# Employment intensity across service and manufacturing expansion

## The number of jobs depends on what expands

Two economies can achieve the same increase in real output and record different changes in employment. The difference may reflect sector composition, productivity, working hours, labor-force conditions, or the definition of a job. A comparison that labels one expansion employment intensive without identifying those mechanisms can mistake a statistical boundary for an economic result.

The United States and India present a particularly important measurement challenge. American payroll records, household employment, and productivity accounts cover different labor concepts. India's factory survey and household labor survey likewise observe different populations and reference periods. Matching a broad national growth rate to whichever employment count is available does not produce a comparable estimate of jobs generated by growth.

Sector labels are also insufficient. Services include highly scalable software, transport networks, professional activities, retail operations, and activities requiring direct personal contact. Manufacturing includes capital-intensive processes and labor-intensive assembly. A useful comparison begins within more specific activities and then examines how their shares shape the aggregate result.

## Published US evidence does not support one service-sector pattern

The Bureau of Labor Statistics' August 26, 2026 release covers thirty selected service-providing industries for 2025. It reports that seven had rising output and falling hours worked. The release also reports widely different productivity changes across those industries. This is evidence of heterogeneity within the selected coverage, rather than a measure of the whole service economy. [BLS, selected service-providing industries, 2025 results](https://www.bls.gov/news.release/prin2.htm).

The corresponding June 24, 2026 manufacturing and mining release reports declining hours in fifty of eighty covered four-digit manufacturing industries in 2025. Its technical note distinguishes real sectoral output from labor hours and explains the combination of employee, self-employed, and unpaid-family-worker information. These measures do not identify a worker's isolated contribution; they reflect the production system in which labor is used. [BLS, manufacturing and mining industries, 2025 results](https://www.bls.gov/news.release/prin.htm).

Neither release supports the general claim that every manufacturing expansion requires proportionately more workers, or that services always absorb labor more readily. An industry's output may rise while its hours fall when productivity increases sufficiently. Conversely, a weak productivity performance can make a modest output increase labor intensive without establishing an improvement in earnings or productive capability.

The employment denominator matters. BLS's establishment-survey guidance counts a person working at different establishments once at each establishment. Those are payroll jobs. The household concept counts employed people rather than adding their jobs together. A move from one to two part-time positions can therefore affect a job count differently from a person count. [BLS, Current Employment Statistics, question 9](https://www.bls.gov/web/empsit/cesfaq.htm).

## Indian factory evidence has a defined perimeter

India's Annual Survey of Industries release for fiscal 2023–24 reports a 5.92 percent increase in total persons engaged within its covered factory sector. Its current-price gross value added rose 11.89 percent. The reference period is April 2023 through March 2024, and the statistical coverage is specified in the release. These are not national employment or all-manufacturing totals. [Ministry of Statistics and Programme Implementation, ASI 2023–24, August 27, 2025](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2161192&lang=2&reg=3).

Dividing the reported employment growth by that nominal value-added growth would not estimate an elasticity with respect to real production. Prices and industrial composition are included in the value measure. Nor should persons engaged be silently relabeled as payroll employees. The factory survey's categories and a household survey's account of individuals must be reconciled before they enter a comparison.

The Indian case is therefore not a reason to disregard measured factory employment growth. It is a reason to state what the evidence covers. A formal factory can add jobs while employment elsewhere falls, and a household can move from a small enterprise into a surveyed factory without increasing the national number of employed people. The transition may still improve wages, stability, or productivity; those outcomes require their own evidence.

For both countries, a comparable study should specify the same production boundary and a consistent labor input. A national real-value-added measure paired with all persons or all hours can answer one question. Establishment jobs paired with matched industry production can answer another. Mixing the boundaries makes the resulting ratio difficult to interpret.

## Exhibit: identical output growth, different staffing requirements

Consider three hypothetical activities, each initially producing one hundred units of real value added. Their production methods and average working schedules differ. Assume constant output per hour and unchanged hours per worker over the short comparison. The figures below are illustrative parameters, not estimates for American or Indian industries.

| Activity | Initial annual hours | Hours per worker | Initial workers | Extra hours for twenty additional output units | Extra workers |
|---|---:|---:|---:|---:|---:|
| Scalable business service | 1,000 | 100 | 10 | 200 | 2 |
| Personal service | 2,000 | 50 | 40 | 400 | 8 |
| Manufacturing process | 1,500 | 100 | 15 | 300 | 3 |

An extra twenty output units require different staffing because both hourly productivity and schedules differ. Expanding personal services appears more employment intensive than expanding the manufacturing activity. Expanding the scalable service appears less intensive. A service-versus-manufacturing dichotomy would miss the larger difference within services.

Now compare two hypothetical growth portfolios. Portfolio A adds twenty output units in the scalable service and twenty in manufacturing. It adds forty output units and five workers. Portfolio B adds twenty in personal services and twenty in manufacturing. It also adds forty output units but eleven workers. The output increase is identical; the employment increase differs by six because of sector mix and working schedules.

Convert labor to equivalents of one hundred annual hours. Portfolio A adds five such equivalents: two in the scalable service and three in manufacturing. Portfolio B adds seven: four in personal services and three in manufacturing. Its headcount increase is eleven, but its comparable-hour increase is seven equivalents. The gap between the portfolios narrows when hours are standardized, though it does not disappear.

