Читать книгу Advanced Portfolio Management - Giuseppe A. Paleologo - Страница 11

In the previous sections we used the example of SYF and WMT. We now analyze in more detail the historical estimates for SYF, using returns for year 2018. Table 3.1 shows the estimates and the 95% confidence interval. The error around the alpha estimate is much larger than the estimate itself, which is not the case for beta. Table 3.2 shows the high degree of variability in the alphas, whereas the betas are relatively stable. This is the case for most stocks: estimates of alphas have large associated confidence intervals, and they seem to vary over time. You can't observe alpha directly, and you cannot estimate it easily from time series of returns. It is the job of the fundamental analyst to predict forward-looking alpha values based on deep fundamental research. The ability to combine these alpha forecasts in non-trivial ways from a variety of sources and to process a large number of unstructured data is a competitive advantage of fundamental investing, and one that will not go away soon. We list some of data sources and processes involved in the alpha generation process.

Table 3.1 Parameter estimates for Synchrony's returns regressed against SP500 returns, 2018. Alphas are expressed in percentage annualized returns.

Parameter	Estimate	95% conf.interval
alpha (%)	−24	(−88, +8)
beta	1.02	(0.84, 1.20)

Table 3.2 Parameter estimates for Synchrony's returns regressed against SP500 returns, 2015–2019. Alphas are expressed in percentage annualized returns.

Year	Alpha (%)	Beta
2016	6	1.45
2017	−24	1.79
2018	−40	1.02
2019	16	1.15
2020	−19	1.69

 Valuation Analysis. This differentiated view comes from a company business model whose data are coming primarily from Cash Flow Statements, Balance Sheets (Statements of Financial Positions), Income Statements and Stakeholder Equity Statements. It is complemented by macroeconomic data affecting the production function of the company as well as demand for its products and services. This is the domain of fundamental analysis [T. Koller and Wessels, 2015] and affects both short-term forecasts, e.g., earnings or the event of a sell-side analyst downgrading or upgrading a recommendation, and long-term forecast, such as the potential long-term unsuitability of a business.

 Alternative Data. Enabled by the availability of transactional data (e.g., credit card transactions) and environmental data (such as satellite imaging), these data complement fundamental analysis and give short-term information about demand, supply, costs, and potential risks. A textbook account of alternative data in finance is [Denev and Amin, 2020]; a reference for data sources, published by J.P. Morgan, is [Kolanovic and Krishnamachari, 2017].

 Sentiment Analysis. This can be interpreted as the subset of alternative data that gives information about demand that is a function of the consumer's expectations of the future state of the economy. Common macroeconomic time series are Conference Board's Consumer Confidence Index, the University of Michigan's Survey of Consumers, and the Purchasing Managers' Index. At the company level, it is possible to gain information from unstructured data [Loughran and McDonald, 2016].

 Corporate Access. Interactions between analysts and company managers occur on a quarterly basis. Although these communications are public information, fundamental analysts can make the best use of them by comparing to previous quarter guidance and by linking them to the vast information accumulated by the analyst on that company.

 Liquidity. A company may enter or exit the composition of an index, may issue a secondary offering, or may go through the expiration of a stock lock-up period. All these events affect demand for the stock at a point in time. Although fundamental investors don't have an edge in these events, they take them into account nonetheless.

 Crowding. It's not only the analyst's thesis about a company that matters, but what is the thesis with respect to the consensus among informed investors and motivated but nonprofessional investors acting in a coordinated way. It is possible, to some extent, to measure this consensus; Section 5.2 is dedicated to this issue. Consensus measures help the analyst gauge the originality of her fundamental thesis, and the risk associated to taking position in a stock that may be enjoying growing popularity, but also fall quickly out of favor with a broad investing base.