A thirty-name price-weighted average as a seasonal regime classroom

A classroom, not a census

The familiar thirty-name industrial average dates from October 1928, when the roster was expanded from 20 stocks to 30. This editorial treats that printed series as a regime classroom, not as a market census.

Editorial framing uses seasonal analysis and seasonality analysis to put a single trade into a diversified or regime-aware context on a weeks-to-months horizon. Seasonal trading is reserved for making entry, exit, and abstention rules testable as one procedure over the system holding period.

How the printed average is built

Early construction added unadjusted closing prices and divided by the number of names. The 1928 version used a divisor of 16.67 rather than a weighted mean. In a price-weighted average, high-priced members move the printed level more than low-priced members.

After 63 years of splits, dividends, and substitutions the divisor had fallen from 16.67 to 0.505, so a small price change in any of the 30 names moves the published average by a much larger amount. Editorial reading: one point is not a stable unit of breadth. It is an artifact of the current divisor and of which names sit on the roster.

Roster changes as era markers

One large technology name entered the roster on May 26, 1932, left on March 14, 1939, and returned on June 29, 1979 after a 40-year absence. The archive presents that path as evidence that membership changes are opaque and era-specific.

This editorial treats composition-substitution as an era marker rather than a transparent, rules-based market definition. A seasonal or valuation reading taken from the printed average is a reading of the roster then in force, not of an unchanging industrial universe.

One close is not the roster

On July 16, 1990 the average closed at 2999.75. An extra 12.5 cents, or one-tenth of 1 percent, in one high-priced name would have printed 3000. Only four names made 12-month highs that day, and one large retailer printed a 12-month low.

Editorial reading: a round printed level can be a thin event. The close belongs next to how many names actually participated, not next to the integer alone.

Volatility, yield, and earnings multiples

Using each year’s percent range from low to high as a rough volatility gauge, only 1974 and 1987 among the prior 16 years approached the extremes of the early decades.

A common overvaluation shorthand of 30 times roster dividends, a 3-1/3 percent yield, sat near 3000 when 1990 dividends were about 100. A 75-year average dividend-multiple was 22.7 times dividends. The August 1987 yield of 2.6 percent was the lowest in 116 years.

From 1966 through 1982 the 1000 level repeatedly attracted approaches that failed to hold. The long earnings-multiple record shows the average can trade well below 10 times earnings before turning up. Editorial reading: those multiples and failed holds are regime context for a single trade, not a standalone forecast of the next move.

Month-of-year extremes

Across 63 years from 1928, August contained only one yearly high and one yearly low. October contained three highs and six lows. January and December hosted 28 and 25 yearly extremes. Nineteen yearly highs, 30 percent of the total, fell in December.

A calendar-extreme is only a count of years in which a given month contained that year’s high or low. Editorial reading: clustering of extremes is a market-regime overlay. It does not, by itself, generate an entry.

Test the reading as portfolio context

The editorial sequence is construction first, then membership, then one close against volatility, dividend-multiples, earnings multiples, and calendar-extremes. Seasonal analysis and seasonality analysis keep that stack in a weeks-to-months market-regime so a single trade sits in diversified or regime-aware context.

If the next step is a rule, seasonal trading folds entry, exit, and abstention into one testable procedure over the system holding period. This case study stops at the classroom: it does not state a trade.