Home › Articles

S&P 500 / SPY: Official Exposure, Valuation Definitions and Empirical Model Research

2026-10-30 · Research note · Series: US index institutional research (part 2 of 2) · Not investment advice; figures are as of the sources and dates cited in the text · 中文
Contents12 sections
  1. Executive summary
  2. Data governance and research timing
    1. Evidence grading
    2. Date and unit matrix
  3. Official definitions of the index and the ETF
  4. Official versus vendor valuation definitions
  5. Exposure and concentration
    1. Official holdings
    2. Sector exposure
  6. Historical total-return performance and risk
    1. Official total-return NAV performance, net of expenses
    2. Verifiable risks and unverifiable gaps
  7. Website model rebuild and out-of-sample diagnostics
    1. In sample: recomputable is not the same as predictive
    2. Purged expanding-window pseudo-OOS
    3. Intervals and the combination model
  8. When the valuation judgment holds and when it fails
    1. Limited judgments that can hold
    2. Conditions that would invalidate the judgment or require reassessment
  9. Monitoring indicators and follow-up evidence
  10. Source register and acceptance
    1. Sources and access limitations
    2. Acceptance for this report
  11. Further reading
  12. Frequently asked questions

Research date / source access date: 2026-10-30 (Taipei time). This note uses the evidence and risk-disclosure format of a research report; it is not a formal rating by a licensed institution, a price target or personalized investment advice. Not all official data dates are 2026-10-29; the actual date is kept for each table below.

About this page: This is one of 2 parts in the "US Index Institutional Research" series (this part: S&P 500 / SPY; the other part covers Nasdaq-100 / QQQ). It is the full text of a research note, republished as written: content, numbers, tables, citations, links and data-gap / limitation paragraphs are reproduced without cuts or rewrites. Only the table of contents at the top of the original has been removed, and links to attachments have been turned into plain text. The original file is SP500 website data analysis (gpt).md (translated name) in the author's research folder (US index research, translated name); research date 2026-10-30 (taken from the original). The data / scripts attachments cited in the text (CSV, JSON, Python) are in the author's research folder and are not on this website. The original itself states that it uses the evidence and risk-disclosure format of a research report and is not a formal rating by a licensed institution, a price target or personalized investment advice. This page is a research record, not investment advice.

Executive summary

  1. Core large-cap exposure is not the same as even diversification. S&P DJI positions the S&P 500 as the US large-cap benchmark, covering about 80% of available US market capitalization; State Street confirms it is weighted by float-adjusted market capitalization. In SPY's official holdings as of 2026-10-28, the top three securities total 21.24% and the top ten securities 38.03%; the single Information Technology sector is 38.62%. These sums are computed from the officially displayed weights and are not 2026-10-29 estimates. O1, O2
  2. Different P/E figures cannot be ranked into cheap / expensive. State Street on 2026-10-28 lists SPY FY1 P/E at 20.48x, explicitly using one-year forward earnings estimates and a weighted harmonic mean; the index Price/Earnings on the same page is 24.48x, but that field's full denominator and treatment of losses are not verified. WorldPeRatio on 2026-10-29 lists 24.06x, with its denominator / aggregation not yet independently verified. The three are not market prices proven to share one definition. O1, V1
  3. Official total-return NAV history is strong, but it is not a guarantee of the future. As of 2026-09-30, SPY's pre-tax ten-year annualized return, net of fund expenses with dividends and capital gains reinvested, is 15.18%; the unmanaged index in the same table is 15.33%. The past one-year NAV return is 15.60%. None of these are price returns, and even less are they the website model's ten-year forecast. O1
  4. The website's −1.31% is only a reproducible output. The coordinator rebuilt the seven models and the combination table from this vendor snapshot, and 44 table fields passed a display-precision check; but the raw P/E history has no point-in-time version evidence. The P/E→ten-year model obtained R²_OOS 0.7927 in a purged pseudo-OOS with a minimum of 60 training points, with only 60 heavily overlapping monthly test start points; with a minimum of 120 points there are no test points. This cannot be used to claim that −1.31% has been fully validated out of sample. Model audit appendix (the author's model audit appendix, not published on this site)
  5. The conclusion is "valuation discipline and concentration risk need managing", not a certain decline. The one-year model's pseudo-OOS actually trails the historical matured-mean benchmark; the five-year model is sensitive to the training threshold. Official maximum drawdown, volatility and a same-definition official valuation for 2026-10-29 were not obtained; this note keeps the gaps and does not fill in guessed values.

