BASIC ECONOMETRICS

UNIT 1 — COMPLETE REVISION NOTES

DSE Course ECON024 | BA (Programme) Economics | Semester II
Reference: Studenmund Ch 1, 2, 4, 5 | 15 Hours

These notes cover every topic in Unit 1 as direct exam-ready answers. Each topic has the theory you
need to write, plus worked numericals taken from previous year question papers.

Topic Marks Weight Type

1. What is Regression & the Basic Model High Theory + Numerical

2. OLS Method — Estimating Coefficients Very High Numerical

3. Interpretation of Results + Prediction Very High Numerical

4. TSS, ESS, RSS and R² Medium Theory + Numerical

5. Assumptions of CLRM (Any Three) High Theory

6. BLUE Property and Gauss-Markov Medium Theory

7. Multiple Regression Medium Theory + Numerical

8. t-test and F-test Very High Numerical

Unit 1 — Basic Econometrics | ECON024 Page 1

 TOPIC 1: WHAT IS REGRESSION? THE BASIC MODEL

1.1 What is Econometrics?

Econometrics literally means 'economic measurement'. It combines three things: economic theory
(which tells us what variables are related), mathematics (which expresses the relationship as an
equation), and statistics (which estimates and tests the relationship using real data).

In simple words: We have an economic idea (e.g., 'more study hours improve grades'). Econometrics
helps us take actual data, estimate HOW MUCH study hours improve grades, and test if the relationship is
statistically real.

1.2 The Simple Linear Regression Model

A regression model expresses one variable (Y) as a function of another variable (X) plus a random error
term. The simplest form is the two-variable regression:

Yi = β0 + β1Xi + εi
This is the Population Regression Function (PRF)
Y■ = Dependent variable (the outcome we are explaining, e.g., Grade, Salary)
X■ = Independent/Explanatory variable (the cause, e.g., Hours studied, Experience)
β■ = Intercept — value of Y when X is zero (the baseline)
β■ = Slope — how much Y changes when X increases by 1 unit
ε■ = Error term — captures everything else that affects Y but is not in the model

What does the Error Term (ε■) represent?

No model can include every factor that affects Y. The error term is a catch-all for factors we have left out.
For example, in a grade regression, the error term captures things like student intelligence, quality of
textbooks, health on exam day, etc.
• Factors not included in the model
• Random variation and unpredictable behaviour
• Measurement errors in the data

PRF vs SRF — Two Key Concepts

Concept Full Name What It Is

The TRUE relationship in the entire population. We

PRF Population Regression Function
NEVER observe this directly.

Our ESTIMATE using a sample. Written as Y■■ = β■■

SRF Sample Regression Function
+ β■■X■

Unit 1 — Basic Econometrics | ECON024 Page 2

 REMEMBER: The 'hat' symbol (■) always means 'estimated'. So β■■ is our estimate of the true β■.
The true values β■ and β■ are unknown. We use sample data to estimate them.

TOPIC 2: OLS — ORDINARY LEAST SQUARES ESTIMATION

2.1 What is OLS?

OLS (Ordinary Least Squares) is the method used to find the best estimates of β■ and β■ from sample
data. It works by finding the line that minimises the sum of squared residuals (errors).

What is a Residual?

Residual = Actual Y − Predicted Y

■■ = Y■ − Y■■
If actual grade = 7 and predicted grade = 6.5, residual = 0.5

The OLS Objective

OLS minimises: Σ■■² = Σ(Y■ − Y■■)²

This is called the Residual Sum of Squares (RSS)
By minimising this, OLS finds the closest possible line to all the data points

2.2 OLS Formulas (MUST KNOW for Numericals)

β■■ (Slope) = [nΣXY − ΣX·ΣY] ÷ [nΣX² − (ΣX)²]

β■■ (Intercept) = Y■ − β■■ × X■
Where: Y■ = ΣY/n (mean of Y) and X■ = ΣX/n (mean of X)

In every OLS numerical, you need to compute five sums from the data table: n (number of observations),
ΣX, ΣY, ΣX², and ΣXY. Then plug into the formulas above.

