multilevel modeling
Definition
Multilevel modeling is a statistical approach designed to analyze data with a nested structure, where observations at one level are grouped within higher-level units. In the study of violent extremist organizations, this means leader-year observations are nested within individual leaders, allowing researchers to separate within-leader change over time from between-leader differences in influence style and power orientation. Because annual lethality varies both across leaders and across years within a single leader's tenure, a flat analytical approach would conflate those sources of variance and produce misleading estimates. Applied to 68 VEO leaders and 645 leader-year observations, multilevel longitudinal analysis revealed that lethality increased by roughly 50 percent per log-time unit across a leader's tenure, and that the effect of power orientation on lethal performance depended on which influence style the leader employed.
Sources: Harms et al. (2026)
Related Terms
- leadership (1 shared article)
- terrorism (1 shared article)
- violent extremism (1 shared article)
- lethality (1 shared article)
Applications
Multilevel Modeling and Leader Influence Style
Leader influence style is treated as a stable leader-level characteristic, making it a natural between-person predictor in a multilevel structure where yearly lethality outcomes are nested within leaders. Because influence style does not vary year to year within a leader's tenure, it can only be tested at the upper level of the model, and its effect on lethality emerged only in interaction with power orientation rather than as an independent main effect.
Sources: Harms et al. (2026)
Multilevel Modeling and Organizational Lethality
Organizational lethality in violent extremist organizations varies both across leaders and across years within a single leader's tenure, a pattern that multilevel modeling is specifically equipped to handle by partitioning these distinct sources of variance. The multilevel longitudinal analysis used in this research tracked annual death counts as the repeated outcome measure, revealing a reliable upward trajectory over time while also identifying the leader-level interaction between influence style and power orientation as the primary driver of overall lethality level.
Sources: Harms et al. (2026)
