@Goodlifer - I don't know, that's precisely the point. Equities have a positive long term EXPECTED (the operative word) return but I would say all 3 events are possible. Invest in the S&P in mid 2000 for example and you're only just above water.
Let me explain it this way and I'll try not to bore you and please don't just write off some of the terms I use as jargon as they do make sense...I promise.
What do all investors ultimately want? Answer: essentially I believe all investors want as much return as they can possibly get for every increment of risk they incur. Risk and return are intrinsically linked, so for a 10% long term return, you will have to assume a risk of at least 10% to achieve it. Sharpe ratio (although it has its limitations for penalising upside or "good" volatility) is one measure of this. Sharpe ratio defines "how well the return of an asset compensates the investor for the risk taken".
Most people therefore want as high a Sharpe ratio as possible agreed? Right, so a long term Sharpe ratio of 1 is both outstanding and also very rare. A better measure in my opinion is the MAR ratio (average annual return divided by the worst peak to valley drawdown)....ie how much gain do you get for how much pain. Equities have a long term average annual return of around maybe 5-6% with a worst peak to valley drawdown of over 50% so if you just invest in equities (and I will come to diversification in a minute), there is a big disparity between the returns you achieve for the risk you're taking on. Does that mean they are a poor investment? No but it does mean you have to combine them with other things to achieve your overall high sharpe ratio or nice stable upward equity curve. And I don't mean other equities either because major market equities are all highly correlated and if the market crashes, by and large they will all go down.
The mistake I believe 99% of investors make, particularly when selecting investment funds, is to think "if I take investment A with a high Sharpe ratio and combine with with investment B, also with a high Sharpe ratio...the result will be the best possible Sharpe ratio or the nicest, most stable upward equity curve". Wrong wrong wrong! The problem with individual high Sharpe ratio assets (this could be a single stock, fund whatever you like) is that on their own, a high Sharpe ratio is not only rare but invariable has a limited shelf life. You are far better to combine assets that have:
A) A positive long term expected return
B) A time series correlation as close to +1 as possible and return series correlation close to -1 as possible
A is obvious but all B means in very basic terms is that if you have two investments that fit the criteria for argument's sake (the exact criteria being impossible to achieve)....is that when instrument A goes up, instrument B will go down not only by the same amount but at the exact same time. This is the key to achieving the highest possible combined Sharpe ratio and you will find that the very best people in the industry who have the best performance records do not select lots of high Sharpe strategies but instead lots of lower Sharpe strategies that are negatively correlated to each other.
How do you put this into practise as a novice investor then? Well first of all, don't JUST invest in equities because on their own, they are not a good investment. You can argue about stock picking and Buffett and all the rest of it but the fact remains that 95% of people / fund managers under-perform the market so if you are someone who likes probabilities then you are far better getting your equity exposure at low cost in the form of an index tracker....even Warren himself says this on youtube. Then you have to find asset classes that have both a long term expected return AND a low (good) to negative (better) correlation to everything else you are investing in. Think commodities, bonds, currencies, managed futures (many are negatively correlated to equities) and then base your portfolio on how much risk you are taking in each sector. I don't have enough space to write about how you do this but google "risk parity" and somewhere there will be a paper explaining what I mean.