Important Properties of OLS Residuals (Theory Question)

• The sum of all residuals equals zero: Σ■■ = 0
• The regression line always passes through the point (X■, Y■) — the means of X and Y
• Residuals are uncorrelated with the X values

2.3 SOLVED NUMERICAL — OLS Estimation

EXAM QUESTION TYPE: This exact question appeared in your exam (Q2b). This template works for
any OLS numerical.

Unit 1 — Basic Econometrics | ECON024 Page 3

 Question: A researcher wants to find the relationship between annual salary (Rs. '0000) and work
experience (years). Data is given below. Estimate the slope and intercept using OLS.

Salary (Y) 20 30 15 80 10 70 120 45 59

Experience
5 7 3 18 2 15 30 10 13
(X)

Step-by-Step OLS Calculation

STEP 1 Count observations: n = 9

STEP 2 Compute the five required sums: ΣX = 5+7+3+18+2+15+30+10+13 = 103 ΣY =

20+30+15+80+10+70+120+45+59 = 449 ΣX² =
25+49+9+324+4+225+900+100+169 = 1805 ΣXY =
(5×20)+(7×30)+(3×15)+(18×80)+(2×10)+(15×70)+(30×120)+(10×45)+(13×59) =
100+210+45+1440+20+1050+3600+450+767 = 7682

STEP 3 Apply the slope formula: β■■ = [nΣXY − ΣX·ΣY] ÷ [nΣX² − (ΣX)²] = [9×7682 −
103×449] ÷ [9×1805 − 103²] = [69138 − 46247] ÷ [16245 − 10609] = 22891 ÷ 5636
= 4.062

STEP 4 Find means: X■ = 103/9 = 11.44 Y■ = 449/9 = 49.89

STEP 5 Apply the intercept formula: β■■ = Y■ − β■■ × X■ = 49.89 − 4.062 × 11.44 =
49.89 − 46.47 = 3.42

EXAM ANSWER
Estimated Equation: Salary = 3.42 + 4.062 × Experience

Intercept (β■■ = 3.42): A person with zero years of experience has a base salary of Rs. 3,420.
Slope (β■■ = 4.062): For every 1 additional year of experience, salary increases by Rs. 4,062.
This confirms the expected positive relationship between experience and salary.

Unit 1 — Basic Econometrics | ECON024 Page 4

 TOPIC 3: INTERPRETING RESULTS & MAKING PREDICTIONS

3.1 How to Interpret Regression Coefficients

After estimating the regression, you must be able to explain what each number means in
economic/real-world language. The professor expects full sentences, not just numbers.

INTERCEPT (β■■): 'When X = 0, Y is expected to be β■■ units. This is the baseline value.'

SLOPE (β■■): 'For every 1 unit increase in X, Y increases/decreases by β■■ units,

holding all other variables constant.'

R²: 'R² × 100% of the variation in Y is explained by the model.

The remaining (1−R²) × 100% is due to factors not included in the model.'

3.2 SOLVED NUMERICAL — Interpretation + Prediction

EXAM QUESTION: This is Q1(a) from your previous year paper. This exact format is asked every
year.

Given: G■■ = 3.2 + 0.07T■ and R² = 0.632

Where G■ = Grade of student in econometrics, T■ = Total hours spent studying

Part A: Interpret the regression coefficients and R².

EXAM ANSWER
INTERCEPT (3.2):
When a student studies for 0 hours, the expected grade is 3.2. This represents the baseline grade a
student gets without studying (perhaps from general knowledge or luck).

SLOPE (0.07):
For every 1 additional hour a student spends studying econometrics, the grade increases by 0.07
points. Since the coefficient is positive, more study time leads to higher grades, which is economically
sensible.

R² = 0.632:
63.2% of the variation in students' grades is explained by the number of hours they study. The
remaining 36.8% is due to other factors not captured in this model, such as student intelligence,
quality of teaching, attendance, or health on exam day.

Part B: What is the predicted grade for a student who studied for 55 hours?

