Chapter-5

TWO-VARIABLE REGRESSION: INTERVAL ESTIMATION AND HYPOTHESIS TESTING

INTERVAL ESTIMATION: SOME BASIC IDEAS

Concept Alignment: Why We Move Beyond a Single Number

In the consumption–income example, the estimated Marginal Propensity to Consume (MPC) is 0.5091. This
value is a point estimate—essentially a single best guess of the true population MPC.

From a decision-making perspective, this number alone does not provide enough assurance. Due to sampling
fluctuations, different samples would produce slightly different MPC estimates. While the estimator is
unbiased (on average it equals the true MPC over repeated samples), any one estimate may deviate from the
actual value.

Measuring Reliability: Standard Error

The reliability of a point estimate is assessed using its standard error.

 Smaller standard error → higher confidence in the estimate

 Larger standard error → greater uncertainty

Instead of relying solely on the point estimate, we create a buffer zone around it using the standard error.

Interval Estimation: Risk-Managed Insight

To better manage uncertainty, we construct an interval around the estimated MPC:

This interval is designed so that it captures the true MPC with a specified level of confidence.

Formally:

Where:

 Confidence Interval → the range itself

 Confidence Coefficient (1 − α) → how reliable the method is (e.g., 95%)
  Level of Significance (α) → acceptable risk of error (e.g., 5%)
 Confidence Limits → lower and upper bounds of the interval

In practice, α and (1 − α) are usually expressed as percentages.

Such an interval, if it exists, is known as a confidence interval; 1 − α is known as the confidence coefficient; and
α (0 <α< 1) is known as the level of significance.2 The endpoints of the confidence interval are known as the
confidence limits (also known as critical values), βˆ 2 − δ being the lower confidence limit and βˆ 2 + δ the
upper confidence limit. In passing, note that in practice α and 1 − α are often expressed in percentage forms as
100α and 100(1 − α) percent

What This Delivers Strategically

If α = 0.05, the interpretation is:

Using this method, 95% of similarly constructed intervals will contain the true MPC in the long run.

This means the interval estimator provides a controlled range of plausible values, not just a single estimate.

Key Governance Principles of Interval Estimation

To avoid misinterpretation, it is critical to align on how confidence intervals should be understood from a risk
and process perspective.

1. Confidence Is About the Method, Not the Number

Equation (5.2.1) does not mean there is a 1−α1-\alpha1−α probability that the true parameter β2\beta_2β2 lies
inside the calculated interval.

Why?

 β2\beta_2β2 is unknown but fixed

 It either lies inside the interval or it does not

What the equation actually communicates is this:

The procedure used to construct the interval is designed to capture the true β2 1−α proportion of the time.

So the confidence level reflects method reliability, not uncertainty about the parameter itself.
2. Why the Confidence Interval Is Random

The confidence interval depends on β^2\hat{\beta}_2β^2, which is derived from sample data.

 Different samples → different estimates

 Different estimates → different intervals

Hence, before observing the data, the interval itself is random and variable.

3. Probability Is a Long-Run Performance Metric

Because the interval is random, probability statements must be interpreted in a long-run, repeated-sampling
framework.

Operational meaning:

If we repeatedly draw samples and construct intervals using the same rules, then approximately 1−α1-\
alpha1−α of those intervals will contain the true parameter.

This is a process-level guarantee, not a one-off assurance.

4. Once Data Is Observed, Probability Collapses

After we:

 Fix a sample
 Compute a specific value of β^2

The interval becomes fixed, not random.

At this stage:

 The true β2\beta_2β2 is either inside the interval or outside

 The probability is therefore 1 or 0, not 95%

For example, if the 95% confidence interval is:

0.4268≤β2≤0.59140

You cannot say:

“There is a 95% chance the true MPC lies in this range.”

The correct interpretation is:

“This interval was produced by a method that captures the true MPC 95% of the time in repeated applications.”
 HYPOTHESIS TESTING: GENERAL COMMENTS

Hypothesis Testing — Simple, clear explanation (with details)

Hypothesis testing is a way in statistics to answer a basic question:

Does our sample evidence support a claim about a population, or not?

Think of it as a formal method to check whether what we observe in data is consistent with what we believe
or whether the data suggests something different.

