A Simulation Study of Effects of Multicollinearity and Autocorrelation on Estimates of Parameters

I. INTRODUCTIOPJ
In attempting t o analytically discover or t e s t economic relationships,
econometricians have 'vailable many computational techniques by
which t o estimate the parameters of t h e i r models. But different soliltion
methods may give unbiased and consistent, biased and consistent, or
biased and inconsistent estimates under varying assumptions. The model
builder i s v i t a l l y interested in how each of these procedures reacts LIEder
varying conditions that nay impinge on his model, but which are conditions
not assumed by the estimation technique.
Three of the more popular techniques are the following single equation
methods: (1) Ordinary Least Squares (LS) , (2) Limited Informatior,
Single Equation (LISE), and (3) Two Sta.ge Lcast Squares (TSLS). Although
there methods estimate parameters one equation a t a two oi"
the procedures do take into account, t o a "1-L~ited"extent, the effects
of the whole system. Al-though Three Stage Least Squares and Full
Information Maximum Likelihood Methods estimate a l l the s t r u c t u r a l equat
i o n simultaneously, they involve extremely lengthy and cmbersme cmput
a t i o n a l procedures.