Dow Theory

Reviewed 12 August 2026 · Sourced from the S&P Dow Jones Indices averages methodology, Hamilton (1922), Rhea (1932), and tests published in Econometrica and the Journal of Finance

What Dow Theory actually claims

The industrial average closes at a record and every headline says so. Then somebody points out that the transportation average has not made a new high since March, and tells you the rally is hollow because the transports aren't confirming. Nobody explains where that rule came from, who decided the transports get a vote, or what you are supposed to do if they disagree for a year.

Dow Theory is that rule, and about five others like it. It is the oldest surviving framework in technical analysis: a set of ideas for telling a real, durable market trend from noise, drawn from newspaper editorials written before 1903. Its most famous requirement is the one above — that two separate stock averages, one of companies that make things and one of companies that move them, both have to reach new highs before a trend is treated as established.

Now the fact that reframes everything else on this page. Charles Dow never wrote “Dow Theory.” He co-founded The Wall Street Journal, wrote its editorials, built the first stock averages, and died on 4 December 1902 without publishing a book, a numbered list, or a system. The theory was assembled out of those editorials by two other men after he was dead. Every version of “Dow's six rules” you will ever be shown is their reading of him.

That is not a debunking. The framework is coherent, it has genuine economic reasoning under it, and most of the vocabulary traders use for trends came from it. But it does change what kind of thing it is. It is not a discovery with a publication date. It is an interpretation, twice removed from its source, and the people who made it said so in their own titles.

Who actually wrote it down

Most trading ideas have vague parentage. This one is unusually well documented, and the documentation is the interesting part, because it keeps pointing away from Charles Dow.

Charles Dow, 1851 to 1902

The dates come from Britannica's biographical entry: born 6 November 1851 in Sterling, Connecticut, died 4 December 1902 in Brooklyn, New York. He founded Dow Jones & Company with Edward D. Jones in 1882, compiled the first average of selected U.S. stock prices in 1884, and first published The Wall Street Journal on 8 July 1889. Britannica's own phrasing is careful and worth borrowing: his writings for the paper form the basis for Dow theory in market analysis. The basis. Not the theory.

The two averages the theory runs on are dated precisely by their current owner. S&P Dow Jones Indices' own averages methodology gives the Dow Jones Industrial Average a launch date of 26 May 1896 at a base value of 40.94, and the Dow Jones Transportation Average 26 October 1896 at 51.72. Dow died six years later.

William Peter Hamilton, 1922

Hamilton took over the Journal's editorial page after Dow's death and wrote it for twenty-six years. In 1922 Harper and Brothers published his book, and its full title is effectively the citation for this whole page: The Stock Market Barometer: A Study of Its Forecast Value Based on Charles H. Dow's Theory of the Price Movement. A study. By Hamilton. Based on Dow's theory. Hamilton is the reason a theory exists at all, and he is also the man whose forecasting record was later tested by two separate sets of researchers who reached opposite conclusions. That is further down the page.

Robert Rhea, 1932

Rhea gave it the name and the structure. Barron's published The Dow Theory: An Explanation of Its Development and an Attempt to Define Its Usefulness as an Aid in Speculation in 1932 — the first book to carry the title. Rhea built it by working through 252 editorials of Charles H. Dow and William Peter Hamilton, and the book's own contents concede the point: one chapter is titled “The Dow Theory Which Hamilton Interpreted.” The man who codified the theory called it an interpretation, and called his own book an attempt.

And the tidy numbering came later still. Nobody wrote six tenets. Rhea's book runs in chapters, not a numbered list, and the enumeration into six rules is a teaching convention with no single author behind it. It is a useful device. It is not a document, and you should be suspicious of any source that presents it as one.

The six tenets, in plain language

Here they are as they are conventionally stated, each with what it actually means and, more usefully, what it does not say. The third column is the one nobody prints.

You will also see the three movements illustrated as tides, waves and ripples. That analogy is the standard teaching image and it is usually credited to Dow's editorials, but every copy of the wording we could reach was secondary, passed down through Hamilton and Rhea, so it is not in quotation marks here. Same for the tenets themselves: these are paraphrases of a paraphrase, and no page — including this one — should present them as Dow's sentences.

