BASIC

ECONOMETRICS
P1
 Introduction
• The term "econometrics" was coined by Ragnar Frisch in the 1920s.
• Econometrics literally means “economic measurement.”
• Econometrics is concerned with the empirical determination of economic laws.

Other interpretation:
1. Mathematical Statistics in Econometrics:
• Econometrics applies mathematical statistics to economic data and provides empirical
support to models constructed by mathematical economics.
2. Quantitative Analysis of Economic Phenomena:
• It integrates theory and observation through appropriate inference methods.
3. Interdisciplinary Approach:
• combines economic theory, mathematics, and statistical inference to analyze economic
phenomena.
4. Empirical Determination of Economic Laws:
• Econometrics is concerned with finding and validating economic laws using real-world data.
5. Improving the Public Image of Economics:
• Econometricians help improve the perception of economics, moving away from abstract
theorizing toward meaningful quantitative analysis.
Econometrics as A Separate Discipline
Economic Theory Mathematical Economics
• Expresses economic theory using
• Provides qualitative statements or
mathematical equations.
hypotheses. • Focuses on theoretical formulation, not on
empirical testing or measurability.
• it does not quantify the relationship

Economic Statistics Mathematical Statistics

• Focuses on data collection, processing, and • Provides tools and methods (e.g., regression,
presentation. probability theory) used in econometrics.
• Produces the raw data used in econometric • Traditional statistical tools assume controlled
analysis. experiments, which are rare in economics.
• Economic statisticians typically do not test
theories using data.

Econometrics
1. Econometrics takes mathematical models from economic theory and reformulates them into
empirically testable forms using real-world data.
2. beyond data collection to test theories.
3. Economic data are usually observational
 Methodology Of Econometrics

1. Statement of theory or hypothesis (hypothesis ≠ theory)

2. Specification of mathematical model (requires: DV & IV, expected sign and the magnitude of
the coefficient of parameters, defining the form of the model i.e. linear or non-linear)
3. Specification of the econometric model (inclusion of stochastic term)
4. Obtaining data
5. Estimation of parameters (OLS, MLM, MM)
6. Testing Hypothesis
7. Forecasting or Prediction
8. Use of method for control or policy purposes
Types of Econometrics

Econometrics

Theoreti
cal Applied

Classic Classic
Bayesian Bayesian
al al
Aspect Classical Bayesian

Random with prior

Parameters Fixed but unknown
distribution

Data Random sample Observed, fixed once given

Due to incomplete knowledge

Uncertainty Due to sampling variability
(beliefs)

Posterior distributions,
Inference tools Confidence intervals, p-values
credible intervals

Incorporates prior info? No Yes (via prior distribution)

Bayes’ theorem + MCMC,

Estimation techniques OLS, MLE, GMM
Gibbs sampling

Output Point estimates + intervals Full posterior distribution

P2
Regression Analysis
Historical Origin
• Term was introduced by “ Francis Galton”

• First one to point out “ regression to mediocrity or regression

to mean”, it is also known as Galton’s law of universal
regression.

• The law was confirmed by Karl Pearson

Modern Interpretation of Regression
Regression analysis is concerned with the study of
dependence of one variable on one or more other variable.
With view to estimate and/or predicting the (population) mean
or average value of the former in terms of the known or fixed
(in repeated sampling) values of the latter.
(gujurati, et al)
Statistical vs Deterministic relationships

Statistical Relationships
• Involve random or stochastic variables, which have probability distributions.
• Statistical models capture relationships that involve uncertainty or randomness

Deterministic Relationships
• Involve non-random variables; the relationship is exact and fixed without random error.
(i.e. predictable and replicable)

Turning Deterministic Becomes Statistical

• If we consider measurement errors in constants or variables, even a deterministic law
becomes statistical.

NOTE* In regression analysis, we focus on statistical (not deterministic) relationships

between variables.
Regression vs Causation
“A statistical relationship, however strong and however suggestive, can never establish
causal connection: our ideas of causation must come from outside statistics, ultimately
from some theory or other.” - Kendall and Stuart

Regression vs Correlation

Feature Correlation Analysis Regression Analysis

Measure strength of
Main Objective Predict or estimate average value
association
Variable
Symmetric Asymmetric
Treatment
Dependent variable is random, independent
Variable Type Both variables are random
Note* one of the key assumption of regression is fixed
in econometrics is explanatory
variable are non-stochastic.
Directionality No cause-effect implication Often implies causal direction (from X to Y)
Types of Data for Economic Analysis

