Case study: the S&P 500

What the index actually is

• A committee's list. S&P's index committee selects ~500 large US companies meeting rules on size, liquidity and profitability (an earnings requirement famously delayed Tesla's entry to late 2020). It is not the 500 largest by decree: judgment is involved, which surprises most people.
• Float-weighted. Each company's weight is its market value of freely tradable shares over the total. Consequence one: winners weight themselves up automatically (a structural momentum tilt, no rebalancing needed to ride a winner). Consequence two: the index's character shifts with leadership eras: energy-heavy in 1980, tech-heavy now.
• A replacement engine. Companies exit by shrinkage, acquisition or failure; new leaders enter. Roughly 20-25 names turn over in a typical year, so the 2026 index shares little with 1980's: the index survives its members, a quiet survivorship machine that individual portfolios do not natively have.
• Price vs total return. The quoted level ignores dividends. Long-run, dividends reinvested are a large fraction of the outcome: always analyze TOTAL return, and distrust any long-horizon chart that does not.

The honest record

long-run CAGR (1926- ):        ~10%/yr    (~7% real, after inflation)
$1 compounding at 10% for 40y:  $45
...but delivered as:
  best years:   +30 to +50%       worst years: -37% (2008), -43% ('31)
  1966-1982:    ~0% REAL for 16 years (the inflation grinder)
  2000-2013:    ~0% price return for 13 years (two 50% crashes inside)
  frequency:    ~-10% most years; -20%+ every ~4-6y; -50% ~2x/generation

Both halves of the record matter equally. The 10% is real, and so are the 13 years of nothing: an investor who needs the money on a schedule the market does not honor experiences the same index entirely differently, which is why the portfolio-construction guide starts from objectives and horizon, not from the average.

Why it grows: the engine, itemized

nominal GDP growth flowing to revenues        ~5-6%
+ margin expansion (secular, esp. post-1990)  ~1%     <- may not repeat
+ buybacks shrinking share counts             ~1-2%
+ dividend yield                              ~2-4% historically, ~1.3% now
+ the replacement engine (losers out, leaders in)
+/- valuation change                          the joker on any horizon

Nothing mystical: aggregate earnings per share compound, and price follows earnings over long horizons even though it ignores them over short ones. The honest caveats professionals attach: today's starting dividend yield is historically low and margins historically high, so decomposition math argues for expecting somewhat BELOW the historical 10% from today's starting point; and the valuation term dominates any single decade, which is why the same machine produced 18%/yr in the 1990s and ~0% in the 2000s.

The crashes, catalogued

Concentration: today's live issue

Float weighting means the index IS its leaders, and by the mid-2020s the top ten names, dominated by the AI-era platforms, reached roughly a third or more of the total weight, with the top name alone at times above 7%: levels at or beyond every prior concentration era (the Nifty Fifty, 2000). Two honest readings coexist: this is the replacement engine WORKING (the biggest weights are the era's most profitable companies, not speculative shells), and simultaneously a buyer of "the diversified market" now holds a substantial single-theme position in AI infrastructure. The equal-weight version of the same 500 stocks has at times diverged from the cap-weighted index by double digits in a year, which IS the concentration, measured. An index holder today should at least know which bet they are carrying; the factor-regression habit from the Quant track answers it in one afternoon.

How people actually own it

• Index funds and ETFs. The core instruments track the index for expense ratios of a few basis points, with tracking differences near zero: the cheapest large financial product in history. The ETF checklist from the funds guide still applies (spread, securities lending, the exact index tracked).
• Dollar-cost averaging vs lump sum. Historically, lump-sum investing beat spreading it out about two-thirds of the time (markets rise more often than not), but averaging in buys regret insurance that keeps real humans invested: a behavioral trade, and often the right one.
• Derivatives around it. The index carries the world's deepest futures and options markets, which is where the hedging ladder's explicit-insurance rung (index puts, collars) is actually implemented.
• Variants. Equal-weight (dilutes the concentration, tilts smaller), the total-market funds (add the small/mid tail; behave nearly identically), and global ex-US complements for the home-bias question the FX guide raises.
• The tax detail. ETF structure (in-kind creation/redemption) makes index ETFs unusually tax-efficient for taxable accounts: part of why the wrapper won.

What it is not

The index is not a savings account (see the drawdown table), not a guarantee (Japan's index needed ~34 years to reclaim its 1989 peak: the US record is one country's sample), not diversification across asset classes (it is one asset class, concentrated in one country's largest firms), and not passive in effect: buying it today is an active bet on continued US large-cap dominance and on its current leaders. It has simply been the best default equity engine available, at the lowest cost, for anyone with a horizon long enough to pay volatility's toll, which is exactly how the portfolio-construction guide slots it: the core, sized by objectives, hedged by the ladder, held through the table above.