Martijn Boermans | De Nederlandsche Bank
Ian Cooper | London Business School
Piet Sercu | KU Leuven
Rosanee Vanpée | KU Leuven


International portfolio choice , home bias; foreign bias , informational advantage , familiarity , behavioral economics , active share

JEL Codes:

G15 , G18 , G30 , G38 , F3

The views expressed in this text do not necessarily reflect the official positions of De Nederlandsche Bank nor the Eurosystem.

We use the European Central Bank’s new database of European equity holdings by households to investigate the international structure of portfolios, including both investments via funds and direct holdings of individual equities. Home bias is smaller than usually believed. For the direct portfolios, the Cremers and Petajisto (2009) ActiveShare measure comes with larger investments in the destination country and better returns, while the fraction of the country sub-portfolio invested via funds has no effect on either returns or the size of the overall. This fits the hypothesis of informational advantages rather than a (behavioral) familiarity bias.


Europe’s financial markets remain strongly bank-oriented, and therefore more vulnerable to banking crises than market-oriented economies. In 2015 the European Commission launched a plan to bolster Europe’s stock and fixed-income markets, making them both larger and more integrated. However, “most indicators that measure the size of capital markets, and the degree to which they are integrated, have moved sideways”, The Economist reports.1 In banks’ bond portfolios, ‘home bias’—investors’ tendency to massively overweight the own-country assets—has improved, but not for a high-debt country like Italy, creating the familiar doom-loop link between banks and governments. In stock markets, home bias in investors’ equities portfolios remains pervasive, after an improvement within the Eurozone at the time the common currency was introduced. Similarly, among foreign markets, those that resemble the home market best tend to be relatively over-represented (‘foreign bias’).

One first question this raises, from a policy perspective, is whether the stock-portfolio biases are as pronounced as some data bases suggest. Commonly cited IMF portfolio survey data are rather imperfect because they lump together all investments into e.g. Luxembourg-run mutual equity funds, whether by home investors or not, with investments in Luxembourg companies’ equity. The ECB’s quarterly Securities Holdings Statistics-Sectoral (SHS-S) data, based on reports by all financial institutions from euro-area countries, offers a welcome and much more precise alternative. The ECB provides detailed data on holdings, for each worldwide stock-listed companies (about +50.000), grouped by the investor’s country of residence and, within the country, by investor class (banks, insurance mutual funds, households) and, within the households, all relevant nationality segments. For instance, we know how many shares of Danone or Citi are held by, say, Irish households living in Portugal. We also know how much each of these groups invest in each of 25,000 mutual funds. We can therefore study, for each of these resident groups in the reporting countries, the size of direct holdings in each company, plus the holdings via funds, and thence their total holdings. Most of the 800m potential [investor-group x company x quarter] positions are of course zero, but we still end up with, on average, 165m non-zero cells per year. We can, of course, aggregate these holdings across all stocks per destination countries—the usual numbers studied in the home/foreign bias literature—but we can also look at how balanced the positions are. For instance, do Chilean investments by French households living in Austria resemble a well-diversified portfolio, or are they focused on a few Chilean companies? We can check whether ‘locals’ (e.g. Dutch living in the Netherlands) differ from ‘immigrants’ (non-nationals living in the Netherlands) or ‘emigrants’ (Dutch living abroad). We can verify whether direct stock holdings and indirect ones (through mutual funds) are differently biased, and we are able to compare total bias with the standard IMF results.

The second question relevant for policy is what should be done, if anything, about the twin biases. Whether policy measures are useful at all depends on what underlies the phenomena. In one view, they reflect a behavioural feature: the natural familiarity investors experience with respect to local companies is said to lead to overconfidence and an illusion of control. A radically different view is that investors feel they have an informational advantage about stocks in the local market, and by extension, about stocks from markets that resemble home: they know more about them, and understand the context better. If that is the mechanism, not familiarity bias, the natural next question is whether that advantage is genuine or imaginary. If the information hypothesis holds, there is no immediate cause for concern, and nowadays little is needed to improve information provision in the first place. Behavioural biases, in contrast, are likely to harm rather than help market efficiency.

