HB 1035 Adds Statistical Complexity to Process of Verifying Statewide Petitions

Last week’s notice from the South Dakota Board of Elections brings to my attention an important change to petition law that escaped my attention during Session. This change means our Secretary of State will need to do more complicated math to certify statewide petitions in this election cycle.

House Bill 1035, as originally presented by the Board of Elections, weakened the petition process in two ways. Pre-HB 1035, proponents and opponents of petitions both had the opportunity to challenge the Secretary of State’s decision on petitions at the Secretary of State level, before having to go to court. HB 1035 takes away the right of petition proponents to challenge a Secretary of State’s rejection of a petition with an appeal to the Secretary of State’s office but leaves intact the right of petition opponents to challenge a petition certification at that level. HB 1035 thus leaves petition opponents with more legal recourse than petition proponents.

Additionally, HB 1035 takes away the opportunity for petition challengers to meet the deadline by submitting their affidavits electronically, then following up with mailed hard copies. By requiring that challengers deliver their notarized paper affidavits to the Secretary of State by the challenge deadline, HB 1035 (a) ignores the 21st century and (b) takes at least half of a working day away from petition challengers to review petitions for irregularities, who must now drive to Pierre to submit challenges. Particularly on candidate petitions, for which statute allows only five business days for challenges, that half-day is a significant loss.

The Legislature had no problem passing these provisions. But on March 6, the first day of the last week of the 2017 Session, Secretary of State Shantel Krebs brought an amendment to Senate State Affairs to add some math to her own petition certification process. NSU statistician Sara Schmidt had advised Secretary Krebs and the Board of Elections that the current method of certifying statewide petitions by sampling 5% of submitted signatures was not statistically rigorous. Instead, Schmidt recommended the following change to our petition-checking statutes:

The random sampling shall be an examination of five percent of the signatures received consisting of a number of signatures that is statistically correlative to not less than ninety-five percent level of confidence with a margin of error equal to not more than three and sixty-two one-hundredths percent [Amendment 1035fa, submitted to Senate State Affairs, 2017.03.06].

95% level of confidence, ±3.62% margin of error—these numbers come from the current standard of sampling 5% of 13,871 signatures, the minimum required for petitions for initiated and referred laws. The math isn’t hard; you can have the Internet do it for you.

Senate State Affairs stared glassy-eyed at math and unanimously approved the amendment and the bill. The only nay in the full Senate came from Senator Jeff Monroe, who probably had an allergic reaction to scientific rigor. He got over that allergy and made the Legislature’s support unanimous in the vote on the conference committee report.

HB 1035’s new statistical requirement reduces the number of signatures the Secretary of State will have to sample on nearly all ballot question petitions but increases the sample size for statewide candidate petitions. Check out the math:

While petitions for referred and initiated laws must have at least 13,871 signatures, initiated amendment petitions must have at least 27,741 signatures. The 5% rule thus compels the Secretary to sample twice as many signatures for a typical amendment petition as she does for a typical referendum or initiative petition. As I noted above, HB 1035’s statistical standard is based on the 13,871-signature threshold, so a 5% sample and the new statistical standard take the same roughly 700 signatures. But move up to 20,000 signatures (a pretty good safety cushion for referendum and initiative petitions), and while the 5% rule requires sapling 1,000 signatures, the statistical rule requires only 708. Jump up to a petition with 35,000 signatures (that’s a 26% cushion for an amendment petition), and the required sample drops from 1,750 to 718.

If this statistical requirement had been in effect in 2015, when the Secretary of State had to review two referendum petitions, four initiated law petitions, and four amendment petitions, the Secretary would have had to review 56% fewer signatures:

petition submitted signatures 5% sample 95%CI/±3.62ME sample
Glodt 52,817 2,641 723
Redistricting 40,400 2,020 720
36% rate cap 20,800 1,040 708
18% decoy 60,028 3,002 725
Anti-Corruption 25,216 1,261 713
Non-Partisan 39,182 1,960 720
Fair Share 30,810 1,541 716
MedMar 16,631 832 702
RL 19 16,715 836 703
RL 20 20,077 1,004 708
16,137 7,138