Data governance and research timing

Evidence grading

Category Evidence and use Must not be extended to
Directly verified official State Street product HTML, S&P DJI official HTML; headers, values and definitions checked in a browser An official guarantee about the future; all fields being of the same date
Third-party estimates carried by an official source FactSet Estimates earnings-growth estimates listed by State Street Realized earnings, an official commitment or this note's own forecast
Saved vendor data The coordinator saved the WorldPeRatio snapshot, values and manifest hashes S&P DJI-certified data or historical versions that were available at the time
Computed in this study Sums of official weights, performance differences; the coordinator's OLS and purged pseudo-OOS A tradable strategy, independent return cases, actual probabilities
Unknown / not verifiable Full historical revisions, same-definition maximum drawdown, some valuation denominator specifications Arbitrary fill-ins, market numbers substituted from common knowledge

Date and unit matrix

The access date is 2026-10-30 throughout. The source code in each later table links to the URL at the end; the data date applies to all observations in that table unless noted otherwise. USD means US dollars; M means million; returns and weights are percentages, and differences are marked separately as "percentage points".

Content Actual as-of Source / verification method
SPY expense ratio, fund and listing information 2026-10-30 page information section O1; HTML and browser headers
NAV, AUM, holdings, sectors, valuation 2026-10-28 O1; HTML and browser headers / tooltips
Official fund / index historical performance 2026-09-30 O1; column order checked in a browser
General index description The page does not mark an effective date for the description; access date 2026-10-30 O2; not real-time market values
Website current P/E and models Valuation date in the page header 2026-10-29; model samples listed separately V1; the coordinator's snapshot and CSV from this run

State Street's simplified HTML text loses column names, making the performance columns appear misaligned; this study read the full headers in a browser and confirmed that "As Of" comes before the return columns before recording performance. A direct fetch of S&P DJI returned 403, but the browser successfully read the description and the methodology document link; search snippets cannot be treated as having read the full methodology. No PDF parsing package was installed, and official PDFs were not saved or reproduced in bulk.

Machine-readable official facts, dates, units, definitions and limitations are in official_sources.json (data/SP500/official_sources.json). Vendor data and program hashes are in manifest.json (data/SP500/manifest.json); the two kinds of sources are stored separately, and official numbers are not used to overwrite website inputs.

Official definitions of the index and the ETF

Official facts, O1 / O2, accessed 2026-10-30:

  • S&P DJI states that the S&P 500 includes 500 leading companies and covers about 80% of available market capitalization; this is an index design description, not a check of exact daily coverage.
  • State Street states that the index is float-adjusted market-capitalization weighted and covers all eleven GICS sectors. It is not equal-weighted, nor is it the complete US equity market.
  • SPY seeks to track, before expenses, roughly the price and yield performance of the index; the index itself cannot be invested in directly.
  • SPY's inception / listing date is 1993-01-22, ticker SPY, trading currency USD; the page lists NYSE ARCA, with quarterly distributions; the information section is as of 2026-10-30.
  • Gross Expense Ratio 0.0945% / year, i.e. 9.45 basis points / year (unit conversion); this is the total annual operating expense ratio before waivers or expense reimbursements, and excludes an individual investor's trading commissions, spreads or taxes.
  • 2026-10-28 NAV USD 771.86 / share; AUM USD 814,718.63 million. These only describe fund size / net asset value, not the closing trade price on the report date.

Methodology boundary: O2 lists the S&P U.S. Indices Methodology, but the PDF request in this study returned 403, so the version, effective date, inclusion thresholds, committee discretion and maintenance rules were not verified item by item; this note does not cite unverified exact thresholds, nor does it claim that the S&P 500 is just a mechanical selection of the top 500 by market capitalization.

