Chad -- From waste product to headline

Part of the Punched Card Collection
See also the Voting and Elections web pages
by Douglas W. Jones
THE UNIVERSITY OF IOWA Department of Computer Science

Copyright © 2000, revised 2002. This work may be transmitted or stored in electronic form on any computer attached to the Internet or World Wide Web so long as this notice is included in the copy. Individuals may make single copies for their own use. All other rights are reserved.

What is Chad?

Chad began as a collective noun, like chaff and wheat. An individual punching from paper or cardstock is a piece of chad, as with a grain of wheat or a flake of chaff. The plural form CHADS emerged very early, however, and implies the use of the singular CHAD to refer to an individual punching. Historical usage within IBM appears to have favored CHIPS as the term for punchings for punched cards, but it seems that paper-tape machines, particularly teletypes, have had chad boxes (to catch the chad under the paper tape punch mechanism) for years, although in the 1920's, the term used in that domain was punchings. U.S. Patent 1,884,755, filed on Oct. 16, 1930 by H. L. Krum for a "Coupon Printer" (actually, a teletype receiver) refers to a "receptacle or chad box ... to receive the chips cut from the edge of the tape." (See page 7, line 106.)

The term became common in patents for teletype machines a few years later. See, for example, U.S. Patent 2,213,475, filed on July 18, 1938 by A. H. Reiber for a "Telegraph Transmitter". Here, the reference is to a "chaff or chad chute" to collect the punchings from punched paper tape. By the 1960's, the term chad was widely used to describe the chips from keypunches, and every keypunch had a chad box fed by a chad chute.

The theory that chad is a back formation from the surname Chadless must have begun a joke, as did the theory that the name is an acronym for "card hole aggregate debris". There were chadless paper tape punches. They were designed so that the chad remained attached to the tape, an example of what later came to be called hanging chad, although in this case, it was deliberate and did not cause read errors. U.S. Patent 2,246,655 for a tape splicer, filed Feb. 20, 1940 by W. M. Bacon, contains this definition of chad on page 1, column 2:

There was never a chadless keypunch! The classic IBM punched card readers of the first half of the 20th century were prone to jams, and a piece of hanging chad, particularly one firmly hinged at one end, would have almost certainly guaranteed a jam. Only with the advent of the vacuum pick, as most fully developed in the Documation card reader in the early 1970's, did hanging chad become something that would not jam the reader and might even survive a trip through the mechanism.

The Votomatic (prescored) ballot

The hanging chad shown in Figure 2 was punched from a pre-scored punched-card ballot using a paperclip. In all cases, the resulting chad hangs in the trapdoor configuration, that is, hinged on two corners. In no case did the punch actually detatch the chad! The cleanly punched hole on the lower left was made by manually stripping the piece of chad from the card, while exactly 3 other holes in this photo were, in fact, punched, and then the chad was pushed back into the holes during subsequent handling.

When pre-scored cards are used with a Votomatic voting fixture, the pin or stylus used to punch the card is centered over the pre-scored piece of chad, so that it pushes down very near the exact middle of the rectangle while both ends of the rectangle are supported from the back by structures called T-strips. As a result, after the punch tears one end of the chad loose, the piece of chad is supposed to fold over the tip of the punch, so that the follow-through of the punch tears the other two corners of the chad loose. An animation of the punching process on a Votomatic machine is available.

**Figure 3:** Another perspective on hanging chad.

Many votes cast on a Votomatic punch may begin as hanging chad that is stripped from the ballot by the followthrough of the voting stylus. When the T-strip mechanism fails to do its job, the chad is usually stripped from the ballot as the ballot is removed from the fixture. The elastomeric T-strips used in a modern Votomatic voting machine provide enough back pressure on the piece of chad that the chad does not usually slide on the T-strips, and as a result, the paper fibers at all four corners usually tear free together. Where dangling chad is found, however, it most likely began as trapdoor chad that was partially removed during subsequent handling.