The two competing views have different testable implications. Familiarity is country-related, not stock-specific, and should therefore be associated with portfolios that (a) heavily load on indirect investments—via mutual funds, that is—while (b) any direct holdings should be well-diversified. Information trading, in contrast, is about specific stocks or perhaps industries, so it should be associated with indirect positions that are smallish and with direct holdings that are quite different from a neutral, market-cap-weighted country portfolio. So if we would see that stronger foreign bias is associated with smaller indirect holdings and less balanced portfolios, that would be consistent with an information explanation. To verify whether the informational advantage is genuine or just wishful thinking, we can then compute excess returns on each group’s country portfolio, over and above the risk-free rate and the part reflecting priced factors, and again check whether those excess returns are stronger when direct portfolios are more specialized and indirect holdings smaller.

The two views also have very different welfare implications. If the bias is based on informational advantage the resulting underdiversification will be compensated by excess returns. In contrast, if it is based on familiarity bias there will be no such compensation, and the bias will lead to a loss of risk-adjusted return compensated only by whatever psychological benefit is achieved by familiarity.


In a full-information, no-frictions world investors would strongly diversify internationally and hold portfolios close to the world-market portfolio, with country weights proportional to the country’s overall ‘cap’ (market capitalisation) scaled by the cap of all the world’s stocks, denoted as wi2. Accordingly the ‘raw’ measure of home bias is commonly defined as the gap between the weight country-i stocks have in the stock portfolio of local investors versus the country’s world-market weight wi. This number is known to be universally positive, and often quite large. We standardise this raw measure by the square root of wi (1- wi), following Cooper, Sercu and Vanpée (2013), but all results reported below remain present when the scaling is omitted.

When measuring raw foreign bias, we eliminate the home-bias effect by looking at the foreign-stock sub-portfolio in itself. So if half of i’s assets are home stocks, all foreign-stock weights are doubled so that the rescaled foreign weights sum to unity. These weights are then again contrasted with the weights the countries would have in the comparable neutral portfolio, i.e. the world weights similarly rescaled after the elimination of country-i (home) stocks. Standardisation is similar to that for home bias.

Now consider the portfolio of country-j stocks chosen by investor group i, like Czech stocks held by Canadians residing in Italy. One relevant metric is funds’ weight, which tells us how much of that is held via mutual finds rather than directly held individual stocks. The second one is the Cremers and Petajisto (2009) active share statistic, defined as half of the mean absolute deviation between stock k’s weight in i’s portfolio and k’s cap-based weight that would have been chosen by a passive investor. Active Share equals zero for a perfectly neutral portfolio and approaches unity when all if i’s money is put into a single, absolutely tiny stock.

Part of our work consists of documenting the size and characteristics of the biases using the new, superior data, including look-through investments via funds, and the differences between direct and indirect position or between portfolios selected by ‘native’, ‘emigrant’ and ‘immigrant’ groups. Lastly we formally investigate the familiarity versus information issue via regressions. We focus on foreign bias, because for home bias we just have one clear (and large) investor group per country; that is, we would have just 13 observations per quarter, and each such number would be aggregated over many investors (viz. all national residents). By looking at e.g. immigrants’ portfolios we have many more observations less subject to aggregation bias. The main analysis is based on two regressions, that relate either foreign bias or excess return to Active Share and Funds’ Weight plus a set of controls commonly used in this literature (like both countries’ sizes and income/capita, bilateral trade links and physical distance, and so on).


Home and foreign bias: new stylized facts

  • ‘Native’ investors’ domestic investments are mostly direct, but the funds they do hold are very diversified internationally. In the direct part, raw home bias in the 13 reporting countries is 70 points, but the overall bias, after adding look-through positions via funds, is just 45 percentage points. Figure 1 shows results per country.