• Time Series

• Cross-section

• Pooled

• Panel, Longitudinal, Or Micro-panel

Different scale to measure variables

• Ratio Scale
• Interval Scale
• Ordinal Scale
• Nominal Scale

* 3 possible properties of a scale

 Distance (X2 – X1 or X1 – X2)
 Ratio (X2 / X1 or X1 / X2)
 Natural Order ( X2 ≤ X1 or X2 ≥ X1)
Terminology and Notation
Symbol/Notation Meaning
Y Dependent variable
X₁, X₂, ..., Xₖ Explanatory variables (Xₖ is the kth explanatory variable)
i Index for cross-sectional observations
t Index for time series observations
Observation of “Xₖ” for “I”th (cross-section) or “t”th (time series)
Xₖᵢ or Xₖₜ
unit
N or T Total number of observations in a population
n or t Total number of observations in a sample
Random / Stochastic Variable A variable that can take on multiple values based on probability

Dependent Variable: explained variables, predicted variable or predictand, Regressand,

response, endogenous, outcome, controlled variables

Independent Variable: explanatory variables, predictor, regressor, treatment or stimulus,

exogenous, covariate, control variable
 Two-Variable Regression
Analysis: Some Basic Ideas
Introduction

A bivariate regression (also called two-variable regression) is a simple linear

regression model that studies the relationship between one dependent variable (Y)
and one independent (explanatory) variable (X).

Simple linear regression model

Y = f(X)
Y=β0​+β1​X

Where,
β0 = intercept,
β1​= slope

β0​and β1 are parameters.

PRF and SRF
The Population Regression Function (PRF) is a theoretical equation that
describes the average relationship between a dependent variable and one or
more independent variables for the entire population.

Where as, The Sample Regression Function (SRF) is the estimated equation
derived from sample data that approximates the Population Regression Function
to describe the relationship between the dependent and independent variables.
PRF

PRF

Deterministic PRF Non-Deterministic

Yi = β0​+ β1​Xi + µi
E(Y / Xi) = f(Xi)
Or, Yi = E(Yi / Xi) + µi
Or, E(Yi / Xi) = β0​+ β1​Xi

Where, µi = Yi - (Y I Xi)
significance of Error term in model
• Vagueness of Theory

• Unavailability of Data

• Intrinsic Randomness Behaviour of Human Being

• Core variable vs Peripheral Variables

• Poor Proxy Variables

• Mis-specification of the equation or wrong functional form

• Error of aggregation of economic data

• Principle of parsimony
SRF

SRF

Deterministic Non-deterministic
i
Comparison between PRF and SRF

Our primary objective is to find and so they are best fit to β0​and β1

Or we can say, we want to estimate Yi = β0​+ β1​Xi + µi on the basis of i

PRF = LRF (linear population regression model)

Here, what linear means?

Linear can be interpreted in two ways i.e. linear in variable or linear in parameters.

Linear in Variable

• the meaning of linearity is that the conditional expectation of Y is a linear function of X. E(Y I Xi)
= β0​+β1​X

Linearity in Parameters

• it means that the conditional expectations of Y, is a linear function of the parameters, the β’s.
• it may or may not be linear in the variable X.
Theory of Estimation

• It deals with the methods by which population parameters (µ, σ, P) from

sampling ( s2, p).
• Estimation(procedure), Estimator(sample statistics), Estimate (specific
observed value)

• Two Types of Estimation (1st Point Estimation, 2nd Interval Estimation)

Ordinary Least Squares
Introduction
• method was developed by Carl Friedrich Gauss, German Mathematician.

• it is used to find out and as estimator of β0​and β1.

​
• the objective of OLS is to minimise (SSR/SSE/RSS) to obtain β0​and β1 which are as close
as β0​and β1 of PRF. [ note* = f(, )

Numerical Properties of OLS Estimators

• terms are expressed in solely in terms of the observable quantities.
• they are point estimators.
• after getting OLS estimators sample regression line can be easily obtained.

Properties of Sample Regression Line

• it passes through sample mean of Y and X.
• the mean value of is equal to mean value of Y.
• the mean value of residual is zero.
• the residual is uncorrelated with the predicted Yi. The residual is uncorrelated with Xi.
Assumptions of OLS

• Linearity: linearity in the parameters

• Fixed Values of X or X values independent of the error term or X is non-stochastic.
• zero mean value of disturbance
• Homoscedasticity or constant variance of residual
• No Autocorrelation between the disturbances
• The number of observation must be greater than the number of parameters to be
estimated.
• Positive Variance of X and no outliers.
• zero co-variance between X and residual.

• Residual is normally distributed

• absence of multicollinearity
• no specification bias