Seen from the back, it is not always easy to tell which pieces of chad have been pushed back into the holes from which they originally came, but from the front, under a 10x loupe, it is sometimes easy to tell. The key thing to look for is loose fibers sticking inward over the corners of the replaced pieces of chad. In a virgin punch position, where the chad has never been torn free, these fibers are tightly integrated into the paper. Once the chad is torn free, it is impossible to weave these fibers back into the body of the sheet of cardstock, so they must sit on the surface when the chad is pushed back into the hole.

It is worth noting that Peter K. Sheerin has also recommended use of magnification for the resolution of questions about chad. His web page, Dimpled Chads Under the Microscope, demonstrates how to distinguish between dimpled chad that was created using a voting stylus and that created using an ad-hoc stylus such as might be done by an election worker intent on fraud. The key elements he notes are the signature pinhole created by the microtip on the stylus and the imprint of the circular edge of the stylus. I have verified that these details are generally visible under a 10x loupe, although the pinhole created by the microtip is very hard to see if it is in the black spot printed in the center of each bit of chad on the standard Votomatic ballot.

None of my experiments with unsupported ballots produced what has come, since the Florida recounts of the year 2000, to be called pregnant chad. The example in Figure 4 was created, with some difficulty, using a Data-Punch machine, a Votomatic-style machine made by Election Data Corp. of St. Charles Illinois, with a microtip stylus. The mechanism of this model is substantially the same as that of the Votomatic machines that were subject to so much controversy in the year 2000.

It is noteworthy that the force needed to dislodge chad from the card was not sufficient to dimple or deform the chad in any of my experiments done outside of Data-Punch or Votomatic machines. Close inspection under a 10x magnifier did not even reveal scratches on the ballot surface! In all cases, the chad ripped free at the corners and remained almost flat.

The Data-Punch machine used to create the pregnant chad shown in Figure 4 was not new, the T-strips had some wear, but there was no chad in the fixture's chad box. Under these conditions, the force required to dimple the ballot (as measured by feel) was greater than the force needed to continue the downstroke and rip the chad free. Because of this, most attempts to create pregnant chad led to clean punches, and only when extreme care was taken to limit the stroke of the punching stylus, for example, by holding it with one fingertip resting on the guard sheet while punching, was it possible to create an example of pregnant chad. In some cases, the easiest ways to control the stroke of the stylus is to push it diagonally so that the friction on the guide plate helps control the depth of the stroke.

At the request of the Palm Beach Post, as reported in the January 13, 2001 edition, Professor Jose Villanueva of Florida Atlantic University conducted experiments on the force required to dimple and cleanly punch Votomatic ballots. He found that it takes, on the average, it takes 11.5 ounces of force on the punching stylus to create a clean punch using a Votomatic fixture, and 12.8 ounces using a Data-Punch fixture. The chad begins to fold over the tip of the stylus at about 6.4 ounces, and at around 10 ounces, the chad begins to tear loose at the corners.

In my experiments with a Data-Punch fixture, the center pin on the stylus began to leave an imprint on the chad at 4 or 5 ounces, and the chad remained bent after being subject to stylus forces of from 8 to 9 ounces. At 11 to 12 ounces, the chad remained bent enough that light was visible around the edges; the force required to bend the chad this far was higher when the stylus was slightly off to the side of the rectangle of chad. At 12 ounces, on all cards tested, it was clearly evident that a crease was starting to develop prependicular to the long axis of the chad through the imprint of the pinpoint. At 15 to 16 ounces 2 corners of the chad tore loose; when the stylus was off center to one side or the other the two corners nearest the stylus tore. When the punch was centered, the other two corners tore loose at 16 ounces.

Because the backing strips on my Data-punch fixture were old and hard, it took 12 to 14 ounces to drive the punch between the T-strips without a card in the fixture. With a card, it took 19 to 21 ounces, depending on the local state of the T-strips.