1. The basic idea

We usually start with a claim or assumption about a population. This claim is called a:

✅ Null hypothesis → (H_0)

 It represents the existing belief, status quo, or theory

 We assume it is true unless evidence strongly suggests otherwise

Example:

 A theory says: income and consumption move one-to-one

So, the slope β₂ = 1
 From data, we estimate β̂ ₂ = 0.5091
 Question: Is 0.5091 “close enough” to 1?
If yes → we do not reject (H_0)
If no → we reject (H_0)

Here “close enough” is measured statistically, not by eye.

2. Alternative hypothesis
If we doubt the null hypothesis, we frame an alternative:

🔄 Alternative hypothesis → (H_1)

It says something different from the null — for example:

 β₂ is not equal to 1 (two-sided test)

  β₂ is greater than 1 (right-tailed test)
 β₂ is less than 1 (left-tailed test)

It can be:

 simple: one exact value (β₂ = 1.5)

 composite: a range (β₂ ≠ 1)

We never try to prove (H_0) is true; we only see whether the sample gives enough evidence to reject it.

3. What does “compatible with hypothesis” mean?

“Compatible” means:

The observed value is not too far from the hypothesized value, according to probability theory.

Because sample data naturally varies, even if the hypothesis is true, estimates will not be exactly equal. So we
ask:

 Is the difference small and explainable by sampling variation?

👉 do not reject
 Is the difference too large to be explained by chance?
👉 reject

This is why probability distributions matter.

4. What hypothesis testing needs

To test hypotheses, we must know the sampling distribution of our estimator.

Example:

 Under normality assumption, β̂ ₂ is normally distributed

 Mean of distribution = true β₂
 Variance known from formula

When we hypothesize β₂ = 1, we are making a statement about:

→ the mean of that distribution

Most economic and econometric tests examine parameters such as:

 means
 variances
  regression slopes
 correlations

Using distributions like:

 Normal
 t
 F
 Chi-square (χ²)

5. Two main approaches to hypothesis testing

✅ (1) Confidence interval approach

Steps:

1. Build a confidence interval (say 95%) for the parameter

2. See whether hypothesized value lies inside it

 If inside → do not reject (H_0)

 If outside → reject (H_0)

Example:

 95% CI for β₂ is (0.40, 0.75)

 Hypothesized β₂ = 1
 Since 1 is outside, we reject (H_0)

✅ (2) Test of significance (p-value approach)

This is more common.

Steps:

1. Assume (H_0) is true

2. Compute test statistic (like t, z, F)
3. Ask: How likely is our observed result under (H_0)?
4. This probability = p-value

Decision:

 p ≤ α (e.g., 0.05) → reject (H_0)

 p > α → do not reject (H_0)

α is the significance level, usually 1%, 5%, or 10%

6. Errors in hypothesis testing
Because we use samples, mistakes are possible.

❌ Type I error

Rejecting (H_0) when it is actually true

Probability = α (significance level)

❌ Type II error

Not rejecting (H_0) when it is actually false

We choose α to control Type I error.

7. What decision language we use (very important)

We never say:

 “Accept the null hypothesis”

Instead we say:

 Reject (H_0) → strong evidence against it

 Do not reject (H_0) → not enough evidence to oppose it

Because we can never be 100% certain — we only work with probabilities.

8. Short practical example

Claim:

 Average monthly mobile bill is 500 BDT

Null hypothesis:

 (H_0: \mu = 500)

Data sample says:

 mean = 560
We test:

 Is 560 close enough to 500?

 Or is the gap too large to be explained by chance?

If gap is statistically large → reject

If it is small → do not reject

HYPOTHESIS TESTING: THE CONFIDENCE-INTERVAL APPROACH

⭐ Hypothesis Testing Using the Confidence-Interval Approach

(Two-Sided / Two-Tail Test — Explained Simply)

🎯 What are we trying to do?

We want to check whether the true value of a parameter (like β₂, the marginal propensity to consume) is
equal to some hypothesized value.

So we set up:

 Null hypothesis (H₀): β₂ = 0.3

→ This is the claim we test
 Alternative hypothesis (H₁): β₂ ≠ 0.3
→ Means β₂ could be either less than or greater than 0.3

Because the alternative allows both directions, this is called a:

✅ Two-sided (or two-tailed) test

This type is often used when we do not have a strong reason to expect the parameter to be specifically higher
or lower — only that it may be different.

