Imagine, each year, after markets close on the last day of June, you form a portfolio based on market equity (market cap). Take the NYSE and find the size decile breakpoints (smallest 10%, smallest 20%, and so on). Then, apply those breakpoints to the NYSE, NASDAQ, and AMEX, form an equally-weighted portfolio of all stocks in that size category.
We can form those portfolios at the end of each June and then look at the average returns (price appreciation as well as reinvested dividends included) from 1926-2012. The results of each strategy are depicted graphically below (click to enlarge).
As one can see, the lowest decile of stocks returns approximately 18%, while the rest return only 11%. We can also see this small-cap premia evolve over time if we look at the yearly returns of these portfolios, rather than taking the average over many years. While yearly returns are volatile, I compare the 10-year results of rebalancing on size each year (click to enlarge). Again, returns include dividends, producing an equally-weighted 14.5% average yearly return from the stock market rather than 12% from price appreciation alone. Further, the returns are 10-year returns and are net and in percent (e.g. 30 on a certain date means that rebalancing to a certain-sized portfolio for the 10 years leading that date would yield a 30% net return (you would possess 1.3*original assets). To be clear about dating, the last date of December 2002 represents the return of a strategy of holding the June 2002 portfolio from December 2002 to end of June 2003, and then reforming the portfolio for every year until liquidation in December 2012: yielding a return around 14% for the small-cap portfolio.
It is important to remember that this is in no way indicative of an "inefficient" market: if some stocks are more risky, they should have more rewards. If the smallest 10% of stocks are illiquid in times when liquidity matters, then they should have a higher average return, to compensate for their bad qualities.