Why it is desirable to remove insignificant variables from a regression?

Questions -

Q1. Why it is desirable to remove insignificant variables from a regression?

Q2. We may consider an example taken from Studenmund (2001: 193-5). The restricted equation for a demand for chicken function is:

Y^= 32.9 - 0.70 PCt+ 0.27 YDt

= (-8.33) (45.91)

SSRR = 185.66 N = 44

Where Y^ is the estimated demand for chicken;

PC is the price of chicken;

t is the time subscript;

YD is disposable income.

The unrestricted equation is estimated as:

Y^= 23.80 - 0.59PCt+ 0.36YDt+ 0.02Y^t2- 0.007Y^t3+ 0.000055Y^t4

= (-0.34) (0.50) (0.29) (-0.68) (1.02)

SSRU= 138.41 N = 44

In this case the unrestricted functional form (specification) is to the fourth power. Calculate RESET test statistic to find that estimated linear form is the correct specification of the model.

Q3. Consider the following simultaneous equations system.

y1t= ?0+ ?1y2t+ ?2y3t+ ?3X1t+ ?4X2t+ u1t(1)

y2t= ?0+ ?1y3t+ ?2X1t+ ?3X3t+ u2t(2)

y3t= ?0+ ?1y1t+ ?2X2t+ ?3X3t+ u3t(3)

a. What do you understand by the term `identification`? Describe a rule for determining whether a system of equations is identified. Apply this rule to equation (1)-(3).

b. Which would you consider the more serious misspecification: treating exogenous variables as endogenous, or treating endogenous variables as exogenous? Explain your answer.