When the stylus was centered, punching forces of over 16 ounces that did not drive the chad through the T-strips almost always tore all four corners, so that the chand could be blown out of the card. Sometimes, a few fibers remained at one corner, producing a very loose dangling bit of chad. When the stylus was off center, though, only two corners generally tore. On the other hand, at forces above 20 ounces, the even where the chad remained in the card, the round imprint of the edges of the stylus was visible under 10x magnification as a pair of "crescent moon" imprints on opposite sides of the central dimple.

It takes well over 30 ounces of force, well above what I could apply with my experimental apparatus, to drive the stylus through the score line around a piece of chad or to drive the stylus through the card away from a pre-scored piece of chad. At 30 ounces, the tip of the stylus was deeply imprinted in the card.

A properly functioning Votomatic punching fixture should fold the chad over the tip of the stylus The pregnant chad in the photograph above illustrates the beginning of such a fold. A simple fold in paper or cardstock generally takes less force to produce than a dimple with radial symmetry, and in fact, no such dimples were produced using the Data-Punch fixture in these experiments.

The Risk of Chad Jams

Given these observations, it seems unlikely that a significant number of voters would create pregnant chad on a properly maintained Votomatic machine. The only abnormal circumstances I can imagine that might lead to this outcome involve some kind of back pressure against the ballot, supporting the rectangle of chad so that it cannot tear free from the surrounding cardboard while the voter pushes from above. The obvious source for such back pressure is a pile of chad behind one voting position on the ballot blocking attempts to punch that position.

**Figure 5:** The mechanism of a Votomatic punch, seen from the back.

A story published in the Fort Lauderdale Sun Sentinel on December 1, 2000, "Minority, senior votes are most rejected" by Stacey Singer, John Maines and Scot Wyman presents some evidence that such piles of chad were problems in some south-Florida polling places during the contested presidential election. The article went on to quote David Beirne, of the Broward County Florida election office, who stated that the chad box on each Votomatic punch should be able to hold on the order of 50,000 bits of chad.

Looking into this, I disassembled a Votomatic voting machine made by Computer Election Systems Incorporated, and found that it has a horizontal steel brace across the back of the mechanism. This brace is behind punch positions 4 and 5 (assigned, respectively, to Pat Buchanan and Al Gore on the now infamous Palm Beach County butterfly ballot). As a result, while there was room for a huge volume of chad behind the other punch positions, probably more than the 50,000 pieces quoted, it seemed that it might be possible to block these positions after only a few hundred votes! This could explain the complaints from voters about their need to repeatedly force the voting stylus before it would punch through the card.

In casting 450 experimental votes in position 5 on a 228 position Votomatic ballot, starting with a clean Votomatic machine, there were no jams that required hammering, but at several points during the experiment, there was a small added resistance to punching as minor chad jams were pushed aside. When the machine was opened after the experiment, there was a large but loosely packed body of chad between the support brace and the bottom sides of the T-strips, and chad was wedged between the T-strips for a distance of several centimeters above and below ballot position 5.

**Figure 6:** A chad jam in a Votomatic punch, seen from behind.

Repeating the experiment using position 119 on a 228 position Votomatic ballot, an impenetrable chad jam developed after 317 votes. Positions 43, 119 and 195 are midway between positions 4 and 5 along the long dimension of a 228 position Votomatic ballot. The metal brace across the back of the machine was not the cause of of this jam, but rather, the plastic brace above the metal brace. This jam is shown in Figure 6, as it looked prior to attempting to clear it by hammering. Both the edge of the metal brace and the plastic brace are visible.

This jam would not yield to hammering, but hammering compacted it enough that the next 39 votes were cast successfully. Inserting or removing the ballot card sometimes pulled the top few pieces of chad off of the jam, and in many cases where it was obvious that the chad punched into the jam was not removed from the card by the punch, it did pull free as the card was pulled out of the machine.