This example does not say that short schedules are inferior or that fewer jobs are desirable. Shorter hours can reflect preference, care responsibilities, cyclical weakness, or an employer's scheduling needs. The point is that workers, jobs, and labor hours describe different aspects of employment absorption. All can matter to an assessment of growth.

## Productivity changes the required hours

Remove the assumption of constant hourly productivity in the manufacturing example. Suppose output rises twenty percent while output per hour increases ten percent. Required hours rise by the factor 1.20 divided by 1.10, or 1.0909. The factory needs approximately 136.36 additional hours rather than three hundred. At unchanged schedules, that is approximately 1.36 additional workers.

If average hours per worker fall five percent at the same time, the required worker count rises by 1.20 divided by the product of 1.10 and 0.95, or 14.83 percent. Starting from fifteen workers, employment becomes approximately 17.22. Headcount rises more than hours because each worker supplies fewer hours. The arithmetic does not explain why productivity or schedules changed; it identifies what a consistent interpretation must account for.

A jobs-to-output elasticity computed over one short interval can therefore be unstable. If output growth is close to zero, a small change in the denominator can produce a very large ratio. If output falls, the sign becomes difficult to summarize with ordinary language about job creation. Longer windows can help, but they can combine several phases of adjustment and structural change.

The relevant horizon also differs between investment and employment. A factory can hire construction workers before producing anything, then employ a smaller permanent operating workforce. A software or professional-service project can invest in development before expanding sales. Comparing employment during one phase with output during another can create an apparent sector difference that reflects timing.

## Outsourcing can move the sector label without moving the work

A manufacturer can employ its own cleaners, logistics staff, and information-technology team, or purchase those services from other firms. Outsourcing reduces the manufacturer's direct employment and increases employment classified elsewhere. The work can remain at the same location and support the same production. A fall in manufacturing payrolls is therefore not always a disappearance of the relevant labor input.

The output boundary changes as well. Value added subtracts purchased intermediates, whereas gross output includes a different set of transactions. A comparison using manufacturing gross output in one country and value added in another would respond differently to changes in outsourcing. Matched production concepts are essential before attributing the employment difference to technology.

Supplier employment should be measured without counting the same workers twice. An input-output model can estimate indirect requirements, but the model's fixed coefficients, import treatment, and demand assumptions must be stated. Adding a firm's own jobs to all jobs at its suppliers would include supplier activity serving other customers. A credible estimate attributes only the relevant share.

The same principle applies to household enterprises and contract work. Changes in registration or contracting can move activity between statistical categories while preserving some underlying tasks. Better formal coverage can improve measurement, but a series break should not be interpreted as an equivalent increase in economic activity.

Net employment change also conceals turnover. Suppose an expanding industry adds one thousand positions at growing establishments while contracting establishments remove eight hundred. Net employment rises by two hundred, but the number of people facing a transition can be much larger. Some workers may move directly between establishments; others may need new skills or relocate. An input-output or aggregate elasticity calculation does not reveal those paths. A comparative employment assessment should distinguish establishment creation and destruction from the change in the stock of jobs, and both from unique people hired. The distinction becomes particularly relevant when expanding service activities require different locations or qualifications from the manufacturing jobs being lost. A positive aggregate employment result can coexist with difficult adjustment for specific workers and communities.

## Employment quality belongs beside quantity

An expansion can create many low-hour positions and little additional income, or fewer positions with higher compensation and more stable schedules. Neither headcount nor productivity alone resolves that comparison. Earnings, working time, employment security, safety, and access to training provide information about the opportunities generated by growth.

For self-employed workers, labor income and business profit are not always separately observed. A rise in persons working in household enterprises can reflect productive entry, distress employment, or changes in reporting. The interpretation depends on income, hours, capital, and demand conditions. Treating all employment additions as equivalent units of welfare would obscure those differences.

Gender, age, and geography can also affect who can take the new work. An industry requiring relocation, long shifts, or credentials may create vacancies without absorbing workers leaving another activity. A national output-employment relationship can conceal that mismatch. The appropriate policy response may involve transport, care services, skills, or local supply conditions rather than an effort to maximize a single employment coefficient.

For businesses, the practical staffing forecast should connect the expected product mix with tasks, operating hours, productivity, and the use of purchased services. For policy analysis, the aggregate comparison should show how much of the result comes from sector mix and how much from changes within matched industries. Those decompositions make claims about employment-intensive growth more useful and more open to verification.

The United States–India comparison consequently requires a bridge between statistical systems before it requires a ranking. The official evidence demonstrates why services, manufacturing, payroll jobs, persons engaged, and hours cannot be collapsed into one undifferentiated measure. Growth can support employment through several channels; identifying the relevant channel is the basis for assessing its durability and economic value.

See [investment and the employment content of Indian growth](/research/investment-and-the-employment-content-of-indian-growth/) for a national structural treatment, and [manufacturing constraints](/research/infrastructure-reliability-and-industrial-utilisation/) for the difference between available capacity and saleable output.

---

American Economic Senate