Prediction Calculation

STEP 1 Write down the estimated equation: G■■ = 3.2 + 0.07T■

Unit 1 — Basic Econometrics | ECON024 Page 5

 STEP 2 Substitute T = 55: G■ = 3.2 + 0.07 × 55

STEP 3 Calculate: G■ = 3.2 + 3.85 = 7.05

EXAM ANSWER
The predicted grade for a student who studies 55 hours is 7.05.

3.3 Interpreting Multiple Regression Coefficients

In multiple regression (more than one X variable), each slope is interpreted 'holding all other variables
constant' (ceteris paribus). This is crucial.

Given: Y■■ = 240.32 + 7.653X■■ + 19.729X■■

Where Y = Price of refrigerator ($), X■ = Energy efficiency rating, X■ = Capacity (litres)

7.653: Holding capacity constant, each 1-unit increase in energy efficiency

rating increases the refrigerator price by $7.653.
19.729: Holding energy efficiency constant, each additional litre of capacity
increases the price by $19.729.
240.32: Base price when both X■ and X■ are zero (theoretical baseline).

Unit 1 — Basic Econometrics | ECON024 Page 6

 TOPIC 4: TSS, ESS, RSS AND R²

4.1 Decomposition of Total Variation

When we run a regression, the total variation in Y can be split into two parts: the part explained by the
regression and the part not explained (residual). This is written as:

TSS = ESS + RSS

TSS (Total Sum of Squares) = Σ(Y■ − Y■)²
Total variation in Y. How much Y varies around its mean.

ESS (Explained Sum of Squares) = Σ(Y■■ − Y■)²

Variation in Y that is EXPLAINED by the regression line.

RSS (Residual Sum of Squares) = Σ■■² = Σ(Y■ − Y■■)²

Variation in Y that is NOT explained — the leftover errors.

4.2 R² — Coefficient of Determination

R² = ESS / TSS = 1 − RSS/TSS

Range: 0 ≤ R² ≤ 1
R² = 0: The model explains NOTHING. X has no linear relationship with Y.
R² = 1: The model explains EVERYTHING perfectly.
R² = 0.80: The model explains 80% of the variation in Y.

4.3 Adjusted R² (R-bar squared)

In multiple regression, R² always increases when we add more variables, even if those variables are
useless. Adjusted R² corrects for this by penalising for adding extra variables.

R■² = 1 − [(1 − R²)(n−1) / (n−k−1)]

n = number of observations, k = number of regressors (excluding intercept)
Adjusted R² can DECREASE if a useless variable is added
Always use adjusted R² when comparing models with different numbers of variables

EXAM TIP: If a question asks 'why is adjusted R² better than R² in multiple regression?', write:
Because R² always increases when more variables are added, even if they add no real information.
Adjusted R² penalises for unnecessary variables and can decrease, making it a fairer comparison tool.

Unit 1 — Basic Econometrics | ECON024 Page 7

 TOPIC 5: ASSUMPTIONS OF CLRM — ANY THREE

5.1 The 10 Assumptions of the Classical Linear Regression Model

The CLRM is the set of conditions under which OLS gives the best possible estimates. There are 10
assumptions. The exam asks you to discuss ANY THREE in detail.

No
Name Mathematical Form Simple Meaning
.

1 Linearity Y■ = β■ + β■X■ + ε■ Model is linear in parameters (the β's)

2 Non-stochastic X X is fixed X values are not random; fixed in repeated samples

3 Zero mean error E(ε■|X■) = 0 On average, errors are zero — no systematic bias

4 Homoscedasticity Var(ε■) = σ² All errors have the same variance (equal spread)

5 No autocorrelation Cov(ε■, ε■) = 0 Errors are independent of each other

6 Zero Cov(X, ε) Cov(ε■, X■) = 0 Error is unrelated to X

7 n>k n>k More observations than parameters

8 X varies Var(X) > 0 X values must not all be identical

Correct
9 No omitted variables Right variables, right functional form
specification

No perfect
10 No exact X■ = cX■ Explanatory variables not perfectly correlated
multicollinearity

5.2 EXAM ANSWER — 'Discuss Any Three Assumptions'

Below are the three best assumptions to write about. Use this exact format.

