Supported Models

PanelSFA implements models ranging from foundational cross-sectional frameworks to advanced multidimensional panel estimators.

Cross-Sectional Models

  • Aigner, Lovell, and Schmidt (ALS 1977): The foundational SFA model. Assumes a Half-Normal distribution for the inefficiency term.

Panel Models (Deterministic)

  • Battese and Coelli (BC 1992): Assumes inefficiency follows a Truncated-Normal distribution that decays (or grows) deterministically over time via an $\eta$ parameter.

  • Battese and Coelli (BC 1995): A single-step estimator where the mean of the inefficiency distribution is a linear function of environmental covariates ($Z$). Eliminates two-step bias.

Panel Models (Unobserved Heterogeneity)

  • Greene (2005) True Fixed Effects (TFE): Profiles out time-invariant firm intercepts ($\alpha_i$) using an internal Newton-Raphson loop. Ensures structural differences between firms are not mistakenly scored as inefficiency.

  • Greene (2005) True Random Effects (TRE): Integrates out a firm-level random effect $w_i$ using Maximum Simulated Likelihood via Halton sequences.

The Four-Component Model

  • Kumbhakar, Lien, and Hardaker (2014): A three-step pseudo-likelihood estimator that completely disentangles firm heterogeneity, random noise, transient (short-term) inefficiency, and persistent (structural) inefficiency.