Interpreting Direct and Indirect Requirements

From a Production Plan to Requirements

In an input-output model, the output of one industry can be an input needed by another. A production plan must therefore cover more than the goods requested directly by final customers: it must also provide the intermediate inputs needed to make those goods, and the inputs needed to make those inputs. The Leontief inverse gathers these requirements across all production rounds.

Suppose there are \(n\) industries. Let \(\mathbf{A}\in\mathbb{R}^{n\times n}\) be the input coefficient matrix, where \(a_{ij}\) is the amount of output from industry \(i\) needed to produce one unit of output in industry \(j\). Assume throughout that \(a_{ij}\geq 0\). Let \(\mathbf{x}\) be the column vector of gross outputs and \(\mathbf{d}\) the column vector of final demands. The production balance equation is \(\mathbf{x}=\mathbf{A}\mathbf{x}+\mathbf{d}\): total output supplies both intermediate use and final demand. If \(\mathbf{I}-\mathbf{A}\) is invertible, this gives \(\mathbf{x}=(\mathbf{I}-\mathbf{A})^{-1}\mathbf{d}\). The matrix \((\mathbf{I}-\mathbf{A})^{-1}\) is the Leontief inverse.

Thus the entry in row \(i\), column \(j\) of the Leontief inverse is not generally the direct input coefficient \(a_{ij}\). It is the total output from industry \(i\) required to meet one unit of final demand from industry \(j\), including all production rounds. The identity term contributes one unit when \(i=j\), because that unit goes to final demand. The term \(\mathbf{A}\) contributes first-round intermediate inputs; the remaining powers contribute further inputs, including inputs that return to an industry through a production chain.

Why the Powers Represent Production Rounds

For any demand vector \(\mathbf{d}\), the vector \(\mathbf{A}\mathbf{d}\) is the intermediate output needed directly to produce \(\mathbf{d}\). To produce that intermediate output, another round of inputs is needed: \(\mathbf{A}(\mathbf{A}\mathbf{d})=\mathbf{A}^2\mathbf{d}\). Repeating the calculation gives the requirements in each successive round. This multiplication order follows the convention that the \(j\)th column of \(\mathbf{A}\) describes inputs for industry \(j\)'s output.

For the infinite sum to describe a finite production plan, the later rounds must become small enough to add to a finite total. The following sufficient condition makes that precise. It is not the only possible convergence condition, but it has a direct interpretation: each column sum is the total amount of intermediate inputs needed per unit of an industry's output.

Proof. Since \(\mathbf{A}\) is nonnegative and its column sums are at most \(q\), each column sum of \(\mathbf{A}^{k}\) is at most \(q^k\). This follows by induction: it holds for \(\mathbf{A}^0=\mathbf{I}\), whose column sums equal \(1\); and multiplying a nonnegative vector by \(\mathbf{A}\) increases its sum by at most a factor \(q\). Therefore every entry of \(\mathbf{A}^k\) is between \(0\) and \(q^k\). Each entry of the series \(\sum_{k=0}^{\infty}\mathbf{A}^k\) converges, because it is bounded term by term by the convergent geometric series \(\sum_{k=0}^{\infty}q^k\). Call the resulting matrix \(\mathbf{S}\). For each integer \(N\geq 0\), telescoping gives \[ (\mathbf{I}-\mathbf{A})\sum_{k=0}^{N}\mathbf{A}^k =\mathbf{I}-\mathbf{A}^{N+1} =\left(\sum_{k=0}^{N}\mathbf{A}^k\right)(\mathbf{I}-\mathbf{A}). \] The entries of \(\mathbf{A}^{N+1}\) tend to \(0\), since they are bounded by \(q^{N+1}\). Taking entrywise limits in these finite matrix products gives \((\mathbf{I}-\mathbf{A})\mathbf{S}=\mathbf{I}\) and \(\mathbf{S}(\mathbf{I}-\mathbf{A})=\mathbf{I}\). Hence \(\mathbf{S}=(\mathbf{I}-\mathbf{A})^{-1}\). Multiplying by \(\mathbf{d}\) gives the stated production plan. \(\square\)

Separating Final Demand, Direct Inputs, and Later Rounds

For a unit of final demand in industry \(j\), the \(j\)th column of \(\mathbf{I}\) is the final-demand unit itself. The \(j\)th column of \(\mathbf{A}\) is the direct intermediate input requirement. The \(j\)th column of \(\mathbf{A}^2\) gives inputs needed to make those direct inputs, and so on. This distinction is useful when explaining why a total requirement exceeds the amount ordered by final customers.

Combining Final Demands Across Industries

For a demand vector with several nonzero entries, each final-demand sector contributes its own column of the Leontief inverse, scaled by the amount demanded. The total plan adds these contributions. Equivalently, \(\mathbf{x}=(\mathbf{I}-\mathbf{A})^{-1}\mathbf{d}\). This is a direct use of matrix multiplication, not a claim that every industry has the same requirement for every demand pattern.

Common Misreadings

Pitfall: treating a total requirement as a direct coefficient. In the two-industry example, the entry in row 1, column 2 of the Leontief inverse is \(\tfrac27\). It would be wrong to report \(\tfrac27\) as the direct amount of industry 1 needed per unit of industry 2 output. The direct coefficient is \(a_{12}=\tfrac14\). The extra amount comes from later rounds: industry 1 is needed to make industry 2's direct input, and the production chain continues. The inverse entry includes all of those rounds.

Pitfall: reversing the row and column interpretation. The entry \(L_{ij}\) answers: “How much output from industry \(i\) is required for one unit of final demand in industry \(j\)?” The column identifies the final-demand industry; the row identifies the industry whose output is being counted. In particular, the \(j\)th column—not the \(j\)th row—gives the requirements for final demand in industry \(j\).

The distinction between direct and total requirements matters when describing production chains, comparing industries, or explaining why gross output exceeds final demand. It also keeps the accounting equation visible: final demand is what remains after intermediate use is subtracted from gross output. The inverse packages that accounting across all industries, while its powers expose how the requirements build up round by round.