In short: A local content percentage is an input to economic impact rather than a measure of it, because two procurement decisions that move the score by the same amount can generate very different domestic activity. The tool that connects the two is the Leontief input-output inverse, set out by Wassily Leontief in the 1930s and still the standard way to trace how spending in one sector pulls output from every other. Impact decomposes into direct spend with the domestic supplier, indirect purchases that supplier makes from other domestic sectors, and induced activity from the wages both generate. The model assumes fixed production recipes, no capacity limits and a static industrial structure, which makes it sound for comparing candidate categories and unreliable as an absolute figure.
Every large resource holder now runs a local content programme with a headline percentage attached. Move from sixty-seven percent to seventy, then to seventy-five by 2030. The number appears in annual reports, ministerial speeches and supplier scorecards, and it gets treated as the outcome.
It is an input. The outcome is gross domestic product, jobs and industrial capability that persists after the capital programme ends, and the relationship between the percentage and those outcomes is not proportional. Two procurement decisions that move the score by the same amount can differ by a factor of three in the economic activity they generate.
Understanding why requires a piece of economics that is eighty years old and still the correct tool for the job.
The Leontief inverse, briefly
Wassily Leontief built input-output analysis in the 1930s and won a Nobel prize for it in 1973. The idea is that every sector of an economy buys inputs from other sectors in order to produce its own output, and those relationships can be written as a matrix.
Call it A, where the entry in row i, column j is the amount of sector i's output required to produce one unit of sector j's output. Steel needs energy, energy needs equipment, equipment needs steel.
Total output x has to satisfy demand from other industries plus final demand f:
x = Ax + f
Rearranged, that gives x = (I - A) inverse times f, and the matrix L = (I - A) inverse is the Leontief inverse, sometimes called the total requirements matrix. It answers the question that matters for local content: if final demand for a sector's output rises by one unit, how much total output does that generate across the entire economy once every round of supplier purchasing has worked through.
The column sums of L are the Type I output multipliers. A multiplier of 1.8 means that a dollar of demand delivered by that sector generates eighty cents of additional activity elsewhere. A multiplier of 1.15 means it generates fifteen.
Miller and Blair's Input-Output Analysis: Foundations and Extensions, second edition 2009, is the standard treatment if you want the derivations, the employment extensions and the variants. The tables themselves come from national statistics offices as supply and use tables, and the OECD publishes a harmonised Inter-Country Input-Output series, which matters when part of the supply chain you are modelling crosses a border.
Why two localisation decisions differ so much
The multiplier depends on how deeply the sector is embedded in the domestic economy, and this is where localisation programmes succeed or produce numbers without substance.
Consider two ways to move a local content score by the same amount.
The first localises assembly of an imported product. Components arrive from abroad, a domestic facility assembles and finishes them, and the value added in-country counts toward the score. The multiplier here is close to one, because almost every input to that facility comes from outside the domestic economy. Very little of the spend circulates.
The second localises manufacture of a product whose inputs are domestically available. Local steel, local fabrication, local engineering services. The multiplier is substantially higher, because each supplier in the chain buys from other domestic suppliers, and the spend goes round several times before it leaks out as imports.
The score treats these as similar. The economy does not. This is the single most useful thing input-output analysis contributes to a localisation programme, and it is computable from published national accounts rather than requiring new data collection.
The same comparison, worked
Two sectors are enough to show the mechanism. Call sector one the localisation candidate, and sector two everything domestic it could buy from, with steel, fabrication and engineering services aggregated together.
Take the assembly case first. Per dollar of its own output, the candidate sector buys 5 cents from itself and 6 cents from the domestic supplier base. Everything else is imported components and value added. The supplier base buys 15 cents of the candidate sector's output and 25 cents of its own per dollar it produces.
So the matrix I - A has rows (0.95, -0.15) and (-0.06, 0.75).
The determinant of a two by two is the diagonal product minus the off-diagonal product. That is 0.95 times 0.75, which is 0.7125, less 0.15 times 0.06, which is 0.009, leaving 0.7035. Inverting swaps the diagonal entries, flips the sign on the others, and divides everything by the determinant, so the first column of L is 0.75 over 0.7035 and 0.06 over 0.7035. That gives 1.066 and 0.085, and the column sum is 1.15.
Now the manufacture case, with the same supplier base and a different recipe. The candidate sector buys 10 cents from itself and 37 cents from the domestic supplier base per dollar of output, because the steel, the fabrication and the engineering hours it needs all exist in-country.
I - A now has rows (0.90, -0.15) and (-0.37, 0.75). The determinant is 0.675 less 0.0555, or 0.6195. The first column of L is 0.75 over 0.6195 and 0.37 over 0.6195, which is 1.211 and 0.597, summing to 1.81.
