Analyzing The Herding Behavior Among Pension Funds
De Nederlandsche Bank
Up 60% In 2020, Axon Capital Says These Tech Stocks Are Value Stocks
Axon Capital was up more than 60% for the first 11 months of 2020 after making some changes to deal with the year's challenges. In his delayed third-quarter letter to investors, which was reviewed by ValueWalk, Axon's Dinakar Singh noted that the year was not only "incredibly stressful" but also "successful." Q4 2020 hedge fund Read More
University of Amsterdam
De Nederlandsche Bank
De Nederlandsche Bank
March 1, 2016
This paper uses unique and detailed transaction data to analyze herding behavior among pension funds. We distinguish between weak, semi strong and strong herding behaviour. Weak herding occurs if pension funds have similar rebalancing strategies. Semi strong herding arises when pension funds react similarly to other external shocks, such as changes in regulation and exceptional monetary policy operations. Finally, strong herding means that pension funds intentionally replicate changes in the strategic asset allocation of other pension funds. Without an economic reason. We find empirical evidence supporting all three types of herding behavior in the asset allocation of large Dutch pension funds.
Analyzing The Herding Behavior Among Pension Funds – Introduction
This paper uses unique and detailed transaction data to analyze herding behavior among institutional investors using a rebalancing model based on Calvet et al. (2009) in combination with a spatial estimation approach. Nofsinger and Sias (1999) define herding as a group of investors trading in the same direction over a period of time. In order to analyze this thoroughly we distinguish between weak, semi strong and strong herding behavior. Weak herding is related to the information motive in the literature, semi strong herding to the regulation motive and strong herding to the reputation motive. We document empirical evidence to support all these types of herding in the asset allocation of large Dutch pension funds. Our findings have potential implications for policy makers who are interested in financial stability. Whereas weak herding can contribute to financial stability, strong herding is a risk for financial stability if pension funds deliberately replicate each other investment strategies without economic reason. Furthermore, regulators need to be aware that semi strong herding might imply that pension funds react in a similar way to regulatory changes.
Global asset portfolios of institutional investors, such as pension funds, have grown substantially over the past decades. Economic and financial policy makers around the globe have therefore become increasingly interested in the factors driving the allocation of these assets. One of the main motivations behind asset allocation decisions that receives increasing attention from global policy-making institutes is investor herding behavior. The IMF does multiple studies on this phenomenon, e.g., Bikhchandani and Sharma (2001); Papaioannou et al. (2013); Jones (2015); Cipriani and Guarino (2014). Also the World Bank analyses herding behavior in Raddatz and Schmukler (2011), as well as the Federal Reserve (Chari and Kehoe, 2002; Cai et al., 2012; Chari and Phelan, 2014) and the Bank for International Settlements (Borio et al., 2001; Nirei et al., 2012).
A key reason why these institutions study herding is its potential implications for financial stability. EIOPA recently provided evidence that pension funds contribute to financial stability as a result of rebalancing strategies (EIOPA, 2016). Since most pension funds aim for a more or less fixed asset allocation within a narrow bandwidth, they typically will buy equities following a period in which the equity allocation decreased. The latter will be driven by relative price effects or exchange rate effects in the prior period(s). The Office of Financial Research in the United States identifies asset manager’s herding as one of the key vulnerabilities to financial stability (Elliot, 2014). If asset managers enter, e.g., into fire sales simultaneously, this can have an amplifying effect on asset price volatility. The Bank of England recently also comments on this phenomenon, relating it to the fact that more pension funds have delegated the management of their assets to external parties (Haldan, 2014). This outsourcing gives rise to the question whether pension fund’s asset allocation decisions are interdependent.
We specifically look at herding behavior among pension funds that, because of their size, are important institutional investors in financial markets. On the one hand, pension funds are long term investors that are able to pursue an optimal long term investment strategy to the best interest of the pension fund’s beneficiaries. This may also contribute to financial market stability as pension funds can offer liquidity in times of financial markets stress. On the other hand, pension funds are typically constraint investors, e.g., by the size and the nature of the liabilities, the risk preferences of the key stakeholders and by external regulation. Pension funds can also feel a constraint from peer group pressure. They may want to invest closely in line with other pension funds to avoid the reputation risk of having to report strongly deviating investment returns.