The Secretary spares herself some work on ballot question petitions, but she shifts that work to increased sampling on candidate petitions. Candidates need far fewer signatures than ballot question promoters to get on the statewide ballot; a 5% sample of candidate petitions thus will be proportionally smaller. The new statistical standard does not decrease linearly with total signature count as the straight 5% sample does. Assuming candidates submit the minimum number of signatures required plus a 20% cushion, the Secretary will have to count 3.6 times as many signatures on independent candidates’ petitions, 4.7 times as many on Republican petitions, and 9.2 times as many on Democratic petitions (remember, Shantel, don’t blame us Dems! you asked for the rule!).

petition signatures 5% sample 95%CI/±3.62ME sample
706 (DEM) 36 360
800 40 383
1,000 50 424
1,250 63 463
1,500 75 493
1,955 (GOP) 98 534
2,000 100 537
2,500 125 567
2,774 (IND) 139 580
3,000 150 590
3,500 175 607
4,000 200 620
13,871 (ref/init) 694 697
14,000 700 697
20,000 1000 708
25,000 1250 713
27,742 (amend) 1388 715
30,000 1500 716
35,000 1750 718
40,000 2000 720
50,000 2500 723
60,000 3000 725
70,000 3500 726

Since candidates have to gather far fewer signatures than ballot measure backers, the only way that HB 1035 forces the Secretary of State to do more work is if we have a cycle with fewer ballot measures and more candidates than usual. As it stands right now, if we get six petitions for initiated and referred laws, three petitions for amendments, three statewide Democratic candidates (let’s speculate on a primary for Governor but not for U.S. House), six statewide Republican candidates (three each for Governor and House), and one independent, I estimate HB 1035 would cut the Secretary’s sampling workload by about 11%.

Notice that neither random sampling provision, 5% or HB 1035 statistical, appears to apply to petitions for organizing a new political party. Those petitions currently require 6,936 signatures (2.5% of the 2014 gubernatorial vote), and if I’m reading the law right, their verification requires checking every line until the Secretary counts 6,936 valid signatures.

2 Responses to HB 1035 Adds Statistical Complexity to Process of Verifying Statewide Petitions

  1. Donald Pay

    Generally, I like putting the signature counting on a better statistical foundation, but I just have to wonder about some assumptions that go along with it.

    What is being sampled by the SOS are the signatures on the petitions. Those signatures are themselves a self-selected sub-set (or sampling) of the “qualified electors.”

    Further, SD has defined “qualified electors” as registered voters, and, I believe, bases valid signatures as being properly matched to voter registration data. But registration data is constantly in flux as people move. Certain subsets of “registered voters” (the poor, people with disabilities, city dwellers, students and young people generally, and the elderly) tend to move quite often compared to other subsets (established families, rural folks). There is a heterogeneous population of qualified electors, and the way the SOS qualifies or disqualifies petitions is biased against some parts of the population and biased for others.

    OK. So what happens when you get an initiative petition that appeals more to poor folks or students? The error rate goes up, not because of cheating or misrepresentation, but because of the bias involved in the registration system. The marijuana measure better secure a solid cushion, because I suspect a lot of young folks will sign, and use of the registration system will be bias the validation process. The new statistical framework might actually work against those initiatives.

    I was involved in the I&R process when every signature was counted up to the required number, and the bias of the registration system was not added in. It seems to me that some political scientist with a good background in statistics could study this, but for people collecting signatures, it means get as big a cushion as you can.

  2. Donald, you get me thinking. We don’t try to ease the burden on the Secretary of State and county auditors on Election Day by having them simply count up ballots, then take some statistically rigorous sample to calculate the outcome to some level of confidence within an acceptable margin of error; we count every ballot.

    Why don’t we count every signature… or at least return to the old system that you describe, of counting every valid signature until we reach the threshold (certified!) or run out of signatures (rejected!). Why should it be up to citizens and lawyers to take questionable results to court? Why not empower and require the Secretary of State to provide a definitive result instead of a statistical guess?

    If the workload is a problem, do what we do with statewide elections: farm the work out to the county auditors. Divide a 40,000-signature petition among 66 county auditors, and each office checks 606 signatures.

    Of course, adopt an electronic petition system like Denver has tried, and we instantly verify every signature.