Interesting and thanks for posting. I remember their first study/survey 7 years ago.
I did some frequentest statistics (proportion hypothesis testing) for data that had sample size values. And all but the increase in non-white demographics are not statistically significant at the 5% level (meaning due to random sampling it's unlikely to tell us anything, like the percentage of men in 2014 being 35% compared to 40% in 2007 - it's very likely that the proportion of men have remained the same in the past 7 years, same goes for proportions of young, poor, and educated. Though at the 10% confidence level, the decrease in young people would be significant, so I suppose more data would be helpful to be more confidence in that finding).
Increase in non-white demographics: p=0.005744
Decrease in men: p=0.1619
Decrease in people aged 18-29: p=0.07628
Increase in those with college education: p=0.1077
Increase in those making less than $30,000: p=0.1431
If anyone wants to check my work, here is my code from R, and you can get the data from this pdf (for the proportion tests I took the percentages and multiplied them by the sample size and rounded, so there is some loss of accuracy in my tests as a resullt.