#8 is correct; you cannot take the microcosm of a single year's income as average, since it will vary by random amounts. Averaging is a specific science, and you must take the entire data before you can make any determination. In this case, the longer you live, and the more income data you have, the better you can predict future income, but it will always come down to averaging all the data instead of just one year's worth.
As with any figure, you will receive a much more reliable and usable figure for research purposes if you look at the average of that figure spread out over a number of years. If we think about it, you might suffer illness one year that might reduce your income. Also, you might work really hard another year and receive a big bonus. Therefore, examining the average is a lot more effective.
As with any experiment or study, the larger the amount of data collected within the bounds of the experiment, the better results you get. As others have said, a lifespan of data makes the occasional anomalies more obvious and easier to spot. You can then take out periods of unemployment, data for when someone is in school or working minimum wage in an interim job and look at real numbers over time.
I have to support all of the posters above. If examining anything, the longer the period examined, the more things tend to equalize. Given that one does not work at the beginning of life and late in life, it would be hard to find realistic numbers which would provide the best numbers.
Most people start their working lives making considerably less per year than they do in the last few years before they retire. By averaging income over a number of years, those changes can be spread over the full range of years, rather than having some years with significantly higher income amounts than other years.
A lifetime distribution extends over a greater period of time, so the variables and variations are equalized. Each specific year is a shorter period of time so every change, like a raise or a pay cut or an illness that reduces income or any earnings/losses from investments that are particular to that year will show as a deviation in the distribution. Many accumulated years of these sorts of varied events equalize the distribution.
Because in any given year, something unusual could happen. You could be unemployed for a little bit one year. Or maybe you get a bonus or you have a really good year in your store or something like that. Over a life time, those sorts of peaks and valleys even out.