Ghost GDP: When Growth and Prosperity Part Ways

The Philippines is reporting rising output while the income base underneath it quietly dwindles.

Two things are both true about the Philippine IT-BPM sector right now.

First: the industry is still growing. IBPAP reports that 2025 closed with roughly $40 billion in export revenues and 1.9 million jobs, and projects about $42 billion and 1.97 million jobs for 2026. GDP attributable to the sector is rising, and the headline numbers look stable.1

Second: the domestic income that made this sector socially consequential — the payroll and the purchasing power and middle class consumption that flows from it — is under real pressure as AI changes the arithmetic of how that output gets produced.

This scenario is the definition of Ghost GDP: a country reporting rising output while the income base that made the model socially durable thins out underneath it.

GDP measures the first part. What it doesn’t tell you is who keeps the money.

Where the dollar goes when labor is replaced

To understand the mechanism, follow a single dollar of value when a foreign-owned BPO operator automates a customer-service process in Manila.

Before automation, that dollar mostly stayed domestic. It paid a wage. The worker spent it on rent, groceries, tuition, transport. Each of those transactions rippled further, and the dollar cycled.

After automation, the dollar splits differently. Some still circulate locally: rent on the office, utilities, remaining staff, local vendors. But a materially larger share now goes elsewhere. The compute spend is billed to a foreign parent or cloud vendor. The productivity gain becomes margin, and the margin gets repatriated — as dividends, as management fees, as transfer-priced IP licensing back to a headquarters in Dublin or New York or Singapore.

GDP registers the output in both scenarios. The national accounts count the value of the service delivered. What they do not separately report is how the income attached to that output has been redistributed. We don’t see how much moved from payroll that circulates domestically to profit that repatriates.

That redistribution is already visible at the margins.2 A Concentrix worker who once handled about 30 calls across a full shift was, by late 2024, getting through that many before lunch with an AI co-pilot. Throughput per worker is rising. And once throughput rises, management stops asking how to hire to fill an account. It starts asking instead how few people it takes to hold service levels. The headcount line bends. The revenue line holds. The gap between them is where Ghost GDP lives.

Why this country is especially exposed

Every country whose service sector faces AI disruption faces some version of this dynamic. The Philippines faces a more acute version because of a structural feature: almost all of the BPO value chain is foreign-owned.

When TCS or Infosys automates a customer-service operation in India, for example, the productivity gain stays in India. It gets reinvested in R&D, in new service lines, in the next generation of workers. The IndiaAI Mission carries $1.25 billion in government funding and over 38,000 GPUs of publicly backed compute.

The Philippines has no TCS equivalent.3 The $40 billion sector is almost entirely controlled by Accenture, Teleperformance, Concentrix, and their peers. When those firms automate, the Philippines absorbs the employment loss while the foreign shareholders collect the productivity upside.

This is the ownership problem at the heart of Ghost GDP. It is not that automation is happening here — it is happening everywhere. It is that the gains from automation in a foreign-controlled sector do not stay here.

The blast radius

Ghost GDP becomes a systemic risk because of how deeply IT-BPM is embedded in the domestic economy.

IT-BPM accounted for 44 percent of Philippine office transactions in 2024.4 Metro Manila office vacancy was already 19.8 percent at year-end.5 BSP data put banking-system real estate exposure at ₱3.3 trillion as of March 2025.6 A sector that anchors office demand, mortgage payments, grocery spending, school fees, and the lion’s share of formal middle-class income does not contract neatly.

Pressure leaks in sequence: BPO headcount → office vacancy → developer revenues and REIT stress → household credit → urban retail and transport → local government tax bases → bank balance sheets. Each link is downstream of the one before it.

The exposure assessments reinforce the concern. An IMF working paper found roughly 14 percent of Philippine workers in high-exposure, low-complementarity roles, with BPO carrying the highest displacement risk.7 ILO analysis put 89 percent of salaried BPO workers in the high-risk category. The directional finding is consistent across methodologies: routine cognitive service work is at the front of the line.

Why Ghost GDP is hard to see in real time

The near-term numbers will obscure the trend. This is what makes the concept analytically important: Ghost GDP does not announce itself.

Reuters reported in 2024 that 8 percent of surveyed IBPAP member firms had already reduced headcount because of AI, while 13 percent reported gains.8 That is the early shape of the wave. It does not arrive as a single employment break. It shows up as narrower teams, higher per-agent targets, slower backfill, and a widening gap between what the firm bills and what the workforce takes home. Revenue and GDP hold. The payroll diminishes quietly.