Official versus vendor valuation definitions

Indicator Value / unit As-of / source Verified definition and what is still missing
SPY fund FY1 P/E 20.48x 2026-10-28, O1 Weighted harmonic mean of each holding's price / one-year forward EPS estimate; the tooltip states that positive and negative extreme values are included in the calculation
Index FY1 P/E 20.48x 2026-10-28, O1 index section The field with the same FY1 name; not equal to an index operating-earnings or GAAP TTM P/E
Index Price/Earnings 24.48x 2026-10-28, O1 index section The field exists and the value was read; the field's full denominator, aggregation and handling of negative earnings were not independently verified, so it is not simply called TTM
SPY fund P/B 5.33x 2026-10-28, O1 The tooltip lists a weighted harmonic mean of each security's closing price / most recently reported book value, and mentions a trailing-twelve-month calculation; not treated as a guaranteed market value of corporate net assets
Website estimated P/E 24.06x 2026-10-29, V1 ETF proxy data; whether the denominator is TTM, FY1 or another estimate, and how it is aggregated, was not independently verified
Holdings' 3–5 year EPS growth estimate 21.73% / year growth-rate estimate 2026-10-28, O1 The official page carries the average of FactSet Estimates for each holding; not realized, not this note's forecast, and not used to generate a price target

20.48, 24.48 and 24.06 come from different fields / times; even though the numbers are close, they cannot be treated as reconciled on the same definition. This note does not compute an "official discount" between them, and does not use them to decide whether the Nasdaq or the S&P is cheaper. No official same-definition long-history P/E distribution was obtained, so the website's Cheap / Fair / Expensive labels cannot be promoted to an official rating.

Vendor historical comparison, V1, valuation date 2026-10-29: the ten-year average of 20.54x and the twenty-year average of 16.80x in the old report and in this snapshot are statistics processed by the website; 24.06 is 17.14% and 43.21% above the two averages respectively (formula used in this study: 24.06 / average −1). The website discloses truncation and normalization of extreme P/Es; the full processing order and historical sources have not yet been independently rebuilt. This only supports "high relative to that vendor's long-term benchmark"; it does not prove that the mean is unchanged across regimes, nor is it equivalent to an expected decline.

Exposure and concentration

Official holdings

O1, SPY Fund Top Holdings, as of 2026-10-28, accessed 2026-10-30; weights are percentages of the fund. Only the top ten weights needed for the research are excerpted; the full holdings file is not reproduced.

Security Weight
NVIDIA 7.81%
Apple 7.52%
Microsoft 5.91%
Amazon 3.74%
Alphabet A 2.86%
Broadcom 2.51%
Alphabet C 2.29%
Meta Platforms A 2.02%
Tesla 1.71%
Micron Technology 1.66%

Computed in this study: the top three total 21.24%; the top ten securities total 38.03%; Alphabet's two share classes total 5.15%. Ten securities are not ten independent issuers; dual share classes are also not two fully independent economic risks. Weights are officially rounded, so sums may have small errors. The fund's holdings count is 504 and the index's on the same page 503 (O1, 2026-10-28), both taken from the page's holdings field and not interpreted as 504 / 503 companies.

Sector exposure

O1, SPY Fund Sector Breakdown, as of 2026-10-28, accessed 2026-10-30; weights are percentages of the fund.

GICS sector Weight
Information Technology 38.62%
Financials 11.83%
Health Care 9.75%
Consumer Discretionary 8.97%
Communication Services 8.92%
Industrials 8.02%
Consumer Staples 4.75%
Energy 3.74%
Utilities 2.00%
Real Estate 1.72%
Materials 1.68%

Computed in this study and interpretation: the displayed weights sum to 100.00%; Information Technology + Communication Services + Consumer Discretionary total 56.51%. This is the sum of three GICS classifications, not an officially defined "tech stocks" or AI exposure. Float-adjusted market-cap weighting gives large companies a larger contribution to the index; holdings in all eleven sectors do not mean that return sources are even.

Full revenue geography, interest-rate duration, factor or derivatives exposure data for the holdings were not obtained, so US-listed large caps are not equated with all revenue coming from the US, and sensitivity to AI, the dollar or overseas revenue is not quantified. The weighted overlap with QQQ was also not computed; that requires complete holdings of both funds on the same date and an exact value cannot be inferred from the shared large stocks alone.