Figure 1

  • What is ‘home’ for migrants? In terms of home bias, ‘emigrants’ do not seem to be much biased towards their country of nationality; relatedly, ‘immigrants’ behave very much like locals of their residence country. So alongside natives we can study the various immigrant groups, with the advantage that those offer less aggregate data on residents’ portfolios.
  • Direct domestic holdings are very actively managed, with Active Share of about 0.45 for the ‘natives’ groups and 0.65 for the (smaller) migrants groups. This suggests information effects, not familiarity.
  • Foreign holdings are mostly through funds, so that overall Active Share in the foreign sub-portfolios is low. This is consistent with any information advantages being mostly home-related.

Foreign bias: information versus familiarity effects?

  • In the Foreign Bias regressions with direct Active Share is the regressor, the coefficient is positive and significant whether one considers over- or underweight portfolios. In addition, the coefficient for Funds’ Weight is significant and negative, in both under- and over-weighted positions. All this is consistent with the information-based choice model.
  • The indirect Active Share does not seem to pick up anything useful, so there is not an informational motivation for the foreign bias within the holdings of funds. Given that these are a large proportion of the international holdings, and they exhibit foreign bias, the informational explanation cannot be a complete explanation for foreign bias for the overall portfolios.
  • In all Foreign Bias regressions, Funds’ Weight is clearly negative. So the shift between funds and direct holdings is consistent with (perceived) information effects whereby investors are overweighting countries in which their positions are actively managed.

Excess returns on foreign investments: information versus familiarity effects?

  • The excess return on direct positions is clearly positively affected by Active Share in all specifications. The relationship with Funds’ Weight is either insignificant or weakly significant and positive. This is in line with information effects.
  • A dummy indicating a positive foreign bias is strongly positively related to excess returns, as the information hypothesis predicts. Puzzlingly, though, the relation between Excess Return and Foreign Bias’s size (rather than its sign) is muddled and hard to interpret under either hypothesis.
  • Two variables that are commonly highly significant in foreign bias regressions, GDP and distance, are insignificant in all specifications. Thus they cannot be proxies for information that is not captured in the level of foreign bias.

Concluding remarks

Portfolio bias should be studied at two levels: the passive part that occurs via mutual finds, and the direct part where stocks are selected by the investors themselves. Our analysis indicates that, in their direct investments, households seem to select stocks based on informational advantages and not so much on the basis of familiarity: overweights (positive foreign bias) are associated with higher Active Share numbers and smaller Funds’ Weight numbers, and these portfolios do provide better returns. In the indirect mutual fund positions in stocks, the excess returns are virtually zero, whereas on the direct stock holdings households gain excess returns, especially for more actively managed portfolios. Whether the profitable active positions are present in most portfolios versus held mostly by a subset of investors or not is hard to tell from our data.

An important part of the context for our results is the continued rise of passive investment strategies in the mutual fund sector where outperformance of the benchmark returns is neither expected nor rewarded. Passive investment strategies via tracker or exchange-traded funds are supported by theoretical and empirical insights from academia: such index funds offer welcome diversification, low ‘churn’ costs and modest management fees. When applied internationally, passive investing can reduce foreign bias if it is used to achieve full international diversification. However, our results show that this effect is incomplete, because funds are also used to hold foreign equities where the holdings exhibit familiarity bias.

Active investment—‘speculation’ in the F. Black (1995) sense—is not naturally done via funds: indirect investment would ignore the end investor’s own information and insights, and empirical studies show that, in funds, trading costs and management fees outstrip any gains from active management. Our finding here is that, despite the rise of passive investment, active selection remains alive, explains part of portfolio bias, and is successful. That success is not just privately relevant, but also socially so: good active management contributes to market efficiency and improves market efficiency.

We are grateful to Pedro Barroso, Harjoat Bhamra, Maurice Bun, Ruth Giimeno, Marcin Kacpercyk, Bharat Parajuli, David Schumacher, Giorgia Simion, and participants at the 11th Portuguese Finance Network Conference, the 34th Australasian Finance and Banking Conference, 2022 European Winter Finance Summit, 2022 Global Finance Conference,2021 World Finance and Banking Symposium, and the DNB 2022 Research Seminar for helpful comments on the Boermans, Cooper, Sercu and Vanpe e (2022) working paper.


