Return and risk of German open‐end real estate funds

Abstract

Open‐end real estate funds (so‐called ‘Offene Immobilienfonds’) play a major role in the German market for securitized real estate investments. Such funds are pools of money from many investors, which are invested in real estate by special investment management companies. This study seeks to identify the risk and return profile of this investment vehicle (before and after income taxes), to compare it with those of other major asset classes, and to provide implications for their appropriate role in a mixed‐asset portfolio. Additionally, an overview of the institutional architecture and role of German open‐end real estate funds is given. Empirical evidence suggests that the financial characteristics of open‐end real estate funds are in many respects similar to those reported for direct real estate investments. Accordingly, German open‐end real estate funds qualify for medium and long‐term investment horizons, rather than for shorter holding periods.

Keywords:

• open‐end real estate funds
• institutional investors
• asset allocation
• long‐term investments

Acknowledgements

This article was developed at The Research Programme for Real Estate Finance, at the University of Frankfurt/Main. Support by the DB Real Estate and the German Investment and Asset Management Association is gratefully acknowledged. The authors thank the participants of the 10th ERES Annual Conference in Helsinki, those of a workshop at the Center for Financial Studies, Frankfurt, and the anonymous referees for their helpful comments. Opinions remain solely those of the authors.

Notes

For the drawbacks of direct property ownership see e.g. Ball et al. (1998), Hoesli and MacGregor (2000), and Seiler et al. (2001).

See e.g. Hoesli et al. (2004).

See e.g. Hoesli and MacGregor (2000), chapter 11.

The first German real estate mutual fund (the iii Fonds Nr. 1) was launched in 1959.

If, however, the fund units are held in a tax‐qualified individual retirement account, the investment income during the accumulation period is tax‐free, whereas the payments in the payout phase will be fully subject to income tax (deferred taxation).

In order to facilitate the taxation, the investment management company is required to annually publish a note on the composition of fund income at the level of the individual unit certificate showing the various taxable and tax‐free income components.

See Hoesli and MacGregor (2000), p. 59.

A similar run happened at the beginning of the last decade in the case of the Dutch Rodamco crisis (see Helmer, 1997, p. 126). In the German market for open‐end real estate funds the A.G.I. Nr. 1 was closed and merged with another open‐end real estate fund in 1993, because unit holders (mostly institutional investors) asked their money back (see Sebastian, 2003, p. 67). However, in contrast to the Australian and Dutch case this was not a crisis of the whole industry for open‐end real estate funds, but merely a problem of a single fund.

See Maurer and Sebastian (2002) for this point.

This was reported in 2004 in the case of a major German open‐end real estate fund. See Die Welt, 15 October 2004.

However, considering the indications for serial correlation in the money market and real estate fund series (to be seen later), the inferences regarding normality of these series should be interpreted carefully.

See in general Geltner (1989, 1991) and for a recent empirical study including German real estate return series Maurer et al. (2004).

See e.g. Maurer and Reiner (2002).

The analyses in the previous sections showed that some of the time series used in this study are non‐normally distributed and/or autocorrelated. Using a standard t‐test for inferring if correlation coefficients are different from zero at a given level of significance is critical in the presence of non‐normally distributed or autocorrelated time series. To address this, each correlation coefficient, for which both of the underlying series proved to be non‐normal, was tested via a bootstrap BCa confidence interval with 10 000 bootstrap replications. To address the item of autocorrelation, a correction for the t‐test suggested by Dawdy and Matalas (1964, p. 8/87) was applied, which at least allows for control for first‐order autocorrelation.

Hoesli et al. (2004) provide detailed analyses of the correlation structure between real estate securities and other asset classes for Australia, France, the Netherlands, Sweden, Switzerland, the UK, and the USA.

Maurer et al. (2004) showed that, for the UK, USA, and Germany, correlation between aggregated level direct real estate investments and the bond markets are near zero.

See e.g. Campbell and Viceira (2002), Kandel and Stambaugh (1996) and Barberis (2000).