■ Assumption 3 — Zero Mean Error

ASSUMPTION 3: Zero Mean of the Error Term
Mathematical form: E(ε■ | X■) = 0

What it means: Given any value of X, the average value of the error term is zero. The model does not
systematically over-predict or under-predict.

Economic example: In a consumption-income regression, Assumption 3 means that on average, the

factors we left out of the model (like family size, age) have no systematic effect on consumption for
any given income level.

What goes wrong if violated: The intercept estimate (β■■) becomes biased. OLS can no longer be
trusted to give the true average effect.

Unit 1 — Basic Econometrics | ECON024 Page 8

 ■ Assumption 4 — Homoscedasticity
ASSUMPTION 4: Homoscedasticity (Equal Variance of Errors)
Mathematical form: Var(ε■ | X■) = σ² for all i (constant for every observation)

What it means: All error terms have the same variance (spread). The errors do not become larger or
smaller as X increases.

Economic example: If we regress household savings on income, richer households have much more
variability in savings than poorer households. As income rises, the spread of the error term also
increases. This VIOLATES homoscedasticity and is called HETEROSCEDASTICITY.

What goes wrong if violated: OLS estimators are no longer EFFICIENT (not minimum variance).
Standard errors are wrong → t-tests and F-tests are invalid.

■ Assumption 5 — No Autocorrelation
ASSUMPTION 5: No Autocorrelation
Mathematical form: Cov(ε■, ε■) = 0 for i ≠ j

What it means: The error in one observation is not related to the error in another observation. This is
mainly an issue in time series data.

Economic example: If GDP in 2020 is over-predicted, GDP in 2021 is also likely to be over-predicted
(because the economy is slow-moving). The errors are correlated over time. This VIOLATES
Assumption 5 and is called AUTOCORRELATION.

What goes wrong if violated: OLS estimators are no longer EFFICIENT. Standard errors are
underestimated → t-statistics are inflated → variables appear significant when they are not (spurious
significance).

EXAM TIP: Always end each assumption answer with: 'When this assumption is violated, OLS
estimators are no longer BLUE (Best Linear Unbiased Estimators), as stated by the Gauss-Markov
Theorem.'

Unit 1 — Basic Econometrics | ECON024 Page 9

 TOPIC 6: BLUE PROPERTIES AND GAUSS-MARKOV
THEOREM

6.1 What Does BLUE Mean?

Under the CLRM assumptions, OLS estimators have a special set of properties captured by the acronym
BLUE. This comes from the Gauss-Markov Theorem.

BLUE = Best Linear Unbiased Estimator

B = BEST: OLS has the minimum variance among all linear unbiased estimators.
It gives the most PRECISE estimates. No other linear unbiased estimator does better.

L = LINEAR: OLS estimators are linear functions of the dependent variable Y.

β■■ = Σk■Y■ where k■ are constants depending only on X.

U = UNBIASED: On average, OLS gives the true population values.

E(β■■) = β■ and E(β■■) = β■
This means the estimates are not systematically too high or too low.

E = ESTIMATOR: It produces estimates of the true population parameters.

6.2 Unbiasedness vs Minimum Variance — In Simple Words

UNBIASEDNESS means: If we took hundreds of different samples and estimated β■ each time, the
average of all those estimates would equal the true β■. The estimates are correct on average, even if any
single estimate might be off.

MINIMUM VARIANCE means: Among all unbiased methods, OLS gives estimates that are closest to the
true value. The estimates are not only correct on average but also as precise (tight) as possible. This is
what 'Best' means in BLUE.

6.3 Gauss-Markov Theorem — The Complete Answer

Unit 1 — Basic Econometrics | ECON024 Page 10

 EXAM ANSWER
GAUSS-MARKOV THEOREM STATEMENT:
Given the assumptions of the Classical Linear Regression Model (CLRM), the OLS estimators β■■
and β■■ are the Best Linear Unbiased Estimators (BLUE) of the true parameters β■ and β■.

