Scribd
Upload
3 views

Assignment 2- 418-with answers - Copy - Copy (2)

The document discusses econometric methods, particularly focusing on the standard normal distribution and its application in estimating economic growth based on trade. It highlights issues such as heterogeneity, autocorrelation, and endogeneity in regression analysis, and introduces the instrumental variable approach to address these problems. Additionally, it explains the differences between fixed effects and OLS models in controlling for unobserved heterogeneity and the implications for variance estimation of beta coefficients.

Uploaded by

Mostafa Allam
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
3 views

Assignment 2- 418-with answers - Copy - Copy (2)

The document discusses econometric methods, particularly focusing on the standard normal distribution and its application in estimating economic growth based on trade. It highlights issues such as heterogeneity, autocorrelation, and endogeneity in regression analysis, and introduces the instrumental variable approach to address these problems. Additionally, it explains the differences between fixed effects and OLS models in controlling for unobserved heterogeneity and the implications for variance estimation of beta coefficients.

Uploaded by

Mostafa Allam
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Assignment II

Econometric Methods (4081)


Answer 1:

- PDF function of Standard Normal Distribution:


No 𝜎 introduced in the function as the standard normal distribution assumes 𝜎 = 1 and 𝜇 = 0
1 1
exp (− (𝑦 − 𝑦̂)2 )
√2 𝜋 2
1 1 2
exp (− (𝑦 − 𝛼̂ − 𝛽̂ 𝑋) )
√2 𝜋 2

𝑀𝑎𝑥𝛼̂ = 𝑌̅ − 𝛽̂ 𝑋̅ → 𝐼𝑓 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑁𝑜𝑟𝑚𝑎𝑙 𝐷𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 𝛼̂ = 0

∑𝑋𝑖 𝑌𝑖
𝛽̂ = →𝜇=0
∑(𝑋𝑖 )2
Answer 2:
a. X=Trade, Y=Economic Growth, Z= Geography
(𝑋 ′ 𝑋)−1 = 0.00833
(𝑋 ′ 𝑌) = 40

𝛽̂𝑂𝐿𝑆 = 0.33
When Trade change by 1 million of US dollars, Economic growth changes by 0.33 percent.

b. Two type of problems:


a. Standard Problems in Variance of Beta: Heterogeneity and Autocorrelation.
b. Problems in Beta iself:
i. Panel data should introduce fixed effect model especially if unobserved
heterogeneity plays a role in determining the slope
ii. Endogeneity problem between X and Y. As higher economic growth affect Y and
vice versa.

c. Instrumental variable approach:


(𝑍 ′ 𝑋)−1 = 0.00233
(𝑍 ′ 𝑌) = 109

𝛽̂𝐼𝑉 = 0.25467
When Trade change by 1 million of US dollars, Economic growth changes by 0.25 percent. OLS
overestimate the impact of Trade on Economic growth.

d. Non-Biased Variance of Beta:


𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍′𝑌
𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍′(𝑋𝛽 + 𝜀)

𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍 ′ 𝑋(𝛽) + (𝑍 ′ 𝑋)−1 𝑍′(𝜀)

(𝛽̂ − 𝛽) = (𝑍 ′ 𝑋)−1 𝑍 ′ (𝜀)


′
𝐸(𝛽̂ − 𝛽)(𝛽̂ − 𝛽) = 𝐸{(𝑍 ′ 𝑋)−1 𝑍 ′ (𝜀) [(𝑍 ′ 𝑋)−1 𝑍 ′ (𝜀)]′}

𝑉𝑎𝑟(𝛽̂ ) = (𝑍 ′ 𝑋)−1 𝑍′(𝜀𝜀)′𝑍 (𝑍 ′ 𝑋)−1

If 𝜎𝑖2 = 𝜎 2 → 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡
8.4
𝑉𝑎𝑟(𝛽̂ ) = 𝜎 2 (𝑍 ′ 𝑋)−1 → × 0.00233
3
If 𝜎𝑖2 ≠ 𝜎 2 → 𝑁𝑜𝑡 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡

𝑉𝑎𝑟(𝛽̂ ) = (𝑍 ′ 𝑋)−1 𝑍 ′ Ω 𝑍 (𝑍 ′ 𝑋)−1

0.259433
𝛀= 1.028212 4.075116
-0.24039 -0.95275 0.222749
-0.99965 -3.96192 0.926282 3.851865

𝑉𝑎𝑟(𝛽̂ ) = −0.000 𝐴𝑝𝑝𝑟𝑜𝑥 = 0

We should be cautious with this derivation as we should do more replacement and estimation in the
matrix itself.

Answer 3:
a. Fixed effects controls for unobserved heterogeneity across individuals, when no heterogeneity is
the rule as in this case, the OLS and Fixed effect will produce almost close results. At least the
sign of the coefficient will be the same.
𝑂𝐿𝑆: 𝑌 = 𝛼 + 𝛽𝑋 + 𝜀
𝐹𝐸: 𝑌 = 𝛼𝑖 + 𝛽𝑋 + 𝜀
𝐼𝑓 𝐻𝑜𝑚𝑜𝑔𝑒𝑖𝑛𝑡𝑦 ℎ𝑜𝑙𝑑: 𝛼𝑖 = 𝛼 𝑎𝑛𝑑 𝑙𝑖𝑘𝑒𝑙𝑦 𝑡𝑜 ℎ𝑎𝑣𝑒 𝛽𝑖 = 𝛽
∑𝜀2
2
b. Yes, number of observations will definitely increase and 𝑅 = 1 − 𝑁−𝐾−1
∑(𝑌−𝑌̅ )2 →With higher N,
𝑁
everything held constant R2 is likely to be Higher, especially that residuals will decrease due to
better explanation of the model itself.
c.
𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍′𝑌

𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍′(𝑋𝛽 + 𝜀)

𝛽̂ = (𝑍 ′ 𝑋)−1 𝑍 ′ 𝑋(𝛽) + (𝑍 ′ 𝑋)−1 𝑍′(𝜀)

𝑇ℎ𝑒 𝑡𝑒𝑟𝑚 𝑜𝑛 𝑡ℎ𝑒 𝑟𝑖𝑔ℎ𝑡 𝑤𝑖𝑙𝑙 𝑐𝑎𝑛𝑐𝑒𝑙 𝑖𝑓 𝐶𝑜𝑣(𝑍, 𝜀) = 0 → 𝐸(𝛽̂ ) = 𝛽

d. Not Evident, the covariance should be tested formally to confirm this. May be this covariance is
not statistically significant.

You might also like