On the lack of correlation between earnings and stock returns "Still other readers objected that stock prices are determined by expectations of earnings rather than the actual earnings themselves. As earnings come in, they might provide little information about future expectations and thus would not be expected to have an impact. We do not buy that for a minute. Revisions in expectations must be highly correlated with the actual reported comparisons of the earnings themselves. Furthermore, numerous academic studies of the impact of earnings revisions on the S&P 500 have been made, and none of these have found any relationship between revisions in earnings and future returns. (One curious result that has been found is that the variability of the revisions in earnings expectations has a weak positive relation to returns several years later.)" Well VN apparently is not aware of this http://www.hussmanfunds.com/rsi/econsurprises.htm There is a correlation between economic surprises and the S&P500, so even if the analyst revisions have been tested and have not been good predictors of returns this could be due analysts having a tendency to be wrong and late in their revisions. Meanwhile the market is updating its own forecast in real-time through closely adjusting its daily value by responding to economic news The idea that earnings for the S&P500 are not meaningful for stock returns makes little sense in theory and even though it shows up in some statistical studies as true, this could be due the fact the market antecipates likely earnings through economic data(where plenty of studies show GDP correlates to earnings). Plus VN shows evidence that the stock market predicts earnings, this could be either through the economic data relationship above or through self-fulfilling prophecies where the stock market performance affects people through sentiment/liquidity/wealth effect. Likely through both The counter-argument is that when the stock market moves based on economic data, its not just about earnings but it also affects expectations of interest rates, taxes, etc But from a practical perspective these VN facts dont change much, if you are bearish in the economy and you believe the market is not aware of certain factors(it has not priced that in), it still makes sense to decrease long-exposure due the link to economic news. Same thing if you are bullish
The reason I believe it makes little sense for some to proclaim 'earnings dont matter' based on correlation studies is because if taken to an extreme that theory would not hold. If all S&P500 companies where to report $0 earnings, the stock market would fall, hard. Yes this doesnt show up in some studies, likely because the market can antecipate such events(through bad economic data), fall in advance then rise when the news hits(sell the rumor buy the fact), hence it appears that earnings dont matter. A similar thing happens with recessions, where the market falls in advance and typicaly starts to perform well before the recession is over
With regards to PE ratios and stock returns he says "The data considered were from 1974 to 2001", and he reaches the opposite conclusion as Robert Shiller, says high PE ratios are followed by good returns, low PE rations by bad returns. The problem is that during that sample you had a stock market bubble where high PE ratios went higher, and it doesnt count a part of the period where the bubble burst VN said this about TF "In general it's the philosophical objection that the followers of long term trends don't take into account one of the fundamental rules of economics, which is that incentives matter. The supply curve moves outward and to the right when prices rise, and inward to the left when prices decline" This also should apply to stock prices, plus VN says many times that stock markets are mean reverting. Since they are mean reverting maybe the PE ratios are a bad way to measure the likelyhood of reversion(And this is probably VN's beef with the PE ratio, although this data sample seems to be distorted by the bubble) but the principle still holds, the higher as a %GDP they go the lower the future returns(even though the trendfollowers will cry all day claming the opposite, its a virtual economic certainty), because of the incentives. IPOs, secondary offerings, option bonuses will pop up all over the place to meet demand
He does mention a different set of data "Kenneth L. Fisher and Meir Statman, in a definitive study covering the years 1872 through 1999, also concluded that P/Es provide unreliable forecasts of future returns. 'There is no statistically significant relation between P/E ratios at the beginning of the year and returns during the following year, or during the following two years,' they wrote.4" "P/E has no significant predictive power for markets so far into the future that it is enough to drive most folks crazy, that is, longer than five years.â I'm not sure I understand Fisher here, it appears that his study shows no short-term correlation but a longer term one although thats not clear
And on the flipside to mean reversion you have reflexivity. A strong move in one direction can bring out forces that enforce and strengthen the move. E.g. a weak market can lead to more selling, and more selling and more selling (and vice versa for a stock bubble), killing a few leveraged mean-reversion players every once in a while. VN is completely oblivious to fat tails in the return distribution of financial markets. Also, it's is very well documented that it is a fallacy to think equity markets were historically "mean reverting" at all times on all time scales. It is simply not true. Momentum has been found to be more predominant on some time-frames (multi-month) whereas mean-reversion was stronger on others (multi-year and multi-day).
The times where reflexive trends get out of control are the exception, most of the time the market corrects itself, the times where they dont bubbles form and one can take advantage of that. Certaintly not through buying every hint of a trend(like TF funds) but by indentifying the market mistake before or during the fact and jumping on board Actually, check Education of a Speculator pg 98, there is a negative correlation in S&P futures from 1 day changes to 2 days, 10 days, monthly, up to 1 year changes
I'm curious how you feel so certain making the above claims, unless it is just a personal opinion. Research/backtesting/rigor can help validate/falsify an opinion, even if the result appears counter-intuitive. I am especially astonished by your claim how TF is in your opinion most certainly the wrong way of riding bubbles up and down and economic analysis is the only way to go. Why can't there be two ways to skin the cat? Why only one? The ironic finding is that TF have historically (obviously this is no guarantee for future returns) have excelled in the big historical market dislocations during boom/bust cycles (1998, 2001/2002, 2008) because they are long volatility. I find this fascinating. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=299107&rec=1&srcabs=1363476 Note the "3-12 months period". There are dozens of other papers on SSRN substantiating the claim. Niederhoffer is wrong. Many markets have (historically!) not been instantly mean reverting on all time frames.
The paper you mentioned might very well have show profitability simply due the small cap effect(the lack of arbing that exists in small cap stocks) that the 'does TF work in stocks' shows. Eric has already said he found a inverse correlation between substantial upmoves and market capitalization in the SP500(which was what I was referring to, mean reversion on large cap indices), VN data doenst show 3-12 month range but I will very shocked if the correlation is not negative as well
I'm curious why VN uses correlation to test for the existence of abnormal returns. You will find no scientific paper on SSRN using autocorrelation tests (like VN) testing for market inefficiencies of price series. I just did a test on 20 years of G8 currency data. The average 12 month autocorrelation is -0.09 (i.e. uncorrelated), yet at the same time significant momentum profits exist.
I'm waiting for the evidence that the TF funds net $ return can beat basic benchmarks or the stock market. I certaintly wont buy the compounded % returns that covel talks about because he's got courses and seminars to sell. If the TF funds indeed have a tendency to have drawdowns with bigger AUM then its possible one could do well by investing on them during bad times and getting out when they are popular. But you talked about reflexivity in financial markets, that is a fundamental concept, if one understands that fundamental concept better than others then he can probably indentify when that is going on and take advantage of that, it makes little sense to short copper if you cant see no way there will be a reflexive sell off. Specially with the supply demand curve adjustment that happens in response to price