Put a 400 million dollar contract through each. The assembly route generates about 460 million of total domestic output, of which 60 million is activity beyond the contract itself. The manufacture route generates about 723 million, with 323 million beyond the contract. Same headline spend, same movement in the score, and more than five times the knock-on.
Two coefficients changed between the cases, and one of them is doing nearly all the work: how much of a dollar of output the candidate sector buys from the domestic supplier base, 6 cents against 37. That number sits in a single cell of your national supply and use table, and reading it before believing a multiplier takes a few minutes.
Direct, indirect and induced
Employment effects come in three layers and the distinction matters because programmes are frequently reported using whichever layer produces the largest number.
Direct jobs are those at the facility receiving the spend. These are countable and defensible.
Indirect jobs are those generated up the supply chain as suppliers increase their own output. These come out of the Leontief inverse using employment coefficients per unit of sector output, and they are as reliable as the coefficients are.
Induced jobs come from household spending. Workers earn wages, spend them, and that spending supports further employment. Including these gives Type II multipliers, and the induced layer is typically a further thirty to forty percent on top of the direct and indirect total.
Type II numbers are legitimate and they are also considerably softer, because they depend on assumptions about savings rates, import propensity and how much of the wage is spent domestically. My preference is to report Type I as the headline and show Type II separately with the induced factor stated, so a reader can see which assumption is doing the work. A programme that reports Type II without saying so is not lying, and it is inviting a challenge it will lose.
What the model assumes, and where those assumptions break
Input-output analysis is a linear model of a non-linear world, and being honest about the assumptions makes the results more useful rather than less.
Fixed technical coefficients. The matrix assumes the recipe for producing a unit of output does not change. In reality, firms substitute inputs when prices move, and a localisation programme is often specifically trying to change the recipe. The model will not capture that change until the coefficients are updated, and national input-output tables are typically published with a lag of several years.
Constant returns to scale. Doubling output doubles all inputs. Real facilities have capacity limits and fixed costs, and a domestic supplier operating below efficient scale has a cost structure the model does not see.
No supply constraints. The model assumes any additional demand is met by additional domestic output. If the domestic sector is already at capacity, the extra demand is met by imports and the multiplier does not materialise. This is the most common reason a modelled impact fails to appear, and it is checkable in advance by comparing the demand increase against sector capacity.
No price effects. A large demand increase in a constrained sector raises prices rather than volumes, transferring income rather than creating output.
Every one of these pushes the same way, which is that modelled impacts are upper bounds. A programme reporting realised impact equal to its modelled impact should be examined rather than congratulated.
Using it to choose rather than to report
Most organisations use input-output analysis after the fact, to attach an impact number to decisions already taken. It is considerably more valuable used beforehand.
Compute the multiplier for every sector in your addressable procurement. Then rank the candidate localisation categories not by how much they move the score, but by economic activity generated per unit of spend, adjusted for how much of that sector already exists domestically.
The ranking that falls out is usually different from the one a score-driven programme produces, and it tends to favour categories with three properties: deep domestic supply chains already in place, sufficient headroom in capacity to absorb the demand, and technical requirements the domestic base can meet without a decade of capability building.
It also identifies the categories where the honest answer is to wait. A sector with no domestic base, no adjacent capability and a small addressable spend will produce a low multiplier and a high cost per point, and funding it produces a facility that survives only as long as the mandate does.
The measurement that matters afterwards
Two things worth tracking that are not the score.
Whether the localised supplier sells to anyone else. A domestic supplier serving one buyer under a mandate is a cost centre with a percentage attached. One that has diversified its customer base has become an industry. Export sales in particular are the strongest available signal that real competitiveness was created rather than purchased, which is why export factors appear in the better-designed scoring formulas.
Whether the capability persisted through a downturn. Capital programmes are cyclical. The test of a localisation programme is what remains after the capital cycle turns and the mandated demand falls. Suppliers that survive that have genuine capability. Those that do not were a transfer payment.
Neither of those shows up in a percentage, and both are more informative about whether the programme worked.
Where I would not push the model
Input-output analysis cannot tell you about capability transfer, which is often the actual objective. A joint venture that produces modest measured multipliers while transferring engineering knowledge to a domestic workforce may be worth far more over twenty years than a high-multiplier assembly operation that transfers nothing. The model has no term for that and it should not be asked to have one.
It also says nothing about quality or reliability. A domestic supplier at a higher price and lower reliability imposes costs across the operation that the multiplier does not net off. Those costs are real and they belong in the decision alongside the economic contribution rather than in a separate conversation.
And I would treat any single headline figure with caution, including the good ones. Cumulative programme impacts running to hundreds of billions in added domestic product are the sum of many modelled estimates over many years, each with the assumptions above embedded. They are useful as an order of magnitude and they are not measurements.
The productive use of all this is comparative. Which of these two categories generates more, and by how much, given what already exists domestically. That question the model answers well, and answering it before the procurement decision rather than after is where the value sits.