پختگی سرمایه گذاران و ریسک پذیری

Using investment policy data of 857 Dutch pension funds during 1999–2006, we develop three indicators of investor sophistication. The indicators show that pension funds’ strategic portfolio choices are often based on coarse and less sophisticated approaches. First, most pension funds round strategic asset allocations to the nearest multiple of 5%, similar to age heaping in demographic and historical studies. Second, many pension funds invest little or nothing in alternative, more complex asset classes, resulting in limited asset diversification. Third, many pension funds favor regional investments and as such do not fully employ the opportunities of international risk diversification. Our indicators are correlated with pension fund size, in line with the expectation that smaller pension funds are generally less sophisticated than large pension funds. Using the indicators for investor sophistication, we show that less sophisticated pension funds tend to opt for investment strategies with less risk.

During the recent financial banking and sovereign debt crises pension funds sustained huge investment losses. The crash in equity prices, coupled with a dramatic decline of long-term interest rates used to discount liabilities, slashed pension funds’ funding ratios (defined as total assets divided by discounted pension liabilities), with only limited relief from increased bond prices. In 2008 alone the market value of total pension assets in the Netherlands dropped by more than 17%. Together with the impact of lower discount rates, the crisis caused the funding ratio to fall in that year by no less than 49% points. Strikingly, however, sustained losses varied considerably across pension funds, illustrating considerable differences among pension fund’s investment policies. These losses have severe consequences since in many countries pension funds play a central role in investing pension savings and providing old age benefits. This is particularly evident in the Netherlands where the assets of pension funds exceed GDP. Most Dutch pension funds now face significant funding gaps and are forced to increase premiums, cut wage or price indexation and, in a number of cases, even to cut pension rights. Evidently, these investment losses have profound implications and have raised questions as to risk taking by pension funds and the quality and sophistication of their investment policies. For pension funds, determining the asset allocation strategy is the most important decision in the investment process. Setting the optimal asset allocation strategy involves two decisions. First, the level of risk preference must be determined in line with the funding ratio and preferences of pension scheme participants and sponsor companies. Second, the allocation of investments to different asset classes should be chosen to maximize expected returns, given a pension fund’s liabilities and its risk preference. Both tasks are highly complex and it is to be expected that the expertise and abilities of different investors in performing them will vary. We examine pension fund investors’ sophistication in setting an optimal asset allocation (task 2) and how this relates to their risk preferences, expressed in terms of risky investments (task 1). A major contribution in the finance literature on optimal asset allocation is the two-fund separation theorem, which prescribes investors to hold an optimal portfolio of risky assets in combination with the risk-free asset (Tobin, 1958). This optimal portfolio should be mean–variance efficient, implying that for a given expected return, no additional diversification can lower the portfolio’s overall risk (Markowitz, 1952). These theorems are building blocks of CAPM, which states that there is only one optimal risk portfolio, that is, the market portfolio (Sharpe, 1964). If this is the correct model, asset allocations for investors with different risk preferences should be simply different linear combinations of the riskless asset and the market portfolio. This implies that investors, including pension funds, should keep the ratio of bonds to equities and other asset classes unchanged across all portfolios and vary allocations to the risk free asset, reflecting varying risk preferences. The finding that investors hold different proportions of risky assets – including the ratio of bonds to equities – conflicts with the two-fund separation theorem and is called the Asset Allocation Puzzle (see also Canner et al., 1997). While we concern ourself with institutional investors, the literature on the sophistication of asset allocation decisions has mostly focused on private investors (individuals or households). Empirical research has shown that private investors invest in ways that are hard to reconcile with standard theory and that have been labeled investment mistakes (Campbell, 2006, Calvet et al., 2007, Calvet et al., 2009a and Calvet et al., 2009b). Private investors often use simple rules of thumb in allocating their wealth across asset classes, resulting in suboptimal investment portfolios. The behavioral finance literature classifies such suboptimal investment decisions as behavioral biases or cognitive errors. Individuals use heuristics, or rules of thumb, because they have limited attention, memory, education, and processing capabilities. A number of papers have shown that individual investors often rely on simple asset allocation rules. Examples of such rules are asset allocations that tend to be either zero or 100% in equities (Agnew et al., 2003) and investor’s use of the 1/n rule to allocate their money among the n funds they invest in (Huberman and Jiang, 2006). Benartzi and Thaler (2001) show that some private investors use the 1/n rule to allocate investments equally among eligible investment funds offered in pension plans and, consequently, that the equity allocation of investors is influenced by the proportion of stock funds offered. The natural conclusion is that the use of heuristics can lead to suboptimal asset allocation by private investors.1 