The research compiled below documents decades of academic evidence that commodity and futures prices exhibit dependence rather than pure randomness — the empirical foundation upon which systematic trend following rests. This evidence spans multiple decades, multiple commodities, and multiple research methodologies, all converging on the same conclusion: price movements are not purely random, and systematic approaches that exploit their non-random characteristics can produce profits after commissions.

1. Working (1928) found a tendency for reversals in wheat prices during the 1927-40 period. He attributed these reversals to scalping and day trading.
2. Brinegar (1954), investigating wheat, corn, and rye futures, also found evidence of price reversals over short differencing intervals. In addition, he found evidence of price continuity over longer time intervals.
3. Brinegar, C.S., (1954), A Statistical Analysis of Speculative Price Behavior, Food Research Institute Studies, Vol. 9, Supplement, 1970.
4. Houthakker (1961) investigated the profitability of various stop order trading rules for wheat and corn futures. He concluded that price changes in these markets were not purely random.
5. Houthakker, H.S. (see reference 7).
6. Smidt (1965) tested the efficiency of the soybean market during the 1952-61 period and found evidence of negative serial correlation in prices. Using a filter rule, Smidt also showed that abnormal profits could be made even after commissions.
7. Smidt, S., (1965), A Test of the Serial Independence of Price Changes in Soybean Futures, as reprinted in Peck (see reference 9), pp. 257-277.
8. Stevenson and Bear (1970) employed both filter tests and formal statistical tests to corn and soybean futures price. They found significant negative dependence for differencing intervals of one and two days and positive dependence over intervals longer than five days. In addition, some of their filter rules yielded profits superior to a buy and hold strategy.
9. Stevenson, R. and R. Bear. (See reference 9).
10. Labys and Granger (1970) applied statistical tests to a number of futures contracts in the 1950-65 period. They concluded that most of the contracts followed a Martingale process.
11. Labys, W. and C. Granger, (1970), Speculation, Hedging, and Commodity Price Forecasts, Heath Lexington Books, 1970.
12. Leuthold (1972) used trading rules and statistical tests to investigate the live cattle market. The trading rules yielded significant profits even after commissions while the statistical tests rejected the independence hypothesis in 17 of 30 contracts.
13. Leuthold, R.M., (1972), Random Walk and Price Trends: The Live Cattle Futures Market, Journal of Finance, Vol. 27, pp. 879-889.
14. Cargill and Rausser (1972) employed several statistical tests to various commodities between 1962 and 1972 and found considerable evidence of non-random behavior.
15. Cargill, T.F., and G. Rausser, (1972), Time and Frequency Domain Representation of Futures Prices as a Stochastic Process, Journal of American Statistical Association, Vol. 67, pp. 23-30.
16. Mann and Heifner (1974) used non-parametric tests to examine nine different commodities in the 1959-71 period. They found significant positive correlation between daily closing prices for 90% of the contracts.
17. Mann and Heifner (see reference 11).
18. Martell and Helms (1978) tested transaction to transaction price changes for a number of commodities and found strong serial correlation.
19. Martell, T.F., and B.P. Helms, (1978), A Reexamination of Price Changes in the Commodity Futures Market, International Futures Trading Seminar, Vol. 5, Chicago Board of Trade, pp. 136-159.
20. It is apparent that most of the previous empirical studies have found some evidence of price dependence, which usually took the form of negative serial correlation over short differencing intervals. However, this is not necessarily surprising. As Smidt (1968) has pointed out, prices are not expected to be perfectly independent or there would be an absence of speculators in the markets to act as risk undertakers for hedgers. He argued that one should expect to find systematic tendencies in price movements, which are just strong enough to attract traders in the activity of eliminating them.
21. Smidt, S., (1968), A New Look at the Random Walk Hypothesis, Journal of Financial and Quantitative Analysis, Vol. 3, pp. 235-261.

What This Research Collectively Demonstrates

Item 20 is the most important on the list. Smidt’s 1968 theoretical argument explains why the empirical findings are not anomalous: prices are not expected to be perfectly independent, because perfectly random prices would offer no compensation for the speculators who provide risk-bearing liquidity to hedgers. The systematic tendencies in price movements are not market inefficiencies waiting to be arbitraged away. They are the necessary compensation for the risk-bearing function that speculators provide. As long as hedgers need to transfer price risk to speculators, systematic tendencies in price movements will exist.

This is the academic foundation for why trend following works and why it will continue to work. The research bibliography above spans from the late 1920s through the 1970s and covers wheat, corn, rye, soybeans, live cattle, and multiple other commodities. Every study found some evidence of price dependence. The pattern is consistent across decades, commodities, researchers, and methodologies. The dependence takes a specific form that trend following exploits: negative serial correlation at short intervals (short-term mean reversion that creates the entry cost of failed breakouts) and positive dependence at longer intervals (the sustained trends that produce the large profits).

Brinegar’s 1954 finding is the clearest statement: reversals over short intervals, continuity over longer ones. This is precisely the statistical environment that trend following is designed to operate in. The short-interval reversals produce the small losses from entries that fail to develop. The long-interval continuity produces the large gains from entries that develop into sustained trends. The edge is not in predicting which entries will succeed. It is in having the correct sizing and exit rules so that the small losses from failed entries are more than offset by the large gains from successful ones.

Monroe Trout, one of the most successful traders documented in Jack Schwager’s Market Wizards series, compiled this research as part of the intellectual foundation for systematic trading. The academic evidence confirmed what practitioners had observed empirically: markets trend, and those trends can be profitably followed with the right systematic approach.

Frequently Asked Questions

What is serial correlation in futures prices and why does it matter?

Serial correlation is the statistical relationship between a price change at one point in time and a price change at a subsequent point. Positive serial correlation means that price moves tend to continue in the same direction. Negative serial correlation means they tend to reverse. The research compiled here consistently found negative serial correlation over short intervals and positive correlation over longer intervals, which means short-term reversals are followed by sustained longer-term trends. This is the statistical environment trend following is designed to exploit.

Why does Smidt argue that prices cannot be perfectly independent?

Because perfectly random prices would offer no compensation for speculators who provide risk-bearing liquidity to hedgers. Hedgers pay a risk premium to transfer price risk. That premium must be large enough to attract speculators. The systematic tendencies in price movements are the form that premium takes. As long as hedging activity exists, systematic price tendencies will exist, which means the opportunity that trend following exploits is structural and durable rather than a temporary market inefficiency.

How many decades of research support the existence of price trends in futures markets?

The research compiled here spans from Working’s 1927-40 wheat study through Martell and Helms’s 1978 work, covering five decades of data across multiple commodity markets. Every study found some evidence of price dependence. The consistency of the finding across different researchers, different commodities, different time periods, and different statistical methodologies constitutes strong cumulative evidence that price dependence is a structural feature of futures markets rather than a statistical artifact of any particular dataset.

Trend Following Systems
Want to learn more and start trading trend following systems? Start here.