Econometrics

An application of statistical methods on an economic data for estimating economic relationships,

creating new models, evaluating economic policies, and testing existing theoretical models.
The data can be obtained as:
1. Experimental and nonexperimental
2. Time series
a. Cross-sectional data – Collected across sample unit on particular point of time
b. Panel Data – Try to follow same sample

Students Performance in Econometrics

⏟
𝑆𝑃𝐸 = 𝑓 (⏟
𝑇𝑖𝑚𝑒 𝑆𝑝𝑒𝑛𝑡 𝐿𝑒𝑎𝑟𝑛𝑖𝑛𝑔, 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐿𝑒𝑐𝑡𝑢𝑟𝑒𝑠 𝐴𝑡𝑡𝑒𝑛𝑑𝑒𝑑 )
Dependent Variable 𝐼𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠

𝑀𝑜𝑑𝑒𝑙 → 𝐷𝑎𝑡𝑎 → 𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑖𝑜𝑛 → 𝑇𝑒𝑠𝑡 → √

𝑆𝑃𝐸 = 𝛽0 + 𝛽1 × 𝑇𝑆𝐿𝐸 + 𝛽2 × 𝑁𝐿𝑆𝐴

• The Data deals with collection and determining the relation of parameters with the
dependent variable whether positive or negative.
• The Estimation is to find the parameters 𝛽0 , 𝛽1 , 𝛽2
• If the Test failed go back to Model phase and recreate or change the mathematical
relationships, add more parameters and collect new data for them

Example: Find statistical data for this 10 grades sample

Econometrics
Grades
2
3
3 Mean = 3.7
3 Mode = 3
4 Median = 4
4 Range = 3
4 Sd = 0.9487
4
5
5

Remark: The standard deviation is an indication of how much the data scattered from the mean
Pearson correlation coefficient
∑(𝑥 𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅)
𝑟𝑥,𝑦 = , −1 ≤ 𝑟𝑥,𝑦 ≤ 1
√∑(𝑥𝑖 − 𝑥̅) 2∑(𝑦𝑖 − 𝑦̅) 2

This coefficient shows the relation and the nature of the two variables

The relation can be:

• Deterministic
• Stochastic
• Independent
r Relation
0.9 < 𝑟𝑥,𝑦 < 1 Very Strong Case Meaning
0.75 < 𝑟𝑥,𝑦 < 0.9 Strong |𝑟𝑥,𝑦 | = 1 Deterministic
0.6 < 𝑟𝑥,𝑦 < 0.75 Moderate 𝑟𝑥,𝑦 = 0 Independent
0.3 < 𝑟𝑥,𝑦 < 0.6 Probability of having −1 < 𝑟𝑥,𝑦 < 1 Stochastic
relation
𝑟𝑥,𝑦 > 0 𝑥↑ 𝑦↑
0 < 𝑟𝑥,𝑦 < 0.3 No probability
𝑟𝑥,𝑦 < 0 𝑥↑ 𝑦↓

Example Assign each correlation coefficient to each drawing bellow:

