The Premise
19 June 2026
Every July, Singapore’s civil servants find out how much they’ll get in their Annual Variable Component (AVC) — the lump-sum bonus tied to the economy’s performance the year before. This year, I wanted to test if AI is able to predict ahead of time: what will the AVC be before the government announces it?
So I turned the question over to AI as a fun experiment (using Deepseek V4 Flash). The brief was simple: use publicly available data, explain your reasoning, and give me a number. No leaks. No inside information — just math, logic, and whatever the government’s own open data tells us. I wasn’t trying to beat Bloomberg or outsmart MTI. I was curious whether a thoughtful AI assistant, given the same public signals any citizen can find, could land in a plausible range.
Here’s the methodology, the numbers, and what we got.
The Methodology
The AVC isn’t a mystery. It’s formula-adjacent. The government has broad discretion, but the NWC (National Wages Council) guidelines and historical precedent make it a rough function of:
- Previous year’s GDP growth — the strongest single indicator (2025: 5% → 0.4-month AVC)
- Current year’s GDP forecast — MTI’s official range
- Inflation — if CPI is high, the government tends to cushion; if low, they have room for generosity
- Fiscal environment — whether Budget had room for special payments (Budget 2026 already included a $200–$600 Cost-of-Living Special Payment)
- Employment and NODX — supporting indicators of underlying economic health
The AI scraped recent MTI press releases, MAS macroeconomic reviews, SingStat GDP data, CNA coverage, and Budget 2026 announcements. No proprietary data. No leaks. Everything below is sourced from what’s publicly available as of June 19, 2026.
The Data
Here’s what the numbers actually say:
| Indicator | 2025 (actual) | 2026 (actual/latest) |
|---|---|---|
| GDP growth | 5.0% | Q1: 6.0% YoY; Full-year forecast: 2–4% |
| Core inflation | ~2.0% | Latest (Apr): 1.4% |
| AVC paid (July 2025) | 0.4 months | ? |
| NODX growth | — | Q1: +9.6% (electronics +57.8%) |
| Global risk | Moderate | High (Iran war, Strait of Hormuz) |
| Budget support | Standard | Enhanced (CDC Vouchers brought forward, Cost-of-Living Special Payment) |
The tricky part: 2025’s 5% GDP (the basis for this year’s AVC) was a bumper year. 2026’s growth is expected to be lower — but Q1 came in well above expectations at 6%. The question is whether the government weights the past (strong 2025) or the future (uncertain second half 2026) more heavily.
The AI’s Reasoning
- Baseline: 0.4 months (what 5% GDP paid last year).
- Downgrade factor: 2026 growth forecast is 2–4% (midpoint ~3%), roughly 60% of 2025’s 5%. → Pure formula gives ~0.24 months. But the government doesn’t mechanically apply this.
- Offset 1 — Q1 strength: 6% YoY in Q1, NODX +9.6%, electronics +57.8%. The economy is not in trouble.
- Offset 2 — Inflation is low: 1.4% core CPI means the government isn’t squeezing households. They have room.
- Offset 3 — NWC norms: The NWC tends to recommend variable bonuses that acknowledge strong past performance while keeping powder dry for uncertainty.
- Offset 4 — Budget 2026 already cushioned: The Cost-of-Living Special Payment ($200–$600) and early CDC vouchers mean the government already spent some political capital on support.
The Prediction
0.3 months.
Not 0.4 (which would assume the strong 2025 GDP alone drives it). Not 0.2 (which would assume the worst). 0.3 is the middle path:
- Rewards 2025’s strong 5% performance
- Acknowledges Q1 2026’s 6% beat
- Leaves room for second-half uncertainty (Iran war, energy costs, US tariffs)
- Consistent with the government’s practice of smoothing rather than cliff-edging bonuses
Comparison to past years
| Year | GDP Growth | AVC (months) |
|---|---|---|
| 2022 | 4.0% | 0.35 |
| 2023 | 1.1% | 0.30 |
| 2024 | 4.4% | 0.45 |
| 2025 | 5.0% | 0.40 |
| 2026 (predicted) | ~3% midpoint | 0.30 |
Will AI Be Right?
We’ll find out in a few weeks. But the interesting thing isn’t whether the AI guesses the exact number — it’s that a reasonably competent prediction is now something anyone can run in about 15 minutes with public data and a capable assistant. The government isn’t opaque. The numbers are there. You just need someone willing to read them and connect the dots.
If the actual AVC comes in at 0.35 or even 0.4, that tells us the government is prioritizing morale and continuity over caution. If it’s 0.2, they’re bracing for a hard landing. 0.3 says they see a strong economy with meaningful, but manageable, headwinds.
(Fingers Crossed!) Let’s see how the AI does.
Disclaimer: AI can be incorrect. This is an exercise using public data. It is not financial advice, and the AVC is ultimately decided by the Singapore government based on factors beyond GDP alone.