The following tables show figures that are direct quotes from the Electoral Commission’s evaluation of the 2003 e-voting pilots, specifically Tables 3 and 7 pp63-65. My response to the Commission’s evaluation is here [PDF]
Turnout at the pilots providing remote e-voting
|Authority||Date of last election||Turnout at last election*||Turnout at 2003 election**||% change||JK’s % change|
|Stratford on Avon||2002||45.0%||36.6%||-8.4%||-8.4%|
|Shrewsbury & Atcham||2002||43.0%||54.8%||+11.8%||+11.8%|
|Averages by JK||-0.71%||-0.86%|
* Pre-2003 turnout figures from the Ofice of the Deputy Prime Minister
** 2003 turnout figures calculated using information gathered from the local authorities and include invalid votes. The e-pilot was held in only 15 of
the 29 Sheffield wards. Turnout as percentage of registered voters.
Turnout and kiosk usage where kiosks were the only method of voting in-person
|Authority||Date of last
|Turnout at last
|Turnout at 2003
|% change||JK’s % change||Kiosk usage|
|Basingstoke & Deane||2002||29.0%||31.3%||+2.6%||+2.3%||75.4%|
|Stratford on Avon||2002||45.0%||36.6%||-8.4%||-8.4%||77.3%|
|Averages by JK||-2.8%||-2.88%|
Comments by Jason Kitcat: This data shows an average change in turnout of -0.71% (mode -12.2%) for the remote e-voting pilots and
-2.8% (mode -8.4%) for the kiosk pilots. Not a great record but the Electoral Commission argue that there is no evidence that introducing new voting
methods reduces turnout in any way. It is possible that the turnout figures would have dropped even further without the pilots, but equally the
figures could have been higher without e-voting. Certainly at first blush these figures contradict the Commission’s view and it is interesting that they
did not include indicators such as averages or modes in their report.
The problem is, as any psephologist will agree, that it is virtually impossible to identify which factors account for changes in turnout. New
candidates, a local scandal, a supportive local paper or a contentious new development can all impact turnout, as can the proportionality of the
voting system. While the figures presented cannot be conclusive it seems that so far e-voting is an extremely expensive way to achieve very little.
2006 update: Looking over these figures again while preparing for the 2007 pilots I’ve noticed some inconsistencies. The changes in turnout have not been correctly calculated, the biggest discrepancy being Stroud and Norwich in the first table. Stroud shows 0% change yet calculates to -1.4%. For Norwich the change is shown as +1.4% when it calculates to +0.8%. Correctly calculating the change in turnouts using the figures in the table gives an average change in turnout of -0.86% for remote e-voting pilots and -2.88% for the kiosk pilots. See the JK’s % change columns.
But that’s not all. For Stroud the Commissions’s Stroud 2003 pilot report (pp15) shows a -3.2% change in turnout between 2003 and 2002 and although these elections were not identical it is this election the Commission claims for use in the table above. Despite being -3.2%, using the table’s figures the change is -1.4% if correctly calculated and only 0% is printed. For Norwich the table shows the last comparable election as being 2002, yet the Commission’s own Norwich-specific report for 2003 (pp17) shows 2001 as the last local election. The figures from this specific report show a 2001 to 2003 turnout change of +0.59% against +0.8% using the above table’s numbers and the 1.4% actually printed.
Just using the two turnout change figures from the detailed reports changes the average turnout for the remote e-voting pilots to -1%. Perhaps there’s a good reason for the figures being the way they are e.g. some adjustments due to boundary changes that weren’t properly noted in the main report, but then why the discrepancies in figures between the council-specific and the main reports all of which were Electoral Commission published?
These errors got me curious so I went through all the Commission’s local reports for the 2003 pilots and put my own table together. I cannot find any rationale for why the figures between the main report and the local pilots reports differ so much. Certainly a small part is that the pre-2003 figures in the table above have been rounded to whole numbers only, but they alone do not explain some huge differences such as Ipswich showing a change of -7.01% against the -0.3% published. The changes aren’t all one way, some areas did better than shown but overall the average change is worse, 23.9% worse: from the published changes in turnout we get an average of -0.71% but from my table we get -0.88%.
Turnout at the pilots providing remote e-voting using local pilot reports
|Authority||Date of last election||Turnout at last election||Turnout at 2003 election||% change||correct % change from published*||published % change**|
|Stratford on Avon||2002||39.18%||36.6%||-2.58%||-8.40%||-8.40%|
|Shrewsbury & Atcham||2002||41.08%||54.80%||+13.72%||+11.80%||+11.80%|
* Shown as JK’s % change column in first table
** Shown as % change in first table