1. 𝑟𝑥,𝑧 = 0, 2. 𝑟𝑥,𝑦 = 0.85, 3. 𝑟𝑥,𝑦 = 1

4. 𝑟𝑥,𝑧 =? , 5. 𝑟𝑥,𝑦 = −0.75, 6. 𝑟𝑥,𝑦 = −0.9

6 1 2

3 4 5
Another formula for Pearson correlation coefficient
∑(𝑥 𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅ )
𝑟𝑥,𝑦 =
√∑(𝑥𝑖 − 𝑥̅) 2∑(𝑦𝑖 − 𝑦̅) 2 ∑𝑥 𝑖
𝑥̅ =
𝑛
∴ ∑𝑥 𝑖 = 𝑛 ⋅ 𝑥̅ , (1)
𝑆𝑡𝑎𝑟𝑡 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑁𝑢𝑚𝑒𝑟𝑎𝑡𝑜𝑟
∑(𝑥 𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅) , 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑦 …
∑(𝑥 𝑖𝑦𝑖 − 𝑥 𝑖𝑦̅ − 𝑦𝑖 𝑥̅ + 𝑥̅𝑦̅ ), 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦
∑𝑥𝑖 𝑦𝑖 − ∑𝑥𝑖 𝑦̅ − ∑𝑦𝑖 𝑥̅ + ∑𝑥̅𝑦̅, 𝑟𝑒𝑝𝑙𝑎𝑐𝑒 𝑢𝑠𝑖𝑛𝑔 (1)
∑𝑥𝑖 𝑦𝑖 − 𝑛𝑥̅𝑦̅ − 𝑛𝑦̅𝑥̅ + ∑𝑥̅𝑦̅, 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦
∑𝑥𝑖 𝑦𝑖 − 2𝑛𝑥̅𝑦̅ + ∑𝑥̅𝑦̅, 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦
∑𝑥𝑖 𝑦𝑖 − 2𝑛𝑥̅𝑦̅ + 𝑛𝑥̅𝑦̅, 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦
∑𝒙𝒊 𝒚𝒊 − 𝒏𝒚 ̅𝒙̅

𝑇ℎ𝑒 𝐷𝑒𝑛𝑢𝑚𝑒𝑟𝑎𝑡𝑜𝑟
∑(𝑥 𝑖 − 𝑥̅ ) 2
∑𝑥𝑖2 − 2∑𝑥 𝑖 𝑥̅ + ∑𝑥̅ 2 , 𝑟𝑒𝑝𝑙𝑎𝑐𝑒 𝑢𝑠𝑖𝑛𝑔 (1)

∑𝑥𝑖2 − 2𝑛𝑥̅ 𝑥̅ + ∑𝑥̅ 2 , 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦

∑𝑥𝑖2 − 2𝑛𝑥̅ 2 + 𝑛𝑥̅ 2 , 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦
∑𝒙𝟐𝒊 − 𝒏𝒙
̅𝟐

∑𝒙𝒊 𝒚𝒊 − 𝒏𝒚
̅𝒙̅
𝑆𝑜, 𝑟𝑥,𝑦 =
√∑𝒙𝟐𝒊 − 𝒏𝒙
̅ 𝟐 ⋅ ∑𝒚𝟐𝒊 − 𝒏𝒚
̅𝟐
Regression Models

- Simple regression model

One independent variable
▪ Linear
▪ Non-linear
- Multiple regression model
Two or more independent variables
▪ Linear
▪ Non-linear

Model Parameters
Independent Var.
Theoretical Line
formula
y=ß ß
Estimated Line
Dependent Var. Error formula
y=ß ß

Note: We take the square of the vertical distances to eliminate the misleading negative sign

Wrong!
Even the sum of errors is 0
But sum of squares is 25
Correct

-5
5

Elasticity Index
∆𝑦 ∆𝑥
𝜀𝑥𝑦 = :
𝑦 𝑥
∆𝑦 𝑥 ∆𝑦 𝑥
= ⋅ → ⋅
𝑦 ∆𝑥 ⏟∆𝑥 𝑦
𝑆𝑙𝑜𝑝𝑒
𝑥
∴ 𝜀𝑥𝑦 = 𝛽̂1 ⋅ ̂ ̂1𝑥 , 𝑛𝑜𝑡𝑒 𝑡ℎ𝑎𝑡 𝑖𝑡 𝑖𝑠 𝑎 𝑣𝑎𝑙𝑢𝑒 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟, "𝑥 𝑣𝑎𝑙𝑢𝑒 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑"
𝛽0 +𝛽
Another formula for Pearson correlation coefficient
2
∑(𝑦𝑖 − 𝛽̂0 − 𝛽̂1 𝑥 𝑖 ) → 𝑚𝑖𝑛 (𝑓𝑖𝑟𝑠𝑡 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒 𝑤𝑖𝑡ℎ 𝑟𝑒𝑠𝑝𝑒𝑐𝑡 𝑡𝑜 𝛽̂0 )

