Sheila Parks, Ed.D.
Copyright © 2008 by Sheila Parks, Ed.D.
INTRODUCTION
In July of 2007, I published a paper titled “ON-SITE OBSERVATIONS OF THE HAND-COUNTING OF PAPER BALLOTS AND RECOMMENDATIONS FOR THE GENERAL ELECTION OF 2008.”[1] In this paper, I described on-site observations of the hand-counting of paper ballots in three elections in two New England States: Rockport and Hudson, MA and Acton, ME. All of these hand-counted paper ballots (HCPB) elections used the read and tally [2] protocol for counting the ballots. With this protocol, teams of two counters are used: one counter reads out loud the name the voter has marked for each contest (or each initiative), and the other member of the team records the vote on a tally sheet that corresponds to the ballot. For a fuller description of the read and tally HCPB protocol, see my “ON-SITE OBSERVATIONS….” paper. [3] The procedure I liked the most was that used by Acton, ME, in which the ballots were counted twice.
During the period of observation of the HCPB elections in Rockport and Hudson, MA and Acton, ME between May 2, 2006 and November 7, 2006, I did not observe an HCPB method called sort and stack, which is used in New Hampshire (NH). I wanted to observe some sort and stack elections in order to have more complete knowledge of the methods used for HCPB.
I called the NH Secretary of State’s Election Division to find out which municipalities in NH use the sort and stack protocol. The woman I spoke with in the Election Division suggested that I call the individual municipalities to find out which ones use sort and stack to hand-count ballots.
NEW HAMPSHIRE MUNICIPALITIES THAT USE SORT AND STACK
There were 236 municipalities on the Secretary of State’s list of city and town clerks. Accuvote electronic voting machines were used by 108 municipalities. That left 128 municipalities that use HCPB. That number was not completely up-to-date, as three places had changed or were going to change over to electronic voting machines. [6] That left 125 municipalities that do not use electronic voting machines but rather HCPB. I spoke with 123 of these municipalities. [7] I did not reach Ellsworth, whose population in 2000 was 87 or Windsor, whose population in 2005 was 237. Since these were very small municipalities, it did not seem to me to be crucial for this paper which HCPB methodology they used.
In most cases, I spoke directly with the Town Clerk. In a few cases I spoke with an assistant to the Town Clerk and in some, to the Town Moderator of the municipality. The moderators run the elections in NH. I said that I was an HCPB advocate, doing research on HCPB protocols. The questions I asked each city or town were: How many registered voters are there in your municipality and what method of HCPB do you use. I then asked them to explain the method to me. [8] The people I spoke with were helpful, friendly and forthcoming. From my conversations, I learned that of the 123 municipalities in NH with which I spoke that use HCPB, 114 use the read and tally protocol (called the ballot-by-ballot method in the Manual of the NH Secretary of State), while only nine use the sort and stack protocol, although not entirely.
The following is a list of the nine NH municipalities using the sort and stack method. The first four use sort and stack in some circumstances, and read and tally in other circumstances. The last five municipalities use sort and stack exclusively.
1) Andover: Number of registered voters - 1615. Uses sort and stack and also read and tally, depending on how many candidates are on the ballot. Does not have a specific rule for when read and tally or sort and stack is used, except that if there are only two candidates, the sort and stack method is used.
2) Greenfield: Number of registered voters - 1000+. In an uncontested town election, sort and stack is used. [9] Generally read and tally is used.
3) Greenville: Number of registered voters - 1200. Uses sort and stack only if there is one yes/no question on the ballot. If in a zoning election, e.g., there were 6-7 questions that were yes/no, would do read and tally. Uses read and tally if there are two or more contestants, including presidential races.
4) Hinsdale: Number of registered voters - 2405. Uses sort and stack and read and tally. Counters decide which method they use.
5) Langdon: Number of registered voters - 447.
6) Marlborough: Number of registered voters - 1522.
7) Orford: Number of registered voters - 800 or 900.
8) Plainfield: Number of registered voters - 1605.
9) Walpole: Number of registered voters - 2626.
