Endogeneity, Exogeneity, and Instrumental Variables in Monitoring and Evaluation

A set of vocabulary flashcards covering key concepts in Monitoring and Evaluation, specifically focusing on endogeneity, exogeneity, the role of instrumental variables, and the function of the error term in statistical modeling causality.

Causality

The ultimate goal in M&E, proving that a program $(X)$ is the direct cause of an outcome $(Y)$.

Randomised Controlled Trial (RCT)

Considered the gold standard of evaluation because random assignment creates exogeneity, ensuring treatment and control groups are comparable.

Instrumental Variables (IV)

A quasi-experimental technique used to estimate causal impact in non-randomised settings by leveraging an external variable that affects programme participation but not the outcome directly.

Endogeneity

A situation where an explanatory variable (such as programme participation) is correlated with the error term (Cov(X, \text{\epsilon}) \neq 0), leading to biased estimates.

Omitted Variable Bias

A source of endogeneity where important factors affecting the outcome are not included in the analysis, such as excluding parental support from an education study.

Reverse Causality (Simultaneity)

A source of endogeneity where the outcome influences the explanatory variable rather than the variable influencing the outcome.

Selection Bias

A form of endogeneity that occurs when participants are not randomly selected into a programme; for example, when more motivated farmers voluntarily join training.

Measurement Error

A source of endogeneity where variables are measured or recorded inaccurately, such as the underreporting of household income in surveys.

Difference-in-Differences (DiD)

A method for addressing endogeneity that compares changes over time between treatment and comparison groups.

Propensity Score Matching (PSM)

An evaluation technique that matches participants and non-participants with similar characteristics to reduce endogeneity.

Fixed Effects Models

A statistical method used to control for unobserved characteristics that do not change over time.

Exogeneity

The condition where an explanatory variable is independent of all unobserved factors influencing the outcome, meaning it is not correlated with the error term (Cov(X, \text{\epsilon}) = 0).

Strict Exogeneity

A condition where the explanatory variable is uncorrelated with the error term in all time periods, commonly assumed in panel data models.

Weak (Contemporaneous) Exogeneity

A condition where the explanatory variable is uncorrelated with the current error term but may be related to past or future errors.

Relevance

The first condition for a valid instrument, requiring the instrument $(Z)$ to be correlated with the endogenous variable $(X)$.

Exclusion Restriction (Exogeneity)

The second condition for a valid instrument, requiring that the instrument $(Z)$ affects the outcome $(Y)$ only through the treatment variable and not directly.

First Stage of IV

The conceptual step where programme participation is predicted using the instrument to extract variation unrelated to unobserved factors.

Second Stage of IV

The conceptual step where the predicted training values from the first stage are used to estimate the impact on the outcome.

Error Term (Disturbance Term)

The term \text{\epsilon} in a regression model representing all factors that affect the outcome variable but are not explicitly included in the model.