2 (∑ 𝑦𝑖 − 𝛽̂0 − 𝛽̂1 𝑥 𝑖 ) (−1) = 0 ∑𝑥 𝑖

𝑥̅ =
𝑛
∑ 𝑦𝑖 = 𝛽̂0 𝑛 + 𝛽̂1 ∑ 𝑥 𝑖 , (2) ∴ ∑𝑥 𝑖 = 𝑛 ⋅ 𝑥̅ , (1)

2
∑(𝑦𝑖 − 𝛽̂0 − 𝛽̂1 𝑥 𝑖 ) → 𝑚𝑖𝑛 (𝑓𝑖𝑟𝑠𝑡 𝑑𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒 𝑤𝑖𝑡ℎ 𝑟𝑒𝑠𝑝𝑒𝑐𝑡 𝑡𝑜 𝛽̂1 )

∑ 𝑦𝑖 𝑥 = 𝛽̂0 ∑ 𝑥 𝑖 + 𝛽̂1 ∑ 𝑥 2𝑖 , (3)

∑(𝑥 𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅ )
𝑟𝑥,𝑦 =
√∑(𝑥𝑖 − 𝑥̅) 2∑(𝑦𝑖 − 𝑦̅) 2
∑𝒙𝒊 𝒚𝒊 − 𝒏𝒚
̅𝒙̅
𝑟𝑥,𝑦 =
√∑𝒙𝟐𝒊 − 𝒏𝒙
̅ 𝟐 ⋅ ∑𝒚𝟐𝒊 − 𝒏𝒚
̅𝟐

𝑈𝑠𝑖𝑛𝑔 (2) 𝑓𝑜𝑟𝑚𝑢𝑙𝑎:

∑ 𝑦𝑖 = 𝛽̂0 𝑛 + 𝛽̂1 ∑ 𝑥 𝑖 , (2) , 𝑟𝑒𝑝𝑙𝑎𝑐𝑒 𝑏𝑦 (1) 𝑡ℎ𝑒𝑛 𝑑𝑖𝑣𝑖𝑑𝑒 𝑏𝑦 𝑛
𝑦̅ = 𝛽̂0 + 𝛽̂1 𝑥̅, 𝑡ℎ𝑒𝑛,

̂𝟎 = 𝒚
𝜷 ̂𝟏 𝒙
̅−𝜷 ̅, (𝟒)

𝑆𝑢𝑏. (4) 𝑖𝑛𝑡𝑜 (3)

∑ 𝑦𝑖 𝑥 𝑖 = 𝛽̂0 ∑ 𝑥 𝑖 + 𝛽̂1 ∑ 𝑥 2𝑖 , (3)

∑ 𝑦𝑖 𝑥 𝑖 = (𝑦̅ − 𝛽̂1 𝑥̅) ∑ 𝑥 𝑖 + 𝛽̂1 ∑ 𝑥 2𝑖 , 𝑠𝑖𝑚𝑝𝑙𝑖𝑓𝑦

∑ 𝑦𝑖 𝑥 𝑖 = 𝑦̅ ∑ 𝑥𝑖 − 𝛽̂1 𝑥̅ ∑ 𝑥 𝑖 + 𝛽̂1 ∑ 𝑥 2𝑖 , 𝑠𝑜𝑟𝑡

∑ 𝑦𝑖 𝑥 𝑖 − 𝑦̅ ∑ 𝑥 𝑖 = −𝛽̂1 𝑥̅ ∑ 𝑥 𝑖 + 𝛽̂1 ∑ 𝑥 2𝑖 , 𝑟𝑒𝑝𝑙𝑎𝑐𝑒 𝑏𝑦 (1)