ON-SITE OBSERVATIONS OF SORT AND STACK ELECTIONS IN NH
Most of the hand-counters came in to count when the polls closed and had not been there all day. There is a polling place in North Walpole, and these ballots are brought to Walpole to be counted, together with the rest of the ballots from Walpole.
In the primary election, there were 11 teams of two people counting the ballots. In the Town Meeting election, there were 13 teams of two counters counting the ballots. Each team of counters sat at a separate table. There were no official observers watching the counters at any of the tables. The Town Moderator paired the counters. He described the process as follows: “We pair the counters by a ‘veteran’ and a new person if needed. We do not consider party affiliation. Many are undeclared. People cannot count if they are on the ballot.” [11]
After all the ballots were taken from the ballot boxes, they were counted into batches of 25. These batches of 25 ballots were put crisscross to make a pile of 100 ballots. The ballots were then distributed to the tables. For the Town Meeting election, there were different color coded ballots, and some of the ballots were more complicated than others. The more complicated the ballot, the less ballots for a table. The less complicated the ballot, the more ballots for a table. The Town Moderator distributed the ballots accordingly. Each counter at the table took half of the total ballots at that table.
Sort and stack is done silently. The ballots are sorted into stacks by the counters, according to the candidate or question/warrant article (yes/no) being voted for. There are also stacks for undervotes, overvotes and write-ins. This is the first time that a pair of eyes sees the vote (and makes sure it goes into the right stack). Then the votes are counted, both by hand and by the eyes of the counter at the same time as counting. That is, not only does the counter manually count the stack of ballots, but also the counter looks at each name (or yes/no of a question) as she or he counts. This is the second time a pair of eyes sees the vote. After the votes are counted and the number recorded, the counters switch stacks and the votes are counted again both by eyes and by hand. This is the third time a pair of eyes sees the vote. [12]
The counters did not look at the number of total votes for each candidate or question that the other counter had recorded. When both counters were finished counting the votes for a candidate or question, they then compared the totals they had written for each candidate or question. If there was a discrepancy, both people counted again. The two counters did not observe each other counting; each counter counted her/his own stack at the same time that the other counter did. The counters counted so fast, as they went from ballot to ballot, that I could not see the names or yes/no on the ballots as they counted.
In the primary election, I observed only tables counting the Democratic votes; I did not observe any tables counting the Republican votes. Also, various tables did the process a little differently. [13] There were two ways the ballots were sorted and stacked and then counted:
1) The ballots were sorted and stacked for the seven Democratic candidates out of 22 who received votes so that there were seven stacks on the table at the same time, one for each of the seven candidates who received votes. There were also stacks for undervotes, overvotes and write-ins. Then each stack was counted, as described above.
2) The ballots were sorted and stacked separately for each of the seven Democratic primary candidates who received votes from the 22 running and for undervotes, overvotes and write-ins - that is, the ballots were sorted and stacked and then counted seven different times, once for each candidate who received votes. As ballots were sorted, a stack was made for each candidate. Then each stack was counted. After each sort and stack and count, the ballots were put together again and then sorted, stacked and counted for another candidate.
DISCUSSION OF SORT AND STACK METHOD
In NH, more than half of the municipalities count ballots by hand – at least 123 of the total 236 municipalities. However, the use of sort and stack surprisingly is not widespread in NH, with only nine municipalities using this method. Of these nine, only five use sort and stack exclusively, while four use sort and stack under certain circumstances, but prefer read and tally in other circumstances.