Historical total-return performance and risk

Official total-return NAV performance, net of expenses

O1, as of 2026-09-30, accessed 2026-10-30; units are %, periods longer than one year are annualized, periods of one year or less are cumulative returns for the period. NAV is pre-tax and net of fund expenses, assuming dividends and capital gains are reinvested; the index is not net of fund expenses. These are not investors' actual after-tax or after-trading-cost returns.

Period SPY NAV Officially listed S&P 500 benchmark
One month −0.35% −0.35%
Quarter to date 2.28% 2.30%
Year to date 12.65% 12.75%
One year 15.60% 15.74%
Three-year annualized 22.73% 22.89%
Five-year annualized 13.65% 13.79%
Ten-year annualized 15.18% 15.33%
Annualized since inception 1993-01-22 10.82% 10.95%

Computed in this study: the ten-year annualized NAV minus benchmark difference is −0.15 percentage points / year; this is the subtraction of two annualized returns, not daily tracking error, and it cannot be attributed entirely to the 0.0945% expense ratio. NAV, market price and the unmanaged index can differ because of cash, expenses and valuation timing.

State Street's Market Value performance is calculated at the midpoint of the bid and ask quotes at the time NAV is calculated, and does not guarantee that an individual can trade at that price. This note does not mix NAV performance with price paths that exclude dividends; nor does it force a reconciliation between the website's past-ten-year input of 15.2339% and the official 15.18%, because their dates and definitions are not yet consistent.

Verifiable risks and unverifiable gaps

  • Official risk disclosure (O1, accessed 2026-10-30, the page does not mark an effective date for the disclosure): there is a risk of loss of principal; the ETF price may be above or below NAV; commissions and ETF expenses reduce returns, and past performance does not guarantee future results.
  • The official 2026-09-30 one-month −0.35% provides a direct example of a negative-return month; it is not a maximum drawdown, a stress-event decline or a loss ceiling.
  • Official maximum drawdown, annualized volatility, worst rolling periods and historical recovery periods: no verifiable same-definition series / statistics were obtained in this study. Remembered financial-crisis or pandemic declines are not filled in, and price-index drawdowns are not passed off as SPY total-return NAV drawdowns.
  • The model RSE in this study is a scale of conditional return residuals, not fund volatility, and even less a maximum drawdown. R² is not a probability of decline.
  • Investors measuring in currencies other than the US dollar bear additional exchange-rate risk; this note has no exchange-rate data and does not estimate home-currency returns.

The official factsheet PDF can be linked, but the current tools only returned the PDF binary and did not parse it into checkable text; so this note does not claim to have obtained calendar-year returns, drawdowns or risk tables from the PDF.

Website model rebuild and out-of-sample diagnostics

This chapter cites the model coefficients and outputs (data/SP500/model_metrics.csv), the purged test summary (data/SP500/purged_oos.csv), the combination diagnostics (data/SP500/combination_diagnostics.csv) and the shared model audit (the author's model audit appendix, not published on this site) produced by the coordinator in this run. The calculations are based on public data from V1 obtained on 2026-10-30, with a page-header valuation date of 2026-10-29; the numbers below are computed in this study, not official S&P / State Street forecasts.

In sample: recomputable is not the same as predictive

Model Paired sample size Buy start-point label range Current output: annualized % In-sample R² Extrapolation status
P/E→one year 348 1996-10 to 2025-09 9.4121 0.0033 x within sample range
P/E→five years 300 1996-10 to 2021-09 1.6825 0.1918 x within sample range
P/E→ten years 240 1996-10 to 2016-09 −1.3120 0.7353 x within the sample's upper edge
P/E→twenty years 120 1996-10 to 2006-09 5.8028 0.6732 x within the sample's upper edge
Past one year→next one year 348 1996-10 to 2025-09 11.2421 0.0031 x within sample range
Past five years→next five years 284 1998-02 to 2021-09 7.6833 0.0704 x within sample range
Past ten years→next ten years 164 2003-02 to 2016-09 6.6217 0.3868 Extrapolation

The ten-year P/E model's training x upper bound is 24.1522x, and the current 24.06x is still within range; but being within range cannot be treated as adequately representing the current environment. The past-ten-year return model's current input of 15.2339% annualized exceeds its training upper bound of 11.9755% annualized. The buy start points are vendor data labels, not trading dates independently verified in this note.