∑ 𝑦𝑖 𝑥 𝑖 − 𝑛𝑥̅ 𝑦̅ = −𝛽̂1 𝑛𝑥̅ 2 + 𝛽̂1 ∑ 𝑥 2𝑖 , 𝑡𝑎𝑘𝑒 𝛽̂1

∑ 𝑦𝑖 𝑥 𝑖 − 𝑛𝑥̅ 𝑦̅ = 𝛽̂1 (∑ 𝑥 2𝑖 − 𝑛𝑥̅ 2 )

∑ 𝑦𝑖 𝑥𝑖 − 𝑛𝑥̅𝑦̅
∴ 𝛽̂1 =
∑ 𝑥𝑖2 − 𝑛𝑥̅ 2

∑( 𝑥−𝑦̅) 2
∴ 𝑟𝑥,𝑦 = 𝛽̂1 ⋅ √∑(
̅) 2
𝑦−𝑦
Price Elasticity of Demand
P
∆𝑸 ∆𝑷
𝜺𝑸,𝑷 = ∶
𝑸 𝑷
∆𝑄 𝑃 32
∴ 𝜀𝑄,𝑃 = ⋅
∆𝑃 𝑄
𝑃 24 ȁ 𝜀ȁ > 1
𝐼𝑓 𝑄 = 32 − 𝑃, 𝑡ℎ𝑒𝑛, 𝜀𝑄,𝑃 = −1 ×
32 − 𝑃
16 ȁ 𝜀ȁ = 1

𝑥̅ 8 ȁ 𝜀ȁ < 1
𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦 𝐼𝑛𝑑𝑒𝑥 𝜀𝑥𝑦 = 𝛽̂1 ⋅
𝛽̂0 + 𝛽̂1 𝑥̅
Q
32

Sum of Squares
𝑦𝑖 − 𝑦̅ = 𝑦𝑖 − 𝑦ො 𝑖 + 𝑦ො𝑖 − 𝑦̅ 𝑦ො 𝑖

𝑆𝑆𝐸
∑(𝑦𝑖 − 𝑦̅)2
⏟ = ∑(𝑦𝑖 − 𝑦ො𝑖 )2
⏟ + ∑(𝑦ො𝑖 − 𝑦̅)2
⏟ 𝑆𝑆𝑇
𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝐸𝑟𝑟𝑜𝑟 (𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙) 𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝑅𝑒𝑔𝑟𝑒𝑠𝑠𝑖𝑜𝑛
𝑺𝑺𝑻 𝑺𝑺𝑬 𝑺𝑺𝑹 𝑆𝑆𝑅
𝑦̅

𝑺𝑺𝑹
𝑺𝑹𝟐 = , 0 ≤ 𝑅2 ≤ 1
𝑺𝑺𝑻
𝑖𝑓 𝑅 2 = 1, 𝑆𝑆𝑅 = 𝑆𝑆𝑇
𝑖𝑓 𝑅 2 = 0, 𝑆𝑆𝑅 = 0

When 𝑅 2 = 0.85 that means: We are able to explain the Y variable around the sample mean by
only 85% using our model

Formula Degrees of Freedom

SSR ∑(𝑦ො 𝑖 − 𝑦̅ ) 2 1
SSE ∑(𝑦𝑖 − 𝑦ො 𝑖 )2 𝑛 −𝑘 − 1
SST ∑(𝑦𝑖 − 𝑦̅ ) 2 𝑛−1
𝑎𝑠 𝒌 𝑖𝑠 𝑡ℎ𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠
𝑑𝑓𝑆𝑆𝑇 = 𝑑𝑓𝑆𝑆𝑅 + 𝑑𝑓𝑆𝑆𝐸
𝑛−1 = 1 + 𝑋