🔍 Using a Confidence Interval Instead of a Test Statistic

We already have from our regression:

 Estimated β̂ ₂ (sample MPC) = 0.5091

And suppose we have a 95% confidence interval for β₂:

📦 95% Confidence Interval = (0.4268 , 0.5914)

This interval means:

If we repeatedly sampled many times, about 95% of those intervals

would contain the true value of β₂.

So this interval represents plausible values for β₂.

🧠 Key Decision Rule (Very Important)

✔ If the hypothesized value (under H₀) is inside the confidence interval

→ Do NOT reject H₀

❌ If the hypothesized value is outside the interval

→ Reject H₀

✅ Apply it to our example

Hypothesized value under H₀:

β₂ = 0.3

But our confidence interval is:

(0.4268 , 0.5914)

Now check:

 Is 0.3 inside this interval?

No. It lies outside.

So the conclusion is:

❌ Reject H₀ at the 5% significance level

→ There is strong evidence that β₂ is not equal to 0.3
📉 What does this mean in plain English?

If the true MPC were actually 0.3, then observing a value as high as 0.5091

 would be very unlikely

 would happen less than 5% of the time just by chance

So we conclude:

👉 The data disagrees with the claim that MPC = 0.3

👉 The result is statistically significant at the 5% level

📌 What does “statistically significant” mean?

✔ When we reject the null hypothesis

We say:

The result is statistically significant

This means:

 the difference is too large to be explained by random chance

 given our chosen significance level (e.g., 5%)

✔ When we do NOT reject the null

We say:

The result is not statistically significant

Meaning:

 the data is still consistent with the null hypothesis

 even if not perfectly equal

🔥 “Highly statistically significant” — what is that?

Sometimes you see phrases like:

 “statistically significant”
 “highly statistically significant”

These usually refer to how small the probability of error is:

 Term Typical Meaning

Statistically significant α = 5%

Highly significant α = 1% or smaller

So:

 5% → moderate evidence
 1% → very strong evidence

But many statisticians prefer reporting the p-value and letting the reader decide.

🎯 Summary (Easy to Remember)

Confidence-interval approach to hypothesis testing:

1. Construct a confidence interval for the parameter

2. Compare hypothesized value with interval
3. Decision:

If hypothesized value… Decision

Inside CI Do NOT reject H₀

Outside CI Reject H₀

✅ In our case

 β̂ ₂ = 0.5091
 95% CI = (0.4268 , 0.5914)
 Hypothesized β₂ = 0.3

Since 0.3 is outside the interval:

👉 We reject H₀ at 5% significance

👉 The result is statistically significant

👉 MPC is not equal to 0.3

Testing the Significance of Regression Coefficients: The t Test

When we run a regression, we estimate coefficients such as β₂ (for example, the marginal propensity to
consume). But estimates from a sample may differ from the true population value. So we must check:

Is the estimated coefficient really different from the hypothesized value, or could the difference just be
due to sampling chance?

The t-test is the main statistical tool used for this purpose.

🔹 1. What is a test of significance?

A test of significance evaluates whether the sample result supports or contradicts the null hypothesis (H₀).

 Null hypothesis (H₀): a specific assumed value for the parameter

Example: β₂ = 0.3
 Alternative hypothesis (H₁): states the parameter is different
Example (two-sided): β₂ ≠ 0.3

We use a test statistic — a number computed from the sample — and compare it to a known probability
distribution. Based on this, we decide:

✔ Do not reject H₀
❌ Reject H₀

🔹 2. The t-statistic for a regression coefficient

Under standard regression assumptions, the statistic

follows a t-distribution with n − 2 degrees of freedom.

Where:

 β^2\hat{\beta}_2β^2 = estimated coefficient from sample

 β2\beta_2β2 = hypothesized value under H₀
 se(β^2)se(\hat{\beta}_2)se(β^2) = standard error of the estimate
 df = n − 2 (because two parameters are estimated in simple regression)

This t-statistic tells us how many standard errors the estimate is away from the hypothesized value.
  If t is close to 0 → estimate is near the hypothesized value
 If |t| is large → estimate is far from the hypothesized value

Large |t| → stronger evidence against H₀

🔹 3. Link to the confidence-interval approach

A 95% confidence interval creates a range of plausible values.
The t-test does the same thing, but from another angle:

 Confidence interval → checks whether the hypothesized value lies inside the interval
 t-test → checks whether the computed t-value lies in the rejection region

These are mathematically equivalent.