Repeating this experiment on position 100 of a 228 position ballot using a Data-Punch machine gave a firm jam that produced a similar quality of dimpled chad after 668 punches. Positions 24, 100 and 176 have the same relationship to one of the plastic cross braces inside the Data-Punch machine as position 119 has to a brace inside the Votomatic machine.

On both machines, some chad occasionally pulled loose from the top of the chad jam as the card was removed; this frequently remained in the mechanism above the T-strips. On inserting the next ballot, an attempt to punch through both the ballot and this loose chad could be more difficult than normal. Professor Jose Villanueva of Florida Atlantic University also speculated about this in his remarks published by the Palm Beach Post, on January 13, 2001, but I did not observe this in my experiments.

Over the next hundred votes, the jam never completely disappeared, and several extreme examples of pregnant chad emerged, as illustrated in Figure 7.

**Figure 7:** Pregnant chad resulting from punching into a firmly packed chad jam.

It is noteworthy that the piece of chad shown in Figure 7 remains well attached at three corners, with a few paper fibers still connected at the fourth corner. This brings into question the most widely used objective standard for what constitutes a vote in a hand recount of Votomatic ballots, that the vote only counts if the chad is detatched by 2 or more corners. Note that the bulge created by the punch is almost spherical, quite unlike the cleanly folded examples of pregnant chad that were produced by pulling back on the punch before it completely dislodged the chad, as in Figure 4.

Figure 8 shows the positions of the internal braces in the Votomatic and Data-Punch voting fixtures, relative to the layout of the 228 position Votomatic ballot. Note that the Data-Punch fixture has more internal braces than the Votomatic fixture, and all but one of these braces are wider than those in the Votomatic fixture. On the other hand, the Data-Punch fixture does not have the broad metal brace behind the top group of plastic braces. Both the metal and plastic braces are visible in Figure 6, and the position of the metal brace is lightly shaded in Figure 8.

**Figure 8:** Positions of obstructions inside the Votomatic (left) and Data Punch (right) mechanisms, relative to a 228 position ballot.

All of these formats conform to the layout for standard punched cards used in data processing, so the best way to describe the ballot locations that are prone to jams is in terms of the universal data processing card and not in terms of the specific ballot format. A standard punched card used in data processing punched card is seen as having 12 rows of 80 columns, where the columns are numbered from left to right. In the voting application, the card is turned 90 degrees, so the column numbers run from bottom to top. These are visible on the right edge of each ballot shown in Figure 8. Table 1 summarizes the problem ballot positions in terms of the standard punched-card column numbering:

**Table 1:** Relationship of punched card columns to internal braces in the punch mechanism
	Votomatic	Data-Punch
Over a brace	64,67	25,29,30,43,47,48,60,61,65
Partly over a brace	33,34,36,37	24,26,42,44,64
Adjacent to a brace	63,65,66,68	28,31,46,49,59,62,66

It should be straightforward to duplicate the chad-jam results obtained here for any of the voting positions directly over a brace, although the actual number of punches required to recreate such a jam should be expected to vary over a fairly wide range from one experiment to the next. Although my experiments did not produce jams for positions partly over a brace or immediately adjacent to a brace, these experiments were far from exhaustive, and is probably a risk of jams in those positions.

Therefore, ballot designers should probably avoid placing voting positions in any of the ballot positions listed in Table 1, unless careful measures are taken to prevent jams, and after a close election, candidates with voting positions adjacent to or over a brace have strong reason to investigate the machines used for evidence of chad jams.