̂𝑖) 2
∑( 𝑦𝑖 −𝑦
𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐸𝑟𝑟𝑜𝑟 𝜎𝑒 = √
𝑛−2
∑( 𝑦𝑖−𝑦̂𝑖 ) 2 𝜎𝑒 ∑𝑥2𝑖
𝜎𝛽̂1 = √( = 𝜎𝛽̂0 = 𝜎𝑒 ⋅ √
𝑛−2) ⋅∑( 𝑥−𝑥̅) 2 √( 𝑥𝑖 −𝑥̅ ) 2 𝑛⋅∑(𝑥−𝑥̅) 2

the average differences between 𝛽̂1 and the sample 𝛽1 is = 0.0000

• 𝑛 ↑ 𝜎𝛽̂1 ↓
• (𝑥𝑖 − 𝑥̅ ) ↑ 𝜎̂𝛽1 ↓
• (𝑦𝑖 − 𝑦̂𝑖 )2 ↑ 𝜎𝛽̂ ↑
1
Formulas Summary
∑(𝑥 𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅)
𝑟𝑥,𝑦 = , −1 ≤ 𝑟𝑥,𝑦 ≤ 1
√∑(𝑥𝑖 − 𝑥̅) 2∑(𝑦𝑖 − 𝑦̅) 2

∑ 𝑦𝑖 𝑥 𝑖 − 𝑛𝑥̅ 𝑦̅
𝛽̂1 =
∑ 𝑥 2𝑖 − 𝑛𝑥̅ 2
𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅,

𝑦̂𝑖 = 𝛽̂0 + 𝛽̂1 ⋅ 𝑥 𝑖 linear correlation coefficient

for the slop will start with negative value until reach one point to get positive due to Bo is
negative

If the GDP increased by 1 EUR then the AGAE will increase by 1,24974935917592

𝑥
𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦 𝐼𝑛𝑑𝑒𝑥 𝜀𝑥𝑦 = 𝛽̂1 ⋅ ̂ ̂1𝑥 , only for 1%
𝛽0 +𝛽

If the X increased by 1% the Y also increases by %

∑(𝑦𝑖 − 𝑦̅) 2
⏟ = ∑(𝑦𝑖 − 𝑦ො 𝑖 )2
⏟ + ∑(𝑦ො 𝑖 − 𝑦̅) 2
⏟
𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝑇𝑜𝑡𝑎𝑙 𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝐸𝑟𝑟𝑜𝑟 (𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙) 𝑆𝑢𝑚 𝑜𝑓 𝑆𝑞𝑢𝑎𝑟𝑒𝑠 𝑅𝑒𝑔𝑟𝑒𝑠𝑠𝑖𝑜𝑛
𝑺𝑺𝑻 𝑺𝑺𝑬 𝑺𝑺𝑹

𝑆𝑆𝑅
𝑅2 = We are able to explain the Y variable around the sample mean by only 00% using
𝑆𝑆𝑇
our model

There is a strong linear association between the dependent and independent variable.
𝑟𝑥𝑦2=0.985232=𝑅2

we know that the association between the two indicators is valid if there is only one independent
variable and one dependent variable.

̂𝑖) 2
∑( 𝑦𝑖 −𝑦
𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐸𝑟𝑟𝑜𝑟 𝜎𝑒 = √ , The average differences between the estimated and the
𝑛−2
observed y is value of sigma e
 ∑( 𝑦𝑖−𝑦̂𝑖 ) 2 𝜎𝑒
𝜎𝛽̂1 = √( = , the average differences between 𝛽̂1 and the sample 𝛽1 is =
𝑛−2) ⋅∑( 𝑥−𝑥̅) 2 √∑( 𝑥𝑖 −𝑥̅) 2

Which go very close to zero indicating that Our estimation of B1 is describe most of the data
from our model

∑𝑥2𝑖
𝜎𝛽̂0 = 𝜎𝑒 ⋅ √
𝑛⋅∑(𝑥−𝑥̅) 2

𝛽̂𝑘
𝑡=
𝜎𝛽̂𝑘

Ex: 0.2096 ≤ β1 ≤ 0.425

As we can see the interval does not include the zero value, it means that the β1 parameter
cannot be zero at 95% confidence level.