Skip to content
We use only essential cookies to keep you signed in and remember your preferences. See our Cookie Policy.
TestStop
MocksPracticeResourcesIndiaBlogPricing
Sign in
Questions › GATE › GATE 2021 › Q14
1 mark, −0.333Econometrics › Endogeneity, instruments and simultaneous equations › Instrumental variables and two-stage least squaresHardabout 3 min

GATE 2021 Question 14: Endogeneity, instruments and simultaneous equations

Consider a regression model y=β0+βx+uy=\beta_0+\beta x+uy=β0​+βx+u where the continuous variable yyy is regressed on a dummy variable xxx, which takes the value either 1 or 0. However, the model was estimated using the instrumental variable (IV) estimation method, wherein the indicator variable zzz is used as an instrument of xxx.

Let:

yˉ1\bar y_1yˉ​1​ and yˉ0\bar y_0yˉ​0​ be the sample averages of yyy when zzz takes the value 1 and 0, respectively

xˉ1\bar x_1xˉ1​ and xˉ0\bar x_0xˉ0​ be the sample averages of xxx when zzz takes the value 1 and 0, respectively

yˉ1\bar y^1yˉ​1 and yˉ0\bar y^0yˉ​0 be the sample averages of yyy when xxx takes the value 1 and 0, respectively

zˉ1\bar z^1zˉ1 and zˉ0\bar z^0zˉ0 be the sample averages of zzz when xxx takes the value 1 and 0, respectively

Then the estimated coefficient of βIV\beta_{IV}βIV​ is

Tap an option to check your answer.

Answer and worked solution

Answer: (C)

This is the classic Wald estimator for IV estimation with a binary instrument and binary endogenous regressor. It equals the reduced-form effect (of the instrument zzz on the outcome yyy) divided by the first-stage effect (of the instrument zzz on the endogenous regressor xxx):

β^IV=Reduced form: effect of z on yFirst stage: effect of z on x=yˉ1−yˉ0xˉ1−xˉ0\hat\beta_{IV} = \frac{\text{Reduced form: effect of } z \text{ on } y}{\text{First stage: effect of } z \text{ on } x} = \frac{\bar y_1-\bar y_0}{\bar x_1-\bar x_0}β^​IV​=First stage: effect of z on xReduced form: effect of z on y​=xˉ1​−xˉ0​yˉ​1​−yˉ​0​​

where, per the problem's notation, yˉ1,yˉ0\bar y_1,\bar y_0yˉ​1​,yˉ​0​ are yyy's sample averages split by z=1z=1z=1 vs z=0z=0z=0, and xˉ1,xˉ0\bar x_1,\bar x_0xˉ1​,xˉ0​ are xxx's sample averages, also split by z=1z=1z=1 vs z=0z=0z=0 (both use the subscript notation tied to zzz, per the stem's definitions).

This matches option (C). (The other options either compare the wrong pair of quantities, such as splitting by xxx instead of zzz in the denominator of (A) and (D), or mix yyy and xxx averages incoherently as in (B).)

The trap: (D) yˉ1−yˉ0zˉ1−zˉ0\dfrac{\bar y_1-\bar y_0}{\bar z^1-\bar z^0}zˉ1−zˉ0yˉ​1​−yˉ​0​​

The four options are deliberately constructed to look extremely similar, swapping which variable is used for the denominator's grouping — a student who correctly recalls 'IV estimator = reduced form over first stage' but doesn't carefully track exactly which sample-average notation (xˉ\bar xxˉ vs yˉ\bar yyˉ​ vs zˉ\bar zzˉ, and whether split by xxx or zzz) corresponds to which stage may pick a plausible-looking but structurally wrong option.

Plain-text version

Consider a regression model y=β₀+β x+u where the continuous variable y is regressed on a dummy variable x, which takes the value either 1 or 0. However, the model was estimated using the instrumental variable (IV) estimation method, wherein the indicator variable z is used as an instrument of x. Let: ȳ₁ and ȳ₀ be the sample averages of y when z takes the value 1 and 0, respectively x̄₁ and x̄₀ be the sample averages of x when z takes the value 1 and 0, respectively ȳ¹ and ȳ⁰ be the sample averages of y when x takes the value 1 and 0, respectively z̄¹ and z̄⁰ be the sample averages of z when x takes the value 1 and 0, respectively Then the estimated coefficient of β_(IV) is (A) (ȳ₁-ȳ₀)/(ȳ¹-ȳ⁰) (B) (ȳ₁-x̄₁)/(ȳ₀-x̄₀) (C) (ȳ₁-ȳ₀)/(x̄₁-x̄₀) (D) (ȳ₁-ȳ₀)/(z̄¹-z̄⁰)

Also written as: Consider a regression model y=beta_0+beta x+u where the continuous variable y is regressed on a dummy variable x, which takes the value either 1 or 0. However, the model was estimated using the instrumental variable (IV) estimation method, wherein the indicator variable z is used as an instrument of x. Let: y_bar_1 and y_bar_0 be the sample averages of y when z takes the value 1 and 0, respectively x_bar_1 and x_bar_0 be the sample averages of x when z takes the value 1 and 0, respectively y_bar^1 and y_bar^0 be the sample averages of y when x takes the value 1 and 0, respectively z_bar^1 and z_bar^0 be the sample averages of z when x takes the value 1 and 0, respectively Then the estimated coefficient of beta_(IV) is

Endogeneity, instruments and simultaneous equations has come up 8 times in 3 exams since 2021.

What this question tests

This question is on Instrumental variables and two-stage least squares, part of Endogeneity, instruments and simultaneous equations in Econometrics.

Using an instrument to recover a causal effect when a regressor is endogenous.

This is the only past-paper question on this subtopic in TestStop's bank.

Take this paper as a timed mockMore on Instrumental variables and two-stage least squares

Similar questions from other papers

  • GATE 2023 · Q48Consider the following simultaneous equations model: Yₜ = β₁+β₂Xₜ+β₃Xₜ₋₁+β₄Zₜ+μ₁ₜ (1) Zₜ = δ₁+δ₂Yₜ+δ₃Wₜ+μ₂ₜ (2) Before estimating the above model, a researcher p

Question from the GATE 2021 paper. Solutions are TestStop's own. Not affiliated with or endorsed by any examination authority.