I live in a vote by mail state (WA). Actually, there is one county that continues to vote at polling stations.
Vote by mail has not significantly increased voter turnout, and manages to disenfranchise many voters. I haven't focused on this issue for quite some time, so I don't have any statistics at my finger tips.
Here are some reasons why I don't support mandatory vote by mail:
1) It is still electronic counting and recording of votes and the counting occurs in a centralized setting. Because it is centralized votes and the recording of votes cannot be truly verified. Now, my county, King County, which is the largest county in the state, appears to do a much better job than many counties or states, of keeping the process as transparent as possible. Observers can watch workers receive, open, and submit the ballots through the counting machines, but anyone who has researched stolen elections knows that is hardly adequate.
2) Part of the process of voting by mail is "verifying signatures" on the ballots. Many voters' signatures are routinely rejected. I have talked with voters whose signatures have been rejected in multiple elections even though their signature has not changed. When this happens, a letter is sent out to the voter and the voter generally must appear at the county election office to sign an affidavit. Only truly motivated voters are willing to take this extra step, and sometimes, even they don't have the time or the physical means to complete this process.
3) Many voters fail to receive their ballots for various reasons. Lost mail, registration isn't current (so ballot goes to the wrong address), or the voter's record is marked or mis-marked in the database as a possible bad address, so the ballot is not mailed at all. In these cases, if the voter realizes he/she has not received her ballot, if time permits another one can be mailed to the correct address, or the voter can go to one of a few inconveniently located voting centers and vote on a touch screen machine (clearly not a good choice).
4) We receive our ballots 3 weeks before the election. Voters lose or forget about their ballot, and either don't vote or need to have a second ballot mailed to them.
5) Ballots that have unusual markings or if the voter changed a vote, must be "duplicated." This means that an election worker must physically duplicate the ballot (possible tampering point though in King County it is done in teams so less likely, plus there are usually observers walking around). King County has revised this procedure with its new machines, and now this "duplication" or "adjudication" is done directly in the database. Again, with electronic voting, we know that data can be altered by people and software, so accuracy and verifiability become a problem.
6) Vote by mail is very costly because ballots must be printed and mailed to all registered voters and requires a lot of costly equipment. The last time I looked at this issue our mandatory vote by mail system, along with the equipment needed to conduct such a system, was much more costly than the previous poll style voting.
7)It is common for the machine vendors (i.e. ES&S) to play a much bigger roll in the all vote by mail process. Again, not a good thing.
8)In our state, ballots are accepted if they are post-marked by election day (a good thing). However, because this delays the results, Sam Reed, the Republican Secretary of State(who is constantly pushing for Internet voting, brought us paperless voting in 2004*, doesn't believe audits are necessary) is pushing for an Oregon type rule that mandates that ballots be received by election day. Because voters cannot accurately guess how long it may take for their ballot to arrive in the mail, voters are disenfranchised if their ballot is not received at the election office by election day.
9) Vote by mail drags out the counting process, unlike poll voting.
For me, I think poll voting, combined with poll-site verification of the count, is the ideal, if our goal is both access and verifiability. Centralized counting removes transparency and verifiability (again, read up on Ohio 2004 election).
*In 2004 a few Washington state counties used paperless touch screen voting machines. That was the year that WA had a close governor's race and we had a recount. The initial tally showed the GOP candidate winning by a small margin. Fortunately WA had some brave Democrats (the governor and the head of the State Democratic Party) who demanded a recount.
One county that used the paperless machines was Snohomish county. Approximately 100,000 votes could not be verified because they were cast on paperless machines. After the election, DU blogger Paul Lehto conducted a study that determined that the votes cast on the touch screen machines went for the GOP candidate by a margin of about 8%, while the votes cast on paper (mail in ballot) went for the Democrat by a margin of about 3%. That is a 5% difference and a 11% total shift. Statistically it was EXTREMELY unlikely that this difference in results would occur (absent tampering). Statistically one would expect the results to have been the same or very close regardless of the voting method. However, there was a 11% shift toward the GOP candidate on the unverifiable machines. (my recollection may be a little fuzzy on the exact percentages, but this is very close - you can read about it at the link provided).
http://mysite.verizon.net/resq4lzq/cvi/id272.html