Other recorded investor mistakes are (i) insufficient diversification ( Calvet et al., 2007 and Goetzmann and Kumar, 2008), (ii) inertia ( Agnew et al., 2003, Campbell, 2006 and Calvet et al., 2009a) and (iii) holding of losing stocks and selling winning stocks ( Dhar and Zhu, 2006 and Calvet et al., 2009a). The tendency to round figures coarsely or to choose attractive numbers is also documented in a number of demographic and historical studies. For instance, self-reported age data in countries or periods characterized by low average levels of education often show high frequencies at attractive, ‘round’ numbers. This phenomenon is called age heaping. Individuals with limited knowledge about their age are found to have a higher propensity to choose a ‘plausible’ number. These individuals do not choose random numbers, but instead have a systematic tendency to choose attractive numbers, particularly those ending in 5 or 0. Age heaping is reported for a number of data sources, including census returns, tombstones, and tax data. Demographic studies have shown that age heaping is correlated to education (e.g. Bachi, 1951), income (e.g. Myers, 1976), illiteracy (Budd and Guinnane, 1991) and, more generally, human capital (A’Hearn et al., 2009). While there is a growing literature documenting behavioral biases of private investors, much less is known about professional parties. Institutional investors are generally considered to be more sophisticated than private investors and are therefore assumed to invest more optimally. A number of theoretical papers argue that more sophisticated investors suffer less from cognitive biases or irrational behavior (e.g. Banerjee, 1992, DeLong et al., 1990, Hirshleifer et al., 1994 and Shleifer and Summers, 1990). However, there is little empirical evidence documenting (i) the investment behavior of institutional investors or (ii) how this behavior is influenced by their level of sophistication. To fill this gap in the literature, we study the investment behavior of institutional investors with varying degrees of sophistication. Scale advantages should enable large pension funds to hire competent experts and consultants and spend more time and resources on optimizing their investment policies. Consequently, large pension funds should have a lower propensity to use heuristics in determining their asset allocation, but should instead use more advanced rules to guide investment policy. The more sophisticated investors are also expected to be more knowledgeable about the range of investment options available to them, and consequently to have a larger proportion of investments in other assets than bonds and equities. These factors should enable more sophisticated pension fund investors to apply better asset class allocation strategies than those of less sophisticated pension funds. The influence of sophistication on risk taking is not self-evident. Less sophisticated investors may underestimate risks and consequently take more risk by investing in high risk-high (expected) return assets. Alternatively, less sophisticated investors may be more risk shy, thus compensating for weaker risk management skills, e.g. the ability to measure and control risk and implement diversification strategies. The latter conjecture is confirmed by previous research, showing that risk tolerance in individuals is negatively correlated with financial knowledge and education (Grable, 2000). We hypothesize that, by analogy, the sophistication of institutional investors correlates also positively to risk taking. We investigate the investment policies of 857 Dutch pension funds during the 1999–2006 period. At the end of 2010, total pension fund assets in the Netherlands amounted to some € 775 billion, or 132% of GDP, ranking the Dutch pension system in terms of the asset-to-GDP ratio as the largest in the industrial world.2 We find that pension funds’ asset allocation policies often seem to be relatively simple and that they vary widely, in line with the asset allocation puzzle. This raises the question whether all pension funds implement optimal asset allocation strategies, given their specific profiles and preferences. To investigate this, we develop three measures of sophistication. The first measure assumes that less sophisticated pension funds are less knowledgeable about their (unpublished) optimal asset allocation, or use human judgment more, and are therefore more likely to choose plausible figures rather than the outcomes of detailed calculations. For example, they may use multiples of 5% to set their strategic asset allocation. The strategic investment allocation reflects pension funds’ (unpublished) investment objectives, which they report to their prudential supervisor, De Nederlandsche Bank. The strategic asset allocation must meet supervisory requirements. The actual asset allocation may depart from the objective as a result of asset price shocks, since pension funds do not continuously rebalance their portfolios (Bikker et al., 2010). We find that most pension funds do, in fact, apply such a coarse approach in allocating wealth to investment classes. This finding is similar to age heaping found in sociological and historical studies, where it is considered an indication of limited education. Our second measure records how much pension funds invest in alternative, more complex asset classes such as commodities and real estate (versus more simple classes such as money market and mixed asset funds), thereby improving asset diversification.3 We find that pension funds that apply rounding to multiples of 5% tend to diversify less to such more complex asset classes. Third, we examine ‘home bias’ and find that many pension funds favor regional investments, thereby limiting international diversification. We also find that all three indicators are correlated to pension fund ‘size’, indicating that smaller pension funds are generally less sophisticated than large funds, which is in line with our expectation. In accordance with the asset allocation puzzle, we observe for Dutch pension funds