In 2021, a shortage of semiconductor chips — produced by a handful of factories in Taiwan, South Korea, and the Netherlands — forced automotive plants in Germany, the United States, and Japan to halt production. Car dealerships ran out of inventory. Used car prices in the United States rose by 45% in a single year. Simultaneously, the shortage affected medical devices, consumer electronics, industrial equipment, and home appliances.
A shortage in one narrow category of product, produced in a handful of geographic locations, cascaded through dozens of industries across dozens of countries. To understand why this happened — and to quantify the likely effects — economists reached for a tool developed eighty years earlier by a Harvard professor with a fundamentally simple insight: in an economy, everyone is both a buyer and a seller.
Wassily Leontief and the Architecture of the Economy
Wassily Leontief was born in St. Petersburg in 1905 and died in New York in 1999. His most important contribution to economics was developed over the late 1930s and 1940s: the input-output model, also called interindustry analysis.
Leontief's central observation was that a modern economy is not a collection of independent industries. It is a network of interdependencies. Steel mills produce steel, but they also buy electricity, coal, machinery, financial services, and transportation. Automobile manufacturers buy steel, glass, rubber, electronics, and labor. Retailers buy automobiles for their fleets, electricity for their buildings, and financial services for their operations.
Every industry is simultaneously a producer and a consumer of other industries' outputs. Leontief's insight was to make this web of relationships quantitative and systematic, capturing it in a mathematical structure that could be used for analysis and prediction.
For this work, Leontief was awarded the Nobel Memorial Prize in Economic Sciences in 1973.
What an Input-Output Table Shows
The basic data structure of input-output analysis is the input-output table — a matrix that systematically records every flow of goods and services between every pair of industries in an economy over a given year.
The table has a specific structure:
Rows: Each row represents an industry as a seller. Reading across a row shows where that industry's output went — how much it sold to every other industry, and how much went to final demand (household consumption, business investment, government spending, exports).
Columns: Each column represents an industry as a buyer. Reading down a column shows where that industry got its inputs — how much it purchased from every other industry, and how much came from imports and primary inputs (labor, capital, taxes).
The intersection: The cell at row i, column j shows the value of goods and services that industry i sold to industry j.
A simplified version might look like this:
| Agriculture | Manufacturing | Services | Final Demand | Total Output | |
|---|---|---|---|---|---|
| Agriculture | 50 | 300 | 20 | 630 | 1,000 |
| Manufacturing | 100 | 500 | 150 | 1,250 | 2,000 |
| Services | 80 | 400 | 300 | 1,220 | 2,000 |
| Value Added | 770 | 800 | 1,530 | — | — |
| Total Input | 1,000 | 2,000 | 2,000 | — | — |
In this simplified table, agriculture sells 300 units to manufacturing (perhaps grain to food processing), 50 units to itself (perhaps seed), 20 units to services, and 630 units to final demand (consumers, exports). Reading down the manufacturing column, it buys from agriculture, from itself (components), and from services, and employs workers and capital (value added).
The Technical Coefficients Matrix
The input-output table records dollar flows. Leontief's key analytical step was to convert these flows into technical coefficients — the fraction of each industry's inputs that come from each other industry.
The technical coefficient a(ij) = how much industry j purchases from industry i per unit of industry j's output. If the automobile industry needs $200 worth of steel for every $1,000 of cars it produces, the technical coefficient from steel to automobiles is 0.2.
These coefficients reflect the production technology of each industry — they describe how each industry converts inputs into outputs. They are assumed to be relatively stable over time (a strong assumption, but workable for many purposes).
With the technical coefficients matrix in hand, Leontief showed how to calculate the total requirements matrix — the amount of output from every industry required (directly and indirectly) to support one unit of final demand for any industry's output.
This is where the model becomes powerful. The total requirements matrix captures not just the direct inputs to production but all the inputs to the inputs, and the inputs to those, through as many rounds of production as necessary. It answers the question: if consumers want one more car, how much more steel, electricity, coal, glass, rubber, and every other input must the economy produce?
Direct and Indirect Effects: Following the Supply Chain
The distinction between direct and indirect effects is the heart of what makes input-output analysis useful.
Direct effects: A car factory needs steel. If car demand rises, steel demand rises directly.
Indirect effects: Steel production requires electricity. More steel means more electricity. Electricity generation requires coal and water. More electricity means more coal and water. And so on through every round of the supply chain.
Induced effects: Workers in expanding industries earn more income and spend it, generating demand across the economy. Including these induced effects (through a social accounting matrix extension of the basic model) captures additional rounds of impact.