CONDITION: All 10 CLRM assumptions must hold for this theorem to apply.
If any assumption is violated, OLS is no longer BLUE. For example:
• If Assumption 4 is violated (heteroscedasticity) → OLS is no longer Best (minimum variance)
• If Assumption 5 is violated (autocorrelation) → OLS is no longer Best
• If Assumption 9 is violated (omitted variable) → OLS is no longer even Unbiased

KEY POINT: Even when OLS loses its BLUE property, it may still be UNBIASED
(as in heteroscedasticity and autocorrelation). It loses 'Best' before it loses 'Unbiased'.

Unit 1 — Basic Econometrics | ECON024 Page 11

 TOPIC 7: MULTIPLE REGRESSION MODEL

7.1 The Model

Multiple regression extends the simple model to include more than one explanatory variable. This is more
realistic because most economic outcomes have multiple causes.

Y■ = β■ + β■X■■ + β■X■■ + ... + β■X■■ + ε■

Each slope coefficient (β■, β■, ...) measures the effect of one variable
holding ALL OTHER variables constant (ceteris paribus).

7.2 SOLVED NUMERICAL — Multiple Regression Interpretation

EXAM QUESTION TYPE: Exam Q2(a) and similar questions. The key is always adding 'holding other
variables constant'.

Given: Y■■ = 240.32 + 7.653X■■ + 19.729X■■

SE = (3.082) (8.992), R² = 0.81, n = 19
Where Y = price of refrigerators ($), X■ = energy efficiency rating, X■ = capacity (litres)

EXAM ANSWER
Intercept (240.32):
When both energy efficiency and capacity are zero, the base price of a refrigerator is $240.32. (This is
a theoretical baseline — a zero-efficiency, zero-capacity fridge does not exist in reality, but the
intercept is needed for the model.)

Slope on X■ (7.653):
Holding refrigerator capacity constant, for every 1 unit increase in energy efficiency rating, the price of
a refrigerator increases by $7.653. More energy-efficient fridges cost more.

Slope on X■ (19.729):
Holding energy efficiency constant, for every 1 additional litre of capacity, the price increases by
$19.729. Larger refrigerators cost more.

R² = 0.81:
81% of the variation in refrigerator prices is explained by energy efficiency and capacity together. The
remaining 19% is due to other factors like brand, design, or warranty.

Unit 1 — Basic Econometrics | ECON024 Page 12

 TOPIC 8: HYPOTHESIS TESTING — t-TEST AND F-TEST

8.1 Why Do We Test Hypotheses?

Just because a coefficient is positive does not mean the variable truly matters. The coefficient might be
positive just by chance (sampling variation). Hypothesis testing tells us whether a result is statistically real
or just noise.

8.2 The t-Test — Testing Individual Coefficients

t-calculated = β■ / SE(β■)
Degrees of freedom (df) = n − k − 1
n = number of observations, k = number of regressors (not counting the intercept)
DECISION RULE:
If |t-calculated| > t-critical → REJECT H■ → coefficient IS significant
If |t-calculated| < t-critical → DO NOT REJECT H■ → NOT significant

One-Tail vs Two-Tail Test

Situation H■ H■ Type When to Use

Variable 'expected to be One-tail

β≤0 β>0 Exam says 'expected positive sign'
positive' (right)

Variable 'expected to be
β≥0 β<0 One-tail (left) Exam says 'expected negative sign'
negative'

Just testing significance β=0 β≠0 Two-tail No direction specified

8.3 SOLVED NUMERICAL — t-Test (from your exam Q2a)

EXAM QUESTION: This exact type appears in every paper. Master this 5-step template.

