximation, so they can perform poorly with large shocks, non-stationary wealth distributions under incomplete markets, or strong non-linearities such as borrowing constraints (p. 7), and “for most models … one cannot provide exact numerical methods” to check their accuracy (p. 7).

The zero-order equity portfolio decomposes into three additive terms: a pure 1/2 diversification term as in Lucas (1982), a hedge of non-tradable labor-income risk that is unambiguously negative (pushing toward foreign equity, i.e., short-selling local stock), and a hedge of real-exchange-rate risk whose sign flips with the elasticity of substitution between home and foreign goods (pp. 12-14). Because labor and dividend income are constant shares of output under Cobb-Douglas production, they are perfectly correlated, so “households should short the local stock to hedge human capital risk” – exactly the term derived by Baxter and Jermann (1997) (p. 14). The real-exchange-rate hedging term generates home bias only when the elasticity of substitution φ is below roughly one (Kollmann’s case); when φ is above one it generates foreign equity bias instead, and when φ = 1 exactly, “both equities are perfect substitutes and there is portfolio indeterminacy” (p. 14).

The authors conclude it is “doubtful that the hedging of real exchange rate risk can account empirically for the equity home bias,” because the mechanism requires local equities to have higher returns when local goods are more expensive, and “as shown by van Wincoop and Warnock (2008), the empirical correlation between excess equity returns … and the real exchange rate is very low, too low to explain observed equity home bias” (p. 17). They add that most real-exchange-rate fluctuations reflect nominal exchange-rate movements, which can be hedged directly through currency forward or bond markets rather than through equities (p. 17-18).

Baxter and Jermann (1997) argued, in a paper titled “The International Diversification Puzzle Is Worse Than You Think,” that the home bias puzzle is worse than commonly appreciated: because labor and capital income are highly correlated within countries and only weakly correlated with foreign equity returns in the data, their estimated optimally diversified portfolio implies investors should hold negative domestic equity positions in every country studied (p. 18). Heathcote and Perri (2008) overturns this by relaxing Baxter and Jermann’s single-good, fixed-capital assumptions: with two goods, home bias in investment spending, and endogenous capital accumulation, a positive home productivity shock lowers the relative price of home goods enough that home dividends (net of the now-larger home investment spending) actually fall relative to foreign dividends, even as home wages rise – generating negative comovement between home labor and capital returns, which “implies home equity bias” (pp. 18-19). The authors note this result “relies on two key elements: endogenous capital accumulation and a strong adjustment of relative prices” (p. 18), and that increasing the elasticity of substitution between goods pulls the model’s predicted portfolio back toward Baxter and Jermann’s (p. 19).

With trade in real bonds as well as equities, the survey’s key result is that hedging becomes “conditional”: bonds absorb the real-exchange-rate hedge almost completely (bond return differentials are “almost perfectly correlated” with the real exchange rate), leaving the equity position to hedge only the part of non-tradable income risk that is orthogonal to bond payoffs (p. 19-20, 23). Formally, “the covariance of local equity returns with returns on non-tradable wealth can be positive [as in Baxter and Jermann], [but] this has no implication for the equity portfolio – only the covariance conditional on bond returns matters” (p. 23), and the authors show this conditional covariance “tends to be negative in the data” (p. 23), unlike the unconditional covariance used in equity-only models. The resulting equity portfolio in their model reduces to the sum of a pure diversification term and this conditional non-tradable-income hedge, which is “unambiguously positive and drives home equity bias in the model” (p. 24-25) – and, notably, is “remarkably stable to changes in preference parameters,” unlike the equity-only portfolio (p. 25).

Using quarterly G7 national accounts data (1980Q1-2008Q3), the authors compute both the unconditional and the bond-return-conditional covariance ratios between relative labor income and relative dividends for each country, and find that “conditioning for exchange rate movements has a strong impact”: unconditionally, wages and dividends comove positively in every country – implying a foreign equity bias in an equity-only model – while conditionally (controlling for bond/exchange-rate returns) they comove negatively in every country, implying home equity bias, “both in the model and in the data!” (p. 26). Results are essentially unchanged whether real or nominal bonds (and real or nominal exchange rates) are used to compute the conditioning variable (p. 26).

The authors flag three main limitations. First, the literature has largely stopped at bonds and equities even though “any asset that is traded publicly could affect the equity portfolio” – housing and corporate debt are natural next steps, and Coeurdacier, Kollmann and Martin (2010) show that under Modigliani-Miller, domestic investors optimally hold corporate debt in the same proportion as equity of the same firms (p. 28-29). Second, these models “cannot replicate realistic moments of asset prices and exchange rates” (the equity premium and excess volatility puzzles) and, more specifically, cannot escape the Consumption-Real Exchange Rate anomaly: efficient risk-sharing implies relative consumption should move almost one-for-one, inversely, with the real exchange rate, but “in the data … the correlation is close to zero and if anything Home relative consumption increases when Home relative prices are higher,” and the authors report it is “extremely hard” to lower this correlation even in models with imperfect risk-spanning (p. 30). Third, most of the theoretical literature is confined to two symmetric countries, limiting its ability to exploit the bilateral and cross-sectional variation now available in data such as the IMF’s CPIS holdings (p. 30-31).