The total effect is typically two to three times the direct effect for mature industrial sectors, though the multiplier varies substantially by industry structure and the stage of economic development.
This multiplier logic explains why governments use input-output models when evaluating infrastructure investments, industrial policy, or economic stimulus. A dollar of spending on construction does not simply generate a dollar of construction output — it generates construction, plus all the construction inputs (cement, lumber, steel, labor, engineering), plus all the inputs to those inputs.
The COVID Pandemic and Real-Time Input-Output Thinking
The COVID-19 pandemic provided an extraordinary demonstration of input-output dynamics in real time, and showed the limitations of economies that had optimized their supply chains for efficiency rather than resilience.
Personal Protective Equipment
In early 2020, hospitals and healthcare systems faced acute shortages of personal protective equipment (PPE): surgical masks, N95 respirators, gloves, and gowns. From an input-output perspective, the question was: why could production not simply ramp up to meet demand?
The answer was a cascade of upstream dependencies. N95 respirators require specialized filtering materials — primarily meltblown polypropylene — produced by a small number of factories with specialized equipment that cannot be quickly replicated. Meltblown polypropylene requires specialized polymer extrusion machines, produced by an even smaller number of manufacturers, mostly in Germany and China. The machines require specialized tooling. Each link in the chain had limited spare capacity.
Input-output analysis revealed that "making more masks" was not a single-industry problem. It was a multi-layer supply chain problem involving polymer chemistry, precision machinery manufacturing, and international trade flows.
The Semiconductor Shortage
The 2021-2022 semiconductor shortage is a case study in second and third-order input-output effects. When automotive manufacturers curtailed orders for chips in early 2020 — expecting a demand collapse — semiconductor foundries reallocated capacity to consumer electronics. When automotive demand rebounded faster than expected in 2021, the allocated capacity was unavailable.
The result: automotive production was constrained by a component that represented a small fraction of the value of a car (typically 2-3% of a vehicle's cost) but without which the vehicle could not be produced. Input-output analysis would call this a bottleneck effect — a critical node in the network with no substitute and limited spare capacity.
The impact cascaded predictably through the input-output network:
- Reduced automotive output affected steel, aluminum, and glass demand
- Reduced new vehicle availability raised used vehicle prices
- Supply-constrained dealers reduced advertising spend, affecting media
- Delayed vehicle deliveries affected logistics and transportation industries
A study using input-output methods estimated that the semiconductor shortage reduced US GDP by roughly $110 billion in 2021, far exceeding the value of the chips themselves.
Industrial Policy and the Input-Output Lens
Governments use input-output analysis to identify strategic industries — sectors with unusually large forward or backward linkages to the rest of the economy. Industries with large backward linkages purchase substantial inputs from many other domestic industries, generating broad economic activity when they expand. Industries with large forward linkages supply critical inputs to many other industries, creating vulnerability when they contract.
Semiconductors, steel, and energy are examples of industries with strong forward linkages — disruptions in these sectors ripple forward into virtually every other sector. This is part of the economic rationale for the US CHIPS and Science Act (2022), which invested $52 billion in domestic semiconductor manufacturing, and for industrial policies focused on "critical industries" in the European Union and elsewhere.
The input-output framework makes explicit what politicians sometimes express more loosely: some industries matter not just for their own output but for what they enable elsewhere in the economy.
"The input-output model reveals that an economy is not a collection of markets operating independently but an integrated network in which the failure of any critical node affects all nodes it connects to, directly or indirectly." — Wassily Leontief, Essays in Economics (1966)
Input-Output Analysis in Climate Economics
Perhaps the most significant contemporary application of input-output analysis is in climate economics, where it provides the foundation for consumption-based emissions accounting.
Why Production-Based Accounts Miss the Full Picture
Most national emissions inventories count emissions where they occur — what is produced within a country's borders. This production-based accounting can be misleading. A country that deindustrializes and imports manufactured goods appears to reduce its emissions, but the global total is unchanged; the emissions have simply moved to the exporting country.
Input-output methods enable consumption-based accounting: attributing emissions to the final consumers of goods and services, regardless of where production occurs. By tracing the full supply chain — using the Leontief inverse matrix to capture all direct and indirect inputs — researchers can estimate the total emissions embedded in each unit of final demand.
The results are revealing. For high-income countries with strong environmental regulations, consumption-based emissions are typically 10-30% higher than production-based emissions. The difference represents emissions generated in other countries to serve domestic consumption — sometimes called carbon leakage.
Climate Policy Design
Input-output models are used to analyze the economy-wide effects of carbon pricing, clean energy transitions, and industrial policy. A carbon tax affects not just the directly taxed activities but all industries that use carbon-intensive inputs.
