>shows the bifurcation of outcomes, you can't just ignore that for half the people incomes haven't been rising Yes, that's how averages work. You similarly can

>shows the bifurcation of outcomes, you can't just ignore that for half the people incomes haven't been rising Yes, that's how averages work. You similarly cannot ignore that for more than half, wages rose, which is why the numbers have increased. And usually bifurcation means bimodal, which in this case is not likely. >Now who's picking advantageous dates? I listed multiple times now what happens when you use all data. The example that there has been recent growth counters your claim of picking two dates from 50 and making claims about the whole. In fact, I suspect that since the last few years have seen median overall wages rising (2015 or so was the highest growth ever recorded for median wage), that once the numbers for the past two years are tallied, new records will be set. >Are we looking at the same graph? No. I quoted exactly what graph the Census data was in with my first post. You keep picking other ones. >I believe mean is obviously going to be a better measure than upper limit of the range Not likely since we're already taking the lower quintile of the range of all income. Mean loses a info due to skew, which is why most discussions about wages should focus on the median. Since we don't have the median (well, you can possibly find the decile cutoffs to get it), looking at cutoffs tells you that more income has flowed into the bottom than before, since the cutoff is higher. The lowest income bracket contains a lot of retirees, who have no income, but have wealth. It also includes students, that have little income, but likely will have wealth in the future. These zeros and low values will pull the mean down, likely much lower than the median. So they count in this case as poor, but they are not. Since the boomers are retiring, this pulls the mean down, but is not representative of what those at the 20% mark of income earn, which is a better measure, since you're not seeing averaged effects. You see the exact measure at a fixed position in the queue. Similarly, at the high end, taking the mean is nor representative because a few high income earners can skew the mean higher than the median. This is why the ends of the distribution are better served by median (which we don't have), or cutoffs (which is basically a median, but at 20% instead of 50%). To demonstrate this quite conclusively, here's [1] BLS data on expenditures by quintile for 2016. The bottom quintile has a mean of $11k as in your numbers, but has mean annual expenditures of $25k, well over twice the mean income. Note the other 4 quntiles have mean expenditures more in line with the income. This is why mean, especially at then ends, is pretty screwy. These people have low incomes, but are not poor. [1] https://www.bls.gov/cex/2016/combined/quintile.pdf https://www.bls.gov/cex/2016/combined/quintile.pdf