Quantitative Methods & Economics

TL;DR

This section covers essential quantitative tools and economic principles used in investment analysis, focusing on regression analysis, time series, and economic indicators. You'll learn how to model relationships between variables, forecast future values, and understand market cycles. Mastering these areas helps you make data-driven investment decisions.

1. The Mental Model

Think of this section as your analytical toolkit. You're learning to use data to understand how financial variables move and why, helping you predict future trends and assess risks more effectively. It's about turning raw data into actionable insights for investment choices.

2. The Core Material

This part of the CFA Level 2 curriculum equips you with the quantitative skills and economic understanding necessary for advanced investment analysis. We'll focus on key areas like regression analysis, time series analysis, and understanding economic principles.

A. Regression Analysis

Photo by Alesia Kozik on Pexels

Regression analysis is a powerful statistical technique that lets you model the relationship between a dependent variable (the one you want to explain or forecast) and one or more independent variables (the factors you think influence it).

• Simple Linear Regression (SLR): This involves one independent variable. The goal is to fit a straight line that best describes the relationship.
  • Equation: $Y_i = b_0 + b_1X_i + \epsilon_i$
    • $Y_i$: Dependent variable
    • $X_i$: Independent variable
    • $b_0$: Y-intercept (value of Y when X is 0)
    • $b_1$: Slope (change in Y for a one-unit change in X)
    • $\epsilon_i$: Error term (unexplained variation)
  • Assumptions: For the regression results to be reliable, several assumptions about the error terms must hold (e.g., normally distributed, homoskedasticity, no autocorrelation). Violations of these assumptions can lead to unreliable standard errors and hypothesis tests.
  • Key Metrics:
    • Coefficient of Determination ($R^2$): Measures the proportion of the variance in the dependent variable explained by the independent variable(s). Higher $R^2$ usually means a better fit, but it doesn't guarantee a causal relationship.
    • Standard Error of the Estimate (SEE): Measures the typical distance between observed Y values and the regression line (a measure of how well the line fits the data).
    • t-statistics and p-values: Used to test the statistical significance of the regression coefficients ($b_0$, $b_1$). A low p-value (typically < 0.05) suggests the coefficient is statistically different from zero.
• Multiple Linear Regression (MLR): This extends SLR to include two or more independent variables. It allows for more complex models that capture multiple influences.
  • Challenges: Multicollinearity (high correlation between independent variables), heteroskedasticity (non-constant variance of error terms), and autocorrelation (correlation between error terms over time) are common issues that need to be addressed.

B. Time Series Analysis

Photo by Rafael Minguet Delgado on Pexels

Time series analysis focuses on data points collected or recorded at specific time intervals. It's crucial for forecasting financial variables like stock prices, interest rates, or economic growth.

• Components of a Time Series:
  • Trend: The long-term direction of the data (upward, downward, or horizontal).
  • Seasonality: Regular patterns of ups and downs that repeat over a short period (e.g., daily, weekly, quarterly).
  • Cyclicality: Patterns that oscillate around the trend but are not of a fixed period (often related to business cycles).
  • Irregular (Random) Component: Unpredictable, random fluctuations.
• Stationarity: A key concept. A time series is stationary if its statistical properties (mean, variance, autocorrelation) don't change over time. Many time series models assume stationarity. Non-stationary series often need to be "differenced" to become stationary.
• Autoregressive (AR) Models: A variable's current value is explained by its own past values.
• Moving Average (MA) Models: A variable's current value is explained by past error terms.
• ARIMA (Autoregressive Integrated Moving Average) Models: Combine AR and MA components and include differencing ('I' for integrated) to handle non-stationary data.

C. Economic Principles and Indicators

Photo by Rafael Minguet Delgado on Pexels

Understanding economic principles helps you interpret macroeconomic data and its impact on financial markets.

• Business Cycles: The natural fluctuation of the economy between expansion and contraction. Key phases: peak, recession, trough, recovery/expansion.
• Leading, Lagging, and Coincident Indicators:
  • Leading: Change before the overall economy does (e.g., building permits, manufacturing new orders).
  • Coincident: Change at roughly the same time as the economy (e.g., GDP, employment).
  • Lagging: Change after the economy has already changed (e.g., unemployment duration, prime rate).
• Monetary and Fiscal Policy:
  • Monetary Policy: Central bank actions (e.g., interest rate adjustments, quantitative easing) to influence money supply and credit conditions.
  • Fiscal Policy: Government actions (e.g., taxation, government spending) to influence the economy.
• Inflation and Deflation: Understanding their causes, measures (CPI, PPI), and impact on asset classes.

