Econometrics Summary
Summary of Key Points from 'Econometrics by Example'
1. Introduction to Econometrics
- Econometrics applies statistical methods to economic data to provide empirical
content to economic relationships.
- The linear regression model (LRM) is a key tool in econometrics, using Ordinary Least
Squares (OLS) estimation.
- Econometric analysis involves:
- Model formulation: Defining relationships between variables.
- Estimation: Determining numerical values of parameters.
- Hypothesis testing: Checking statistical significance.
- Post-estimation diagnostics: Ensuring model validity.
Example:
Consider a model predicting wage (Y) based on education (X1) and experience (X2):
Y = β0 + β1X1 + β2X2 + u
Where u is the error term.
2. Regression Models
- **Linear Regression Model:** Uses a straight-line relationship between dependent and
independent variables.
- **Functional Forms of Regression Models:** Includes log-linear, polynomial, and
reciprocal models to capture different relationships.
- **Qualitative Explanatory Variables:** Uses dummy variables to incorporate
categorical data.
- **Multinomial and Ordinal Regression Models:** Extends regression for dependent
variables with multiple categories.
Example:
A company's sales (Y) depend on advertising (X), modeled as:
Y = β0 + β1 X + u
If the relationship is nonlinear, we use log-linear:
log Y = β0 + β1 X + u
3. Regression Diagnostics
- **Multicollinearity:** High correlation between independent variables affects
coefficient reliability.
- **Detection:** Variance Inflation Factor (VIF)
- **Remedy:** Remove redundant variables or use Principal Component Analysis.
- **Heteroscedasticity:** Non-constant variance of errors.
- **Detection:** Breusch-Pagan Test, White Test.
- **Remedy:** Weighted Least Squares (WLS), robust standard errors.
- **Autocorrelation:** Serial correlation in time-series data.
- **Detection:** Durbin-Watson Test, Breusch-Godfrey Test.
- **Remedy:** Cochrane-Orcutt procedure, ARMA models.
- **Model Specification Errors:** Includes omitted variables and incorrect functional
forms.
- **Detection:** Ramsey RESET Test.
Example:
A time-series model predicting GDP growth may suffer from autocorrelation:
Y_t = β0 + β1 X_t + u_t
where residuals u_t are correlated over time.
4. Advanced Econometric Models
- **Logit and Probit Models:** Used for binary outcomes like "employed" vs
"unemployed".
- **Limited Dependent Variable Models:** Includes Tobit and Heckman selection
models.
- **Panel Data Regression Models:** Accounts for cross-sectional and time-series
dimensions.
- **Time-Series Models:** Includes ARIMA, VAR, and cointegration techniques.
- **Asset Price Volatility Models:** ARCH and GARCH models for financial data.
---
Key Econometric Concepts
ARCH Test (Autoregressive Conditional Heteroscedasticity)
- Tests for changing variance in residuals over time.
- Formula:
u_t^2 = α0 + α1 u_{t-1}^2 + e_t
- Application: Financial time-series with volatility clustering (e.g., stock prices).
Example:
If stock returns exhibit periods of high and low volatility, ARCH tests detect this pattern.
GARCH Test (Generalized ARCH)
- Extends ARCH by including past variances:
u_t^2 = α0 + α1 u_{t-1}^2 + β1 σ_{t-1}^2 + e_t
- Used in risk modeling and economic forecasting.
Example:
GARCH models predict future volatility in forex markets, helping in risk assessment.
Heteroscedasticity
- Occurs when the variance of errors is not constant.
- Consequences: Inefficient estimators, biased standard errors.
- Remedies: Log transformations, robust standard errors.
Example:
A wage model may have higher variance for higher education levels, indicating
heteroscedasticity.
Breusch-Pagan Test
- Tests for heteroscedasticity.
- Formula:
u_t^2 = γ0 + γ1 X1 + ... + γk Xk + e_t
- Application: Regression models where residual variance changes with predictors.
Example:
Testing income inequality effects in an economic growth model.
Ordinal Regression Model
- Used when dependent variable is ordered (e.g., customer satisfaction: low, medium,
high).
- Formula (Ordered Logit Model):
P(Y ≤ j) = e^{α_j - β X} / (1 + e^{α_j - β X})
- Application: Surveys, credit ratings, health assessments.
Example:
Predicting likelihood of customer satisfaction levels based on service quality scores.
---
This document now provides a more detailed explanation of key econometric concepts,
with relevant formulas and practical applications.
