Michigan’s election system does not count every vote

Alan Fox has helped run Michigan political campaigns for over 40 years. He is on the staff of Lansing-based Practical Political Consulting and has been a Democratic member of the Ingham County Board of Canvassers since 1993.

 

The partial recount of the 2016 presidential election in Michigan did not uncover any hacking or fraud. It uncovered the small human errors that accompany any large undertaking run by thousands of people who don’t get to practice very often. What it did reveal is something that everyone involved in Michigan elections has known for the past decade: The system we have does not count every vote cast.

According to data released Tuesday by the Michigan Bureau of Elections, the partial recount examined and retabulated nearly 2,100,000 ballots. The six candidates for president each gained votes. The net increase was 1,648 votes, or one of every 1,275 votes cast. All of the candidates lost votes in some precincts and gained in others so the gross change from the official total is far greater.

There is no evidence that ballots for any candidate were systematically uncounted or miscounted. Candidates gained additional votes roughly in proportion to their counted votes. But between human error and the shortcomings of the optical scan systems used in Michigan votes routinely go uncounted.

I won’t go into the sources of human error – they exist no matter what tabulation system is used. There are far fewer such errors now than there were when punch cards, lever machines or paper ballots were used.

The optical scan systems fall short for a variety of reasons. Some failures can be blamed on the age of the equipment; most scanners in the state have been in use for over a decade. The equipment failures, however, do not generally affect the scanners’ ability to count votes. They lead instead to misfeeds and jams that result in the widely reported count discrepancies that could render a precinct unrecountable under Michigan law.

The scanners, even when new, fall short because they cannot be programmed to spot every mark in the “target area” of a ballot and also to properly distinguish intentional markings from stray marks. When ballots are examined in a hand recount candidates generally gain votes because the human eye can spot and the human brain can interpret marks that would be ignored or misinterpreted by any scanning system.

In Ingham County’s presidential recount, the county Board of Canvassers was called upon to decide 21 ballots where the recounters’ determination had been contested by one of the observers.

Ingham County uses a connect-the-arrow form of the scanned ballot; other counties use a fill-the-bubble system. The legal standard is that for a vote to be counted there must be a mark of any sort in the area used to connect the arrow and the mark must be found to have been intentionally placed by the voter.

Nearly every one of the small number of ballots we examined had marks that had clearly been made intentionally, that did not connect the pieces of the arrow, but that did have portions of the marking in or very near the “target area.” In one case, the voter had circled the head of each arrow. Most of the circles went completely around the head, but the circle in the presidential vote was off-center and missed the target area by a pen-width. That is not a vote.

Several voters consistently colored over the head of each arrow in a scribble. If even a tiny portion of the scribble went beyond the arrow head and into the area between the parts of the arrow the vote was valid; if it did not we could not count it. The whole process involved use of magnifying glasses, pocket lights and straightedges. It could not possibly have been programmed.

These ballots, however, were less than 10 percent of the votes added to the candidates’ totals in Ingham County. In about 200 other cases the recount workers found and identified without controversy valid votes for president that must have been missed by the scanners. Put another way, about one out of every 700 votes cast for president in Ingham County was missed by our equipment but was easily identified and tabulated by human counters. That rate of failure by the equipment is unacceptable.

Every recount of optically scanned ballots in Michigan has found this pattern. Just this summer a recount in the Democratic primary for Redford Township Treasurer counted 32 more votes than the 4,981 originally tabulated. A recount in an Oakland County legislative district Republican primary added 42 votes to the 12,891 counted. In each case, both candidates gained votes.

Next year, local units of government will start purchasing new voting equipment. The intent is that new scanners will be in wide use for the 2018 elections. The new equipment is overdue. Whether it will solve the problem remains to be seen.