Black, F. (1995). Hedging, speculation, and systemic risk. Journal of Derivatives, 2(4).

Boermans, M. A., Cooper, I. A., Sercu, P. and Vanpée, R. (2022). Foreign bias in equity portfolios: Informational advantage or familiarity bias? DNB Working Paper 742.

Cooper, I., Sercu, P., Vanpée, R., et al. (2013). The equity home bias puzzle: A survey. Foundations and Trends in Finance, 7(4):289416.

Cremers, M. K. J. and Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. The Review of Financial Studies, 22(9):3329–3365.

About the authors

Martijn Boermans

Martijn Boermans is Senior Economist at the Dutch Central Bank (DNB) and holds a PhD in Economics from Utrecht University. His main fields of research include banking and finance, international trade and sustainable finance. His current work focuses on analyzing granular portfolio holdings data from the ESCB securities holdings statistics (SHS) with various applications, recent working papers and publications.

Ian Cooper

Ian Cooper is Professor of Finance at London Business School. He currently teaches corporate finance and valuation. He has also held visiting positions at the University of Chicago and Australian Graduate School of Management. Ian carries out research on corporate finance, valuation, and international finance. His research has appeared in top academic journals such as the Journal of Finance, Journal of Financial Economics, and Review of Financial Studies. He disseminates the practical implications of his work through publications in leading practitioner journals such as Journal of Applied Corporate Finance and Oxford Review of Economic Policy, and in published consulting reports. He has taught across the MBA, EMBA, and Masters in Finance programs, as well as Executive Education. For many years he directed and taught the Corporate Finance Evening and Modular Programs. The quality of his teaching has been recognised by multiple teaching prizes. Ian has more than 40 years of international consulting experience on valuation, corporate finance, and cost of capital. He has advised companies, financial institutions, regulators, and governments around the world. He has appeared as an expert witness in international tribunals and high courts .

Piet Sercu

Piet Sercu is Professor of International Finance at KU Leuven. He holds the degrees of Business Engineer, Master of Business Administration, and Doctor in Applied Economics from KU Leuven. He also taught at the Flemish Business School in Brussels and held visiting-faculty appointments at New York University, Cornell University, the University of British Columbia, London Business School, the Université Libre de Bruxelles, TIAS-NIMBAS (Utrecht), and KU’s sister university in Louvain-la-Neuve, Université Catholique de Louvain. His early research focused on International Asset Pricing with real exchange risk and inflation risk. He also did some work on corporate take-over models and lending but has recently returned to international finance and hedging. He has published in the Journal of Finance, Review of Finance, Journal of Banking and Finance, Journal of International Money and Finance, European Economic Review, Journal of Empirical Finance, Journal of Business Finance and Accounting, and many other journals. He is a member of the standing reading committees for WFA, EFA, EFMA and WFC. Jointly with Raman Uppal, LBS, he won the 1995 Sanwa Prize for a monograph, International Finance, Exchange Rate Volatility, Trade, and Capital Flows under Alternative Currency Regimes, published by Cambridge University Press in 2000 and 2006. In 2010 he won two more best-paper awards, related to Risk Management and Corporate Governance, respectively. Sercu and Uppal have also produced International Financial Markets and The Firm (International Thomson Publishers, Cincinnati-London, 1995), the forerunner Piet’s 2009 solo book, International Finance, Theory to Practice (Princeton University Press). On and off, he works on a text on Asset Pricing. Emeritus `in pay only’ since October 2016, he continues teaching and research much as before, and works on/off on a new edition of International Finance.

Rosanee Vanpée

Rosanee Vanpée holds a PhD in Applied Economics from KU Leuven. She has worked in the financial industry and as professor of International and Sustainable Finance at KU Leuven. Rosanne is an expert in valuation, optimisation problems and econometrics. Based on her interest in mobility and transport issues, she followed courses on environmental and transportation economics. At TML she works on social benefit analyses and discrete choice modelling. She investigates the social, economic and financial impact of investments in transportation. Since february 2021 she works as a researcher for TML.

More on these topics