Chad Jams in Election 2000

Were chad jams a problem in the contesed Florida general election in 2000? There is no conclusive evidence. The Palm Beach Post, has conducted a fairly careful survey of the dimpled and hanging chad found on ballots in Palm Beach. Their evidence is summarized in Table 2:

**Table 2:** Dimples per vote in the Palm Beach County 2000 General Election
Candidate	Ballot Position	Card Column	Dimples d	Votes v	Rate d / v
Bush	3	70	2,476	152,951	1.6%
Buchanan	4	66*	188	3,411	5.5%
Gore	5	62*	3,260	269,732	1.2%
McReynolds	6	58	37	302	12.3%
Browne	7	54	74	743	10.0%
Phillips	8	50	126	190	66.3%
Nader	9	46*	103	5,565	1.9%
Moorehead	10	42*	64	104	61.5%
Harris	11	38	30	45	66.7%

Hagelin	13	30*	14	143	9.8%

The card columns marked with stars in Table 2 are adjacent to or over braces in one or another of the voting mechanisms used in Palm Beach County. Note that the Dimples column is a count of the total number of dimpled or hanging chad found for that candidate.

If dimpled and hanging chad in this race was largely due to chad jams of the type I have observed, we would expect the rate of problems for Buchanan, Gore, Nader, Moorehead, and Hagelin to be relatively high compared to the others. No such trend is visible in this data! Instead, we see the largest number of dimpled or hanging bits of chad for the candidates that received the most votes, and at the same time, we see the lowest rates for the popular candidates.

In fact, the pattern we observe here is quite interesting! Voters casting votes for relatively unpopular candidates had the greatest trouble with dimpled or hanging chad! It looks as if fully 1/3 of the few voters attempting to vote for Phillips were unable to properly punch their voting position, despite the fact that there was no brace to obstruct this voting position, while voters preferring the two popular candidates, Bush and Gore, were almost always successful.

This leads to the hypothesis that voters who prefer minor-party candidates are generally inept, or alternately, that regular and frequent use of a voting position clears problems with that voting position. If the latter is the case, it is possible that there were some kinds of obstructions in the machines other than the brace-related chad jams discussed here, but we have insufficient evidence to do more than speculate about such jams.

The Palm Beach Post reported, on January 13, 2001, that the Data-Punch mechanism was far more prone to dimpled chad than the Votomatic mechanism. Unfortunately, I have not had a chance to find and study the Palm Beach vote totals and dimple counts broken down by machine type. This would be interesting.

Prior to each election using any voting technology, it is prudent to perform a pre-election test to make sure that the system works. With the Votomatic system, the prudent test is to insert a test ballot in the voting fixture and verify that each voting position can be punched. This was done in Palm Beach County, and the Palm Beach Post has examined the test ballots. Remarkably, there were many test ballots with dimpled or dangling chad, but apparently, none of the election workers had been instructed to check for this. The result of this study is summarized in Table 3:

**Table 3:** Dimpled and hanging chad from the Palm Beach County 2000 pre-election testing
Candidate	Ballot Position	Card Column	Votomatic Dimples	Data-Vote Dimples
Bush	3	70	12	31
Buchanan	4	66	near 9	near 29
Gore	5	62	8	near 26
McReynolds	6	58	15	29
Browne	7	54	9	24
Phillips	8	50	12	27
Nader	9	46	11	near 22
Moorehead	10	42	7	partly 22
Harris	11	38	4	14

Hagelin	13	30	8	over 18

The notes in the Dimples columns indicate the relationship between this punch position and a brace within the mechanism. Note again that the Dimples columns count both dimpled or hanging chad found on the test ballots positions for that candidate.

Table 3 includes data from 4,867 test ballots, although there is evidence that, on some test ballots, the Palm Beach Post reported in the December 9, 2001 edition, that there were 65 ballots showing no evidence of any attempt to punch any of the 10 presidential voting positions. Only 76 precincts used Data-Punch mechanisms, while 455 precincts used Votomatic mechanisms. Unfortunately, I do not have the number of machines themselves, but even without this, it is extremely clear from Table 3 that the Data-Vote mechanism was far more likely to lead to problems than the Votomatic mechanism.

The pattern in Table 3 is quite different from that in Table 2, but we expect this! Exactly one test ballot was voted on each voting machine, and each test ballot was supposed to have been punched in all 10 presidential voting positions, as well as in all of the local office voting positions. Therefore, if a machine was badly jammed in some voting position, it would contribute only one dimple or hanging chad to Table 3, while it could contribute hundreds of dimples to the totals in Table 2.