The seven single models plus the combination table, 44 fields in all, match the page's display precision, which proves arithmetic consistency. The P/E→ten-year −1.3120% is the conditional output of an OLS straight line at the current vendor P/E; it is not a certain ten-year loss, nor a verified official capital-market assumption.

Purged expanding-window pseudo-OOS

Research design parameters, not market data: test start point t may only use matured labels that satisfy training start + forecast months < t; training expands month by month. The mean of the then-matured training Y is used as the benchmark, with minimum training thresholds of 60 / 120 points. The data is the currently rebuilt version and there is no data version available at the time (point-in-time), so it is called pseudo-OOS, not true real-time OOS.

Minimum 60 points; units: RMSE, MAE and benchmark RMSE are all in percentage points of annualized return. R²_OOS = 1 − sum of the model's squared test errors / sum of the benchmark's squared errors; it is not an ordinary training R², nor a win rate.

Model Test n / start-point period RMSE MAE Benchmark RMSE R²_OOS
P/E→one year 276; 2002-10 to 2025-09 18.2591 13.8437 15.9976 −0.3027
P/E→five years 180; 2006-10 to 2021-09 8.9027 8.3240 9.6256 0.1446
P/E→ten years 60; 2011-10 to 2016-09 4.6006 3.8447 10.1055 0.7927
Past one year→one year 276; 2002-10 to 2025-09 18.2012 13.3697 15.9976 −0.2945
Past five years→five years 164; 2008-02 to 2021-09 10.1867 9.7841 10.2308 0.0086

Sensitivity with a minimum of 120 points:

Model Test n / start-point period RMSE: percentage points Benchmark RMSE: percentage points R²_OOS
P/E→one year 216; 2007-10 to 2025-09 18.9496 16.9797 −0.2455
P/E→five years 120; 2011-10 to 2021-09 9.6884 9.2726 −0.0917
Past one year→one year 216; 2007-10 to 2025-09 17.8148 16.9797 −0.1008
Past five years→five years 104; 2013-02 to 2021-09 10.5316 8.7098 −0.4621

Untestable range: P/E→ten years has no test points at the 120-point threshold; P/E→twenty years and past ten years→next ten years have no testable points at either threshold, because under the strict maturity condition the usable training history is insufficient, not because the models tested with zero error.

Empirical reading: the one-year models trail the matured-mean benchmark under both designs; the five-year improvement disappears at the higher threshold, showing sensitivity to the sample window / research design. The ten-year 60-point result beats the benchmark, but the ten-year holding periods of 60 monthly investment start points overlap heavily, and they are not 60 independent ten-year periods. There are no pre-revision historical inputs, and no results from genuinely future start points from 2026 on, so it cannot be claimed that the current −1.31% has been fully validated, and even less can 0.7927 be read as 79.27% accuracy.

Intervals and the combination model

  • The website's seven models use 1.645 but label it 80%; a central two-sided normal 80% should be about 1.282, and 1.645 corresponds to about 90%. This is a mathematical check of the label, not a re-certification of the website's interval as 90%; sample overlap, error distribution and regime change remain unresolved. NIST
  • The range reproduced with 1.645 for the current ten-year P/E model is about −5.66% to 3.03% annualized; it does not have an actual coverage of 80% or 90% proven in this study.
  • An inverse-variance-weighted combination form can be numerically backed out, but it is not a vendor-confirmed method. The ten-year center is about 2.8452% annualized, and the combined interval about −0.1528% to 5.8432% annualized; it cannot be called more reliable just because it is narrower.
  • The same-time in-sample residual correlations for one / five / ten years are 0.9959 / 0.9440 / 0.8303 respectively, with paired counts of 348 / 284 / 164; the high positive correlation argues against the simplifying assumption that "the two models provide independent information".
  • The combined error variance should include 2w(1−w)Cov(e1,e2); the backed-out narrow-interval form does not show this term. The weighted R² is also not a re-estimated, validated multi-factor R². The combination model has not received independent, calibrated-interval OOS certification in this note.

