Here c = empirical 0.50 quantile (median) for sum of n Paretos:

And we're doing 10 million simulations of each sum.
Really 10 million; i.e. when n=5, we're actually 
generating 50 million Paretos altogether, 5 for each sum.


n =  2  a =  0.5  c =  14.94038 
n =  2  a =  0.6666667  c =  8.624624 
n =  2  a =  1  c =  5.109139 
n =  2  a =  1.5  c =  3.664906 
n =  5  a =  0.5  c =  89.23954 
n =  5  a =  0.6666667  c =  37.2884 
n =  5  a =  1  c =  16.89877 
n =  5  a =  1.5  c =  10.60077 
n =  10  a =  0.5  c =  351.0256 
n =  10  a =  0.6666667  c =  111.268 
n =  10  a =  1  c =  40.4931 
n =  10  a =  1.5  c =  23.00139 
n =  20  a =  0.5  c =  1392.249 
n =  20  a =  0.6666667  c =  327.3599 
n =  20  a =  1  c =  94.74584 
n =  20  a =  1.5  c =  48.97829 
n =  50  a =  0.5  c =  8654.798 
n =  50  a =  0.6666667  c =  1345.801 
n =  50  a =  1  c =  283.0764 
n =  50  a =  1.5  c =  130.0214 
n =  100  a =  0.5  c =  34628.25 
n =  100  a =  0.6666667  c =  3882.272 
n =  100  a =  1  c =  636.17 
n =  100  a =  1.5  c =  268.7346