Advanced Portfolio Management. Giuseppe A. Paleologo
Читать онлайн книгу.
attractive to investors, and prudent is perhaps the most important decision they face.
This is the subject of Chapter 10.
2.8 How to Analyze New Sources of Data
New sources of data that go far beyond standard financial information become available every day. The portfolio manager faces the challenge of evaluating them, processing them and incorporating them into their investment process. The ability to transform data and extract value from them will become an important competitive advantage in the years to come. The range of methods available to an investor is as wide as the methods of Statistics, Machine Learning and Artificial Intelligence, and experimenting with them all is a daunting task. Are there ways to screen and learn from data so that the output is consistent with and complementary to your investment process?
This is the subject of Section 8.4.
Notes
1 1 Among them, The Journal of Portfolio Management, The Journal of Financial Data Science, and the Financial Analysts Journal.
2 2 Joe Armstrong, a leading computer scientist and the inventor of the computer language Erlang, uses an effective metaphor for the lack of separation between the object of interest and its environment: You wanted a banana but what you got was a gorilla holding the banana and the entire jungle [Seibel, 2009].
Chapter 3 A Tour of Risk and Performance
What will you learn here: A very gentle introduction to factor models, starting with the simplest example of a model, which you probably already know. And how risk estimation, performance attribution and hedging can be performed using this simple approach.
Why do you need it: Because the themes I introduce here will return over and over again throughout the book, from simple heuristics to advanced optimizations.
When will you need this: Always. This will become your second nature. You will break the ice at cocktail parties mentioning how much risk decomposition helped you in your life.
3.1 Introduction
On July 3, 1884, the Customer's Afternoon Letter (owned by Dow Jones & Co.) began publishing the first stock index: a simple price average of nine transportation companies and two industrial ones. In 1886 it published the first Dow Jones Industrial Average. In 1889, the newspaper became The Wall Street Journal, and over time more indices were created. Indices provided a benchmark against which to compare one's investment; and they are a summary of the overall behavior of the market or of a specific sector. A typical benchmarking exercise: if we hold a stock, on any given day we first look at the overall market return, as provided by the index, and then we compute the out- or underperformance of the stock compared to the market. When we look at indices as market summaries, we implicitly know that they describe most, or at least some, of the stock returns for that market segment. In a very real way, having an index gives us a way to describe performance and variation of stock returns. Factor models capture these two intuitive facts, make it rigorous, and extend them in many directions.
A first extension aims to offer more flexibility in the relationship between stock and benchmark. For example, a cyclical stock in the financial sector like Synchrony Financial (ticker: SYF) moves more in sync (lame pun) with the market than, say, Walmart (ticker: WMT), a large, stable, defensive stock. Figure 3.1 bears this out. We take the daily returns of Synchrony and Wal-Mart and regress them against the daily returns of SPY.1 The regression coefficient is denoted “beta”. If the market returns an incremental 1%, the stock returns an incremental (beta) * (1%), everything else being equal. It is a measure of market sensitivity that differs from stock to stock. Figure 3.1 shows the regression of SYF daily returns against SP500 futures returns, for the period January 2, 2018, to December 31, 2019. Let us go with the assumption that we can estimate the true beta of a stock to the market; i.e., the estimation error of the beta doesn't really matter. Then a simple decomposition of market return + stock-specific return gives us a great deal of information. On the benchmarking side, we now know what fraction of the return is attributable to the market. It may seem that SYF is outperforming the market in a bull market and Wal-Mart is underperforming. However, after decomposing returns, it may be the other way around: SYF is just a leveraged bet on the market, and after we remove the market contribution, SYF has underperformed, and WMT has outperformed. Another benefit from the linear relationship between market and stock returns is that it establishes a relationship among all stock returns. The market is the common link. For the overwhelming majority of stocks, the beta to the market is positive, but it can take values in a wide interval; several stocks exhibit betas higher than two. This relationship has implications for expected returns and risk as well. We will delve deeper into both later in this chapter. But before we proceed, we introduce a new term, “alpha”. In your daily job, it's alpha that will pay your salary. Beta, on the other hand, can get you fired. This explains why so many portfolio managers have the symbol
Figure 3.1 Linear regression of Synchrony's (SYF) and Wal-Mart's (WMT) daily returns against the SPY daily returns, using daily returns for the years 2018 and 2019.
3.2 Alpha and Beta
Consider the regression line for SYF again. The complete formula for a linear regression includes an intercept and an error term, or residual. We write this explicitly. For a given stock,
Visually, this relationship is shown in Figure 3.2. We already introduced the term