We may also allow the parameter of the capital-labor ratio to vary across groups. If this is the case,
the system of equations (9) may be written as:
it
it
∣+ β +γAge + μ(Temporary -σ,φδZ)+¾, i∏>, =1
l. I L J,-2 εω Φ(δZ,) 1it 1 (9')
I + β +Y2 Age, + μ (Tm→∖ .,τφ''Z' + U2it ifDι= 0
2 2 , 2 εω 2it 1
llt I L Jit-2 1 1-φ(δZt )
Clearly, if none of the parameters varies across groups, equation (9) reduces to the simple original
form (4):
4,Γ Y I *,Γ K I . . , ( Temporary I
Δ 2 Inl — ∣ = α 2lnl — ∣ + β + γigei + μ∖---t-—∣- + εlt
∖L j it ∖ L j it ∖ L j it-2
This is the constrained equation (constraint being equal parameters in all groups) used in the former
Chow test. Leaving aside the potential endogeneity of the capital-labor ratio, the problem of
endogenous group formation in the latter case disappears altogether.
As to the potential endogeneity of the capital stock, we assume that capital accumulation depends
on past investment intensities and initial levels of capital/labor ratio, conditional on size, sector and
group of the firm, whether or not it has introduced innovations, and on firm age. Because of this, we
=1
= α∆ 2 lnl — ∣ + <β1 + γ1 Agei + μ1
V L J it I
* P(D1 = 1) +
=0 ∣*P(D1 =0)
TemporaryI
L J it-2
Φ(δ zit )
σεω----— (Φ(δ Zit ) +
ειω Φ(δ zi )) lt
+ ∖β2 + Y2 Agei + μ2
TemporaryI
∣ + σεω
L J it-2 2
φ(δ Zt ) Г e l
-----it— ⅛ -Φ(δ' Zt )]
1 -Φ(δ' Zt ) it
Rearranging terms, we would obtain an estimable specification which allows to perform Wald tests of parameters
instability. When the coefficients of the interactions are equal to zero, this procedure is a convenient way to impose
cross-equation restrictions on the two-regime specification:
( Y I (KI ( TemporaryI
δ2 lnl - I = α2 lnl I + β2 + γ2Agei + μ21 I + (β1 -β2^φ(δZt) +
V L Jit V L Jit V L Jit-2
+ (Y1 - Y2)Ageiφ(δZt )+ (μ1 -μ2)f τemp°rary∖ φ(δZt) + +φ(δZt)(σε ω - σεω)+ ξit
V L J it-2 2 1
where ξit has a standard normal distribution.
13