Infrastructure reliability and industrial utilization

A factory operating below capacity may lack electricity, materials, workers, or customers. Each condition can reduce measured utilization, but each calls for a different response. Reliable infrastructure enlarges the set of production schedules that a firm can deliver. It does not ensure that buyers will purchase the resulting output. Conversely, a strong order book cannot make an unavailable connection, congested freight corridor, or interrupted power supply disappear.

South Africa and India provide useful evidence on these distinctions. Their published manufacturing surveys report utilization alongside other business information, while energy statistics describe separate aspects of power availability. The countries differ substantially in industrial structure, electricity institutions, geography, and the coverage of individual surveys. The comparison can identify how constraints interact without assigning an identical institutional cause to similar observed outcomes.

The appropriate question is consequently conditional: how much additional saleable output could a specific infrastructure improvement support, given demand, materials, and production technology? That question requires both an engineering assessment and a commercial one. It cannot be answered by subtracting a national utilization rate from one hundred.

Statistics South Africa's November 6, 2025 release reports capacity utilization of 77.9 percent among large manufacturers in August 2025. Its corresponding underutilization measure is 22.1 percent. The stated reasons include insufficient demand at 12.4 percentage points, raw-material shortages at 3.1, labor shortages at 1.4, and other reasons at 5.2. These figures belong to that release and its weighting basis. They are reported allocations of unused capacity, rather than an experiment measuring independent causal effects. Statistics South Africa, P3043, August 2025, Table A.

The same publication introduces weights based on 2022 value added and provides a bridge to the preceding weighting system. This matters when reading changes across releases. A different aggregate can reflect a changed industrial composition as well as changed utilization within individual factories. The published division tables also show why a national average should not become an operating assumption for every plant.

In particular, the residual category does not provide a direct estimate of electricity losses. It includes several reasons, including maintenance. Allocating every residual percentage point to grid performance would be unjustified. Nor can an observed demand shortfall establish that supply improvements have no value. Reliable delivery can help a producer retain customers, qualify for contracts, or avoid precautionary inventories even when aggregate orders remain subdued.

Eskom's 2025 financial statements report a substantial reduction in system hours of load shedding and curtailment relative to the preceding reporting year. That is evidence about system operations within the utility's stated measure. It is not a record of every interruption experienced by each manufacturing customer. Local distribution faults, planned work, and a plant's own equipment can produce different exposure. Eskom, Annual Financial Statements 2025, operational performance.

The two publications can therefore be read together without being merged into a single causal series. One describes reported manufacturing capacity use; the other describes a utility operating outcome. Establishing their relationship would require matched periods, locations, customer connections, and a design that accounts for concurrent demand changes.

The Reserve Bank of India's October 9, 2024 OBICUS release provides an instructive measurement case. For April–June 2024, unadjusted capacity utilization fell to 74.0 percent from 76.8 percent in the preceding quarter, while seasonally adjusted utilization rose to 75.8 percent from 74.6 percent. The same underlying survey thus supported different descriptions of the sequential movement, depending on whether normal seasonal patterns were removed. These are historical survey results, not a claim about current utilization. Reserve Bank of India, OBICUS for Q1:2024–25, Table 1.

The release identifies voluntary company responses and publishes separate respondent counts for different indicators. Capacity, orders, and inventories should therefore not be assumed to describe an identical panel merely because they appear in one publication. Survey evidence can reveal business conditions without providing a census of all factories or a common cross-country production frontier.

India's Ministry of Power separately reports an all-India energy deficit of 0.1 percent for fiscal 2024–25. That balance compares energy requirements and availability at the system level. It does not establish that each industrial feeder supplied uninterrupted electricity or that every new project obtained its requested connection on time. Ministry of Power, February 5, 2026 parliamentary reply, Annex I.

The distinction is especially relevant for firms that require continuous processes, stable voltage, or a larger contracted load. A small aggregate energy gap and a material site constraint can coexist. Comparing that energy gap with South African load-shedding hours would mix an energy proportion with a duration measure. Comparing either directly with manufacturing utilization would introduce a third denominator.

Consider a hypothetical factory with technical capacity to produce 100 units during a period. Firm customer orders support sales of 80 units. Electricity interruptions restrict achievable production to 70 units. Material deliveries permit 90 units, and the available skilled workforce can support 85. Assume a fixed product, no inventories, no substitution between constraints, and no ability to sell above the order book. Actual production equals the smallest of these limits.

Increasing the electricity limit from 70 to 95 raises actual output by 10, not 25, because orders then become binding. Subsequent demand growth raises output by only another 5 because the workforce limit binds. Relieving the workforce constraint permits output of 90, at which point materials and orders jointly limit production. The sequence demonstrates complementarity without claiming that actual factories follow a rigid minimum function.

