Creator Dynamics: Replication Package for "From Exposure to Followers: A Stock-and-Flow Closed-Loop Framework of Creator Dynamics"

This archive provides the replication package for the article “From Exposure to Followers: A Stock-and-Flow Closed-Loop Framework of Creator Dynamics” (Information Processing & Management, https://doi.org/10.1016/j.ipm.2026.104677). It contains the full Python 3.9.6 pipeline we use to construct datasets, estimate the model primitives, validate the closed loop in and out of sample, and run robustness exercises across the main YouTube sample, subsamples, and holdout checks. The package includes processed per-account descriptives, estimated parameters that are sufficient to reproduce the published tables and figures once raw inputs are reconstructed, and systematic-sampling scripts for rebuilding the account–day panels from public webpages and trackers, but it does not redistribute bulk scraped pages or API payloads. Platform terms restrict redistribution of bulk-scraped pages and API payloads. Accordingly, the public replication package does not include raw pages or API payloads, but instead provides stable identifiers and a systematic sampling procedure so that external researchers can recreate the input panel from public webpages and trackers. This design promotes reproducibility and transparency while remaining consistent with the relevant platform terms of service. For details on the directory structure, required input format, and step-by-step usage, please see the accompanying README.md and the article.