This paper distinguishes between three types of herding. We define weak herding as the result from the fact that pension funds have similar rebalancing strategies. Most pension funds operate this way (Bikker et al., 2010; Gorter and Bikker, 2013; Calvet et al., 2009). This behavior is inherent to the investment strategy of pension funds and the transactions resulting from the rebalancing strategy are not obviously a form of herding in the sense that pension funds deliberately mimic the transactions of other pension funds. This unintentional or spurious form of herding occurs because groups face similar decision problems and information sets and take similar decisions (Bikhchandani and Sharma, 2001). Semi strong herding arises if pension funds react similar to external shocks, e.g., changes in pension fund regulation. Sias (2004) and Andonov et al. (2013), e.g., show that regulation can have a significant impact on pension fund’s investment decisions. We define strong herding as a case in which pension funds intentionally copy the investment decisions of other pension funds, without an economic reason. This could, e.g., be the case if a group of pension funds follow changes in the strategic asset allocation of another pension fund or a group of pension funds. In this type of herding, an informed agent follows the trend even though that trend is counter to his initial information about the asset value (Avery and Zemsky, 1998). Whereas weak herding can contribute to financial stability, strong herding is a risk for financial stability.
This paper seeks to shed light on herding behavior among Dutch defined benefit funds. The Dutch pension system is an interesting case study for several reasons. First, it is relatively large in terms of size: total assets represent roughly twice the size of GDP of the Netherlands. The investment behavior of these pension funds is therefore of significant importance to Dutch financial stability. Second, during the recent financial crisis, most pension funds in the Netherlands suffered considerable decreases in their funding ratios. Indeed, pension fund’s funding ratios (as defined by the ratio of total assets over liabilities) moved largely in tandem. This is fuelled by the impact of changes in the term structure of interest rates on the value of the liabilities. But also the assets have been hit in a similar way as pension funds all have very broadly diversified investment portfolios. The returns will therefore be very similar.
We examine the extent to which these pension funds follow one another in terms of changes in their asset allocation. We use a unique dataset from De Nederlandsche Bank (DNB), containing monthly transaction data of large Dutch occupational pension funds across a period from January 2009 until January 2015. To test our hypotheses, we employ an econometric specification based on a rebalancing model in combination with a spatial estimation approach. The latter, although common in the political economy literature (see, e.g., Beck et al. (2006); Franzese and Hays (2007)), is to the best of our knowledge a novelty in the pension economics literature. This approach enables us to estimate the spatial dependence of pension fund’s equity and bond allocations. We also check the robustness of our results using an alternative model specification based on the Error Correction Model (Engle and Granger, 1987).
The remainder of this paper is organized as follows. Section 2 reviews motivations in the literature for herding behavior among asset managers. Section 3 introduces the hypotheses we will test, while Section 4 describes our data. In Section 5 we lay out the model for our empirical analysis. The results are discussed in Section 6. In Section 7, we replicate the analysis using an alternative regression model to check for robustness of the obtained results. Section 8 concludes the current paper.
Motives for herding behavior
There is an extensive body of theoretical and empirical literature on institutional herding behavior. Institutional investors may exhibit herding behavior for a number of reasons. Bikhchandani and Sharma (2001) mention three motives for herding behavior: information-based herding, compensation-based herding and reputation-based herding. We present an almost similar classification of motives, distinguishing between an information motive, a regulation motive and a reputation motive. Moreover, we apply an ordering to these motives, reclassifying the information motive as weak herding, the regulation motive as semi strong herding and the reputation motive as strong herding behavior. Weak herding is unintentional, while strong herding is intentional. All are discussed in more detail below.
Information motive (weak herding)
We define weak herding behavior as the result from the fact that pension funds have similar rebalancing strategies. Investors typically rely on similar sources of information when they make investment decisions. The information could for instance be market signals such as the returns on different asset classes. This can lead to herding behavior, which we classify as weak because it is an unintentional consequence of being exposed to similar information. Typically, pension funds have a rebalancing strategy, by aiming for a fixed asset allocation (Calvet et al., 2009; Bikker et al., 2010; Gorter and Bikker, 2013; Rubbaniy et al., 2012). Recently, Blake et al. (2015) report short-term mechanical portfolio rebalancing by UK pension funds. Also EIOPA recently published evidence that pension funds typically have rebalancing strategies (EIOPA, 2016). This way pension funds counteract changes in the asset allocation due to valuation changes in the different asset classes. Since pension funds are exposed to similar market risks, this results in trades into similar directions. Hence, this unintentional herding occurs because pension funds face similar decision problems and information sets (Bikhchandani and Sharma, 2001). For example, Rauh (2006) identifies the dependence of investments for defined benefit pension plans, particularly when financial constrained. Very similar, the rising popularity of “index tracking” also leads to herding behavior among institutional investors. Gleason et al. (2004); Chen et al. (2011); Shek et al. (2015) and Shek et al. (2015) document herding behaviour in the market for exchange traded funds (EFTs).
See full PDF below.