Académique Documents
Professionnel Documents
Culture Documents
yt = xt’β1 + εt if t = 1,…,τ
= xt’β2 + εt if t = τ+1,…,T
where
1t = 0 if t < τ, 1t = 1 if t > τ,
and
γ = β2-β1
Assume that the x’s are weakly exogenous and the ε’s
are spherical (i.e., homoskedastic and serially
uncorrelated).
Notes -
The paper is not general enough to cover testing for
structural change in a GMM setting, although
subsequent work by Andrews and others have
provided tests that appropriate in a GMM setting when
the breakpoint is unknown (the known breakpoint
being a special case).
These tests can easily be modified to only allow for
structural break in a subset of the parameters
Although I call these structural break tests, they are
actually test of parameter constancy or homogeneity.
III. Testing for a Structural Break with an
Unknown Break Point
Suppose
yt = xt’β1 + εt , t = 1,…,τ
= xt’β2 + εt , t = τ+1,…,T
QLRT max FT ( )
{ min ,..., max }
sup B (r )' Bk (r )
QLRT ( k )
D r [rmin , rmax ] r (1 r )