When the valuation judgment holds and when it fails

Limited judgments that can hold

  • If the vendor's P/E definition, historical processing and constituent splicing are comparable across periods, 24.06 being high relative to its longer-term mean can serve as a valuation-discipline warning; it is not a single-day trading signal.
  • If, under the same definition, earnings estimates are revised down, discount rates rise or concentrated holdings stall together, a high valuation and concentrated weights may amplify asset losses; this is a transmission mechanism, and no arbitrary earnings-decline rate or price decline is filled in.
  • The official FY1 P/E provides another clearly defined forward-earnings measure; investment judgment must track both earnings-estimate revisions and the degree to which they are realized, not just the multiple.

Conditions that would invalidate the judgment or require reassessment

  • The website's P/E denominator / loss treatment, or retroactive revisions to history, change so that long- and short-term means are no longer comparable.
  • Sector and earnings structures change permanently so that the old mean is no longer a reasonable anchor; conversely, unrealized forward earnings would also make a seemingly low Forward P/E misleading.
  • New regimes, interest-rate or profit-margin states outside the sample destabilize the historical linear relationship; this note has no causal identification or regime-switching model.
  • Official and website data are updated out of sync, or NAV, price / total return, forward / realized EPS are mixed up.
  • The advantage disappears when evaluated with different sample thresholds, or future newly matured data shows the benchmark continuing to win.

Therefore this note sets no price target, gives no arbitrary optimistic / base / pessimistic earnings growth, lists no buy / sell probabilities, and does not carry over the teaching scenarios of the old report as empirical forecasts. A relative verdict against the Nasdaq requires first unifying data definitions and evidence; the two websites' long-term outputs cannot be compared directly.

Monitoring indicators and follow-up evidence

The frequencies below are research-process suggestions, not market forecasts or trading rules.

Monitoring item Suggested check method / frequency When to reassess / gaps
Official FY1 P/E and EPS revisions Save the original as-of and field definition at each update Estimates revised down or the aggregation method changes; old EPS must not be joined to new prices
Website P/E version Save a snapshot and SHA at each update Rerun the models when old history is revised; currently no point-in-time history
Top three / top ten and sector concentration Monthly and at major constituent changes, against same-date holdings Distinguish dual share classes from issuers; do not set unverified safety thresholds
Total-return NAV versus index difference Official monthly / quarterly updates, keeping periods consistent When the difference widens, check expenses, cash and valuation timing; do not treat it directly as tracking error
Expense ratio, spreads and premiums / discounts Expense announcements and public pre-trade data Individual trading costs are not in historical NAV returns
Official risk series Fill in same-definition NAV total-return history Only then can maximum drawdown, volatility and recovery periods be reported; currently left blank
Incremental model validity Rerun the purged design after new labels mature Report threshold sensitivity, overlap and the benchmark; do not look only at training R²

Source register and acceptance

Sources and access limitations

Code URL / document Access date Use in this note and status
O1 State Street SPY official product page 2026-10-30 Directly verified via HTML and browser; valuation / holdings 2026-10-28, performance 2026-09-30, fund information 2026-10-30
O2 S&P DJI S&P 500 2026-10-30 web_fetch 403; the browser successfully verified the general description and the methodology link; the description has no effective date
O3 SPY factsheet PDF 2026-10-30 Linked from the official product page; a request to the generic path retrieved the PDF but it was not parsed, not used as numerical evidence
O4 S&P U.S. Indices Methodology PDF 2026-10-30 Document link listed on the official HTML; PDF 403; version, date and full rules not verified
V1 WorldPeRatio S&P 500 2026-10-30 The coordinator saved this snapshot, page-header valuation date 2026-10-29; not an official definition
M1 Model audit appendix (the author's model audit appendix, not published on this site) / model_metrics.csv (data/SP500/model_metrics.csv) / purged_oos.csv (data/SP500/purged_oos.csv) Computed 2026-10-30 Produced by the coordinator; this report read the CSVs and cites their actual results
M2 combination_diagnostics.csv (data/SP500/combination_diagnostics.csv) / manifest.json (data/SP500/manifest.json) Computed 2026-10-30 The combination form is backed out; the vendor snapshot, calculations and hashes are traceable
M3 NIST normal distribution table Originally checked 2026-10-30 Mathematical reference for the interval label, not market data; the original report recorded this check

