Virtually all of the money spent taking the Census was on enumerating non-responding households, hiring people to go door-to-door to count people. Only a few million were spend mailing out forms.

The problem with online submissions is that it's quite difficult to uniquely identify people online, and a large portion of the population doesn't use computers. It's also expensive to go through records of people who submitted both online and mail-in forms with slightly different entries and determine what is and isn't a duplicate.

Government software projects are also very expensive. I know you could throw this thing together in about an hour, but this isn't the same thing as the US Census. The Census doesn't get to assume that everyone uses a Google account that's tied directly to their real-life identity, and they need assurance that their software is extremely secure. It doesn't justify the obscene prices many contractors charge, but keep in mind that most corporations overpay for software, too (e.g. Oracle database)

There's a reason the government doesn't offer to pay people to turn in their census forms. It incentivizes people to turn in multiple forms, making the count less accurate.

There's also not much utility to a form that anyone can customize. Sometimes, when you're dealing with 310 million people, it's easier to treat them as cattle even if they really are all unique snowflakes.

The problem with online submissions is that it's quite difficult to uniquely identify people online,

Does the census form ask the social security number of the person? Social Security number would uniquely identify individual online or offline.

and a large portion of the population doesn't use computers.

According to this http://www.google.com/publicdata?ds=wb-wdi&met=it_net_us... over 75 per cent of the population is online. And this is probably more in New York City with easy access to internet cafes.

It's also expensive to go through records of people who submitted both online and mail-in forms with slightly different entries and determine what is and isn't a duplicate.

My proposal is for online form only.

The Census doesn't get to assume that everyone uses a Google account that's tied directly to their real-life identity, and they need assurance that their software is extremely secure.

Again, can this be solved with asking social security number of people submitting their forms?

If you do online-only and only 75% of people are online, you have to hand-count 25% of the population, which means you'd need about a 70% response rate with the people online to match what you have with paper forms. That's not realistic, I don't think. Plus, the people who aren't online tend to be in remote areas, which are much more expensive to hand-count.

The Census is inefficient in a whole lot of ways, but one thing that makes it really efficient is standardization. Every one of the ~700k enumerators gets the same exact kit, every manager gets the same boxes, and everything is basically made so an idiot can do it. There are two reasons for this: the first is that it's cheaper, but the second, more important reason is that the data is consistent from place to place. If people in NY respond online and people in North Dakota respond by mail, you'll get lots of little statistical differences that make the data less easy to work with. This the big reason they need to homogenize everything.

As the other commenter said, the social security number only works for people who are citizens, but the intent of the census is to count everyone who is regularly living in the area at the time. It's a tricky thing to define, and I think a silly one, but it's what the constitution says to do so we do it.

http://citizencensus.appspot.com/about