Table 3 shows no particular preponderance of problems on the ballot positions near or over the braces in the machine; in fact, at first glance, it appears to show a fairly uniform distribution of problems over the entire ballot. In fact, however, the frequency of hanging or dimpled chad is unformly higher for voting positions at the top of the ballot than for positions at the bottom of the ballot! Overall, for both the Votomatic and Data-Punch mechanism, problems were about 35% more common for the 5 candidates located at the top of the ballot relative to the 5 candidates at the bottom. This appears to be significant, but we have no evidence to explain it.

The most significant finding coming out of the investigation conducted by Palm Beach Post as reported in the December 31, 2001 edition, is that the maintenance standards for the Votomatic and Data-Vote machines were different! Votomatic machines were opened up once every election cycle (possibly only every 2 years) and vacuumed, while Data-Vote machines were not opened.

Clearing Chad Jams

I experimented with clearing out the chad jams I made, and found that following the instructions molded onto the inside of the Data-Punch mechanism were not sufficient! On this mechanism, there is a small door on the back for chad removal. On opening this door, the message "SHAKE WELL TO REMOVE CHADS" is exposed. In fact, shaking the mechanism does not remove any of the chad wedged in the slots between the supports for the T-strips, and it certainly does not remove chad that is jammed above the cross braces that connect these supports.

Thus, following these instructions does not remove any of the chad that conributed to the chad jams that I observed! Even a few hard thumps does nothing! Only when I held the Data-Punch mechanism in my hands like a hammer and banged it as hard as I could on my desk did the "loose" chad between the T-strip supports begin to come loose, and it took a series of 3 or 4 hard bangs to release most of this chad. Continued hard banging began to release the jammed chad over the cross braces, but the jams were not fully released until 8 to 10 hard bangs.

My experiments strongly suggest that a few hundred votes will not jam these mechanisms, and that is the most that would be expected in any one election. However, also I conclude that if a machine is routinely well shaken to remove chad, following the Data-Punch instructions, chad will accumulate in the jam prone positions from one election to the next until impenetrable chad jams form!

Recommendations

Therefore, I strongly recommend a more vigorous approach to cleaning these machines. Chad jams can be raked out from the front by inserting the voting stylus (or any similar tool) between the top ends of the T-strips of the machine and dragging it down to the bottom. A rake-shaped tool with 12 tines would be a handy improvement, cutting the time taken to rake out the chad by a factor of 12. If the chad is raked out before each election, chad jams should not be a problem!

In light of my findings, it seems that dimpled chad should generally be counted in a hand recount, particularly if the dimple looks like a section of a sphere, and this should not depend on whether votes cast on other positions on the same ballot were also dimpled, since an obstruction providing back pressure to one part of the ballot may not provide comparable back pressure to another. Special consideration should be given to dimpled chad found in positions near the horizontal braces on the underside of the punch block, and ballot layouts should generally avoid those positions!

It would seem to me that those involved in hand recounts, including observers and judges, should be given opportunities to punch and attempt to dimple ballots using clean, properly maintained voting fixtures and ballots identical to those used in the election, and they should be shown the inside of the mechanism and the relationship between the horizontal braces and the voting positions on the ballot that was used, and they should be given access to magnifiers or low-power microscopes in order to inspect the dimples and to look for loose fibers around bits of chad that may have been pushed back into the ballot.

Furthermore, if a significant number of ballots show dimples after an election, it would seem essential to investigate the voting fixtures on which those ballots were cast, the procedures for cleaning the machines, and the procedures for assuring that they remain free of obstructions during the election. The above-cited Sun Sentinel article suggests that, during the contested Florida election, the standards for voting machine maintenance and cleaning were far from uniform. Some precinct workers even mentioned routine shaking or banging to remove accumulated chad, but jams such as that shown in Figure 6 were difficult to dislodge by such banging!