The history of technical analysis, and Dow Theory
The answer
Dow Theory is a set of 6 principles about how markets trend, assembled by other people from newspaper editorials that Charles Dow wrote between 1899 and 1902. Dow never called it a theory and never proposed it as a trading system. Almost every modern idea about trends comes from it, and 2 of the 6 tenets no longer carry the economic logic that made them work.
Why this costs you money
You are told that the averages must confirm each other. So you watch for a divergence. One index makes a new high and a related index does not, and you sell everything, because Dow Theory says a non-confirmation warns of a reversal.
Then the market rises for 14 more months.
Here is what went wrong. Non-confirmations are common. They resolve upward far more often than most traders assume, and Dow Theory was never a timing signal in the first place. It was a description of the primary trend, measured over months and years, written by a man who was trying to build a barometer of business conditions rather than an entry rule.
The cost of misusing history is quiet. You do not blow up. You sit out good markets holding a 100-year-old rule you never checked, and you attribute the loss to bad luck rather than to the rule.
Almost every technical idea in circulation is a simplified version of an older idea whose original conditions no longer apply. Knowing where an idea came from is the fastest way to know where it stops working.
How it works
Where it came from
Charles Dow co-founded Dow Jones and Company with Edward Jones in 1882 and launched the Wall Street Journal in 1889. He created the Dow Jones Transportation Average in 1884 and the Dow Jones Industrial Average in 1896, which began with 12 companies.
His purpose was not trading. He wanted a single number that summarised the state of the market, in the way a barometer summarises the state of the weather. He wrote about how that number behaved in editorials in his own newspaper. He died on 4 December 1902.
Three other people turned the editorials into a theory.
S. A. Nelson collected and organised Dow's editorials in The ABC of Stock Speculation in 1902, and is generally credited with first using the phrase "Dow's Theory".
William Peter Hamilton succeeded Dow as editor of the Wall Street Journal and applied the ideas in editorials for 20 years, then set them out in The Stock Market Barometer in 1922.
Robert Rhea produced the version taught today in The Dow Theory in 1932, which numbered and codified the principles.
So the theory has 3 authors and Dow is not one of them. That is not a criticism. It is the reason the tenets sound more precise than anything Dow actually claimed.
The 6 tenets
1. The averages discount everything. Every known fact and expectation is already in the index level, except acts of God. Note that Dow applied this to the averages, not to individual stocks. He believed a broad index summarised collective judgement in a way a single company could not.
2. The market has 3 trends. The primary trend lasts a year or more and is the tide. The secondary trend lasts weeks to a few months and moves against the primary, typically retracing between one third and two thirds of the previous primary move. The minor trend lasts days and is mostly noise. Every modern discussion of "trend" descends from this 3-level structure.
3. Primary trends have 3 phases. Accumulation, when informed buyers buy quietly against a background of bad news. Public participation, the long middle section when the trend is obvious and most of the move happens. Distribution, when the news is excellent, participation is broad, and the same informed buyers sell into it.
4. The averages must confirm each other. A new high in the Industrial Average should be matched by a new high in the Transportation Average. If it is not, the signal is not confirmed.
5. Volume confirms the trend. Volume should expand in the direction of the primary trend and contract on moves against it. Rising prices on falling volume is a warning.
6. A trend continues until a definite reversal signal. The burden of proof sits on the reversal. This is the most useful of the 6 and it is the one people abandon first.
The economic logic under tenet 4, and why it has weakened
Tenet 4 was not a chart rule. It was an argument about the real economy in 1900.
The Industrials made goods. The Railroads carried those goods to buyers. If the Industrial Average was rising because factories were producing more, then the Railroads had to be earning more to carry that output. If production rose and shipping did not, then the goods were sitting in warehouses and the production was not meeting real demand.
That is a genuine mechanism and it deserved its reputation.
Now look at the same 2 averages today. The Dow Jones Industrial Average contains software, payments and healthcare companies whose output does not travel by rail. A large part of modern economic value has no physical shipment at all. The Transportation Average still measures railroads, airlines and freight, which matter, but the tight physical link between the 2 indexes is much weaker than it was in 1900.
The tenet survives. The mechanism that justified it has thinned. That is exactly the kind of distinction this cluster exists to teach, and it applies to far more than Dow Theory.
What it tells you, and what it does not
Dow Theory tells you to think in 3 timeframes at once, and this is its most durable contribution. A fall that is alarming on a daily chart may be an ordinary secondary reaction inside an intact primary trend. A trader who cannot separate those 2 things will trade the noise and miss the tide.
It tells you that trends have social phases, and the accumulation and distribution idea is a genuinely useful description of how sentiment and price travel together.
It does not give you an entry price. Rhea's own signals are slow by design. By the time a primary trend change is confirmed under a strict reading, a large part of the move is over. Dow Theorists have always accepted this. Anybody selling you Dow Theory as precise timing has changed the product.
