A whole lot of both, most likely.
Warren Buffett just finished the annual Berkshire Hathaway shareholders meeting, and apparently doled out his normal dose of insights. Should people care what he has to say?
There is one school of thought that says that markets are sufficiently efficient that investment returns versus the market average are luck. This is saying that investors are like people who know nothing about football playing the office pool. The point spread on each game already incorporates any available information about line-ups, home-vs.-away and so on, and those playing the game in the office might as well go over versus under on any given game. Whoever wins the pool does it by just the sheer luck of somebody happening to get more lucky guesses than anybody else. By this analogy, a stock picker is equally likely to pick a basket of stocks that over-perform vs. under-perform the market in any given quarter, so over-performing or under-performing the market in any given quarter is just "coin flipping". If enough people flip enough coins for enough quarters, some small number will get an incredible run of heads through sheer random chance. The money managers who have great investment track records are just those who happened to be the luckiest.
But of course, sequential investment decisions are (at last for many investors) very different than sequential coin flips in one important sense. Each coin flip is independent of all prior flips, but most investors have underlying theories (explicit or implicit) about the economy, so that sequential returns are very far from independent.
Warren Buffett, for example, is famously a "value investor", which crudely means that he expects stocks that are trading "below intrinsic value" will have much higher odds of future price appreciation than stocks which are not. So, in part, the success of his series of bets will tend to be higher in periods where this turns out to be true than in periods when it is not.
The way value investing is often expressed operationally is to buy stocks with a low price-to-earnings (P/E) ratio. Lots and lots of very prominent people advocated buying stocks over the past few months, because the market P/E had gotten low enough that it was at a point where prior investments into the market at similar P/Es had tended to work out well, as long as you were willing to hold the investment for a decade or more. Of course, what Buffett means in practice by value investing is naturally far more sophisticated than just this. If you want to listen to a real investment genius, read this article from ten years ago that is a near-transcription of one of the only presentations I've known Buffett to give on the overall market.
But the basic idea of buying when markets have low P/E seems like just plain-old common sense, and is often marketed this way. You're buying stocks when they're cheap, after all. But is it true, and if so, to what degree? One obvious problem is common to all such investment advice: if it's so obvious, why don't enough investors immediately buy a bunch of stocks as soon as market P/Es drop low enough that they are at a level that tends to predict rapid future price appreciation, and thereby bid up the price of stocks enough so that the market P/E gets back up high enough that it is no longer at a level that reliably tends to predict future price appreciation? Because there are competing theories of valuation in contention, and none can be shown inductively to be correct for future periods.
To take an extreme case, what if one had the theory that in late 2008 market P/E had dropped to a level that had for decades tended to predict rapid price appreciation, but that the U.S. was entering a period of rapid, terminal economic decline, so that the average P/E ought to be generally lower over the next century than it had been in the 20th century? One could conclude that the market was undervalued versus historical norms, but that it was still over-valued versus rational future expectations (which are of course what matters for this purpose).
This raises the question: over what period of time should we evaluate a long-term approach like value investing in order to believe that it is a reliable guide for future behavior? There is good data for prices and earnings for the New York Stock Exchange going back to the 1870s. I looked at every month from 1881 to 1998 and identified it as a High-P/E (> 15 P/E) or Low P/E (<15 P/E) market. I then looked at the annual inflation-adjusted returns (without dividends) for an investor who purchased a representative basket of stocks in that month, held it for ten years and sold it. For each decade for the 1880s to the 1990s, I calculated the simple average returns for this investment strategy in Low-P/E markets and divided it by the simple average returns for this investment strategy in High-P/E markets. This, in essence, created a simple measure of the advantage created by investing in Low-P/E markets for that decade.
Here are the results for the past 120 years: