This post summarizes the results of our survey on voting experience for the September 1 primary . The data from this survey are imperfect in a number of ways, and the specific stories offered by voters show ways that the vote-by-mail system can go wrong, but the results do not suggest a pattern of malfunction in the postal service.
Overall, my takeaways are:
- The vote-by-mail system worked pretty well over all, but it is more complicated than voting in person so there were a few problems of different flavors.
- Voters should use the mail system, but they should always request ballots by mail as soon possible and then return the ballots as soon as possible. Mail delays are not a consistent problem, but they do happen and also clerical delays can occur.
- Using drop boxes avoids any concern about mail delivery in returning a ballot.
- It is wise to check and make sure your ballot has been accepted using the online system. Even if the systems works perfectly, it is possible that there could be a problem with your submission — for example, you forget to sign it.
- Taking all these steps proactively assures that if there is a problem there is time to vote in person if necessary.
Survey Methods and Limitations
We sent an email on Thursday, September 10 to approximately 6300 registered voters in my senate district — Back Bay, Fenway, Allston, Brighton, Watertown, and Belmont. We ran into email delivery volume limits as the mail went out and approximately 1400 of the emails to Boston recipients were bounced by our email system. We did not resend the emails.
The recipients had all at some point corresponded with my senate office by email (supporting or opposing my positions or requesting assistance) and therefore may be considered to be a politically active subset of voters. They may have forwarded the survey to others. There were no tracking mechanisms included in the email so the results were anonymous.
We closed the survey in the evening of Friday, September 11. The total number of responses was 1339 — approximately 27%. Of those responses, 98% were from Boston, Watertown or Belmont.
The survey was written with the expectation that we would show only relevant questions and require responses to them. So, for example, if a person answered the first question and said they were not a registered voter, they would not be presented any further questions. However, it turned out that our form tool had bugs in its conditional logic, so we had to present all questions and make the responses optional. As a result, not all of the responses were complete and logically consistent. This further limited the volume of usable responses.
What mechanism did people use for voting?
In this sample, a majority of people voted by mail and did so at slightly higher rates in Belmont and Watertown than in Boston.
|How Voted (Response count)||Belmont||Boston*||Watertown||Total|
|Early in person||28||36||73||137|
|On 9/1 in person||83||72||85||240|
|Tried to vote by mail had to vote in person||9||36||17||62|
|How voted (%)||Belmont||Boston*||Watertown||Total|
|Early in person||5%||12%||16%||11%|
|On 9/1 in person||15%||23%||19%||18%|
|Tried to vote by mail had to vote in person||2%||12%||4%||5%|
How long did by-mail voters wait for ballots?
The average time elapsed from ballot request sent to ballot received was a little over two weeks in all three communities.
|Days elapsed from ballot request to receipt||Belmont||Boston*||Watertown||Total|
|Count of respondents with complete data||210||135||184||529|
|Average days elapsed||13.5||16.9||14.9||14.9|
Certainly, date responses are imprecise — most people do not have a fully accurate memory of exact dates. However, an understandable pattern does emerge from the data: delays were longest during the early days of the response to the postcards sent by the Secretary of State. In mid-July, voters received post cards telling them how to request ballots. There was a surge in requests then which resulted in backlogs of work for elections staff — the assembly of vote-by-mail packets is very labor intensive. Offices staffed up to respond and then the peak passed by. By mid-August, the turnaround dropped dramatically. This pattern does not support a concern about mail delays. If delays were caused primarily by mail, one would expect them to be roughly the same through the whole period.
|Week that ballot was requested||Count of respondents||Days elapsed from request to receipt of ballot|
|July 13 to July 19||88||23.4|
|July 20 to July 26||64||20.7|
|July 27 to August 2||147||15.4|
|August 3 to August 9||107||12.0|
|August 10 to August 16||77||9.1|
|August 17 to August 23||36||6.9|
|August 24 to August 26 (Weds)||5||3.8|
How did voter behavior affect delays?
The data in the preceding table show that many voters did not request ballots until mid August, but it is unclear when they got postcards from the Secretary of State. It is also unclear how much difference a swifter request would have made in terms of when they got their ballot back, since backlogs were longest at the start of the process.
It does appear that some voters did delay in returning ballot once they had them. Boston voters turned their ballots around more quickly, perhaps because they waited a little longer for them.
|Days elapsed from ballot received to ballot returned||Belmont||Boston*||Watertown||Total|
|Count of respondents with complete data||282||119||192||593|
|Average days elapsed||7.9||2.4||5.3||5.6|
How did return date affect success?
The survey asked people how they returned their ballot and whether they checked the online system to determine whether the ballot had been accepted by elections staff. This gave us a measure of whether mailed ballots were properly received.
|Week that ballot was returned to clerk by mail||Count respondents with complete data||% of those who checked who saw that ballot was accepted|
|August 5 to August 11||27||96.3%|
|August 12 to August 18||62||95.1%|
|August 19 to August 25||63||98.4%|
|August 26 to September 1||26||96.1%|
The consistency of results across weeks suggests that mail delays were not a major factor in ballots not being accepted. If mail delays were a problem, one would expect to see the acceptance rate decline for later submissions.
Any number less than 100% is troubling, but there are many ways for the surveyed rate to be below 100% other than actual failures to record a mailed vote. There were only six in this sample who were unable to confirm their ballot had been accepted. None for them found that their ballots had been rejected for being late and only 2 found that their ballot had not been received. A finding that the ballot had not been received could reflect the time that the online check was made or it could reflect an error in using the check system or it could reflect an error in the clerk’s office in updating the check system. The other four out of six did not complete the checking process to a conclusion or did not respond to the question.
147 comments were submitted while the survey was still outstanding. They appear immediately below. Some voters report specific problems that they experienced involving either mail or the balloting process, although many report a smooth experience. The comments highlight the additional complexity that voting by mail creates.