In a talk given by NH Assistant Secretary of State Anthony Stevens at Democracy Fest in NH on June 10, 2007 titled “Hand Counting Paper Ballots,” Stevens stated that the “Secretary of State [of NH] indicates a preferred method [of hand-counting paper ballots] in [the] NH Election Procedure Manual.”[14] This Manual is presented on the website of the NH Secretary of State. [15] In the Manual, sort and stack is described as follows: “This model [sort and stack] is presented as a best practice in hand counting, based on the secretary of state’s experience with hand recounts.”[16] Thus, Gardner’s preference is based on hand recounts and not necessarily with sort and stack as the HCPB method of counting votes on election night. The NH Manual further states: “…This ballot sort and stack method is considered the faster and easier method, even though each mark is seen more times than the method using ballot reading and tally marks. Counters who have tried other methods express more pleasure with the sort and stack method because (a) it is simpler to count, and (b) counters are more confident in the results.”[17]
In his talk, Stevens also said that the “Sort-and-stack method may not be used widely in New Hampshire on election night.”[18] That indeed turned out to be the case, as only nine municipalities use sort and stack in any elections. In several telephone conversations with me, Stevens was very helpful. On August 26, 2008, he noted that he prefers the sort and stack method on election night; however, he stated that he would not and could not push the sort and stack method on any municipality. [19] In NH each Town Moderator, by law, chooses the way votes are counted. [20] In our conversation, Stevens further said that he is currently reaching out to municipalities in NH to teach the sort and stack method. It will be interesting to see if Stevens is successful in teaching the sort and stack method and getting a larger number of municipalities to adopt it. The small number of municipalities using sort and stack was surprising, given the strong endorsement by the Secretary of State and Assistant Secretary of State of NH.
Sort and stack worked well in Walpole, NH. Ernie Vose, the Town Moderator, was very skilled and effective in the implementation of this method. As far as I know, he is the only Town Moderator in NH who uses sort and stack in all elections, even ones in which there are multiple seat races. It is important to note that the NH Manual states that the Secretary of State’s preference for sort and stack is “based on the secretary of state’s experience with hand recounts [emphasis mine].” [21] Therefore, Gardner could be extrapolating from less complicated recounts to elections that could be more complicated.
Both sort and stack elections that I observed in Walpole were done efficiently and finished in a timely manner. Vose noted that it takes a lot of time after the counting to record the data into the computers. [22] The effectiveness of the sort and stack method in Walpole makes it potentially a model that could be followed by other municipalities in NH and across the country if they so wished – at least in simple elections.
This paper has described the sort and stack method of HCPB used in NH, while a previous paper described the read and tally method used in ME and MA. As noted in the present paper, the NH Secretary of State and Assistant Secretary of State prefer sort and stack, but most municipalities using HCPB use read and tally. Only nine of 123 municipalities use sort and stack, and not in all circumstances. What can be said about the two methods?[23]
There are certain differences between the two. First, in the read and tally method, there is a written record of each individual vote cast, as well as the total number of votes each candidate or question received. [24] In the sort and stack method, there is only a total vote count for each candidate or initiative on the tally sheet – no written record of each vote that was counted. On the other hand, with sort and stack the ballots are always counted twice (once by each counter) and this is not necessarily so with read and tally. In the three read and tally elections I observed, only one used an HCPB protocol that hand-counted the ballots twice. Second, as noted earlier, the NH Manual states that sort and stack is faster than read and tally. [25] However, sort and stack, while fast, may sometimes be too fast in some instances. In the sort and stack elections that I observed, the counting of the stacks was done so rapidly that there was no way for me as an unofficial observer to see the name or yes/no on the ballot in each stack as it was counted. In read and tally, it is possible to see the name or yes/no on each ballot as it is counted and then recorded. [26] A third difference between the two methods, is that sort and stack is done silently, so the room is quieter than with read and tally.
Beyond these differences, there are important similarities. Both sort and stack & read and tally were used efficiently in all the elections I observed. Moreover, and most significantly, in the read and tally election in Acton, ME and in both of the sort and stack elections in Walpole, NH, the ballots were counted twice. I strongly recommend that in all HCPB elections, the ballots be counted twice - the second counting immediately following the first. [27] A second counting guarantees greater accuracy and could be considered an audit. The question remains: Why don’t we do it right the first time – i.e., hand-count the ballots on election night. Hand-counting the ballots twice on election night (the second counting immediately after the first) would do away with the fraud and error rampant with the use of electronic voting machine counting.
