Signal Processing - May 2017 - 15

Many SP researchers are eager to apply SP techniques to stock
price prediction or profitable trading strategy by analyzing
price time series. However, without economic justifications
and rigorous out-of-sample tests, such analysis may easily
fall into the category of technical analysis, which is not mainstream in finance academia or among financial professionals.
The knowledge of the major streams of finance theories is
very useful for SP researchers to position and formulate their
research before applying SP techniques to financial data analysis. Economic behaviors of people are often about individual
choices. The foundational theory (or hypothesis) of economic
choices is the expected utility theory, which explains the relationship between investment risk and expected return. Almost
all asset-pricing models are based on the expected utility theory.
To understand the aggregated market behavior, a fundamental
and insightful theory in finance and economics research, the
efficient market hypothesis (EMH) [4], was developed. The
EMH assumes that all information is incorporated into price
and that the market reaches equilibrium through rational decisions of market participants. Stock returns are compensation for
risks investors take in markets. The fundamental asset-pricing
model and portfolio theory are formulated on the basis of the
EMH [5]. Overwhelming empirical evidence supports the
EMH, suggesting that it holds most of the time.
Although the EMH is a foundational theory in academic
finance research, most stock analyses used by financial practitioners, such as technical analysis [6], which is popular among

individual investors, and fundamental analysis, which is popular
among professional investors, contradict the EMH [4] because
the current price does not reflect all available market information. Instead, these analyses share some common ground with
behavioral economics theories [7]-[9]. Although powerful,
behavioral economics theories rely on empirical psychological
and qualitative arguments to explain economic events and have
not led to fully developed asset-pricing models. In contrast with
the EMH-based models, it is often difficult to develop rigorous
mathematical formulations for psychological effects, making SP
methodology difficult to apply to behavioral economics studies.
The knowledge of extant data techniques, results, and issues
in finance research will greatly help SP researchers to exert useful research effort by identifying correct problem formulations
more effectively. For example, SP researchers are often interested in market correlation structures, an area where modern
asset-pricing and portfolio theories [5] have well formulated the
correlation structure within stock markets under the normality
assumption. In contrast with SP research in which controlled
experiments or simulations are conducted and ground truth is
often the basis for testing and verifying the models, model testing and verification in finance pose many traps and difficulties.
Nevertheless, with the tremendous increase in the amount
of economic data in digital form, the demand for applying SP
techniques to finance, economics, and marketing research is
increasing, presenting huge opportunities to SP researchers.
Many economics applications (including finance and marketing)

IEEE SIgnal
Signal ProcESSIng
Processing MagazInE
Magazine

May 2017

Table of Contents for the Digital Edition of Signal Processing - May 2017