Consider the following structural equation model assuming Y’s as endogenous and X’s as pre-determined variables :
Y₁ = α₁₀ + α₁₂Y₂ + α₁₃Y₃ + β₁₁X₁ + u₁
Y₂ = α₂₀ + α₂₃Y₃ + β₂₁X₁ + β₂₂X₂ + u₂
Y₃ = α₃₀ + α₃₁Y₁ + β₃₁X₁ + β₃₂X₂ + u₃
Y₄ = α₄₀ + α₄₁Y₁ + α₄₂Y₂ + β₄₃X₃ + u₄
where u₁, u₂, u₃ and u₄ are stochastic terms. Use rank and order conditions to identify the equations.
20 marks · directive: Identify
The core of a good answer
With G = 4 endogenous and K = 3 predetermined variables, every equation excludes exactly G − 1 = 3 variables, so by the order condition all four are just identified; but the rank condition fails for equations 1, 2 and 3 (the 3 × 3 matrix of excluded-variable coefficients in the other equations has rank 2, because Y₄ and X₃ appear only in equation 4), so equations 1, 2 and 3 are under-identified, while equation 4 has a rank-3 matrix and is exactly identified.
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
Apply the order condition (a necessary condition: excluded predetermined variables K − k ≥ m − 1, equivalently total excluded variables ≥ G − 1) and then the rank condition (necessary and sufficient: the matrix of coefficients, in the other equations, of the variables excluded from the equation must have rank G − 1). Show the coefficient table and each sub-matrix. The trap is to stop at the order condition.
Answer outline
Introduction
Endogenous Y₁, Y₂, Y₃, Y₄ (G = 4); predetermined X₁, X₂, X₃ (K = 3); the intercept appears in every equation and is ignored. Write each equation with all variables on one side; table of coefficients (columns Y₁ Y₂ Y₃ Y₄ X₁ X₂ X₃):
Step 1: Coefficient table
- Eq 1: −1, α₁₂, α₁₃, 0, β₁₁, 0, 0
- Eq 2: 0, −1, α₂₃, 0, β₂₁, β₂₂, 0
- Eq 3: α₃₁, 0, −1, 0, β₃₁, β₃₂, 0
- Eq 4: α₄₁, α₄₂, 0, −1, 0, 0, β₄₃
Step 2: Order condition (K − k vs m − 1)
- Eq 1: m = 3 (Y₁, Y₂, Y₃), k = 1 (X₁): K − k = 2 = m − 1, just identified by order.
- Eq 2: m = 2 (Y₂, Y₃), k = 2 (X₁, X₂): 1 = 1, just identified by order.
- Eq 3: m = 2 (Y₃, Y₁), k = 2 (X₁, X₂): 1 = 1, just identified by order.
- Eq 4: m = 3 (Y₄, Y₁, Y₂), k = 1 (X₃): 2 = 2, just identified by order.
- Order condition is necessary only, so the rank condition must be checked.
Step 3: Rank condition for equations 1, 2, 3
- Eq 1 excludes Y₄, X₂, X₃; rows of Eq 2, 3, 4: [0, β₂₂, 0], [0, β₃₂, 0], [−1, 0, β₄₃]. Rows 1 and 2 are proportional, rank 2 < 3: under-identified.
- Eq 2 excludes Y₁, Y₄, X₃; rows of Eq 1, 3, 4: [−1, 0, 0], [α₃₁, 0, 0], [α₄₁, −1, β₄₃]. Rank 2 < 3: under-identified.
- Eq 3 excludes Y₂, Y₄, X₃; rows of Eq 1, 2, 4: [α₁₂, 0, 0], [−1, 0, 0], [α₄₂, −1, β₄₃]. Rank 2 < 3: under-identified.
- Reason: Y₄ and X₃ occur only in equation 4, so excluding them gives no information about equations 1 to 3. Within the block of equations 1 to 3 (3 endogenous, 2 predetermined), each equation excludes only one variable where two are needed.
Step 4: Rank condition for equation 4
- Eq 4 excludes Y₃, X₁, X₂; rows of Eq 1, 2, 3: [α₁₃, β₁₁, 0], [α₂₃, β₂₁, β₂₂], [−1, β₃₁, β₃₂].
- Determinant = α₁₃β₂₁β₃₂ − α₁₃β₂₂β₃₁ − α₂₃β₁₁β₃₂ − β₁₁β₂₂, non-zero for general parameter values, so rank 3 = G − 1.
- Rank condition holds and the order condition holds with equality: exactly identified (estimable by ILS, IV or 2SLS).
Conclusion
Equations 1, 2 and 3 are under-identified (they satisfy the order condition but fail the rank condition), and equation 4 is exactly identified. Ranks and the determinant were checked symbolically with sympy.
Examples and data to cite
- The order condition is necessary but not sufficient; the rank condition is necessary and sufficient.
- This example shows why the order condition alone is not enough: a block-recursive structure can make it misleading.
Facts to remember
- Order condition: K − k ≥ m − 1 (or excluded variables ≥ G − 1); necessary only.
- Rank condition: rank of the (G − 1) × (excluded variables) coefficient matrix from other equations = G − 1; necessary and sufficient.
- Equality in the order condition with rank satisfied means exact identification; strict inequality means over-identification.
- An under-identified equation's structural parameters cannot be recovered from the data by any technique.