Quick answer
Economic Survey 2025-26 finds weaker output responsiveness of U.S. service employment after December 2022; this is structural evidence, not AI causation.
AI’s employment effect is the net result of task substitution, worker augmentation and demand expansion. Box XIV.1 finds a changed employment–output relationship in U.S. professional, business and information services (PBIS), warning India against both displacement alarmism and complacency.
Concept diagram
Redesign tasks, not just job titles
A human reviews AI-assisted work: augmentation depends on task design and skills.
- Task exposure is not a forecast of equivalent job losses.
- Training and organisational redesign shape augmentation versus displacement.
Basics: tasks are not whole occupations
Augmentation and substitution
- Augmentation helps a worker perform tasks; substitution shifts tasks to technology. Both can occur within the same occupation.
- An LLM drafting a response may complement an agent who verifies it, even if some routine drafting work disappears.
Employment elasticity
- Employment elasticity is the percentage change in employment associated with a percentage change in output. A decline need not mean employment itself falls.
- Training models creates integration work, while repeated inference can change the demand for routine tasks. Neither establishes a net employment effect by itself.
ARDL and causality
- An autoregressive distributed lag model relates current outcomes to their own lags and current or lagged explanatory variables.
- The Survey uses March 2016–July 2025 observations and a post-December 2022 indicator; it explicitly tests structural change, not a causal AI treatment.
Table XIV.1: Short-run dynamics — coefficients only
| Variable | PBIS | Non-PBIS |
|---|---|---|
| D(Regime indicator) | 25.74 | 1.58 |
| D(Interaction) | −2.572 | −0.00007 |
| Error correction term | −0.217 | −0.07 |
Table XIV.2: Long-run dynamics — coefficients only
| Variable | PBIS | Non-PBIS |
|---|---|---|
| AI / post-2022 regime indicator | 9.409 | 1.90 |
| Interaction with real output | −0.938 | −0.00795 |
Read Box XIV.1 without overclaiming
Source and specification
- Source: Economic Survey 2025-26, Box XIV.1 and Tables XIV.1–2; author’s calculations. Tables above reproduce coefficients, not standard errors.
- Controls include wages, the federal funds rate, monthly real GDP and CPI; the bounds-test F-statistic is significant at 5%.
The two results can coexist
- The PBIS regime indicator has a positive association with employment, while its interaction with output is negative and statistically significant in both horizons.
- Employment can be supported during organisational adjustment even while the marginal employment response to output becomes weaker.
Sectoral comparison
- Non-PBIS employment shows a near-zero, statistically insignificant long-run regime association, unlike PBIS.
- This comparison supports a sector-specific structural interpretation; it does not identify task-level automation or eliminate every alternative explanation.
Implications for India’s labour transition
Monitor entry-level work
Track tasks, hiring and output together rather than relying only on headline employment.
Build complementary skills
Support domain expertise, verification, judgement and workflow redesign.
Enable reallocation
Pair adoption with retraining and pathways into higher-value services.
Common exam traps
Statements examiners like to twist — learn the correct version.
A positive regime coefficient proves AI created jobs.
The regime is a time indicator, not a measured causal AI treatment.
−0.938 means AI destroyed 0.938% of jobs.
It is the reported interaction coefficient; no such job-loss interpretation is warranted.
U.S. PBIS estimates are India’s employment forecast.
They are early-warning evidence from a different labour market.
Prelims focus
- PBIS means professional, business and information services.
- Study window: March 2016–July 2025; regime begins after December 2022.
Mains angle
- Discuss how employment growth can coexist with declining labour intensity of output.
- Use Box XIV.1 to distinguish association, structural change and causality.
Flip & recall
Answer in your head first, then tap the card to check.
Practice MCQs
Original Prelims-style questions with explanations.
0 / 3 correct · 3 left
Q1.Box XIV.1 is explicitly designed as:
Q2.A lower employment response to output necessarily implies:
Q3.The PBIS interaction with output after 2022 is reported as:
