{
  "name": "Number-superstition replication scorecard",
  "description": "Every published price or market effect attributed to number superstition, with the baseline each effect is measured against and whether it has been independently reproduced. Compiled from the primary papers; sixteen of the studies were read first-hand and the rest are marked as reported at one remove.",
  "url": "https://numeluna.com/economics-number-superstition/#scorecard",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "creator": "Numeluna",
  "temporalCoverage": "1949-01-01/2009-12-31",
  "updated": "2026-08-11",
  "fields": {
    "number": "The digit or date the effect is attributed to.",
    "market": "Market and period studied.",
    "sample": "Sample size as the paper states it.",
    "effect": "The effect in the authors' own units and wording.",
    "coefficient": "The raw table value, where the published figure is a transformed one. The Hong Kong percentages are log coefficients ×100, not exact price effects; convert with exp(beta)-1 if you need the proportional figure.",
    "plate": "Hong Kong digit rows only: the same two coefficients as numbers ({four, three}), so a consumer does not have to parse them back out of the `effect` string. Absent on the digit 7, which is the omitted category and has no coefficient by definition.",
    "baseline": "What the effect is measured AGAINST. Never empty — a figure without its baseline is not interpretable.",
    "replication": "replicated | mixed | failed | untested | unreported. 'failed' means a test that did not reproduce the claimed effect: read it as 'not established', never as 'shown to be zero'.",
    "provenance": "first-hand (the paper was downloaded and its tables read directly) or one-remove (reported through a survey of the literature).",
    "note": "The condition a citer needs in order not to misuse the row."
  },
  "rows": [
    {
      "number": "8",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+63.5% (4-digit) · +94.8% (3-digit)",
      "coefficient": "β = 0.635 / 0.948 on ln(price)",
      "plate": {
        "four": 63.5,
        "three": 94.8
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "replicated",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "note": "The published percentages are log coefficients ×100, not exact price effects. Convert with exp(β)−1 if you need the proportional figure."
    },
    {
      "number": "1",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+33.2% (4-digit) · +46.9% (3-digit)",
      "coefficient": "β = 0.332 / 0.469 on ln(price)",
      "plate": {
        "four": 33.2,
        "three": 46.9
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "3",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+32.5% (4-digit) · +55.4% (3-digit)",
      "coefficient": "β = 0.325 / 0.554 on ln(price)",
      "plate": {
        "four": 32.5,
        "three": 55.4
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "2",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+32.1% (4-digit) · +42.7% (3-digit)",
      "coefficient": "β = 0.321 / 0.427 on ln(price)",
      "plate": {
        "four": 32.1,
        "three": 42.7
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "6",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+29.2% (4-digit) · +36.5% (3-digit)",
      "coefficient": "β = 0.292 / 0.365 on ln(price)",
      "plate": {
        "four": 29.2,
        "three": 36.5
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "9",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+21.5% (4-digit) · +29.4% (3-digit)",
      "coefficient": "β = 0.215 / 0.294 on ln(price)",
      "plate": {
        "four": 21.5,
        "three": 29.4
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "0",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+12.4% (4-digit) · +30.0% (3-digit)",
      "coefficient": "β = 0.124 / 0.300 on ln(price)",
      "plate": {
        "four": 12.4,
        "three": 30
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "compact": true
    },
    {
      "number": "7",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "— (the baseline itself)",
      "baseline": "n/a — 7 is the omitted category every other digit is measured against",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "note": "On the chart deliberately. Seven is not \"no effect\" — it is the reference point, and every Hong Kong percentage here is relative to it. This row is why the others are interpretable."
    },
    {
      "number": "4",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "−11% (4-digit) · −27.3% (3-digit)",
      "coefficient": "β = −0.110 / −0.273 on ln(price)",
      "plate": {
        "four": -11,
        "three": -27.3
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "mixed",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "note": "Reproduced in Vancouver (−2.2%) but NOT significant in Auckland. The lucky-8 premium replicates everywhere it has been looked for; the unlucky-4 discount does not."
