ECON2206 Assignment

Unit Name: Introductory Econometrics

Student ID:
Student Name:
Due Date:
Question 1

Part a - Uncertain

Suppose that economic theory indicates a causal relationship between 𝑦 and 𝑥. As 𝑥 causes 𝑦 then

the econometric model becomes 𝑦 = 𝛽0 + 𝛽1 𝑥 + 𝑢 . In this instance, there is no need to include

extra explanatory variables in order to obtain a good estimate of the causal effect, 𝛽1 . This

statement is uncertain because in most economic models there are two or more explanatory

variables that influence the dependent variable y. For example, in a demand equation the quantity

demanded of a commodity depends on the price of that commodity, the prices of substitute and

complementary goods, and income. Output in a production function will be a function of more

than one input. Aggregate money demand will be a function of aggregate income and the interest

rate. Investment will depend on the interest rate and on changes in income.

Part b - False

Ordinary Least Squares applied to the model log(𝑦) = 𝛽0 + 𝛽1 𝑥 + 𝛽2 𝑥 2 + 𝑢 will not be BLUE –

Best Linear Unbiased Estimators, because it does not satisfy the necessary linearity assumption.

This statement is false because each observation on the dependent variable 𝑦 depends on the

random error term 𝑢, each 𝑦 is also a random variable. The statistical properties of y follow from

̂1 and 𝛽
those of 𝑢. The estimator 𝛽 ̂2 are best when compared to similar estimators, those that are

̂1 and 𝛽
linear and unbiased. The theorem does not say that estimator 𝛽 ̂2 are the best of all possible
 ̂1
estimators. In other words, under the assumption of the linear regression model, the estimators 𝛽

̂2 have the smallest variance of all linear and unbiased estimator of 𝛽1 and 𝛽2 . They are the
and 𝛽

best linear unbiased estimators (BLUE) of 𝛽1 and 𝛽2 .

Question 2

Part a.

Demand model for the ecolabelled apples:

𝑒𝑐𝑜𝑙𝑏𝑠 = 𝛽0 + 𝛽1 𝑒𝑐𝑜𝑝𝑟𝑐 + 𝛽2 𝑟𝑒𝑔𝑝𝑟𝑐 + 𝛽3 𝑓𝑎𝑚𝑖𝑛𝑐 + 𝛽4 ℎℎ𝑠𝑖𝑧𝑒 + 𝑢

𝛽1 and 𝛽2 provide the marginal effects of hypothetical prices of the ecolabelled apples and regular

apples on the quantity of ecologically friendly apples the family would demand given a set of

̂1 and 𝛽
hypothetical prices. The estimator 𝛽 ̂2 should be expected to have negative value, and

positive value, respectively. It is because when the hypothetical prices of the ecolabelled apples

go down, it is expected that the quantity of ecologically friendly apples the family would demand

would increases (Law of demand). However, when the price of regular apples goes up, it is

̂2 to be positive, since two of the goods could be treated as substitute goods.

expected for 𝛽

𝛽3 and 𝛽4 provide the marginal effects of family income and household size on the quantity of

̂3 and ̂
ecologically friendly apples the family would demand. Both estimators 𝛽 𝛽4 are expected to

have positive value. It is because when the family income and household size go down, it is

expected that the quantity of ecologically friendly apples the family would demand would
decreases; in contrast, when the family income and household size go up, the demand for

ecolabelled apples would increase.

Part b.

STATA Command:

regress ecolbs ecoprc regprc faminc hhsize

STATA Output is shown below:

Source SS df MS Number of obs = 660

F(4, 655) = 6.71
Model 165.420807 4 41.3552018 Prob > F = 0.0000
Residual 4038.71602 655 6.16597865 R-squared = 0.0393
Adj R-squared = 0.0335
Total 4204.13682 659 6.3795703 Root MSE = 2.4831

ecolbs Coef. Std. Err. t P>|t| [95% Conf. Interval]

ecoprc -2.902999 .5882107 -4.94 0.000 -4.058005 -1.747993

regprc 3.030618 .7108334 4.26 0.000 1.634831 4.426405
faminc .0028292 .002727 1.04 0.300 -.0025254 .0081839
hhsize .0536591 .0637884 0.84 0.401 -.0715953 .1789134
_cons 1.629504 .4503931 3.62 0.000 .7451154 2.513892

