In johngil/mscidata: MSCI USA and Min Vol Data Analysis

Abstract

The low-risk anomaly has created opportunities for arbitrage in the financial markets. As Baker et al. discuss in “Benchmarks as Limits to Arbitrage: Understanding the Low-Volatility Anomaly,” low-volatility and low-beta portfolios outperform and high-volatility and high-beta portfolios by a factor of several times due to benchmarking and lottery-preferences. The iShares MSCI USA Minimum Volatility (USMV) is an ETF tracking a minimum volatility index that was used to find data and will be used for trading arbitrage. Frazzini et al. discuss arbitrage opportunities by quantitative focused funds like AQR in “Betting Against Beta”, and this thesis explores a more advanced type of index front-running as a potential arbitrage opportunity. Data was collected from USMV from its inception in October 2011, and from EUSA, the parent ETF of USMV, from the same period until December 2016. 52-week trailing beta, 52-week trailing volatility, lagged price/book, and current index membership were calculated, and a regression model was run to quantify the relationship between current index membership and these four variables. In the model, a probabilities of index membership were calculated and an optimal cutoff was calculated to which the model would be 95% accurate of its findings of a stock to be in or out of USMV, given the historical data. Backtesting with prior data showed with a model accuracy of 95%, arbitrage opportunities of X% could be collected after each rebalancing.

Introduction

The iShares MSCI USA Minimum Volatility (USMV) Exchange Traded Fund (ETF) is designed to track the investment results of the MSCI Minimum Volatility USA index, which is composed of stocks with a lower volatility than the general market. This can provide investors with exposure to a portfolio with less risk than many alternatives, and historically has declined less in value than the broader market during economic downturns. The ETF is comprised of 189 holdings, and is rebalanced two times per year. The purpose of this dissertation is to create a logistic regression model that can accurately predict which stocks will be added or removed from this ETF before rebalancing occurs, and understand what factors are involved. The model will take into account volatility attributes of each stock, as well as others potentially significant predictor variables from prior studies. An accurate model will allow for arbitrage investment opportunities.

Exchange Traded Funds

An Exchange Traded Fund (ETF) is a collection of stocks and/or bonds in a single portfolio, that is traded on a major exchange just like a stock is (http://www.investopedia.com/terms/e/etf.asp). As a result, the price of an ETF fluctuates on a regular basis. Exchange Traded Funds generally have more liquidity and less fees when compared to other alternatives instruments like mutual funds. Owning an ETF can allow investors to minimize risk, since owning an ETF is comparable to owning a little bit of many different stocks. This diversification comes at lower costs and less effort for investors as well.

ETFs can also track an index, commodity, bonds, or basket of all of the above. Unlike an ETF, which is publicly traded, an index is not. The goal of the USMV ETF is to track the MSCI Minimum Volatility USA index, and this is more complicated than it seems. In addition to tracking this index, the ETF aims to mirror returns of the index and any difference is called tracking error. Many times, the tracking error is often very small, and can be around a tenth of a percent. This error can come from indices being market capitalization weighted, meaning that for each price fluctuations of each stock lead to the weighting being changed by a ratio of its market cap against the market cap of all stocks in the index (http://www.investopedia.com/articles/exchangetradedfunds/09/tracking-error-etf-funds.asp). With these stocks weightings in the index constantly changing and people buying in and out of ETFs constantly, it is hard to track performance entirely accurately. However, ETFs very closely follow indices, as their tracking errors are generally quite small. Thus, although ETF data is not the same as index data, they are very similar.

iShares MSCI Min Vol USA ETF

The iShares MSCI Min Vol USA ETF (USMV) is a Blackrock-managed ETF that tracks the investment results of the MSCI Minimum Volatility USA index. The MSCI Minimum Volatility USA index constituents come from the MSCI USA Index, which are roughly comprised of the top 600 US stocks by market cap. This minimum volatility index is intended to have a lower beta, lower volatility, lower cap bias, and contain more stocks with less risk than its parent index, which contains US mid-cap and large-cap stocks. The index is rebalanced twice a year, on the last trading days of May and November. The index typically has around 180 constituents, with an average of 20 new additions and 14 deletions every 6 months when rebalancing occurs. Over the last five, years, the number of additions has ranged from 12 to 25, while the deletions have been between 10 and 19. Changes to the index are usually announced nine trading days before they are set to take place.

Using the Barra Open Optimizer, USMV creates a minimum variance portfolio of low risk stocks, as a subset from its parent index of USA large-cap and mid-cap stock. Using this estimated security covariance matrix, the MSCI Minimum Volatility Index is the product of the lowest absolute volatility, considering the constraints. Moreover, these additions are simply a relabeling of existing stocks in the parent index, and do not include new additions to the parent index. The low-risk stocks chosen to be in USMV are determined by a set of constraints, like maintaining a certain sector or country weight relative to the parent index.

There are many specific constraints to this index. The first is that an individual stock cannot exceed 1.5% or 20 times the weight of the stock in the parent index. The minimum weight of a security in the index is also capped at 0.05%. USMV also aims to keep the weight of specific countries within a 5% range of the weight in the parent index, or 3 times the weight of the country in the parent index. Sector weights of USMV also cannot deviate more than 5% from the sector weights in the parent index. One way turnover of the index is also maxed at 10%. Thus, taking into account these constraints, the Barra Open Optimizer creates the lowest absolute volatility portfolio possible (https://seekingalpha.com/article/3964639-understanding-ishares-msci-usa-minimum-volatility-etf)

Purpose

As mentioned, the purpose of this thesis is to create a model to that will predict rebalancing of stocks in the Min Vol index, and thus the USMV ETF, before it actually happens. There is significant price movement whenever a stock is added or removed from a large ETF, like USMV. When a stock is added to the index, the ETF will buy large amounts of that stock, increasing the demand, and consequently market price for that stock. If the stock is bought in advance of this large purchase, then the investor can enjoy pretty immediate price appreciation in the stock. Moreover, if a stock is removed from the Min Vol index, the USMV ETF will sell all current holdings of the stock, which would increase the supply of the stock, driving down market price of the stock. If one were to short this stock before that happened, he/she can also profit from that event.

A phenomena known as ETF front-running has been around for a long time and is similar to what this paper hopes to accomplish, but is one step behind. ETF front-running involves traders buying or selling stocks in advance of ETF managers after they announce an exit or entrance of a position (https://seekingalpha.com/article/165877-how-traders-are-front-running-etfs). There is typically a slight lag between an announcement of an ETF to add or remove a position, and the actual purchase or sale of this position. By acting quickly, traders can scalp profit by buying a stock before an ETF does, and selling it to them at a slight profit, or short-selling a stock before an ETF exits the position, and then buying it back at the lower price. The thesis will take this one step farther, and try predict the stock addition or deletion before announcement. This will allow traders to similar front-run the index, but they will do so before the market is able to react, leading to larger profit opportunities.

Model

These goals of this paper will be achieved by creating a logistic regression model, which will be transformed to calculate a probability of a stock being in the out of the index. The predictor variables will include 52-week trailing volatility, 52-week trailing beta, price/book ratio, and whether or not the stock was in the index 6 during the previous rebalancing. These attributes were chosen after looking at the historical literature and understanding of the minimum volatility index.

johngil/mscidata documentation built on May 19, 2019, 5:14 p.m.