Abstract
How does accounting for differences in students' nominations change the evidence for structural balance? We study friendship-enmity networks in 137 Spanish classrooms (3,512 students) and 78 Salvadoran classrooms (1,830 students). For each of four directed rules, we measure how often the tie closing a two-step path has the predicted sign. We compare two benchmarks that keep the observed ties and redraw their signs, using either a common positive probability or probabilities matching, on average, each student's positive and negative nominations sent and received. Under degree adjustment, excess for a friend's friend falls from 13.43 to 2.68 percentage points in Spain and from 11.81 to 1.31 in El Salvador. The contrast for an enemy's enemy changes from deficit to excess. Both mixed-sign rules show positive adjusted mean excesses, but excess for an enemy's friend exceeds that for a friend's enemy by 4.24 points in Spain and 4.15 in El Salvador in paired comparisons. Refitting simulations produce departures under N4 itself, with EE departures reaching 14.71 points and negative FE-EF contrasts in known-probability scenarios on all 215 classroom supports. The empirical contrasts therefore describe benchmark-specific excesses. Assessing the four rules separately reveals rule-specific responses to degree adjustment that a single balance score conceals.