Why the transports had to agree in 1900

Tenet four is the one worth spending time on, because it is the only idea in classical technical analysis with a genuine causal story rather than a pattern. Everything else on a chart is a shape. This one is an argument about the physical economy.

Think about what the American economy was in 1896. It made things — steel, coal, sugar, tobacco, cotton oil, lead — and it moved them on railroads. Those were not two sectors among many. They were the two halves of the whole machine. And the link between them was mechanical: if factory output was genuinely rising, somebody had to be hauling the raw materials in and the finished goods out. Railroad revenue was very nearly a direct function of tonnage carried.

So a rise in industrial share prices with no corresponding rise in railroad share prices was a contradiction that demanded an explanation. Either the factories were not really producing more, or the railroads were mispriced. Two independent markets, priced by different investors on different information, were being asked to agree about the same underlying fact. When they did, you had corroboration. When they did not, you had a question.

That is a real epistemic move, and it is the same move this site makes elsewhere. Breadth on Markets · Technical asks whether the average is being carried by the whole list or a handful of names. The cross-asset overlay on Markets · Economic asks whether bonds and commodities agree with stocks. In each case the value is not in the indicator, it is in the independence of the second series. A price chart cannot contradict itself. Two charts can.

What the two averages actually measure

Before you can judge whether the confirmation rule still works, you have to know what it is comparing. Most people picture two broad economic indices. That is not what these are.

From the S&P Dow Jones Indices averages methodology: the industrial average is 30 stocks, the transportation average is 20 stocks, and both are price-weighted. Members are chosen by an Averages Committee made up of three representatives of S&P Dow Jones Indices and two representatives of The Wall Street Journal. The document is direct about the basis for selection: “While stock selection is not governed by quantitative rules, a stock typically is added only if the company has an excellent reputation, demonstrates sustained growth and is of interest to a large number of investors.” It adds that companies should be incorporated and headquartered in the U.S. with a plurality of revenues derived from the U.S.

Read that again. The two series Dow Theory compares are 30 and 20 hand-picked companies, chosen partly on reputation and investor interest, by a committee that includes newspaper staff. That is not a criticism of the committee — it is a well-run, transparent, century-old process. It is a caution about what a “new high” in one of them can possibly tell you about the economy.

Price weighting, and why it distorts the picture

Price-weighted means exactly what it sounds like. A company's influence on the average comes from its share price, not from how big the company is.

The divisor starts at the number of members and is adjusted over time for splits and substitutions, so the series stays continuous. The consequence is easiest to see with round numbers.

None of that makes the averages useless — they are the longest continuous daily record of American share prices in existence, which is worth a great deal on its own. But it does mean that when the industrial average makes a new high, the honest description is “the sum of 30 hand-picked share prices reached a new high,” and the leap from there to “American industry is expanding” is a lot longer than the tenet implies.

Whether the confirmation rule still means anything

This is the honest open question about Dow Theory, and it deserves both sides rather than a verdict. Read both columns before you decide what to do with the next “transports aren't confirming” headline you see.

The case that it has gone stale

• The industrial average is not industrial. The methodology says so itself: apart from the transportation industry group and the utilities sector, which have their own averages, “the index includes constituents from a variety of sectors.” It is a 30-stock cross-section of large, well-regarded American companies. A new high in it is not evidence that factories are busy, because a large share of its members do not run factories.
• The causal link has thinned. Software, financial services, health care, media and intangible assets are a large and growing part of U.S. output, and none of it moves on a truck. A software company can double revenue with essentially no incremental freight. The 1896 mechanism — more output means more tonnage means more railroad revenue — simply does not reach that part of the economy.
• The sample is tiny and hand-picked. Twenty transportation stocks, selected on judgment rather than rules, standing in for the movement of goods across a $25-trillion-plus economy. A change in an average can also reflect a committee substitution rather than anything happening in the world.
• The rule survived partly because it is memorable. “The transports must confirm” is a great sentence. Great sentences outlive their evidence, which is not the same as being right.

