While it is only one data point and doesn't directly compare to voter registration pools (and hence is only a partial answer), a Master's Thesis comparing juries in 2000 and 2010 reached this conclusion (raw Chi square statistics omitted without indication) regarding jury pools in a Tennessee county in informative of the general trends:
[C]ompared to the 2008 Census data, the jury panels have a
significantly higher percentage of White/Caucasians (83.2% vs. 74.1%,
p < .01), and a higher percentage of males (53.1% vs. 46.8%, p =.015).
Additionally, the participants were older (p < .01), had higher levels
of income (p < .01), higher levels of education (p < .01). . . .
[C]ompared with the 2000 Census, the jury panels in 2000 had
significantly more White/Caucasians (83.3% vs. 75.5%), p=.046.
Additionally, the potential jurors from the previous studies tended to
be older (p < .01), have higher levels of income (p < .01), and higher
levels of education (p < .01). . . .
[T]he present jury pools are composed of citizens who are older, more
male, more White/Caucasian, more likely to be married, have higher
education levels, and have higher levels of income than 10 years ago.
. . . In 2000, the typical juror was a 39-year-old White, female, who
was married, with some college, and whose income was about $55,000 per
year. In 2010, the typical juror is a 50-year-old White, male, who is
married, with an associate‟s degree, and whose income is approximately
$59,000 per year.