First of all, [30-33] does NOT match 30, 31, 32; it matches any of the single characters 3, 0 - 3, 3, which is the same as [0-3].  (The expression [...] is a character class, meaning that it matches only a single character).

Secondly, when you do /50[0-2]|5[30-33]/, you are trying to match 4 characters in a row:  5, 0, 0-2 or 5, 0-3.

I think what you want instead is:

if ($tag =~ /^(50[0-2])|(53[0-2])$/)
[download]

or more simply:

if ($tag =~ /^(50|53)[0-2]$/)
[download]

or even:

if ($tag =~ /^5(0|3)[0-2]$/)    # Note the parentheses ( ) are optiona
+l
[download]

As above, I'd recommend using number equality instead of regular expressions if that's what you really mean. The range operator and grep are handy for this, like in this (untested) snippet:

while (<DATA>){
      chomp;
      $tag = $_;
      if (grep {$tag == $_} 500..503,531,532) {
           print "$tag\n";
      }
}
[download]

And of course, if you're strictly dealing with numerical data, you don't necessarily have to mess around with a regex -- you might be able to select the numbers you want based on a mathematical rule, as in this (untested) snippet:

  while (<DATA>) {
    my $tag = $_;
    if (($tag % 10) < 3) {
      print "$tag\n";
    }
  }
[download]

I just want to capture 500, 501, 502, 531, 532

Currently you want to do something like /5(?:0[012]|3[12])/, although in the development version of Perl, which will one day be released as 5.10, you are better off (thanks to the work of Yves Orton) by just listing the alternations, and let the regexp engine build a trie out of the alternatives: /500|501|502|531|532/. This is fast.

$h{$_} = () for (500, 501, 502, 531, 532);

while (<DATA>) {
    chomp; print "$_\n" if exists $h{$_};
}

__DATA__
500
501
502
503
504
508
522
531
532
505
[download]