It does not work well on a single stock. The whole framework is about broad averages, where individual company noise cancels out.
And it does not have strong statistical support as a trading system. The evidence on that is in the cases below, and it is genuinely mixed rather than damning.
The decision rule
Before you act on a divergence between 2 indexes, ask what the mechanism is that links them.
If you can state the mechanism in 1 sentence — one index measures production and the other measures the shipping of that production — the divergence is informative and worth watching.
If you cannot state the mechanism, the divergence is a coincidence between 2 lines on a screen. Indexes made of different companies will diverge constantly for reasons that have nothing to do with the economy.
And a second rule, from tenet 6, which is worth more than the other 5 combined:
Assume the trend continues. Require evidence to conclude it has ended. Whatever you use as that evidence, write it down before you need it, because you will not think clearly on the day.
Try this now
This takes about 6 minutes and it measures the confirmation tenet on live data, rather than on the 2 examples that appear in every book.
- Open a daily chart of your main index. In India use the Nifty 50. In the United States use the Dow Jones Industrial Average. Set the range to 2 years.
- Open a second chart of a related average. In the United States use the Dow Jones Transportation Average, which is the original pairing. In India use the Nifty Bank or the Nifty Midcap 150, because there is no Indian transport average with a comparable role.
- Find every date where 1 index made a new 3-month high and the other did not make one within the following 10 sessions. Each of those is a non-confirmation. Write down the dates. In 2 years you will usually find between 3 and 8.
- For each date, look at the main index over the next 40 trading sessions. Write yes if it fell more than 5% at any point, no if it did not.
- Count your yes answers against the total.
What you should see. Most readers find that a minority of non-confirmations were followed by a meaningful decline. Something like 1 or 2 out of 5 is typical, and the exact number depends heavily on which 2 years you chose.
Now do the thing almost nobody does. Ask what the base rate is. Over any random 40-session window in the last 2 years, how often did your index fall more than 5% at some point? If the answer is similar to your non-confirmation number, the signal added nothing.
That comparison is the whole exercise. A signal is only worth something if it beats what would have happened anyway, and the phrase "what would have happened anyway" is missing from almost every pattern book ever printed.
Three real cases
1. Charles Dow, the Wall Street Journal, 1899 to 1902 — the theory its author never wrote Dow published a series of editorials describing how the averages behaved. He compared the market to the tide, with waves and ripples on top of it, which is the origin of the 3-trend structure. He did not publish a system, did not number any tenets, and by the accounts of his contemporaries was more interested in the market as an indicator of business conditions than as something to trade. The 6 tenets were assembled by Nelson in 1902, developed by Hamilton in 1922 and codified by Rhea in 1932. Every time you read "Dow said", check whether Dow said it. Frequently, Rhea did.
2. William Peter Hamilton, October 1929 — the famous call, and 2 audits of it Hamilton published an editorial in the Wall Street Journal in late October 1929, usually titled "A Turn in the Tide", identifying a change to a primary bear market. It appeared within days of the crash. Hamilton died a few weeks later, in December
- The call became the single most cited piece of evidence for Dow Theory.
Then it was audited twice, with opposite results. Alfred Cowles examined Hamilton's full record of calls in 1933 and concluded it did not beat holding the market. Stephen Brown, William Goetzmann and Alok Kumar re-examined the same record in the Journal of Finance in 1998 and found it did add value once adjusted for risk, largely because Hamilton was out of the market during volatile periods.
The right lesson is not that Dow Theory works or that it fails. It is that a single famous call proves nothing, that a full record can be audited, and that the answer depends on the benchmark you audit it against.
3. The Nifty and the Nifty Midcap index, 2018 (India) — a real non-confirmation, and what it did and did not tell you Indian mid-cap and small-cap indexes peaked in January 2018 and declined through that year. The Nifty 50 recovered and made new highs around August 2018 while the broader market remained well below its January levels. That is a textbook non-confirmation: the headline average advancing without the broader market.
In September and October 2018 the Indian market fell sharply, with the IL&FS defaults from late August 2018 as the trigger. So the divergence was followed by a decline, which looks like a success for the tenet.
Read it carefully before you accept it. The narrow index had been diverging for 7 months. A trader acting on the non-confirmation in February 2018 would have sat out most of a rally. A trader acting in August 2018 was right within weeks. The signal did not contain the timing, and a signal without timing is not a trade. It is a reason to be careful, which is a real thing but a much smaller one.
The question that resolves it
A novice asks: what does Dow Theory say about the market now?
An expert asks: which of the 6 tenets is a statement about human behaviour, and which is a statement about the economy of 1900?
Tenets 2, 3 and 6 are about how people behave in a crowd, and people have not changed. Tenet 4 is about the physical relationship between manufacturing and railways, and that relationship has changed a great deal. Tenet 5, on volume, sits between the 2 and is discussed in the volume articles of this cluster.