Klarna is instructive if you read the full sequence rather than the headline.9 In early 2024, the company reported its AI assistant was handling the equivalent of 700 full-time agents. Total headcount fell from roughly 5,000 to 3,800. By September 2025, Klarna was walking some of that back after quality problems surfaced — and then iterating with better models. The lesson is in the cycle: firms cut labor when the economics look good enough, spend the next phase fixing quality, and return with tighter systems. Each iteration compresses the window in which human labor at volume retains its advantage.

For the Philippines, the risk is that aggregate statistics (sector revenue, GDP) will stay reassuring long enough to delay the policy response past the point where it matters.

The right scorecard

GDP was not designed to answer the question this transition is actually asking. It measures the value of output produced within the country’s borders. It does not distinguish between output whose income stays and output whose income leaves.

Three metrics would give a clearer picture of Ghost GDP in real time:

  • Gross National Income versus GDP. GNI adjusts for income flows across borders; it subtracts profits repatriated by foreign firms and adds remittances and overseas earnings. In an economy where automation is shifting income from domestic wages to foreign profits, the GDP-GNI gap will widen before the employment numbers signal a crisis.

  • Remitted profits as a share of sector revenue. As the payroll-to-revenue ratio falls, the profit-repatriation-to-revenue ratio rises. Tracking this directly inside IT-BPM would make visible what aggregate GDP conceals.

  • The current account trajectory under automation scenarios. The scenario modeling in the underlying analysis projects that broad AI adoption could shift the current account from ~3.4 percent of GDP to as low as ~7.5 percent as BPO export demand restructures.3 That deterioration would arrive gradually, well before any nominal GDP decline. It is the financial signature of Ghost GDP made concrete.

Against this, the current policy response is not only weak, it’s calibrated to the wrong instrument. Project UNLAD carries ₱740 million over four years — roughly $1.70 per IT-BPM worker per year in dedicated government transition funding.10 Training programs matter. But they are a labor-market response to what is also a balance-of-payments problem and an ownership problem. Reskilling workers for a sector whose profit pools are foreign-controlled addresses one symptom while leaving the structural condition intact.

The Philippines can reskill every worker in the sector and still lose most of the upside if the profit pools, client control, and product decisions remain in foreign hands. The right scorecard would make that visible. The right policy response would treat it as the problem it is.

GDP will keep reporting that the economy is growing. But if the old model is simply defended until the margin pool migrates, GDP may hold up while the middle class hollows out.

The Philippines built one middle class on rented labor. The next one has to be built on something it actually owns.

Footnotes

  1. “Philippine IT-BPM Industry Closed with Growth, Moves Forward with Optimism,” IBPAP, 2026. https://ibpap.org/news-room/35 ↩

  2. “AI is reshaping call center work in the Philippines,” Rest of World, 2024. https://restofworld.org/2024/ai-reshaping-call-center-work-philippines/ ↩

  3. “From Rented Labor to Owned Capability,” Bernd’s Blog, March 2026. https://blog.brennebeck.com/posts/from-rented-labor-to-owned-capability/ ↩ ↩2

  4. “IT-BPM industry sees strong growth and strategic shifts in 2024,” Colliers Philippines, BPO Situationer Part 5. https://www.colliers.com/en-ph/news/bpo-situationer-part-5-it-bpm-strong-growth-strategic-shift-2024 ↩

  5. Colliers Philippines, Quarterly Property Market Report: Office Q3 2025. https://www.colliers.com/en-ph/research/colliers-quarterly-property-market-report-office-q3-2025-philippines ↩

  6. Bangko Sentral ng Pilipinas, Quarterly Update on the Banking System, Q1 2025. https://www.bsp.gov.ph/Media_And_Research/Publications/QUBS-2025-Q1.pdf ↩

  7. “Artificial Intelligence and the Philippine Labor Market,” IMF Working Paper WP/25/43, February 2025. https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025043-print-pdf.pdf ↩

  8. “Philippine outsourcing to grow 7% this year despite AI threat, industry group says,” Reuters, October 2024. https://www.reuters.com/technology/artificial-intelligence/philippine-outsourcing-grow-7-this-year-despite-ai-threat-industry-group-says-2024-10-02/ ↩

  9. “Klarna’s AI assistant does the work of 700 full-time agents,” OpenAI, 2024. https://openai.com/index/klarna/ ↩

  10. “IT-BPM exec optimistic of sector’s growth amid challenges,” Philippine News Agency, 2026. https://www.pna.gov.ph/articles/1267882 ↩