You method will copy the entire @a array back and forth, while the builtin splice will do as little work as possible (and probably will add to the array by shuffling around with pointers instead of copying the contents).

A simple benchmark:

use Benchmark "timethese";

my @c = (1,2,3,4,5,6,7,8,9);
my @s = (1,2,3,4,5,6,7,8,9);
my $pos = 3;
my $new_elem = 100;

timethese(5000, {
COPY => sub {@c = (@c[0..$pos-1],$new_elem,@c[$pos..$#c])},
SPLICE=> sub {splice (@s, $pos, 0, $new_elem)},
});
print "\nAfter iters: size(\@copy): ", scalar @c, " and size(\@splice)
+: ",
  scalar(@s), "\n";
[download]

Benchmark: timing 5000 iterations of COPY, SPLICE...
      COPY: 12 wallclock secs (11.88 usr +  0.00 sys = 11.88 CPU) @ 420.76/s (n=5000)
    SPLICE:  0 wallclock secs ( 0.05 usr +  0.00 sys =  0.05 CPU) @ 100000.00/s (n=5000)
            (warning: too few iterations for a reliable count)

After iters: size(@copy): 5009 and size(@splice): 5009