    },
    {
      "number": "5",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "−12.5% (4-digit) · −19.5% (3-digit)",
      "coefficient": "β = −0.125 / −0.195 on ln(price)",
      "plate": {
        "four": -12.5,
        "three": -19.5
      },
      "baseline": "the digit 7 (omitted category)",
      "replication": "unreported",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "note": "Discounted about as heavily as four, significant in every column the authors ran — and the word \"five\" appears nowhere in their paper."
    },
    {
      "number": "13",
      "market": "Hong Kong licence plates, 1997–2009",
      "sample": "41,069 plates, 292 auctions",
      "effect": "+22.1% (3-digit ordinary) · +70.6% (4-digit special)",
      "coefficient": "β = +0.221 / +0.706 on ln(price)",
      "baseline": "plates without the 13 pattern",
      "replication": "untested",
      "source": "Ng, Chong & Du 2010",
      "provenance": "first-hand",
      "note": "A PREMIUM — the same numeral priced unlucky in the West. Not uniform: it is −0.337 on 3-digit specials. The authors report the coefficient without interpreting it."
    },
    {
      "number": "8",
      "market": "Greater Vancouver housing, 2000–2005",
      "sample": "~117,000 sales",
      "effect": "+2.5%",
      "baseline": "addresses ending in any other digit",
      "replication": "replicated",
      "source": "Fortin, Hill & Huang 2014",
      "provenance": "first-hand",
      "note": "ONLY in census tracts where ethnic Chinese residents exceed the Greater Vancouver average of 18%. There is no market-wide effect."
    },
    {
      "number": "4",
      "market": "Greater Vancouver housing, 2000–2005",
      "sample": "~117,000 sales",
      "effect": "−2.2%",
      "baseline": "addresses ending in any other digit",
      "replication": "mixed",
      "source": "Fortin, Hill & Huang 2014",
      "provenance": "first-hand",
      "note": "Same 18% conditioning as the 8 row. Agrees with Hong Kong in sign; does not reproduce in Auckland."
    },
    {
      "number": "13",
      "market": "Greater Vancouver housing, 2000–2005",
      "sample": "~117,000 sales",
      "effect": "−2.8%",
      "baseline": "comparable properties",
      "replication": "failed",
      "source": "Fortin, Hill & Huang 2014",
      "provenance": "one-remove",
      "note": "Reported via Larsen's summary of the paper, not read in the original. Larsen then retested 13 in Ohio and found nothing."
    },
    {
      "number": "3, 6, 8, 9",
      "market": "Auckland housing, 1989–1996",
      "sample": "2,164 sales (+2,369 comparison)",
      "effect": "+2.4% (≈ NZ$7,100)",
      "coefficient": "β = 0.0119 [SE 0.0064], p = 0.063",
      "baseline": "all other last digits, mean price NZ$293,200",
      "replication": "replicated",
      "source": "Bourassa & Peng 1999",
      "provenance": "first-hand",
      "note": "The oldest study of the effect. A clean placebo: the premium is absent in a comparable nearby area with few Chinese residents."
    },
    {
      "number": "4",
      "market": "Auckland housing, 1989–1996",
      "sample": "2,164 sales",
      "effect": "no significant effect",
      "coefficient": "β = −0.0171 [SE 0.0114], p = 0.134",
      "baseline": "all other last digits",
      "replication": "failed",
      "source": "Bourassa & Peng 1999",
      "provenance": "first-hand",
      "note": "The most-dropped fact in this literature. ⚠️ Underpowered — the interval is consistent with a discount larger than the premium the same paper reports, so this is 'not established', NOT 'shown to be zero'."
    },
    {
      "number": "13",
      "market": "Montgomery County, Ohio housing",
      "sample": "3,701 sales (+80,000 in robustness)",
      "effect": "no effect",
      "coefficient": "β = 0.00785 [SE 0.08784], t = 0.09, p = 0.929",
      "baseline": "all other house numbers",
      "replication": "failed",
      "source": "Larsen 2015",
      "provenance": "first-hand",
      "note": "Designed for the question and adequately powered — the paper's headline result, not an incidental null."