Here's a diagram illustrating the steps for conducting a basic regression analysis:

graph TD
    A["Define Research Question"] --> B["Identify Dependent (Y) & Independent (X) Variables"]
    B --> C["Collect & Prepare Data"]
    C --> D{"Check Data Assumptions\n(Linearity, Normality, etc.)"}
    D -- "Violated" --> E["Transform Data or Choose\nAlternative Model"]
    D -- "Met" --> F["Run Regression Model"]
    F --> G["Interpret Coefficients & Statistical Significance\n(R^2, p-values, t-stats)"]
    G --> H["Check Model Assumptions\n(Residual Analysis)"]
    H -- "Violated" --> E
    H -- "Met" --> I["Use Model for Prediction or Explanation"]

3. Worked Example

Let's say you want to predict a company's sales (dependent variable, $Y$) based on its advertising expenditure (independent variable, $X$). You collect historical data for both.

Data:

Using a statistical software (like Python with statsmodels or R), you run a simple linear regression.

Python Example:

import statsmodels.api as sm
import pandas as pd

# Data
data = {
    'Advertising': [10, 15, 12, 18, 20],
    'Sales': [120, 150, 130, 165, 175]
}
df = pd.DataFrame(data)

# Add a constant for the intercept term
X = sm.add_constant(df['Advertising'])
y = df['Sales']

# Fit the OLS model
model = sm.OLS(y, X).fit()

# Print the summary
print(model.summary())

Partial Output Interpretation (Focus on key parts):

                            OLS Regression Results
==============================================================================
Dep. Variable:                  Sales   R-squared:                       0.985
Model:                            OLS   Adj. R-squared:                  0.980
Method:                 Least Squares   F-statistic:                     199.3
Date:                Tue, 23 Jan 2024   Prob (F-statistic):            0.000490
No. Observations:                   5
Df Residuals:                       3
Df Model:                           1
Covariance Type:            nonrobust
==============================================================================
                 coef    std err          t      P>|t|      [0.025      0.975]
------------------------------------------------------------------------------
const         70.3807      4.041     17.417      0.000      57.519      83.242
Advertising    5.2000      0.368     14.117      0.000       4.032       6.368
==============================================================================

Interpretation:

• R-squared (0.985): This means 98.5% of the variation in Sales can be explained by Advertising expenditure. This is a very strong fit.
• Constant (coef = 70.3807): If advertising expenditure were zero, predicted sales would be $70,380.70.
• Advertising (coef = 5.2000): For every additional $1,000 spent on advertising, sales are predicted to increase by $5,200.
• P>|t| (for both const and Advertising = 0.000): Both coefficients are highly statistically significant (well below 0.05), meaning they are unlikely to be zero by chance.

4. Key Takeaways

• Regression analysis helps model relationships between variables, allowing for prediction and understanding influences.
• The R-squared value indicates how much of the dependent variable's variance is explained by the independent variable(s).
• Statistical significance (p-value) tells you if a coefficient is reliably different from zero.
• Time series analysis is essential for forecasting and understanding patterns in data over time.
• Stationarity is a critical concept in time series, often requiring data transformation if not met.
• Economic indicators (leading, lagging, coincident) provide insights into current and future economic conditions.
• Monetary and fiscal policies significantly impact economic activity and financial markets.

Common Mistakes to Avoid:
- Assuming causation from correlation: Just because two variables move together doesn't mean one causes the other.
- Ignoring model assumptions: Violating regression assumptions can lead to unreliable results and flawed conclusions.
- Overfitting a model: Including too many variables or complex models can fit past data perfectly but perform poorly on new data.
- Using non-stationary time series without transformation: This can lead to spurious regressions and invalid forecasts.

5. Now Try It

Find a simple dataset online (e.g., S&P 500 daily returns and VIX daily returns, or GDP growth and interest rates for a country). Using a tool like Excel, Python (with statsmodels), or R, perform a simple linear regression with one variable attempting to explain the other.

What to do:
1. Define your variables: Clearly identify your dependent (Y) and independent (X) variables.
2. Collect data: Get at least 30 observations for each variable.
3. Run the regression: Use your chosen tool to perform the linear regression.
4. Interpret the results: Look at the $R^2