Yesterday on Twitter, I sent an inquiry to the official Crimea Referendum account for information about the publication of referendum data, especially polling station level data. I have searched for the data and have not found it online, and my request was unanswered. I cobbled together a bit of information based on press releases, and have a couple of observations about the course of the elections.
The proportion of reported invalid ballots (0.72%) is low compared to earlier elections in Crimea (for example, earlier elections yielded invalidation rates of 2.9% in 1998, 3.6% in 1999, 3.7% in 2002, 1.5% in 2004, and 1.2% in 2010 for Crimean voters). Voters make mistakes, and they also intentionally invalidate ballots to send messages. While voter education, ballot complexity, and other factors influence the accuracy of ballot casting, it would be interesting to see the distribution of invalid ballots as 0.72% is quite low.
Turnout data also illustrate the exceptional nature of this vote. The figure below shows reported Crimean turnout in elections held from 1998-2014 at various reporting periods throughout the day. The referendum vote (red squares) even exceeds the falsified second round of the 2004 presidential election in turnout at every reporting period. Crimean voters were either incredibly mobilized to participate - far beyond any past election - or the results were artificially inflated.