Acceptance for this report

Acceptance item Result
Physical new-version file, Traditional Chinese, clickable table of contents Completed, located in the designated reports folder; not merely suggestions
Old report unchanged The old file remains in place; baseline SHA-256: 1d0d6bf1d1d66fee218ec59bb1f8e056c42fef09b85d5e3cf68d02e736252cd9
Dates / units / sources / definitions of official numbers Listed in this note and in official_sources.json; 2026-10-29 was not passed off as the official date
Separation of official and vendor data Completed; P/E, timing, NAV and price are not mixed, and different indexes are not ranked directly against each other
Rebuild and OOS results Actual CSVs cited; the 44-field rebuild passed; only pseudo-OOS, with overlap and untestable horizons stated
Full official risk / methodology data Partially completed: qualitative risks verified; maximum drawdown / volatility and the full methodology PDF not verified
No arbitrary price targets / probabilities / growth assumptions Complied with; the official 21.73% estimate is marked separately as a third-party estimate and is not used to derive a price target
Shared appendix and quantitative attachments Checked that all relative file links in this report exist; JSON parsing, holdings / sector sums, table-of-contents anchors and the unchanged old-file SHA all passed

Research bottom line: reproducible is not the same as predictive; official sources do not imply a single valuation definition; longer history and more monthly points do not imply more independent experience. The existing evidence supports continued monitoring of valuation and concentration; it does not support a certain decline or a guaranteed long-term return.

Further reading

Frequently asked questions

What is the S&P 500 P/E ratio right now?

It depends on the definition. State Street on 2026-10-28 lists SPY FY1 P/E at 20.48x (one-year forward earnings, weighted harmonic mean); the index Price/Earnings on the same page is 24.48x, but that field's full denominator and treatment of losses are not verified; WorldPeRatio on 2026-10-29 lists 24.06x, with its denominator / aggregation not yet independently verified. The three come from different fields / times, are not numbers proven to share one definition, and cannot be ranked into cheap / expensive.

Is the S&P 500 overvalued right now?

This note gives only a limited judgment: on the vendor's data, 24.06x is 17.14% and 43.21% above its ten-year average of 20.54x and twenty-year average of 16.80x respectively, which only supports "high relative to that vendor's long-term benchmark" and can serve as a valuation-discipline warning; it does not prove that the mean is unchanged across regimes, nor is it equivalent to an expected decline. This note sets no price target and lists no buy / sell probabilities; the conclusion is "valuation discipline and concentration risk need managing", not a certain decline.

How much do SPY's top ten holdings account for?

Based on State Street's official holdings (as of 2026-10-28), the top three securities total 21.24% and the top ten securities 38.03%, and the single Information Technology sector is 38.62%; Information Technology + Communication Services + Consumer Discretionary total 56.51%. Ten securities are not ten independent issuers (Alphabet's two share classes total 5.15%), and the sum of three sectors is not an officially defined "tech stocks" or AI exposure.

What has SPY returned over the past ten years?

As of 2026-09-30, SPY's pre-tax ten-year annualized NAV return, net of fund expenses with dividends and capital gains reinvested, is 15.18%, and the unmanaged index in the same table is 15.33%, a difference of −0.15 percentage points / year (the subtraction of two annualized returns, not daily tracking error). This is official total-return historical performance, not a price return, and not a guarantee of the future.

The website says the S&P 500 will return −1.31% a year over the next ten years. Is that credible?

This note considers it only a reproducible output. The P/E→ten-year model obtained R²_OOS 0.7927 in a purged pseudo-OOS with a minimum of 60 training points, but there are only 60 heavily overlapping monthly test start points, and with a minimum of 120 points there are no test points; the raw P/E history also has no point-in-time version evidence. This cannot be used to claim that −1.31% has been fully validated out of sample, and even less can 0.7927 be read as 79.27% accuracy.

What is SPY's expense ratio?

The Gross Expense Ratio is 0.0945% / year, i.e. 9.45 basis points / year; this is the total annual operating expense ratio before waivers or expense reimbursements, and excludes an individual investor's trading commissions, spreads or taxes.

More articles

← All articles Open today's market dashboard →

Contents