🔹 4. The acceptance and rejection regions

For a two-tailed test at significance level α:

 Critical values are ± tα/2

 The middle area = acceptance region
 The two outer tails = rejection (critical) regions

If

👉 Do NOT reject H₀

If

👉 Reject H₀

This means the sample estimate is too far from the hypothesized value to be explained by chance alone.

🔹 8. Two-tailed vs One-tailed t-tests

✅ Two-tailed test
Used when alternative is:

H1:β2≠0.3

We consider both tails of the distribution.

✅ One-tailed test
Used when prior theory suggests direction.

Example:

 If theory says MPC should be greater than 0.3

Then:

 H₀: β₂ ≤ 0.3
 H₁: β₂ > 0.3

Now we only use the right tail.

Critical value is:

t0.05

(not t0.025)

So:

 Only large positive t values lead to rejection

 Negative values do not matter

Choice of test depends on theory and prior evidence.

 Testing the Significance of σ2: The χ2 Test

⭐ The Chi-Square Test of Significance — Explained Simply (and in Detail)

So far, you’ve seen how hypothesis testing works using the t-test for regression coefficients. Now here’s
another important example of the test-of-significance approach — this time involving the variance (σ²).

The test used here is called the:

✅ Chi-Square (χ²) Test of Significance

This test is used when we want to check:

Is the population variance σ² equal to some hypothesized value, or not?

🔹 1. The Chi-Square Test Statistic

The “Zero” Null Hypothesis and the “2-t” Rule of Thumb
 The Choice between Confdence-Interval and

Test-of-Signifcance Approaches to Hypothesis Testing

1️⃣ Test-of-Significance Approach

Main idea:
Focuses on whether we can reject the null hypothesis using statistical tests (like the t-test).

How it works:

 Set up a null hypothesis (e.g., β₂ = 0)

 Treat it like a “straw man” and try to reject it
 If the result is statistically significant, we say:

The coefficient is different from zero

Problem highlighted in the text:

 In economics, most null hypotheses are unlikely to be exactly true

 Yet researchers still focus on rejecting them
 Journals often prefer results that reject the null
 This shifts attention away from how large or meaningful the coefficient really is

So the focus becomes:

✔️“Is it zero or not?”
❌ rather than “How big is it, and does it matter?”

2️⃣ Confidence-Interval Approach

Main idea:
Focuses on the range of plausible values for a parameter (like β₂).

How it works:

 Instead of just testing β₂ = 0

 We estimate an interval (e.g., 0.5 to 0.9)
 This tells us:

What values the data rule out

And what values remain plausible

What the authors argue:

 Confidence intervals help us judge economic meaning, not just statistical detection
  They show:
✔️The size of the effect
✔️The uncertainty around it
✔️Which parameter values are excluded by evidence

This supports the broader scientific goal:

understanding which models are good approximations to reality

✅ Key Difference Summarized

Approach Main Question Focus Limitation (from text)

Test-of- Is the estimate statistically different Rejecting or not Overemphasizes rejection;

Significance from the null value (usually 0)? rejecting H₀ ignores size & meaning

Confidence- What range of parameter values are Magnitude & precision

Requires interpretation effort
Interval consistent with the data? of estimates

how to report regression result and evaluate regression result

1. Reporting Regression Results

When reporting regression results, the objective is to clearly communicate the model, estimates, and their
implications.

First, the regression model should be stated, identifying the dependent variable and the independent
variables, along with the sample size and data period.

Second, the regression output should be presented in a table including:

 Estimated coefficients
 Standard errors
 t-statistics or p-values
 R² and Adjusted R²
 Number of observations

Each coefficient should be briefly interpreted. A coefficient shows the expected change in the dependent
variable due to a one-unit change in the independent variable, holding other variables constant.

Third, the statistical significance of coefficients should be reported using t-tests and conventional significance
levels (1%, 5%, or 10%). This indicates whether an explanatory variable has a statistically meaningful effect on
the dependent variable.
Finally, the overall fit of the model should be reported using R² and Adjusted R², which measure how much of
the variation in the dependent variable is explained by the model.

2. Evaluating Regression Results

Evaluation focuses on determining whether the regression results are reliable and meaningful.

