Skip to contents

sube is organized around a workflow API instead of a script pipeline. That design choice is tied to the research problem developed in the companion paper, but the package documentation treats the operational workflow as primary and the paper as motivation. Wrapping input-output multipliers in confidence intervals.

The workflow is source-agnostic: it operates on any supply-use data in the canonical long format. WIOD and FIGARO are shipped as example importers, but custom national supply-use tables work the same way once reshaped into the canonical format.

Why this package exists

The package sits between two traditions:

  • classic Leontief multiplier analysis on supply and use data
  • econometric work that compares multiplier behavior across countries, products, and years

The paper motivates why those should be brought together. sube turns that motivation into reusable software by keeping benchmark matrices, econometric models, comparison tables, and export helpers in the same package workflow.

Design choices

  • Inputs and outputs are explicit function arguments and return objects.
  • Shipped example data is intentionally small so examples and vignettes stay CRAN-safe.
  • File writing is optional and separated from in-memory computation.
  • Historical paper scripts are not part of the package source tree; they are kept locally in an ignored archive for reference only.

Public research workflow

The public package API is organized around a research sequence:

  1. import supply-use data,
  2. build domestic matrices,
  3. compute Leontief benchmark results,
  4. estimate SUBE models, and
  5. compare and export both layers from package objects.

That last step is important. The package is not only about producing model coefficients; it is also about making the Leontief-versus-SUBE comparison easy to inspect, communicate, reproduce, and export without a separate script pipeline.

Public objects

The package returns structured objects that are meant to be inspected and then converted into tidy tables as needed:

  • sube_suts
  • sube_domestic_suts
  • sube_matrices
  • sube_results
  • sube_models

Paper companion role

The package should be useful even without the paper, but the paper gives the best motivation for why the comparison helpers exist. If you want to understand why sube exposes Leontief matrices, model tables, and paper-style comparison plots in one workflow, the article is the place to start.