# Case study SMARTBETA-1: data-licence decision

**Recorded:** 2026-09-26 (UTC), before the preregistration was sealed and before any return series (ETF prices, index levels or factor returns) was loaded by this study.
**Question:** which sources may a commercial website (holdoutlabs.io) use to publish derived statistics about the "before" (backtest-era factor premia) and "after" (live ETF returns) periods?
**Rule applied:** a source is used only if its terms permit publishing *derived statistics* on a commercial site. Raw series are never republished. Where no permissive source exists, the item is dropped rather than published under doubtful terms.

## Evidence

| Source | What its terms say (fetched 2026-09-26 unless noted) | Decision |
|---|---|---|
| **MSCI index levels** (MSCI USA Momentum / Sector Neutral Quality / Minimum Volatility / Enhanced Value / Low Size, and their predecessors) | MSCI index data are licensed products; the MSCI index pages we tried returned 404/403 to our fetcher, and MSCI's standard terms prohibit redistribution of index data without a licence. | **Not used.** No MSCI index level or return series is loaded. We do not reconstruct the pre-launch backtest. |
| **S&P Dow Jones Indices** (S&P 500 Low Volatility Index) | spglobal.com returned HTTP 403 to our fetcher. S&P DJI index data are licensed; redistribution needs a licence. | **Not used.** No S&P index series is loaded. |
| **Issuer and index-provider documents** (iShares product pages; SEC filings on EDGAR; MSCI Research Insight "Foundations of Factor Investing", Dec 2013; an HSBC structured-note free-writing prospectus quoting S&P DJI marketing text, SEC EDGAR, 2012-02-03) | Public documents. | **Used only for facts and short quotations** (fund names, inception dates, index names, index-change dates, and one- or two-sentence claims), each with a citation. **No backtest performance figure from these documents is reproduced** in the preregistration or the results. |
| **AQR data sets** (e.g. Betting-Against-Beta factors) | AQR Terms of Use: "You may not reproduce, modify, copy, alter in any way, distribute, sell, resell, transmit, transfer, license, assign or publish any information obtained from this Website without AQR's express prior written consent." | **Not used.** BAB is therefore not the low-volatility proxy. |
| **Kenneth R. French Data Library** (Fama/French 3 and 5 factors, Momentum factor, univariate sorts on variance, book-to-market, operating profitability, prior 12-2 return) | The page carries "Copyright Eugene F. Fama and Kenneth R. French" and no licence or use restriction; the library is published for free public download and is cited in thousands of academic and practitioner publications. | **Used, derived statistics only** (annualised means, t-statistics, volatilities, drawdowns), with citation. Raw files are kept out of the public repo (git-ignored) and are not republished. **Flag for the lawyer hour**: confirm that publishing derived statistics from the library on a commercial site is acceptable. |
| **Yahoo Finance public chart endpoint** (daily adjusted close for MTUM, QUAL, VLUE, SIZE, USMV, SPLV, SPY) | Same basis as every published Holdout Labs batch (`samples/README.md`, `samples/batch3/README.md`): derived statistics only, never raw prices, no redistribution. | **Used, derived statistics only.** |
| **Issuer-reported fund total returns** (iShares / Invesco fact sheets) | Issuer pages. | **Internal data-quality cross-check only** (after the seal): the published output is limited to "maximum absolute difference between our Yahoo-derived calendar-year total return and the issuer's", never the issuer's figures themselves. |

## Consequences for the design

1. The "before" evidence is the **academic factor premium** measured on French's free data, not the MSCI or S&P backtest itself. It is a proxy: French factors are long-short (or long-only quintile) portfolios over all US stocks, while the ETFs are long-only tilts on large and mid caps. All "before vs after" comparisons are therefore **descriptive only**.
2. The launch-era **claims** are quoted from issuer and index-provider documents in one or two sentences with citations; their backtest numbers are not reproduced.
3. The "after" test uses ETF total returns (Yahoo adjusted close), the investable product as sold, which is what investors actually received.

**Terms to confirm (owner / lawyer hour):** French Data Library commercial derived-statistics use; Yahoo chart-endpoint use (already on the list from earlier batches); short quotations from MSCI and S&P DJI marketing text under fair use.