ECS4863
ASSIGNMENT 1 2025
UNIQUE NO.
DUE DATE: 16 MAY 2025
, Question 1: Understanding Key Econometric Concepts
1.1 Omitted Variable Bias: Positive vs. Negative
Omitted variable bias occurs when a relevant factor that affects both the dependent and
independent variables is excluded from a regression model. This can cause the
estimated effect of the independent variable to be inaccurate.
Positive bias happens when the missing variable is positively related to both
variables in the model, making the estimated effect larger than it should be.
Negative bias occurs when the omitted variable has a positive relationship with
one variable and a negative one with the other, which reduces or even reverses
the estimated effect.
Example:
If we study how education affects income but leave out work experience (which affects
both), the result may not reflect the true impact of education alone.
1.2 Testing for Serial Correlation with Strictly Exogenous Variables
Serial correlation means that a model’s errors are related over time, which can affect
the accuracy of results.
To test this when variables are strictly exogenous (not influenced by error terms), the
Durbin-Watson test is commonly used.
Steps:
1. Estimate your model and collect the residuals.
2. Use the Durbin-Watson statistic to check if there’s a pattern in how residuals
follow each other.
3. Interpret:
o A value near 2 suggests no serial correlation.
o A value below 2 indicates positive serial correlation.
ASSIGNMENT 1 2025
UNIQUE NO.
DUE DATE: 16 MAY 2025
, Question 1: Understanding Key Econometric Concepts
1.1 Omitted Variable Bias: Positive vs. Negative
Omitted variable bias occurs when a relevant factor that affects both the dependent and
independent variables is excluded from a regression model. This can cause the
estimated effect of the independent variable to be inaccurate.
Positive bias happens when the missing variable is positively related to both
variables in the model, making the estimated effect larger than it should be.
Negative bias occurs when the omitted variable has a positive relationship with
one variable and a negative one with the other, which reduces or even reverses
the estimated effect.
Example:
If we study how education affects income but leave out work experience (which affects
both), the result may not reflect the true impact of education alone.
1.2 Testing for Serial Correlation with Strictly Exogenous Variables
Serial correlation means that a model’s errors are related over time, which can affect
the accuracy of results.
To test this when variables are strictly exogenous (not influenced by error terms), the
Durbin-Watson test is commonly used.
Steps:
1. Estimate your model and collect the residuals.
2. Use the Durbin-Watson statistic to check if there’s a pattern in how residuals
follow each other.
3. Interpret:
o A value near 2 suggests no serial correlation.
o A value below 2 indicates positive serial correlation.