Separating those 2 kinds of claim is the skill. It applies to every old idea in markets, not only to this one.
What would make this wrong
If markets moved with no persistence, the 3-trend structure would be a way of describing noise after the fact, and tenet 6 would have no content. The evidence in article 2 of this cluster says there is some persistence, so the structure is not empty. It is also nothing like as precise as the language suggests.
The honest limits are 3.
First, Dow Theory is defined loosely enough that its signals are frequently disputed among people who follow it. Two Dow Theorists can look at the same chart and disagree about whether a primary trend change has occurred. A method whose practitioners cannot agree on what the signal was cannot be tested cleanly, and an untestable method cannot accumulate evidence.
Second, the confirmation tenet is now applied to index pairings chosen by convenience rather than by mechanism. Nifty and Nifty Bank are not an industrial and a transport average. They overlap in constituents and both are dominated by large financial and technology weights. Divergences between them tell you about sector rotation, which is useful, but it is not the tenet Dow wrote.
Third, the historical record has been polished. The 1929 call is repeated endlessly and the calls that failed are not. That is survivorship bias operating on a reputation rather than on a database, and it is harder to see.
In India
India has no direct equivalent of the original Dow pairing, and pretending otherwise is the most common error in Indian Dow Theory writing.
The averages. The BSE Sensex launched in 1986 with a base period of 1978-79 set at 100, covering 30 companies. The NSE Nifty 50 launched in 1996 with a base date of 3 November 1995 set at 1,000, covering 50 companies. Both are free-float market-capitalisation weighted. Neither is price weighted, which is a genuine difference from the Dow Jones averages and is discussed below.
The confirmation problem. The nearest Indian analogues to "the averages must confirm" are these, and each measures something different.
- Nifty 50 against Nifty Midcap 150 or Nifty Smallcap 250. This measures breadth: whether the advance is broad or concentrated in the largest companies. It is the most useful Indian substitute and the 2018 case above is an example.
- Nifty 50 against Nifty Bank. Financial companies are a very large weight in the Nifty itself, so this pairing is closer to a check on internal consistency than an independent confirmation.
- The advance-decline line. The count of rising stocks minus falling stocks. This is not a Dow tenet, but it delivers what tenet 4 was reaching for, which is a check on whether the average represents the market.
A practical warning. Both Indian headline indexes are heavily concentrated. A small number of large companies can move the index while most stocks do not move at all. Article 11 of cluster 1 covers this. It means the phrase "the market is up" needs checking in India more often than in the United States.
In the United States
The original apparatus still exists and can be watched directly.
The averages. The Dow Jones Industrial Average, now 30 companies, and the Dow Jones Transportation Average, 20 companies covering railroads, airlines, trucking and delivery. Both are price weighted, meaning a company with a higher share price has more influence regardless of its size. This is an accident of 1896, when the calculation had to be done by hand, and it produces effects that surprise people. A company with a $500 share price moves the Dow more than a company with a $50 share price, even if the second company is 10 times larger.
The S&P 500 is capitalisation weighted and is the index most professionals actually use. Dow Theory's tenets are usually applied to the Dow averages for historical reasons rather than because they are better measurements.
Where the mechanism still shows. Freight and rail volumes remain a real economic signal, and the Transportation Average still responds to them. The link to the Industrial Average is what has weakened, not the information in transport itself. Some analysts now compare transport to the S&P 500 for this reason.
Data. Both averages have long, clean, publicly available histories going back into the 19th century. This is why so much of the academic literature tests on the Dow. That is a convenience of data, not a claim about importance, and it is worth remembering when you read that a rule "worked on the Dow since 1897".
Where they differ, and what that tells you
The difference is the weighting method, and it changes what a divergence means.
The Dow averages are price weighted. The Sensex and the Nifty are free-float capitalisation weighted. When 2 price-weighted averages diverge, part of that divergence can be an artefact of share prices rather than company performance. A stock split in a high-priced Dow component reduces that company's influence overnight, with no change in the business at all.
What that tells you is where to look when your indexes disagree. In the United States, before treating a Dow divergence as economic information, check whether a large-priced component had a corporate action. In India, before treating a Nifty divergence as economic information, check the weight of the top 5 constituents, because a capitalisation-weighted index concentrated in a few very large companies can move without the market moving.
The second difference is the availability of the original mechanism. The United States still has an independent, investable transport average with a genuine economic role. India does not, and the honest substitute is breadth rather than sector confirmation. If you import the tenet without noticing the substitution, you will draw conclusions from 2 indexes that share half their constituents, which is a statistical exercise in confirming yourself.
Carry this
- Dow wrote editorials. Nelson, Hamilton and Rhea wrote the theory.
- Tenets 2, 3 and 6 are about people. Tenet 4 was about railways.
- Before acting on a divergence, name the mechanism linking the 2 indexes.
- Assume the trend continues until you have written-down evidence that it has not.