The case that it still holds

• Physical goods still move, in staggering volume. If anything, ecommerce made delivery more central to household consumption than it was in 1970, not less. Packages, ports, warehouses and last-mile freight are a bigger part of ordinary economic life now, not a smaller one.
• The mechanism is about margins as well as tonnage. Transport companies feel fuel costs, driver pay, capacity gluts and freight-rate collapses early and directly, because they have thin margins and no ability to hold inventory. Their share prices can register a slowdown before it shows up in anybody's earnings report.
• The reframed version may still be useful. Practitioners today mostly use the industrials as a large-cap proxy and the transports as a cyclical, economically sensitive proxy. The check becomes “do cyclicals agree with large caps,” which is a legitimate question about market internals. It is just not the question Dow's readers were asking.
• An independent second series costs nothing to watch. Whatever its record, adding a series that can contradict the first is methodologically better than staring at one chart. That is the same reason this site carries breadth.

We are not going to resolve that for you, because it is not resolved. What can be said without picking a side is narrower and more useful: the rule gets repeated as though it were a mechanical test on two indices that still measure what they measured in 1896, and neither half of that is true. The transports are 20 companies picked by a committee. The industrials contain constituents from a variety of sectors by their own methodology. Anyone who quotes the rule without saying that is quoting a sentence, not making an argument.

What this site charts, and why it isn't the averages

The Dow Theory overlay on Markets · Technical plots two lines on one scale so you can see whether they agree, which is the whole analysis. Neither line is a Dow average. They are the tickers DIA and IYT — two ETFs. That substitution is deliberate, and it is worth understanding rather than glossing over.

Why ETFs instead of the averages

The averages are licensed intellectual property. The methodology document states it plainly: “DOW JONES, DJIA, THE DOW and DOW JONES INDUSTRIAL AVERAGE are trademarks of Dow Jones Trademark Holdings LLC,” and “A license is required from S&P Dow Jones Indices to display, create derivative works of and/or distribute any product or service that uses, is based upon and/or refers to any S&P Dow Jones Indices and/or index data.” A free chart embed does not carry that license. There is also a second, better reason: an ETF is a thing a reader can actually look up, price and buy. An average is a number nobody can own.

An ETF tracks its index. It is not its index.

This is the part that matters if you are reading the overlay closely. A large, liquid ETF follows its benchmark tightly, but it is a separate security with its own life. It has its own share price, set by its own supply and demand in the market. It can trade at a small premium or discount to the value of the holdings behind it. It charges a fee that comes out of returns. And it holds real positions that have to be traded, so it can drift from the index it is chasing. The ETF page covers the creation and redemption machinery that keeps the gap small.

DIA is the State Street SPDR Dow Jones Industrial Average ETF Trust, structured as a unit investment trust, holding 30 positions, with a gross expense ratio of 0.16% and an inception date of 14 January 1998. On a $10,000 position that fee is $16 a year — small, and it is a steady drag on top of the index rather than a distortion of the shape you are looking at.

Also note the overlay is a picture of agreement or disagreement, not an instruction. It does not generate a signal, this site does not publish signals, and a divergence on the chart is a question to go investigate on Markets · Economic — where freight-sensitive series like industrial production and retail sales live — not a conclusion.

What the record shows, and what is only convention

Dow Theory has something almost no other classical charting idea has: two named, published, peer-reviewed tests of the same body of forecasts. They reach opposite conclusions. Both are worth knowing precisely, because the internet is full of confident hit rates for this theory and none of them come from either study.

Cowles, 1933: it underperformed

Alfred Cowles 3rd examined Hamilton's Wall Street Journal editorials in “Can Stock Market Forecasters Forecast?”, Econometrica vol. 1, no. 3 (1933), pp. 309–324. His window was December 1903 to December 1929, and he counted 255 editorials containing forecasts across Hamilton's 26 years in the chair.

• Following the forecasts on the industrial averages returned 12% a year, against 15.5% for staying continuously invested in the same stocks.
• On the railroad averages: 5.7% against 7.7%.
• Of 90 announced changes of position, 45 were successful and 45 were not — an exact coin flip.

Cowles's own conclusion: “Hamilton therefore failed by an appreciable margin to gain as much through his forecasting as he would have made by a continuous outright investment in the stocks composing the industrial averages.”