First, the economic or theoretical consistency of the results should be checked. The signs and magnitudes of
coefficients should be consistent with economic theory and logical expectations.

Second, statistical reliability should be assessed. Individual coefficients are evaluated using t-tests, while the
overall significance of the model is tested using the F-test.

Third, the goodness of fit of the model is evaluated using R² and Adjusted R². A higher value indicates greater
explanatory power, though a low R² may still be acceptable depending on the nature of the data.

Fourth, diagnostic tests should be conducted to verify regression assumptions, including:

 Multicollinearity
 Heteroskedasticity
 Autocorrelation (in time-series data)
 Normality of error terms

Violations of these assumptions may require model re-specification or corrective measures.

Finally, the practical usefulness of the model should be considered by assessing whether the results are
meaningful for prediction, policy analysis, or decision-making.

Normality Tests

Although several tests of normality are discussed in the literature, we will consider just three: (1) histogram of
residuals; (2) normal probability plot (NPP), a graphical device; and (3) the Jarque–Bera test.

1. Histogram of Residuals

A histogram of residuals is a simple graphical tool used to assess the shape of the probability distribution
function (PDF) of a random variable, particularly the OLS residuals.

In this method, the values of the residuals are divided into appropriate class intervals on the horizontal axis. For
each interval, a rectangle is drawn whose height represents the frequency of observations within that interval.

By visually comparing the histogram with a bell-shaped normal distribution curve, one can evaluate whether
the residuals approximately follow a normal distribution. If the histogram appears symmetric and bell-shaped,
the normality assumption of the regression model is likely to be reasonable.
Therefore, plotting a histogram of residuals is considered a practical and preliminary diagnostic technique
for checking the normality assumption in regression analysis.

[Link] the consumption–income regression example, we test the hypothesis that the residuals are normally
distributed. This hypothesis is essential for the validity of statistical inference in the classical linear regression
model.

Figure 5.7 illustrates the parameters of the normal distribution fitted to the residuals. The estimated mean of
the residuals is approximately zero, which is consistent with the OLS assumption, and the standard
deviation is about 6.12, indicating the dispersion of the residuals around the mean.

These characteristics suggest that the residuals reasonably conform to a normal distribution, thereby supporting
the normality assumption required for hypothesis testing.

3. Jarque–Bera (JB) Test of Normality

The Jarque–Bera (JB) test is an asymptotic (large-sample) test used to examine whether the OLS residuals
are normally distributed. The test is based on the skewness (S) and kurtosis (K) measures of the residuals.

The JB test statistic is defined as:

where

 nnn = sample size

 SSS = coefficient of skewness
 KKK = coefficient of kurtosis

For a normally distributed variable, skewness equals 0 and kurtosis equals 3. Therefore, the JB test evaluates
the joint hypothesis that S=0S = 0S=0 and K=3K = 3K=3. Under this condition, the JB statistic is expected to
be close to zero.

Under the null hypothesis of normality, Jarque and Bera demonstrated that, in large samples, the JB statistic
follows a chi-square distribution with 2 degrees of freedom. If the computed p-value is sufficiently low, the
null hypothesis of normality is rejected. Conversely, a high p-value implies that the normality assumption
cannot be rejected.

In the consumption–income example, the sample size is relatively small; therefore, the JB test should be
interpreted with caution. Nonetheless, applying the JB formula yields a test statistic of 0.7769. The
corresponding p-value from the chi-square distribution with 2 degrees of freedom is approximately 0.68, which
is relatively high.

Hence, we do not reject the null hypothesis of normality for the residuals in this example, while recognizing
the limitation imposed by the small sample size.

give me a hypothetical regression and interpret them

Hypothetical Regression and Interpretation
Assume the following OLS regression model:

Interpretation
The intercept (20) represents baseline consumption when income and household size are zero.

The income coefficient (0.6) indicates that, holding household size constant, a one-unit increase in income
increases consumption by 0.6 units. This effect is statistically significant and economically meaningful.

The household size coefficient (0.4) suggests that an additional household member increases consumption by
0.4 units, ceteris paribus, and is statistically significant.

The R² value of 0.75 shows that the model explains 75% of the variation in consumption expenditure. Overall,
the regression model is statistically sound, theoretically consistent, and suitable for interpretation.