    },
    {
      "number": "7, 25, 27, 34, 40, 54, 59",
      "market": "Montgomery County, Ohio housing",
      "sample": "98 re-estimations, 3,701 sales",
      "effect": "7 of 98 significant; 91 null",
      "baseline": "all other house numbers",
      "replication": "failed",
      "source": "Larsen 2015",
      "provenance": "first-hand",
      "note": "The author's own verdict is coincidence — about the crop you would expect from running the same test a hundred times. On the chart as the multiple-comparisons control, not as findings."
    },
    {
      "number": "Friday the 13th",
      "market": "US equities (CRSP), 1962–1985",
      "sample": "24 years of index returns",
      "effect": "returns below other Fridays",
      "baseline": "all other Fridays",
      "replication": "failed",
      "source": "Kolb & Rodriguez 1987",
      "provenance": "first-hand",
      "note": "Found 1987, not reproduced since. Famous ever since, and the single most-propagated claim in this literature."
    },
    {
      "number": "Friday the 13th",
      "market": "19 national indices",
      "sample": "19 countries",
      "effect": "returns HIGHER, in 11 of 19",
      "baseline": "all other Fridays",
      "replication": "untested",
      "source": "Lucey 2000",
      "provenance": "first-hand",
      "note": "No support for the negative effect anywhere, and the opposite sign in a majority. Unexplained. ⚠️ 19 tests at a 10% threshold yield ~2 false positives by chance; 11 is not that."
    },
    {
      "number": "4",
      "market": "US mortality, 1973–1998",
      "sample": "US death certificates",
      "effect": "cardiac deaths peak on the 4th (ratio 1.13, 95% CI 1.06–1.21)",
      "baseline": "the average of surrounding days",
      "replication": "failed",
      "source": "Phillips et al. 2001",
      "provenance": "first-hand",
      "note": "The Hound of the Baskervilles effect. Failed twice — see the two rows below."
    },
    {
      "number": "4",
      "market": "US mortality, 1969–1988 and 1999–2001",
      "sample": "same ICD category, wider windows",
      "effect": "nothing out of the ordinary",
      "baseline": "days 3 and 5, and the omitted heart-disease codes",
      "replication": "failed",
      "source": "Smith 2002",
      "provenance": "first-hand",
      "note": "Phillips used 5 of the 36 available heart-disease codes, having used all of them in earlier work. The omitted categories show no day-4 excess."
    },
    {
      "number": "4, 14, 24",
      "market": "Hong Kong mortality, 1995–2000",
      "sample": "17,346 cardiac deaths",
      "effect": "no difference from any other day",
      "coefficient": "P = 0.88 (Gregorian) · P = 0.46 (lunar)",
      "baseline": "all other days, on both calendars",
      "replication": "failed",
      "source": "Panesar et al. 2003",
      "provenance": "first-hand",
      "note": "The same five ICD codes, in a >95% Cantonese-speaking population — the effect is missing from the population that holds the belief hardest."
    },
    {
      "number": "lucky listing codes",
      "market": "Chinese IPOs (Shanghai, Shenzhen), 1991–2005",
      "sample": "not stated in the survey",
      "effect": "initial premium, gone within three years",
      "baseline": "non-lucky listing codes",
      "replication": "untested",
      "source": "Hirshleifer et al. 2018, via Borowski 2019",
      "provenance": "one-remove",
      "note": "Read through a survey of the literature, not in the original. On the chart because a premium that decays is what a coordination effect looks like."
    },
    {
      "number": "4",
      "market": "Tokyo Stock Exchange, 1949–2001",
      "sample": "52 years of returns",
      "effect": "returns low on the 4th, fading from the mid-1980s",
      "baseline": "all other days of the month",
      "replication": "untested",
      "source": "Pinto 2015, via Borowski 2019",
      "provenance": "one-remove",
      "note": "Read through a survey, not in the original. The fade is attributed to internationalisation — foreign capital arriving with no opinion about the fourth."
    }
  ]
}