Ansar,

I think Arindam's answer is very useful, but I would caution you
on two points:
1) industry aggregate numbers are often misleading; and
2) when implementing any analysis of data, make sure you know what the data represents.

I've run into these problems many times but can illustrate with a simple example that we've used at a past SEPG:

How Many Logical Source Statements are Here?

if A then B else C endif;

When polled, the audience responded:
1 statement : 20%
2 statements: 30%
3 statements: 45%
4 statements: 5%

There's no right answer - it depends on the counting rules you use.
And if you are collecting size from many teams/people, you really have
to make sure everybody is using the same definition of size.

Similarly with productivity, except now you have at one more
definitional hurdle.
For example,

76 staff hours go into the production of a unit that's 539 lines of code
= 7.1 LOC/hr

The automated coder counter, however, only counts executable LOC at 398
which
yields 5.2 LOC/hr.

BUT, the effort value of 76 is dependent upon who is included.
And I've seen data used for comparisons without regard for types,
language, etc. etc. These considerations are often
overlooked when using aggregated industry data, for instance.

There are a few places that sell highly regarded benchmark data, which
can be useful especially to compare against best in class. ISBSG andPMG
Benchmarking
come to mind.

hope this helps,
Jim McCurley