The dominant finding, going back to French and Poterba (1991), is that transaction costs “must be much larger than the one typically observed” to rationalize equity home bias – their own estimate and Jeske’s (2001) subsequent estimate both imply implicit costs of several hundred basis points, an order of magnitude the authors flag as “too big to be true” (p. 31-32). A partial counter-finding is Sercu and Vanpee (2008), who find much smaller implicit costs (0.10-0.20 percent per annum) once currency risk, inflation hedging and other factors are controlled for, though the authors caution that all such indirect estimates “suffer from potential statistical uncertainty” given the high volatility of stock returns (p. 32). Further evidence against a pure transaction-cost story comes from Tesar and Werner (1995), who find turnover is actually higher, not lower, for foreign equity holdings than for domestic ones – the opposite of what a simple trading-cost friction predicts (p. 32).

Following Martin and Rey (2004) and Coeurdacier and Guibaud (2009), the authors note that “even small transaction costs may lead to sizable home bias when Home and Foreign stocks are close substitutes: any small transaction cost is amplified if the benefits of diversification provided by foreign assets are small” (p. 32). This connects to Cole and Obstfeld’s (1991) classic result that gains from international risk-sharing can themselves be small, because terms-of-trade movements already share risk internationally even under portfolio autarky; consumption-based welfare calculations of the cost of underdiversification (e.g., Van Wincoop 1999, estimating 1.1-3.5% of permanent consumption over 50 years) are typically much smaller than stock-return-based calculations (e.g., Lewis 2000, estimating 10-30% of wealth), a gap the authors call “an open question” (p. 33).

Building on finance-literature models (Gehrig 1993; Brennan and Cao 1997) in which investors receive a less precise signal about foreign than domestic assets, informational asymmetry raises the perceived riskiness of foreign equity and induces home bias, and generates a “return-chasing” pattern – less-informed foreign investors respond more strongly to price signals – that the authors note finds empirical support. However, Glassman and Riddick (2001) quantify the scale of perceived-riskiness adjustment needed to match observed US home bias and find investors “would have to scale up standard deviations of returns by a factor from 2 to 5,” which the authors flag as “implausibly high” (p. 35). Macro-DSGE applications, such as Hatchondo (2008) and Tille and van Wincoop (2009), extend this logic into general-equilibrium open-economy models, the latter showing that dispersed private information “disconnects stock prices from the currently observed fundamental values” and helps international capital flows forecast future fundamentals (p. 36).

Using Sims’s (2001) rational-inattention framework, Van Nieuwerburgh and Veldkamp (2009) show that “a tiny information advantage is enough to generate significant home bias if investors have a limited capacity to process information,” because learning is self-reinforcing: “the more of an asset the agent owns, the more attractive it becomes to learn about the asset,” so that costly, endogenous information acquisition “amplifies the initial small informational advantage” rather than eroding it – “learning turns out to amplify information asymmetries instead of reducing them” (pp. 36-37). This model class also predicts that assets “learnt about a lot by investors should have lower returns” than a standard CAPM implies (p. 37), and related work (Mondria 2010; Mondria and Wu 2010) extends this to explain both index-based learning strategies and the gradual, decades-long decline in home bias following financial liberalization.

Behavioral accounts range from overconfidence about domestic-asset returns (already flagged by French and Poterba 1991, and documented cross-nationally by Shiller et al. 1991, who find Japanese investors expect relatively higher returns on Japanese stocks and US investors the reverse) to “familiarity” effects (Huberman 2001; Benartzi 2001) and self-assessed investment “competence” (Graham, Harvey and Huang 2009). The authors are explicit that “it remains difficult to disentangle empirically informational frictions linked to distance and/or institutional differences from behavioural biases such as ‘familiarity’ and/or ‘competence’ effects” (p. 38). The most fully worked-out model is Dumas, Lewis and Osambela (2009), in which investors share the same information but differ in how much they trust local versus foreign public signals; the authors note this framework is “observationally equivalent to existing models of segmented markets due to asymmetric information” (p. 39), underscoring how hard it is to separate behavioral from purely informational channels empirically.

Beyond equities, the authors construct parallel home-bias measures for cross-border bond holdings (BHBi) and cross-border bank loans (LHBi) using IMF CPIS/IFS and BIS data, finding bond home bias “of a slightly larger magnitude than the one documented for equity,” averaging 0.75 across developed countries in 2008 (versus 0.63 for equities) and even higher and less declining in emerging markets (pp. 39-41). Bank-loan home bias is “similar” in magnitude to equity home bias and likewise lowest in Europe and highest in Latin America (p. 41). At the fund level, using Thomson Financial Securities data on mutual fund equity holdings across countries (1997-2002), they document “a great deal of heterogeneity both across countries and within country,” with the distribution of domestic-holding shares typically bimodal – funds cluster near fully domestic or fully international, yet a non-trivial share sit in between – which the authors attribute partly to fund mandates that are themselves “clearly not exogenous” (pp. 42-44).

The authors’ concluding assessment is that “the home bias puzzle is now less of a puzzle,” given the combined progress of Open Economy Financial Macroeconomics and rational-inattention models of endogenous information acquisition, and they suggest the field should move toward testing these models against a “broad array” of other portfolio facts rather than fixating solely on the equity home bias statistic (p. 44). They flag several priorities for future work: modeling firms’ capital structure and its interaction with corporate-debt home bias; modeling monetary and fiscal policy jointly with endogenous portfolio choice, since gross cross-border holdings likely affect policy transmission; modeling the portfolio choices of central banks and sovereign wealth funds, given their growing share of international capital flows and their bearing on currencies’ international role; introducing heterogeneous-agent structures to help resolve the Consumption-Real Exchange Rate anomaly; and, especially, modeling financial intermediaries and delegated investment explicitly, since “a large share of household investment is not direct portfolio holdings but intermediated” (pp. 44-46) – a share the authors document rose from 46% of US household equity in 1989 to 62% in 2001.