For example, a tax on steel (which is carbon-intensive to produce) raises costs not just for steel producers but for every industry that uses steel: construction, automotive manufacturing, industrial equipment, and many others. Input-output analysis can trace these cost increases through the economy, estimating effects on prices, output, employment, and trade competitiveness in each sector.
The Global Trade Analysis Project (GTAP) database, widely used in academic and policy research, integrates input-output tables for over 140 countries in a framework that allows modeling of trade flows alongside domestic economic linkages.
Limitations of the Model
Input-output analysis is powerful but rests on assumptions that limit its accuracy in certain contexts.
Fixed coefficients: The model assumes that production technology (the mix of inputs per unit of output) does not change. In reality, firms substitute inputs as relative prices change, adopt new technologies, and change production processes over time. The model works best for short-run analysis and is less reliable for projecting long-run structural changes.
No capacity constraints: The basic model assumes that any industry can expand output to meet demand given sufficient final demand, without supply constraints. In practice, industries have capacity limits, and expansion takes time and investment.
Linear relationships: Input-output is a linear model. Real economies have nonlinearities: economies of scale, threshold effects, and tipping points. The model handles proportional changes well but not step-change disruptions.
Static structure: Input-output tables are typically constructed from national accounting data that is several years old and updated infrequently. Dynamic processes — new industries emerging, old ones declining — are captured only slowly.
No price dynamics: The basic model does not explicitly model how prices change in response to supply and demand shifts. Extensions (the Leontief price model) address this, but standard applications focus on quantities.
Despite these limitations, input-output analysis remains one of the most widely used tools in applied economics, precisely because its core insight — that economic sectors are interdependent in ways that can be systematically measured — is both correct and practically important.
Conclusion
Wassily Leontief built the input-output model from a simple observation: you cannot understand any part of an economy in isolation, because every industry is simultaneously a buyer and a seller. Changing any one part changes every other part, through chains of inputs and outputs that can extend across dozens of industries and many countries.
This observation, formalized in a mathematical framework and applied to empirical data, produced one of economics' most useful tools. It explains why semiconductor shortages ground auto assembly lines, why carbon taxes affect everything made of steel, and why government investment in one sector generates income and employment across the economy.
In a world of increasingly integrated global supply chains — and increasingly complex policy challenges involving energy transitions, industrial strategy, and pandemic preparedness — the input-output model's essential insight is more relevant than ever. The economy is not a collection of independent parts. It is a network, and networks have properties that their individual nodes do not.
Frequently Asked Questions
What is the input-output model in economics?
The input-output model, developed by Wassily Leontief in the 1940s, is a quantitative framework that maps the interdependencies between different sectors of an economy. It shows how the output of each industry becomes an input to other industries — how steel production feeds car manufacturing which requires transportation which uses fuel. By capturing these linkages in a matrix of coefficients, the model allows analysts to trace how a change in demand or supply in one sector cascades through the entire economy.
Why did Wassily Leontief win the Nobel Prize?
Wassily Leontief was awarded the Nobel Memorial Prize in Economic Sciences in 1973 for his development of the input-output method and its application to economic problems. His work provided the first systematic empirical tool for analyzing the structure of entire economies, enabling governments, planners, and researchers to understand how industries depend on each other and to estimate the full economic effects of policy changes, price shocks, or demand shifts.
How does the input-output model explain supply chain disruptions?
When a sector fails to produce — because of a natural disaster, pandemic, geopolitical shock, or other disruption — input-output analysis can estimate the downstream effects on every sector that uses that sector's output, and the sectors that supply those sectors, and so on. The 2021 semiconductor shortage, for example, disrupted not only electronics but automotive production, home appliances, and medical devices, as the input-output relationships between these industries made them all dependent on the same upstream bottleneck.
What is an input-output table?
An input-output table is a matrix where rows represent industries as sellers and columns represent industries as buyers. Each cell shows the value of goods and services sold by the row industry to the column industry in a given year. Final demand (consumption, investment, government, exports) appears in additional columns, and value added (wages, profits, taxes) appears in additional rows. Reading down a column shows where an industry gets its inputs; reading across a row shows where its output goes.
How is input-output analysis used in climate economics?
Climate economists use input-output models to trace the full carbon footprint of goods and services — not just direct emissions but the emissions embedded in all the inputs that produced them. A t-shirt's carbon footprint includes the cotton growing, the fabric manufacturing, the dyeing, the transport, and the retail — all captured through input-output linkages. These 'consumption-based' emissions accounts show that countries with strong environmental regulations often import emissions-intensive goods from other countries rather than eliminating them.