We extend the impact decomposition proposed by LeSage and Thomas-Agnan (2015) in the spatialinteraction model to a more general framework, where the sets of origins and destinations can bedifferent, and where the relevant attributes characterizing the origins do not coincide with those of thedestinations. These extensions result in three flow data configurations which we study extensively: thesquare, the rectangular, and the non-cartesian cases. We propose numerical simplifications to computethe impacts, avoiding the inversion of a large filter matrix. These simplifications considerably reducecomputation time; they can also be useful for prediction. Furthermore, we define local measuresfor the intra, origin, destination and network effects. Interestingly, these local measures can beaggregated at different levels of analysis. Finally, we illustrate our methodology in a case study usingremittance flows all over the world.
Impact decomposition; local effects; spatial interaction autoregressive models; non-cartesian flow data;
- C13: Estimation: General
- C31: Cross-Sectional Models • Spatial Models • Treatment Effect Models • Quantile Regressions • Social Interaction Models
- C46: Specific Distributions • Specific Statistics
- C51: Model Construction and Estimation
- C65: Miscellaneous Mathematical Tools
Christine Thomas-Agnan, Paula Margaretic, and Thibault Laurent, “Impact computations in the general spatial interaction model”, TSE Working Paper, n. 22-1357, September 2022.
TSE Working Paper, n. 22-1357, September 2022