At a contribution margin of 20 currency units per additional saleable unit, the immediate operating benefit of the electricity intervention is 200. Valuing all 25 units of newly feasible electrical production at that margin would produce 500 and overstate the immediate benefit by 300. An intervention can have additional reliability, safety, or contract benefits, but those require separate measurement rather than being inserted into unused technical capacity.

Now assume orders had instead supported 95 units initially. With the same workforce and material constraints, improving electricity would raise output from 70 to 85. The operating contribution would be 300. The same technical intervention thus has a different short-run economic return under a different demand condition. Neither result estimates a return in South Africa or India; the example isolates the logic a plant assessment must establish.

Two plants can lose the same number of production hours while suffering different economic damage. A continuous process may need a lengthy controlled restart, discard work in progress, or operate below normal quality after power returns. A flexible assembly line may recover some hours through an additional shift. The relevant loss is consequently not a universal hourly multiple of annual revenue.

Interruption frequency and advance notice matter alongside duration. Ten short unexpected interruptions can be more damaging than one scheduled stoppage of the same total length if each requires cleaning and recalibration. Alternatively, a long interruption can exhaust backup fuel or breach a customer deadline that shorter events would not affect. Engineering specifications and order commitments determine which case applies.

Backup investment also changes the observed relationship between grid reliability and output. A factory may maintain production by purchasing generators, batteries, or redundant connections. Stable output can then coexist with a substantial cost of unreliability. A study that looks only for lost production will miss the resources spent preventing that loss. The counterfactual must include the production and costs that would have occurred without those measures.

The adaptation itself should be measured net of other benefits. A battery might provide tariff management as well as backup. A generator may serve an isolated construction project regardless of grid conditions. Assigning the entire purchase cost to a particular electricity failure would overstate its attributable burden. Operating records can distinguish use for resilience from other commercial functions.

A manufacturing bottleneck can occur outside the factory. An exporter with reliable electricity may still be unable to obtain inputs or secure a vessel departure. If material deliveries constrain production, the factory survey can report a raw-material shortage even when the underlying cause is a transport interruption. That reporting category does not identify the responsible institution.

The same issue arises when a plant has expanded nominal capacity but lacks a complementary rail connection, water supply, or trained maintenance team. Installed machinery increases the denominator before saleable output increases. Utilization can fall during expansion even while the firm's production and expected long-run capacity rise. The analyst must distinguish a commissioning period from a persistent inability to use established assets.

Cross-country comparisons should also examine the product mix. Capacity in a steel process, food-production line, and electronics assembly operation is defined through different practical operating conditions. A weighted national average combines these realities. Harmonizing sector coverage improves comparability, but does not remove differences in maintenance practices, shift conventions, or reported feasible output.

Matched establishment evidence would be more informative than a national scatterplot. It would connect interruptions and delivery times with orders, production, backup costs, investment, and employment at the same firms. A credible causal analysis would then need variation that separates the infrastructure change from other influences on those outcomes. Before-and-after movement alone cannot supply that separation.

Recovery after an interruption should be assessed over the complete order cycle. If a factory restores production through overtime in the next week, its annual output may be preserved while labor costs rise. If the customer cancels an order before recovery is possible, the same physical interruption can reduce annual sales. An accounting of downtime therefore needs the subsequent schedule and customer outcome. It should distinguish permanently lost production, delayed production, additional recovery costs, and damage to materials. Summing all initially interrupted output with later recovery costs would overstate the loss when some of that output is subsequently produced and sold.

For a new industrial project, the most useful reliability measure is the one tied to the promised operating schedule. That includes connection capacity, delivery date, interruption characteristics, and the cost of feasible backup. For an established producer, actual downtime logs and rejected orders can reveal where additional capacity would become commercially productive.

Financing can become binding even after the physical supply constraint is removed. A manufacturer may need to pay for additional materials and wages before collecting revenue from a larger order. Reliable power expands its feasible production set, but a working-capital limit can prevent immediate use of that opportunity. The project model should therefore connect the operating schedule with payment terms and available liquidity.

There is also a distinction between aggregate spare capacity and a suitable supplier's available capacity. An economy may have idle facilities producing the wrong specification or located far from the buyer. Qualifying an alternative supplier can require tooling, testing, and customer approval. A national utilization average cannot establish that spare machinery is commercially interchangeable across products or regions.

Policy appraisal needs a corresponding distinction. Removing a physical bottleneck may protect existing output, permit new production, or reduce adaptation costs. These benefits can occur at different times and accrue to different firms. Adding them requires consistent assumptions and attention to double counting, particularly when avoided production losses already include the value of preserved customer relationships.

South African and Indian evidence therefore supports a disciplined comparison of constraint types, rather than a single index of industrial readiness. Demand, infrastructure, skills, and materials jointly determine the use of productive assets. A useful assessment identifies the next binding limit after an intervention succeeds, because that is where an engineering improvement becomes an economic outcome.

See energy reliability and industrial location for a cost and financing treatment, and investment and Indian employment for the distinction between productive capacity and jobs.