Topic 5 Notes:

Fundamental Financial Applications of Regression

Konark Saxena ∗

Abstract

These notes highlight fundamental applications of regression models in finance, show-

casing their role in both financial analysis and academic research. Univariate regres-
sions are employed to estimate key metrics like alphas, betas, and cumulative abnor-
mal returns (CARs), forming the basis for implementing financial theories such as
the CAPM. The CAPM and Single-Index Model (SIM) provide frameworks for cal-
culating expected returns and assessing systematic risk, while abnormal returns and
CARs measure value creation or destruction after accounting for the cost of capital.
Applications, such as event studies analyzing M&A announcements, demonstrate
how regression models can be used to understand changes in firm value due to cor-
porate actions.

∗
Associate Professor, ESCP Business School. Email: [email protected].
1 Estimating Expected and Abnormal Returns
Univariate regressions are an essential statistical tool in finance for analyzing relation-
ships between variables. They play a critical role in estimating metrics such as alphas,
betas, and cumulative abnormal returns (CARs). These metrics are foundational in test-
ing financial theories like the Capital Asset Pricing Model (CAPM), identifying abnormal
returns, and decomposing systematic and unsystematic risks.

1.1 Regressions to Estimate Expected Returns

In this section, we apply regressions to analyze fundamental finance applications,
specifically focusing on the relationship between industry excess returns and market ex-
cess returns. Using data from the Fama-French (FF) website, we calculate key metrics
such as alpha, beta, and R2 to evaluate the proportion of idiosyncratic components in
industry returns compared to those explained by the market.

Data Preparation
The process for preparing the data used in class is as follows:

1. Download the monthly industry returns and market excess returns from the Fama-
French website.

2. Compute the industry excess returns by subtracting the risk-free rate (Rf ) from the
raw industry returns (Rindustry,t ).

3. Align the time series data for industry and market excess returns to ensure consis-
tency.

1
Regression Model
We run a univariate regression for each industry using the following equation:

Rindustry,t − Rf = α + β(Rm − Rf ) + ϵt

where:

• α: Intercept, representing alpha.

• β: Slope coefficient, representing market beta.

• ϵt : Residual, capturing the idiosyncratic component of returns.

Sample Interpretation of Output

• A higher α suggests significant industry-specific outperformance.

• A higher β indicates a strong dependence of industry returns on market movements.

• A lower R2 implies a greater idiosyncratic component, meaning industry-specific

factors are important drivers of returns.

Through this analysis, we gain insights into the relationship between market and
industry returns, allowing us to quantify how much of an industry’s performance is driven
by market-wide effects versus unique, industry-specific factors.

1.2 CAPM and Expected Returns

The CAPM provides a theoretical framework to estimate the expected return of an
asset or portfolio based on its exposure to systematic risk as estimated from the above
regression model. According to the CAPM, the expected return is given by:

E(Ri ) = Rf + βi · [E(Rm ) − Rf ]

2
 Where:

• E(Ri ): Expected return of asset i,

• Rf : Risk-free rate,

• βi : Asset’s sensitivity to market returns (systematic risk),

• E(Rm −Rf ): Expected excess return of the market portfolio. This is often estimated
using the average historical excess return or risk premium of the market portfolio.

If CAPM holds, the abnormal return (denoted as α) should theoretically be zero.

Higher β industries should have higher average returns, corresponding to higher expected
returns.

1.3 Single-Index Model (SIM)

The Single-Index Model (SIM) simplifies return relationships by relating the excess
return of an asset to the excess return of the market:

(Ri − Rf ) = αi + βi · (Rm − Rf ) + ϵi

Where:

• αi : Intercept, capturing abnormal return,

• βi : Slope, measuring systematic risk,

• ϵi : Error term, representing idiosyncratic risk.

Regression analysis estimates the following:

• α̂i : Average abnormal performance (e.g., due to unobserved factors),

3
 • β̂i : Sensitivity to market movements.

The regression model produces estimates of alphas and betas that can be used to
calculate the expected return using SIM:

E[Ri ] = Rf + α̂i + β̂i · E[Rm − Rf ]

1.4 Abnormal Returns (AR)

Abnormal returns (AR) measure the deviation of actual returns from the expected
returns implied by CAPM or SIM. This is expressed as:

ARi = Ri − R̂i

Where:

• Ri : Actual return,

• R̂i : CAPM-expected return, calculated as:

R̂i = Rf + β̂i · (Rm − Rf )

• Alternatively, SIM-expected return, calculated as:

R̂i = Rf + α̂i + β̂i · (Rm − Rf )

Example: Assume the actual return of an industry portfolio is 8% and the CAPM-
implied expected return is 6%. The abnormal return is:

4
 AR(CAPM) = 8% − 6% = 2%

If the SIM estimated α is 1%, the abnormal return measured with respect to SIM will
be:
AR(SIM) = 8% − (1% + 6%) = 1%

2 Estimating Expected Returns for Industries

The table below summarizes the empirical analysis of returns for 12 industries by
calculating alpha and beta, and their implications for expected returns under the CAPM
and the Single-Index Model (SIM). It also examines average returns and abnormal returns
for 2024.