Chapter-1
 1. what is econometrics about? Briefly discuss its methodogies

✅ What is Econometrics?
Econometrics is a branch of economics that uses mathematics, statistics, and data to test economic theories
and measure economic relationships.
In simple words:

Econometrics = Economics + Statistics + Data

✅ Main Methodologies of Econometrics (Brief)

1️⃣ Formulating an Economic Model

Start with an economic theory

Example: Higher income increases consumption.

Write it mathematically
Consumption = f(Income)

2️⃣ Specifying the Econometric Model

Convert the theory into a statistical equation

Example:

(where uuu is the error term)

3️⃣ Collecting Data

Use:

 Time series data (over time)

 Cross-sectional data (many people/firms)
 Panel data (both)

4️⃣ Estimating Parameters

Use statistical tools (like OLS regression) to estimate unknown values such as β0, β1.

5️⃣ Hypothesis Testing

Check whether relationships are statistically significant

Example: Is income really affecting consumption?

6️⃣ Model Evaluation

See if the model fits well:

 R²
 Residual analysis
 Diagnostic testing

7️⃣ Forecasting & Policy Use

Use the model to:

 Predict future values

 Support economic decision-making

2. define econometrics and use of econometrics

Econometrics is a branch of economics that uses mathematics, statistics, and real-world data to measure
and analyze economic relationships.
It helps test economic theories and estimate how one economic variable affects another.

✔️Uses of Econometrics
Econometrics is used to:
 1. Test economic theories
e.g., Does higher income increase consumption?
2. Estimate economic relationships
e.g., How much does demand change when price changes?
3. Forecast future trends
e.g., Predict GDP, inflation, unemployment, sales, etc.
4. Evaluate government and business policies
e.g., Impact of tax cuts or subsidies
5. Aid decision-making for businesses and policymakers
e.g., pricing, investment, budgeting

3. explain the concept of central limit theorem or monte carlo experiment

✅ Central Limit Theorem (CLT) — Simple Explanation

🌟 Main Idea

When you take many samples from any population and calculate their averages, the distribution of those
averages becomes approximately normal (bell-shaped), even if the original data are not normal — as long as
the sample size is large enough.

🔹 Key Points

 Works for any population distribution (normal or not)

 Sample size usually n ≥ 30 is enough
 The mean of sample means = population mean
 The spread (standard error) decreases as sample size increases

🎯 Why it matters?

Because of CLT, we can:

✔ use normal distribution
✔ construct confidence intervals
✔ perform hypothesis testing

—even when the population is not normal.

🧠 Simple Example

Suppose exam marks are very irregular (not bell-shaped).

1. Take many random samples of 30 students each

 2. Calculate the average mark for each sample
3. Plot those averages

👉 The graph will look bell-shaped.

✅ Monte Carlo Experiment — Simple Explanation

🌟 Main Idea

Monte Carlo experiments use repeated random simulations on a computer to study how a statistical
method or model behaves.

It’s like running thousands or millions of “what-if” experiments.

🔹 How it works

1. Create a model
e.g., Y = 2 + 3X + error
2. Generate random data for X
3. Add random error
4. Estimate the model
5. Repeat thousands of times
6. Study the results

🎯 Why it is useful?

Monte Carlo helps economists:

✔ test how methods work in practice
✔ understand sampling variation
✔ check estimator accuracy
✔ study situations where theory is hard

[Link] the methodology of econometrics

Same as no 1
 5. distinguish between time series and cross sectional data? What is MPC and elasticity

Point Time Series Data Cross-Sectional Data

Data collected on one variable (or unit) Data collected on many units (people, firms,
Meaning
over different points in time states, etc.) at one point in time
GDP from 1990–2025, daily stock prices, Income of households in 2025, egg production
Examples
monthly unemployment rate across 50 states in 1990
Time Dimension Yes — varies over time No — same time period
Common
Daily, weekly, monthly, quarterly, yearly One-time surveys, censuses, polls
Frequency
Typical Use Trend, forecasting, economic cycles Comparing differences across groups/regions
Stationarity (mean & variance may change Heterogeneity (units differ a lot in
Main Issue
over time) size/behavior)

✅ What is MPC?
MPC = Marginal Propensity to Consume

👉 It measures how much consumption changes when income changes by 1 unit.

Example:
If income increases by $100 and spending increases by $80:

MPC = 80 / 100 = 0.8

So people spend 80% of extra income.