Estimated demand model for the ecolabelled apples:

̂ = 1.630 − 2.902 ∗ 𝑒𝑐𝑜𝑝𝑟𝑐 + 3.03 ∗ 𝑟𝑒𝑔𝑝𝑟𝑐 + 0.003 ∗ 𝑓𝑎𝑚𝑖𝑛𝑐 + 0.054 ∗ ℎℎ𝑠𝑖𝑧𝑒

𝑒𝑐𝑜𝑙𝑏𝑠

 Sample size: 660

 𝑅2 : 0.0393, or 3.93%

Part c.
From the regression output above, it could be observed that only two of the estimators ecoprc and

regprc are statistically significant with p-value = 0.000. It indicates that when the price of

ecolabelled apples increase by $1, the quantity demand of ecolabelled apples would decrease by

approximately 2.902. In contrast, when the price of regular apples increases by $1, the quantity

demand of ecolabelled apples would increase by approximately 3.03. Both estimators faminc and

hhsize are statistically insignificant, with p-value of 0.3, and 0.4, respectively. It means that family

income and household size have no impact on the quantity demanded for ecolabelled apples.

Hypothesis Testing

𝐻0 : 𝛽1 = 0 – The price of ecolabelled apples have no impact on its quantity demanded.

𝐻0 : 𝛽1 ≠ 0 – The price of ecolabelled apples have an impact on its quantity demanded.

Calculate t-statistics

̂1 − 𝛽1 −2.903 − 0
𝛽
𝑡 − 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐𝑠 = = = −4.94
𝑠𝑒(𝛽̂1 ) 0.588

Degree of freedom (DF) = Sample size (n) – Number of parameters (k)

DF = 660 – 4 = 656

By choosing level of significance = 5%

It is provided for t-critical value of ±0.96

Since t-statistics = -4.94 < t-critical value = -0.96

Hence, we reject the null hypothesis and conclude that the price of ecolabelled apples have an

impact on its quantity demanded.

Part d.

Step 1

regress ecolbs ecoprc regprc faminc hhsize

Step 2

predict ecolbsp if e(sample)

Step 3

corr ecolbs ecolbsp if e(sample)

The output is shown below:

ecolbs ecolbsp

ecolbs 1.0000
ecolbsp 0.1984 1.0000

Step 4

di r(rho)^2

The output is shown below:

0.03934715

Conclusion:
It could be confirmed that the squared correlation between the predictions and actual values of

ecolbs is equal to the 𝑅2 reported, or 0.19842 = 0.0393.

𝑅 2 = 0.0393, or 3.93%

The combination of all of the explanatory variables in the Model 1 could explain for

approximately 3.93% the variation in the quantity demanded of ecolabelled apples.

Part e.

Zero conditional mean implies the error has an expected value of zero given any values of the

independent variables. It should be noted that the zero conditional mean is about the population,

not the model in the sample. In other words, the residuals always add up to zero and that the

covariance, and thus correlation, between the residuals and the explanatory variables is zero.

Running regression for Model 2, the output is shown below:

Source SS df MS Number of obs = 660

F(2, 657) = 12.41
Model 153.086023 2 76.5430116 Prob > F = 0.0000
Residual 4051.0508 657 6.16598296 R-squared = 0.0364
Adj R-squared = 0.0335
Total 4204.13682 659 6.3795703 Root MSE = 2.4831

ecolbs Coef. Std. Err. t P>|t| [95% Conf. Interval]

ecoprc -2.926466 .5879291 -4.98 0.000 -4.080913 -1.77202

regprc 3.028913 .7108311 4.26 0.000 1.633138 4.424688
_cons 1.965303 .3800793 5.17 0.000 1.218987 2.71162

From the output above, the assumption of zero-conditional-mean would be expected not to be

hold for two reasons. First, the model is not fitted well with R-squared value for only 3.64%.

Second, the value of Root Mean Square Error is reported for 2.4831, which is different from 0.