Given: Y■■ = 240.32 + 7.653X■■ + 19.729X■■

SE = (3.082) (8.992), n = 19, R² = 0.81
The variables X■ and X■ are expected to have positive coefficients. Test at 5% level of significance.

t-Test — Five Step Template

STEP 1 Since 'expected positive' → use ONE-TAIL test For X■: H■: β■ ≤ 0 vs H■: β■ > 0
Hypotheses For X■: H■: β■ ≤ 0 vs H■: β■ > 0

STEP 2 t(β■■) = 7.653 / 3.082 = 2.484 t(β■■) = 19.729 / 8.992 = 2.194

t-statistic

Unit 1 — Basic Econometrics | ECON024 Page 13

 STEP 3 df = n − k − 1 = 19 − 2 − 1 = 16 (n=19 observations, k=2 regressors X■ and X■)
degrees of
freedom

STEP 4 Critical From t-table: t-critical (df=16, one-tail, 5%) = 1.746

Value

STEP 5 For X■: t = 2.484 > 1.746 → REJECT H■ → Significant ✓ For X■: t = 2.194 >
Decision 1.746 → REJECT H■ → Significant ✓

EXAM ANSWER
CONCLUSION:
Both X■ (energy efficiency) and X■ (capacity) have statistically significant positive effects on
refrigerator prices at the 5% level of significance.

The data supports the prior expectation that more energy-efficient and larger refrigerators are more
expensive.

8.4 The F-Test — Testing the Overall Model

The F-test checks if the regression model as a whole is significant, i.e., if at least one explanatory variable
has a non-zero coefficient. It tests all slope coefficients jointly.

H■: β■ = β■ = ... = β■ = 0 (none of the variables matter)

H■: At least one β■ ≠ 0

F = (R² / k) / [(1 − R²) / (n − k − 1)]

Large F value → small p-value → REJECT H■ → model is jointly significant
n = observations, k = number of regressors (excluding intercept)

Quick Example: If R² = 0.81, n = 19, k = 2: F = (0.81/2) / (0.19/16) = 0.405 / 0.011875 = 34.1. This is a
very large F → reject H■ → the model is jointly significant.

Unit 1 — Basic Econometrics | ECON024 Page 14

 QUICK REFERENCE: FORMULAS, RULES & EXAM
STRATEGY

All Unit 1 Formulas at a Glance

Formula Name Formula Key Notes

β■■ = [nΣXY − ΣX·ΣY] / [nΣX² −

OLS Slope Compute 5 sums first: n, ΣX, ΣY, ΣX², ΣXY
(ΣX)²]

OLS Intercept β■■ = Y■ − β■■ × X■ Always calculate slope first, then intercept

R² R² = ESS/TSS = 1 − RSS/TSS Range: 0 to 1; higher = better fit

Adjusted R² R■² = 1 − [(1−R²)(n−1)/(n−k−1)] Penalises for adding useless variables

TSS Identity TSS = ESS + RSS Total = Explained + Residual

t-statistic t = β■ / SE(β■) Compare with t-critical from table

Degrees of freedom df = n − k − 1 For t-test in regression

F-statistic F = (R²/k) / [(1−R²)/(n−k−1)] For overall model significance

Most Important Rules to Remember

Interpreting slope Always say 'holding other variables constant' in multiple regression. Without this,
interpretation is incomplete.

Interpreting R² Always write '___% of variation in Y is explained by the model. The remaining __% is
due to factors not included.'

One-tail or two-tail? One-tail when the question says 'expected positive' or 'expected negative'. Two-tail
otherwise.

df for t-test df = n − k − 1. In the refrigerator example: 19 − 2 − 1 = 16.

When is OLS BLUE? When ALL 10 CLRM assumptions hold. Then Gauss-Markov theorem applies.

What OLS minimises OLS minimises the Residual Sum of Squares (RSS = Σ■■²), not the sum of
residuals.

PRF vs SRF PRF is true (unobservable). SRF is our sample estimate (with hats on β's).

Unbiased vs Efficient Unbiased = correct on average. Efficient = minimum variance. BLUE needs both.

END OF UNIT 1 NOTES — Basic Econometrics (DSE ECON024)

These notes cover the complete syllabus for Unit 1 as per the course handout.

Unit 1 — Basic Econometrics | ECON024 Page 15