NYSE, HKEX & SGX Size Premium Data by Market-Cap Decile

The same CAPM-alpha-per-decile methodology, applied independently to three markets — NYSE-listed, HKEX-listed, and SGX-listed common stocks, each split into 10 market-cap deciles (D1 = largest, D10 = smallest) and equal-weighted, for direct comparability across tabs. SGX is the market the Calc page's own build-up is based on, though its original study used 5 value-weighted quintiles instead — see that tab for why. All three are static snapshots of one-off analyses, not live calculators, and are entirely independent builds — different universes, different benchmarks, and (as the HKEX and SGX tabs below show) very different robustness results.

Rolling 10-year size premium by decile

0% 4% 8% 12% 16% 20% +1.3% D1 (largest) +1.6% D2 +0.8% D3 +3.3% D4 +3.6% D5 +5.3% D6 +6.3% D7 +7.8% D8 +9.0% D9 +16.6% D10 (smallest)

Firm size by decile (2026 market capitalization)

What "largest" and "smallest" actually mean in dollar terms — each company's market cap at the start of 2026 (its Dec-31-2025 snapshot, the same figure that determined that year's decile assignment). Note the huge range within D1 alone: it spans from the smallest mega-cap in the bucket to the very largest company in the whole universe.

Decile Min Median Max
D1 (largest) $52.50B $100.74B $1.58T
D2 $20.43B $31.80B $52.28B
D3 $10.27B $13.87B $20.23B
D4 $6.01B $7.80B $10.26B
D5 $3.66B $4.76B $6.01B
D6 $2.14B $2.80B $3.66B
D7 $1.22B $1.62B $2.14B
D8 $603.2M $837.1M $1.21B
D9 $295.1M $411.2M $596.9M
D10 (smallest) $12.4M $155.2M $294.7M

Full detail

Decile Size premium Actual return Beta Companies Rolling-window history
D1 (largest) +1.3% 11.4% 1.12 179 9 window-end-years (2018–2026)
D2 +1.6% 12.0% 1.17 178 8 window-end-years (2019–2026)
D3 +0.8% 10.4% 1.11 178 8 window-end-years (2019–2026)
D4 +3.3% 12.7% 1.12 178 8 window-end-years (2019–2026)
D5 +3.6% 12.9% 1.13 178 7 window-end-years (2020–2026)
D6 +5.3% 15.0% 1.24 178 5 window-end-years (2022–2026)
D7 +6.3% 17.3% 1.35 178 5 window-end-years (2022–2026)
D8 +7.8% 17.2% 1.24 178 5 window-end-years (2022–2026)
D9 +9.0% 17.3% 1.02 178 5 window-end-years (2022–2026)
D10 (smallest) +16.6% 23.8% 0.92 178 4 window-end-years (2023–2026)

Methodology

  • Universe: 1,880 NYSE-listed common stocks (NASDAQ Trader's official symbol directory, exchange code N, ETFs/preferred/warrants/units filtered out).
  • Decile construction: every company is assigned a decile each year by its prior Dec-31 market cap (NTILE(10), D1 = largest, D10 = smallest). Membership is annual; monthly returns within a year use that year's decile assignment.
  • Benchmark: NYSE Composite Index (^NYA) as the market proxy. Risk-free rate: US 10-Year Treasury yield (FRED series GS10, monthly).
  • Size premium = CAPM alpha for a decile: actual return − [Rf + β×(Rm−Rf)], using that decile's own expanding-window, real-time beta — beta for year Y is estimated only from months strictly before January of Y, never look-ahead.
  • Rolling 10-year figure = trailing 120-month mean of monthly alpha, annualized ×12 (linearly, not geometrically). The window shown here is 2016-08 through 2026-07 for every decile — the most recent fully-populated 120-month window in the dataset, not a hand-picked period.
  • Weighting: equal-weighted (simple average across companies in a decile each month) — the version shown here, and, per a robustness check described below, the one that held up best.

Limitations

  • Survivorship bias. The 1,880-company universe is the current NYSE listing — companies that delisted, were acquired, or went bankrupt over 2006–2026 are not included. This likely biases returns upward, especially for the smallest, highest-risk decile.
  • Shorter effective history than it looks. Price data spans a full 20 years (2006–2026), but Yahoo Finance's shares-outstanding history for US-listed companies is much shallower — usable market-cap coverage across the universe only becomes robust from 2016 onward. The rolling 10-year window above is the earliest one that could be fully populated.
  • Uneven confidence across deciles. The 2016–2026 window itself is identical for every decile, but how many independent prior rolling-window readings back it up is not: D1 has 9 window-end-years of history (back to 2018), D10 only 4 (back to 2023). Treat the smallest deciles' premiums as resting on a much shorter track record.
  • Beta is per-decile, not per-company. A company's alpha is computed against its decile's typical systematic risk, not its own — the same approximation the SGX study uses.
  • Checked and passed: a liquidity filter (require nonzero trading volume on >50% of a month's trading days before that company-month counts) changes every decile's premium by less than 0.04 percentage points — this result is not an illiquid-stock artifact.
  • Checked and failed to help: value-weighting (by prior month-end market cap, instead of equal-weighting) was tried specifically to smooth the one interior dip in the ladder (D2→D3) — it fixed that dip but introduced two new ones (D4→D5, D8→D9), making the ladder less monotonic overall. Equal-weighting (shown above) is the more defensible construction here.
  • Not yet built: no cost-of-debt side for NYSE, no alternate-benchmark check, no continuous cross-sectional regression (which would be monotonic by construction, unlike a 10-bucket sort).
  • This page is a static snapshot from a specific analysis run — it does not update automatically and is not connected to a live database.