Publication companion
Network Meta-Analysis for Renewable Energy
Synthesizing more than a decade of behavioral intervention trials to rank the most effective levers for cutting household electricity consumption.
Objective
Provide policymakers with a single ranking of demand-side interventions by harmonizing 200+ experimental and quasi-experimental studies from 2010–2024.
- Convert heterogeneous statistics into standardized mean differences.
- Preserve study-level nuance (region, opt-in, delivery channel, pricing).
- Support both single-intervention and bundled program design.
Evidence base
Each arm in the network aggregates effect sizes for feedback, information, motivation, monetary incentives, and social comparison—plus every observed pairwise or triple combination.
- Meta-data captured study design, sample size, evaluation window, weather.
- Risk-of-bias scoring filtered outliers and low quality quasi-experiments.
- Created transparent provenance sheets for journal submission.
Modeling approach
Ran a random-effects NMA in R's netmeta with design-by-treatment
interactions to surface inconsistency, then replicated results via Bayesian HMC
in Stan for full posterior draws. Moderator terms handled geography, opt-in
status, and pricing structure.
- Meta-regressions (metafor) for single-arm summaries.
- Sensitivity checks excluding pre-2012 pricing studies.
- Probability rankings (SUCRA) exported for policy memos.
Key findings
Bundles that mix feedback + social pressure + monetary nudges cut consumption by 6–9%, roughly double the impact of single-channel information campaigns.
- Opt-in pilots underperform mandated rollouts by ~1.8 percentage points.
- Monetary incentives only dominate when paired with disaggregated feedback.
- Rural programs benefit most from social comparison arms.