of the committee’s preference scores should differ from the floor’s median. To determine
the existence of outliers, I use a Monte Carlo Wilcoxian difference of medians test to
determine whether there is a statistically significant difference between the W-
NOMINATE median of the budget committee members and the W-NOMINATE median
of the Duma floor for each session of the Duma. This non-parametric approach allows us
to get around the assumption of a normal distribution of W-NOMIATE scores and the
false rejection of a hypothesis when violations of symmetry are extreme (Groseclose
1994). The Monte Carlo approach randomly selects with replacement 10,000 committees
for each session from among the observed data. A difference in median test is conducted
and recorded between the floor and each of the 10,000 randomly selected committees.
This procedure produces an approximate density of the median of the budget committee,
assuming the committee was selected randomly (Groseclose 1994). If the committee is a
preference outlier, then the p-value from this test should be less than or equal to 5
percent. In other words, a committee would be composed of preference outliers when
less than or equal to 5 percent of the 10,000 simulated committees had median
differences from the floor equal to or greater than the original committee’s median
difference from the floor.
Table 2 About Here
Table 2 reports the results of the Monte Carlo Wilcoxian difference of medians
tests for each session of the Duma. The All Deputies column reports the p-values for the
difference of medians tests, which compare the W-NOMINATE median of the entire
budget committee with the floor’s median. The results reveal an interesting pattern—
committee outliers are clustered in the Sixth Duma. Of the four sessions of the Fifth
18
Convention