EPILOGUE
ENDNOTES
[1] http://www.opednews.com/articles/opedne_sheila_p_070718_on_site_observations.htm , retrieved from the web January 4, 2008.
[2] The read and tally HCPB protocol is also called read and mark. The two terms seem to be used interchangeably. The web site of the Secretary of State of NH calls this method “ALL OFFICES, BALLOT-BY-BALLOT METHOD,” http://www.sos.state.nh.us/FINAL%20EPM%208-30-2006.pdf , page 147, retrieved from the web January 4, 2008.
[3] op.cit. http://www.opednews.com/articles/opedne_sheila_p_070718_on_site_observations.htm , retrieved from the web January 4, 2008. See especially the sections on Acton, ME, Rockport, MA, Hudson, MA.
[4] http://www.sos.nh.gov/clerks.htm, retrieved from the Web June 6, 2008. The table I worked with is no longer on the website; it was labeled “2008 (updated July 12, 2007)”. I have a hard copy of this table. There have been some changes in town clerks, but otherwise the two lists are the same.
[5] http://www.sos.nh.gov/voting%20machines2006.htm, retrieved from the Web, June 6, 2008. The table I worked with is no longer on the website; it was labeled “(as of July 2007)”. I have a hard copy of this table. These two tables are the same, except for the three jurisdictions that had already changed over to machines or were going to do so, when I spoke with them. The table I used did not have these municipalities listed as using Accuvote. The new table does.
[6] Conway, East Kingston, Plymouth. As noted in Endnote 5, these three towns are now included in the table listing the jurisdictions that use Accuvote.
[7]Acworth, Albany, Alexandria, Alstead, Andover, Antrim, Bartlett, Bath, Bennington, Benton, Bethlehem, Boscawen, Bradford, Bridgewater, Bristol, Brookfield, (Campton, which was not on SOS list for machines, used machines), Caroll, Center Harbor, Charlestown, Chatham, Chesterfield, Chichester, Clarksville, Colebrook, Columbia, Cornish, Croyden, Dalton, Danbury, Deering, Dixville, Dorchester, Dublin, Dummer, Dunbarton, (East Kingston switched to machines), Easton, Eaton, Effingham, Errol, Francestown, Franconia, Freedom, Gilsum, Goshen, Greenfield, Greenville, Groton, Hancock, Harrisville, Harts Location, Haverhill, Hebron, Hill, Hinsdale, Holderness, Jackson, Jefferson, Kensington, Lancaster, Landaff, Langdon, Lempster, Lincoln, Lisbon, Lyman, Lyme, Lyndeborough, Madbury, Marlborough, Marlow, Mason, Middleton, Millsfield, Monroe, Mont Vernon, Nelson, New Castle, Newington, Newport, Northfield, Northumberland, Orange, Orford, Piermont, Pittsburg, Plainfield, (Plymouth going to machines), Randolph, Richmond, Rollinsford, Roxbury, Rumney, Salisbury, Sandwich, Sharon, Shelburne, South Hampton, Springfield, Stark, Stewartstown, Stoddard, Strafford, Stratford, Sugar Hill, Sullivan, Surry, Sutton, Temple, Thornton, Troy, Tuftonboro, Unity, Walpole, Warner, Warren, Washington, Waterville Valley, Webster, Westmoreland, Whitefield, Wilmot, Wilton, Woodstock.
[8] I learned a lot about different nuanced ways to do a read and tally hand-count.
[9] They have to use sort and stack even with only one name on the ballot, because of undervotes, overvotes and write-ins.
[10] Thanks to Nancy Tobi of Democracy for New Hampshire who had strongly suggested that I contact the Town Moderator of Walpole, NH about sort and stack elections.
[11] Ernie Vose, Moderator of Walpole, email correspondence, January 12, 2008, in response to my question.
[12] This paragraph is a compilation of: a conversation with an election official at the Walpole election on why she liked sort and stack so much; my own observations of the hand-counting at the Walpole elections; phone conversations with Anthony Stevens, NH Assistant Secretary of State, August and September, 2008.
[13] Conversation with Ernie Vose , January 1, 2008.