that there are large differences in asset allocation strategies across pension funds. Specifically, relative holdings of bonds and equities, investments in more complex asset classes and international diversification all vary significantly. Whereas specific conditions such as size (reflecting scale economies with respect to e.g. asset management and risk management), funding ratio, age distribution of participants, type of pension plans or type of pension fund contribute to this spread (Bikker et al., forthcoming-a), the variation remains largely unexplained. An important question is whether pension fund investors’ sophistication influences risk taking. It would be a rational risk-management strategy for pension fund investors with less financial expertise to reduce exposure to risks that are not well understood. We investigate the impact of sophistication on risk taking by estimating a model for the strategic bond allocation, where our measures of sophistication are added as explanatory variables. The empirical results indicate that less sophisticated pension funds have a significantly lower risk profile, investing more in bonds and less in equities. There are at least two reasons why the pension sector in the Netherlands provides an ideal setting to study the impact of investor sophistication on risk taking. First, total assets under administration, our measure of the size of pension funds, which may be related to sophistication, varies widely. Pension funds range in size from small institutions – with assets below 100 million euro (almost two-thirds of the funds) – to very large institutions with assets of more than 100 billion euro. The variation in terms of participants is also wide, from less than 100 participants (5% of institutions) to more than a million participants. Large institutions include industry-wide pension funds such as ABP and PFZW, which are among the biggest in the world. Small institutions are mostly company funds that provide pensions for the employees of a single company. Second, De Nederlandsche Bank collects comprehensive data on the investment policies of all these institutions, which allows us to study their asset allocation strategies. This article is organized as follows. Section 2 describes our dataset, while Section 3 develops three measures of sophistication in pension funds’ investment behavior and examines their mutual connection and relationship to size and other characteristics of pension funds. Section 4 investigates the influence of investment sophistication on risk taking. Section 5 provides an update of our approach for 2007–2010 as a robustness test, while the last section concludes.

We examine the impact of investor sophistication on risk taking. We focus on pension funds since their size and other characteristics vary widely and comprehensive data is available. To measure investor sophistication, we construct three measures of the sophistication of pension funds’ investment policies. The first indicator gauges the use of attractive, but imprecise, numbers for the strategic allocation of assets to both equities and bonds. Most pension funds in the Netherlands apply such rule of thumb, using particularly multiples of 5%. This supports the observation that in current practice, asset allocation does not follow directly from optimization of ALM models. Rather, it is determined by human judgment, given results from ALM studies. The second indicator is the use of alternative, complex investments other than equities and bonds, as an instrument to diversify the investment portfolio. We observe that many pension funds invest little in alternative, more complex asset classes, suggesting suboptimal portfolio diversification. The third indicator is home bias in the equity investment portfolio. We show large differences in terms of relative investments in the euro area, suggesting suboptimal international diversification in many pension funds. We find that these three measures correlate with pension fund size indicating that smaller pension funds tend to be less sophisticated than larger ones. Nevertheless, investment sophistication contributes independently to the explanation of risk aversion, showing that investment expertise also varies among pension funds in the same size class. These results suggest that the asset allocation policies of many pension funds, particularly small ones, are suboptimal. A notable finding is the huge variation in asset allocation practices across pension funds, in a broader context also referred to as the asset allocation puzzle. Part of this variation can be explained by the pension fund’s size, its type of pension plan, its preference indicators, such as assets per participant, participant age distribution and the funding ratio and, finally, its governance type. In addition, we find that all our indicators of investor sophistication are highly statistically significant. Even when controlling for size, sophistication and other fund-specific variables, pension funds make significantly diverging portfolio choices. We believe that this reflects widely varying views regarding the optimal investment mix. It seems likely that differences in risk-return assumptions for the various asset classes, in the level of expertise of pension fund investment managers and in personal preferences of pension boards also play an important role. The analysis of this latter phenomenon is outside the scope of this article, but is suggested as an interesting topic for future research. Our findings suggest that further consolidation of the Dutch pension sector, by mergers or increased cooperation (e.g. in so-called general pension institutions, which can administer pensions plans for several companies or industries) may contribute to improve the sophistication of pension funds’ investment policies. Such benefits of consolidation are in line with previous studies, in which we find a negative correlation between the size of pension funds and the administrative and investment costs per participant ( Bikker and de Dreu, 2009 and Bikker et al., forthcoming-b).