Requires an IOT and one scenario. Choose a satellite account, a factor of production or "total value added" as the indicator.

import mario

db = mario.load_test("IOT")
paths = db.calc_spa(
    indicator="CO2", item="Agriculture",
    final_demand_region="Reg1",
    max_depth=3, cutoff=0.0, top_n=10,
)
paths[["path", "depth", "contribution", "share_of_total"]]

Result · One row per reported path, ranked by absolute contribution. paths.attrs contains the total footprint and reported coverage.

Path contributions

Upstream supplier → Producing sector → Final demand

The row’s contribution is the selected indicator generated at the upstream end of that path. depth=0 is the producing sector’s direct contribution; each additional depth adds one upstream transaction.

Depth and pruning

Depth

max_depth

Number of upstream rounds. More depth can create many more paths.

Threshold

cutoff

Absolute contribution in the indicator’s unit. Paths below it are neither retained nor expanded.

Displayed paths

top_n

Number of rows returned after traversal. It does not limit traversal time. None returns all retained paths.

Footprint coverage

paths.attrs["total_footprint"]
paths.attrs["reported_contribution"]
paths.attrs["coverage_share"]

The reported paths are a partial decomposition. Depth, pruning and top_n can omit contributions. A small contribution at one node can also prune important upstream contributions; coverage is not an error bound. Direct final-demand emissions in EY are outside this production-chain expansion. Signed flows can cause cancellation, so shares need not behave like percentages of a positive total.

Show or save a plot

paths = db.calc_spa(
    indicator="CO2", item="Agriculture",
    final_demand_region="Reg1", max_depth=3, top_n=10,
    plot="sankey", show_plot=True,
)

Choose "paths" for ranked paths, "sankey" for their network or "depth" for totals by depth. Pass save_plot="spa.html" to save a figure. Setting plot alone does not build it.

What is being expanded?

The analysis expands \(e(I-z)^{-1}y\) into \(ey + ezy + ez^2y + \cdots\), where y is the selected final-demand bundle. The infinite series converges when the spectral radius of z is less than one. An item’s selector filters the rows of Y; region and category selectors filter its columns. When item is omitted, sector units must be homogeneous. The current implementation accepts one IOT scenario at a time.

Follow the EXIOBASE motor-vehicles example for a real-data workflow. All arguments: db.calc_spa().