[14] http://www.democracyfornewhampshire.com/files/Hand_count_training_D-fest_July_5_2007.pdf retrieved from the Web November 4, 2007. I was not present at his talk and saw his words on his power point presentation posted on the Internet.
[15] op. cit. http://www.sos.state.nh.us/FINAL%20EPM%208-30-2006.pdf , r etrieved from the Web January 4, 2008. The NEW HAMPSHIRE ELECTION PROCEDURE MANUAL 2006-2007, PUBLISHED BY THE NH SECRETARY OF STATE.
[16] ibid. http://www.sos.state.nh.us/FINAL%20EPM%208-30-2006.pdf , retrieved from the Web January 4, 2008. Page 144 deals with sort and stack, called by the Manual, “SORT AND STACK BY CANDIDATE METHOD.” Page 147 deals with read and tally, called by the Manual, “ALL OFFICES, BALLOT-BY BALLOT METHOD.”
[17] ibid. Page 144. Retrieved from the Web November 4, 2007.
[18] op.cit. http://www.democracyfornewhampshire.com/files/Hand_count_training_D-fest_July_5_2007.pdf retrieved from the Web November 4, 2007.
[19] Telephone conversation with Anthony Stevens, August 26, 2008.
[20] op.cit. http://www.sos.nh.gov/FINAL%20EPM%208-30-2006.pdf , retrieved from the Web, January 4, 2008, page 144: “…State law authorizes the moderator to choose the system of hand counting to be used and to supervise the counting. RSA 659:60…However, neither state law nor the Secretary of State require that any particular system of counting be used. Moderators should ensure that that system of counting they adopt is accurate and efficient.” And again, page 149 “NOTE: This is a model describing how some moderators have chosen to count ballots. State law authorizes the moderator to choose the system of hand counting to be used and to supervise the counting. RSA 659:60. This model [ballot-by-ballot] is presented as one example of an acceptable practice, however, neither state law nor the secretary of state require that any particular system of counting be used. Moderators should ensure that the system of counting they adopt is accurate and efficient.”
[21] See Endnote 17.
[22] Telephone conversation with Ernie Vose, July 2, 2008.
[23] http://car.elpasoco.com/NR/rdonlyres/60904F6C-F63B-48AC-93A4-DBC9356F6982/0/Final_Report_.pdf , retrieved from the Web,June 30, 2008. A study called the Hand Counting Project was conducted by the Election Department of Colorado between January 6, 2008 and January 24, 2008, examining the read and tally & sort and stack methods. The study did not recommend either HCPB protocol - read and tally or sort and stack - but rather wanted readers to come to their own conclusions. Repeated phone calls and a requested email were never answered by the Town Clerk who directed this study
[24]Chain of custody of the ballots is very important in any HCPB method. However, should anything happen to the ballots themselves anywhere along the way, this written record would be an exact replication of the votes on the ballots themselves.
[25] op.cit. http://www.sos.nh.gov/FINAL%20EPM%208-30-2006.pdf, retrieved from the Web January 4, 2008, page 144.
[26] op. cit. http://www.opednews.com/articles/opedne_sheila_p_070718_on_site_observations.htm , retrieved from the Web January 4, 2008. In the three read and tally elections I observed, I was able to see the names of the individuals or yes/no as the counters made their tallies.
[27]From Maine Revised Statutes Annotated (MRSA), CONDUCT OF ELECTIONS, Chapter 9, page 3, (Title 21-A §695): “The team counts each lot together; 1 member reads and the other member tallies. The team members then switch roles, so that the tally is done a second time. If they agree, that count is completed. If there is a discrepancy, the team must recount the race or races where the count was off ….” The NH Manual, op. cit. page 144 http://www.sos.nh.gov/FINAL%20EPM%208-30-2006.pdf, retrieved from the Web January 4, 2008, states that “…This process [sort and stack] enables team members to simultaneously examine each mark on each ballot at least once, and to keep things simple by identifying choice in a single race at a time. If one team member makes a mistake, the other can catch it…”