We use a novel sample of exchange-traded security baskets managed by users of an online platform to provide new evidence on the risk-taking behavior of portfolio managers under option-like incentives. The unique feature of our setting is that it combines explicit-ly convex incentives with detailed portfolio holding data. We find that managers subse-quently increase portfolio risk when approaching their high-water mark (HWM), even when using portfolio holdings to separate between actively and exogenously driven risk-changes. There is no significant difference in risk-shifting behavior between portfolios operated by institutional asset managers and those managed by individual investors. Risk-shifting can especially be observed over the second and third quarter of the year, and is typically implemented by reducing both cash holdings and diversification likewise. Be-sides, portfolio managers are becoming more active when close to their HWM, increasing both the number of trades and portfolio turnover.

This paper studies relative price gaps between pairs of nearly identical Exchange-Traded Funds (ETFs) listed on US exchanges. Prices usually move in lockstep, but sometimes diverge from parity by larger amounts. Over the period 2010-2016, a simple pairs trading strategy produced abnormal returns of up to 1.8 percent per year net of fees. Price gaps cannot be justified by fundamental differences in liquidity, replication methodologies, or security lending activities, but are related to both proxies for cross-sectional and time-varying limits to arbitrage. Prices typically diverge following a sequence of days with abnormally low liquidity in both the relatively over- and underpriced fund.

Several studies have demonstrated that pairs trading with single stocks achieved significant excess returns, with decreasing profits in recent years. However, practitioners expanded pairs trading to ETFs. I discuss potential benefits of ETF over stock pairs and compare their risk and return characteristics using NYSE data over a time span of January 2001 to June 2016. I find that like stock pairs, ETF pairs also yield significant excess returns and alphas, albeit at a much lower level, since arbitrage opportunities occur less frequently and are smaller in magnitude. Nevertheless, ETF pairs outperform stock pairs with respect to downside risk, which is due to a smaller share and magnitude of non-converging, loss-making pairs. A sub-period analysis reveals that this is especially the case for recent periods. However, in recent times arbitrage opportunities are scarce for ETF pairs and probably too tiny to cover transaction costs.

Social trading networks provide access to an innovative type of delegated portfolio management. This paper provides a rationale for how these platforms are organized and gives some very first empirical insights to make social trading more tangible for both academics and practitioners alike. First, we discuss the basic mechanics and institutional aspects in light of the arising agency relationships between signal providers (users sharing their in-vestment ideas, i.e. portfolio managers) and signal followers (users subscribing to the sig-nals of other users, i.e. investors). We argue that as an intermediary, the platforms may generally reduce information asymmetries between both groups. However, due to the de facto non-existent entry barriers, it is reasonable to expect a high number of uninformed charlatans among signal providers. Second, using a unique dataset comprising transactions from the four major platforms, we analyze the returns generated during the year 2012. We find that signal providers typically engage in active trading rather than buy-and-hold strategies, which ultimately results in non-normal return distributions, implying that mean-variance analysis is an insufficient framework for performance analysis and portfolio selection in the context of social trading. More precisely, we argue that signal providers typically follow directional approaches and thus may exhibit substantial market risk at any point in time. Hence, far more than the documented 25 percent of the signal providers in our sample can be assumed to bear systematic risk.