✅ What is Elasticity?
Elasticity measures how responsive one variable is to a change in another variable.

Most common:

 Price elasticity of demand

how much quantity demanded changes when price changes

Example:
If price rises by 10% and demand falls by 20%
Elasticity = –2

Meaning → demand is highly responsive.

 [Link] between mathematical equation and statistical model?

Point Mathematical Equation Statistical Model

Nature Exact and deterministic Based on probability and uncertainty
Shows a perfect relationship between Shows a real-world relationship with
Meaning
variables randomness
Error Term No error term Includes an error term (u or ε)
Example ( y = 2x + 5 ) ( y = 2x + 5 + u )
Predictability Same output for same input always Output may vary even with same input
Use Pure math and theory Real-world data analysis (like econometrics)
Assumptions Variables are exact Variables are random/stochastic
Application Physics, engineering, algebra Economics, finance, social sciences

Types of data

✅ Types of Data

1️⃣ Time Series Data

📌 Meaning

Time series data are observations on a variable collected over different points in time.

⏳ Time intervals may be:

 Daily → stock prices

 Weekly → money supply
 Monthly → unemployment rate, CPI
 Quarterly → GDP
 Yearly → government budget
 Every 5 or 10 years → census

Even real-time (continuous) data exists, such as live stock prices.

⚠️Special Issue: Stationarity

A time series is usually assumed to be stationary, meaning:

 mean does not change over time

 variance does not change over time
But in reality, many economic time series trend upward or downward, so they may be non-stationary
(example: money supply).

2️⃣ Cross-Section Data

📌 Meaning

Cross-section data are observations on one or more variables collected at a single point in time.

Examples

 Census of population in a given year

 Consumer expenditure survey
 Public opinion polls
 Egg production and price in 50 U.S. states in 1990

So, data are collected across individuals, firms, households, states, etc. at one time.

⚠️Special Issue: Heterogeneity

Different units may be very different in size or behavior

(e.g., large vs small states).
This scale difference must be considered in analysis.

3️⃣ Pooled Data (Combined Data)

📌 Meaning

Pooled data combine both time series and cross-section data.

Example from the text:

 Egg production & prices in 50 states

 For two years

So we have:

 50 cross-section observations per year

 Repeated for more than one year

Total = 100 observations

Another example: CPI of several countries over many years.

4️⃣ Panel (Longitudinal / Micropanel) Data
📌 Meaning

A special type of pooled data where

the same units are followed over time.

Example

Surveying the same households every few years about:

 income
 housing
 spending

So:

 Same people/firms/families
 Observed repeatedly over time

🎯 Why useful?

It helps study changes in behavior over time.

The Sources of Data

✅ Sources of Data

Data used in empirical (real-world) economic analysis can come from many places, such as:

✔ Government Agencies

Example: Department of Commerce

They collect data like GDP, unemployment, inflation, trade, etc.

✔ International Organizations
Examples:

 IMF (International Monetary Fund)

 World Bank

They collect and publish data on:

 global economies
 development indicators
 financial statistics

✔ Private Organizations

Example:

 Standard & Poor’s (S&P)

They collect data such as:

 stock prices
 credit ratings
 financial market indicators

✔ Individuals or Researchers

Sometimes data is collected through:

 surveys
 experiments
 academic research

So, data may come from public, private, or individual sources.

🌐 The Role of the Internet

The internet has revolutionized data collection.

Now you can:

 search topics (e.g., “exchange rates”)

 find multiple data sources instantly
 download most data easily and often free

Websites with economic & financial data are widely available.

✅ Types of Data Based on Collection Method

1️⃣ Experimental Data

These are collected under controlled conditions.

The researcher tries to hold some factors constant while changing others.

Example

Studying the effect of obesity on blood pressure while keeping:

 smoking
 drinking
 eating habits

constant.

👉 This helps isolate the true effect.

Experimental data are common in natural sciences.

2️⃣ Non-Experimental (Observational) Data

These are not controlled by the researcher.

The researcher simply observes existing real-world data.

Examples

 GDP
 unemployment
 inflation
 stock prices

These are typical in economics and social sciences.

⚠️Problem

Since the researcher cannot control conditions, it becomes hard to determine cause-and-effect relationships.

Example question:

Does money supply cause GDP to rise, or does GDP growth increase money supply?

This uncertainty creates econometric challenges.