Brown, Goetzmann and Kumar, 1998: risk-adjusted, it looks better

Sixty-five years later, Stephen J. Brown, William N. Goetzmann and Alok Kumar re-examined exactly that record in “The Dow Theory: William Peter Hamilton's Track Record Reconsidered,” The Journal of Finance vol. 53, no. 4 (1998), pp. 1311–1333. They trained a recurrent neural network on Hamilton's 1902–1929 editorials to reconstruct the rules he was actually applying, then priced the resulting strategy properly for risk. Their finding, in their words: “Hamilton's timing strategies actually yield high Sharpe ratios and positive alphas for the period 1902 to 1929,” with a Jensen alpha of roughly 290 basis points a year, and “Adjustment for systematic risk appears to vindicate Hamilton as a market timer.” The reason the two studies differ is that Hamilton's approach spent stretches out of the market, so it carried less risk than buy-and-hold — a difference Cowles's raw return comparison did not price.

Their caveats matter as much as their finding, and they state them themselves: “Normal trading frictions would preclude using the Theory to generate large excess returns, particularly in the most recent period,” and, on the era since, “Lack of reliable daily return data and trading cost data over the period since 1930 prevents us from precisely calculating return earned.”

So the honest summary of the evidence is this. The best-documented test of Dow Theory is a test of one man's editorial judgment, a century ago, on two averages that no longer hold what they held then — and the two most serious readings of that record disagree with each other. That is a genuinely interesting historical result. It is not a hit rate, it cannot be converted into one, and no reliable published success rate for the confirmation rule as a mechanical test was located for this page. None is printed here.

The three criticisms that stick

It is hard to falsify. Tenet six says the primary trend continues until a clear reversal, and never defines clear. That gives the framework an escape hatch: any call that went wrong can be reclassified afterward as a secondary reaction inside an intact primary trend. A rule that can absorb its own counterexamples is comfortable to hold and impossible to test. Compare the way this site treats a breakout or a level of support and resistance — the same after-the-fact problem, named out loud.

It is late by construction. Confirmation requires both averages to have already made new highs or new lows. By definition that happens after the move is under way, often well into it. The framework was designed to identify a trend in progress, and it says so; the problem is that it gets quoted by people who want it to call a turn, which it was never built to do.

The phases are retrospective. Accumulation, public participation and distribution are cleanly separable in a chart of the past and very rarely in the present. Labeling the current market as “distribution” is a prediction wearing the costume of a description.

The ledger

House rule on this site: claims of different confidence get labeled differently rather than blended together.

What trips people up

• Attributing the tenets to Charles Dow. He wrote editorials and died in 1902. Hamilton wrote the theory in 1922, Rhea named and structured it in 1932, and the six-item list came later than both.
• Reading confirmation as an instruction. It describes agreement between two averages. It is not a signal, it does not say what to do, and treating it as one is how a description becomes a trade you cannot justify.
• Expecting it to be early. Confirmation happens after both averages have already made new extremes. Anyone selling Dow Theory as a way to catch turns is selling something the framework explicitly does not do.
• Reclassifying every wrong call as a secondary reaction. Once you allow that move, the theory can never be wrong — which also means it can never be right, and you have stopped learning anything from it.
• Assuming the averages still measure what they measured. The industrial average includes constituents from a variety of sectors, by its own published methodology. The transports are 20 hand-picked companies. Neither is a census of the economy.
• Forgetting price weighting. A $400 share moves the average ten times as hard as a $40 share on the same percentage change. A new high in the average is a fact about share prices, not about company size or output. Run the arithmetic above if that still feels abstract.
• Confusing the ETF with the average. Especially here: IYT tracks a different index than the Dow transports, with 43 holdings and different weighting. A chart of two ETFs is a good picture of the underlying question and a loose implementation of the strict rule.
• Quoting a hit rate. The two published studies of the same record disagree with each other, and neither tested the confirmation rule as a mechanical system. Any percentage you see attached to Dow Theory was invented by whoever typed it.
• Mistaking age for evidence. The framework is roughly 130 years old and taught everywhere. That is evidence about teaching. Work through the reasoning in the Technical Analysis course and decide for yourself what it is worth.

The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.