Table 1: Alphas, Betas, Expected Returns, and Abnormal Returns by Industry.

Alpha and Beta for Each Industry

• Alpha (α): Measures the industry’s average return not explained by the market’s
overall movement. Industries with high alphas, such as NoDur (0.17) and Hlth
(0.24), indicate consistent outperformance relative to their risk-adjusted market
returns.

• Beta (β): Measures sensitivity to market returns:

5
 – Low beta industries (e.g., Telcm (0.67), Utils (0.76)) are less volatile than the
market.

– High beta industries (e.g., Durbl (1.28), BusEq (1.26)) are more sensitive to
market fluctuations.

Observation: Low beta industries tend to have higher alphas, consistent with the
idea that low-risk investing can yield high returns.

2.1 Expected Returns Using CAPM and SIM

• Market Risk Premium: The excess return required for taking market risk is 0.69.

• CAPM Expected Excess Returns: Calculated as:

Expected Excess Return = β × Market Risk Premium

Example for NoDur:

0.75 × 0.69 = 0.51

High beta industries, like BusEq (β = 1.26), have higher expected returns (0.86)
due to their greater exposure to market risk.

2.2 Average Industry Excess Returns and Abnormal Returns

• Average Industry Excess Returns: Observed average excess returns above the risk-
free rate for a specific month (e.g., 0.69 for NoDur, 0.90 for BusEq).

• Abnormal Returns (CAPM): Abnormal returns measure deviations of actual returns

from expected returns:

Abnormal Return = Observed Return − Expected Return

6
 Example for NoDur (CAPM):

0.69 − 0.51 = 0.17

• Abnormal Returns (SIM): Expected returns under the Single-Index Model (SIM)
adjust for alpha. As for the entire sample, alpha adjusts for any difference between
expected and average returns. Thus, the abnormal returns with respect to the SIM
are always zero in the table.

2.3 Returns for 2024

• Average Returns in 2024: Average returns for each industry (e.g., NoDur: 0.75,
Hlth: 1.63) are calculated and reported in the table.

• SIM Abnormal Returns in 2024: Abnormal returns under SIM for 2024 are not
mechanically zero. For example, NoDur: 0.07, Hlth: 0.86. This occurs because
alpha is calculated using the full sample, not just for 2024. Therefore, it does not
mechanically ensure that average returns in 2024 are equal to the SIM-expected
returns.

3 Event Study: An Overview

An event study analyzes the impact of a specific event (e.g., an M&A announcement)
on the stock price of a company. The goal is to measure how much the event created or
destroyed firm value using abnormal returns (ARs), which reflect the deviation of actual
stock returns from expected returns during the event window.

Steps in an Event Study

1. Define the Event: Identify the event (e.g., an M&A announcement) and the event
date (t = 0), when the information becomes public.

7
2. Specify the Event Window: Define the event window, such as [−2, +5] trading days,
where t = 0 is the event date. The pre-event period captures possible information
leakage, while the post-event period reflects the market’s reaction.

3. Estimation Window: Specify a window before the event (e.g., [−120, −20] days) to
estimate the stock’s expected returns (for example, using the SIM/Market Model)
in the absence of the event. The estimation window should avoid contamination
of expected return estimates from the abnormal returns realized during the event.
The output of this estimation will be estimates of a stock’s expected returns and its
corresponding αi and βi .

4. Calculate Abnormal Returns (ARs): Compute abnormal returns as:

ARit = Rit − E(Rit )

Where:

• ARit : Abnormal return for stock i on day t,

• Rit : Actual return,

• E(Rit ): Expected return, estimated using a market model:

E(Rit ) = αi + βi Rmt

Here, Rmt is the market return.

5. Cumulative Abnormal Returns (CARs): Cumulative Abnormal Returns (CARs)

aggregate abnormal returns over a specified time horizon, such as [−2, +5]. They
provide a measure of abnormal returns over a longer duration.

8
 Compute cumulative abnormal returns (CARs) over the event window:

t2
X
CARi = ARit
t=t1

Where [t1 , t2 ] defines the event window.

Example: Suppose the abnormal returns over three days are:

• Day 1: AR = 1%,

• Day 2: AR = 0.5%,

• Day 3: AR = 0.7%.

The cumulative abnormal return over this period is:

CAR = 1% + 0.5% + 0.7% = 2.2%

Positive CARs indicate persistent overperformance, while negative CARs suggest

underperformance.

6. Statistical Testing: Test whether CARs are significantly different from zero to infer
the event’s impact on stock prices.

7. Interpret Results: A significantly positive CAR indicates that the M&A announce-
ment created shareholder value, while a significantly negative CAR suggests that
the market perceives the M&A as value-destroying.