This paper uses Monte Carlo simulation to examine the size and power p
roperties of two modified Wald tests for non-linear restrictions. We f
ind that the modifications can sometimes reduce power and need not res
ult in improved size properties in small samples.