从乔治·华盛顿时代起，美国人就一直对选举预测着迷。一个多世纪以来，他们求助于博彩市场的预测，其中最大的一个市场就位于纽约证券交易所外。尽管人们普遍认为这些市场从不出错，但人们常常担心它们可能被操纵以影响选民，或者会给投注者提供直接影响选举的动机。
Americans have been fascinated with forecasting elections since George Washington’s time. For over a century they turned to predictions from betting markets, the largest one located just outside the New York Stock Exchange. While these markets were widely perceived never to be wrong, there was often a concern that they could be manipulated to sway voters or that they would give bettors an incentive to influence the election directly.
随着卡尔希和波利市场等在线预测市场的兴起，这些担忧再次出现。有人推动对它们进行监管甚至禁止，其中呼声最高的是州监管机构、像“更好的市场”这样的利益集团，以及参议员伊丽莎白·沃伦和约翰·柯蒂斯等政界人士。
These concerns have returned with the rise of online prediction markets like Kalshi and Polymarket. There has been a push to regulate or even ban them, with the loudest voices from state regulators, interest groups such as Better Markets, and politicians such as Sens. Elizabeth Warren and John Curtis.
目前缺少的是任何表明预测市场损害选举公正性的证据。据我们所知，没有这样的例子。我们最近研究了1900年至2024年的预测市场交易者，发现尽管参与者的人口结构存在偏差，但这些市场对政治偏见具有韧性。而且，任何有能力影响重要选举的人，在传统证券或衍生品交易中能赚到的钱要多得多。
Missing is any evidence that prediction markets diminish election integrity. To our knowledge there is no such example. We recently studied prediction market traders from 1900 to 2024 and found that these markets are resilient to political bias despite skewed demographics among participants. And anyone who has the means to affect important elections can make far more money trading in traditional securities or derivatives.
奇怪的是，民调——另一个主要的选举预测工具——却没有受到如此严格的审视。然而，与预测市场不同，我们现在已经有两个当前选举周期中虚假民调的例子。
Curiously, polls, the other leading forecaster of elections, haven’t received anything like this level of scrutiny. Yet unlike prediction markets, we now have two examples of fraudulent polls from the current election cycle.
今年夏天，民调机构Median Strategies被发现在伪造民调，最突出的是夸大了威斯康星州民主党州长初选中一名极左翼候选人的胜算。该公司称其正在进行一项“短期社会实验”，以考察“未经独立核实的所谓民调信息如何进入政治信息生态系统并传播”。上周，Wedgewood Polls承认其大部分民调数字也是一项“实验”。
Over the summer the pollster Median Strategies was found to be fabricating polls, most prominently overstating the chances of a far-left candidate in the Democratic primary for Wisconsin governor. The firm said it was conducting a “short-term social experiment” to examine “how purported polling information could enter and spread through the political information ecosystem without independent verification.”Last week Wedgewood Polls admitted most of its polling numbers were also an “experiment.”
而这一实验至少有一项有记录在案的影响，即改变了选举结果。州参议员马洛里·麦克莫罗退出了密歇根州民主党联邦参议员初选，原因是Wedgewood的假数字显示她落后。这导致捐款枯竭，工作人员士气低落。“这害死了我们，”她在X上写道。阿卜杜勒·埃尔-赛义德赢得了提名，目前正在为这一席位展开激烈角逐。Wedgewood的假民调可能就是原因之一。
This one had at least one documented effect in changing election outcomes. State Sen. Mallory McMorrow dropped out of Michigan’s Democratic U.S. Senate primary when Wedgewood’s fake numbers showed her lagging. This led donations to dry up and demoralized staffers. “It killed us,” she wrote on X. Abdul El-Sayed won the nomination and is in a tight race for the seat. Wedgewood’s phony poll may be one reason why.
受影响的不仅是政客。《纽约时报》在其民调汇总中使用了Wedgewood的数字，颇具影响力的民调分析师内特·西尔弗也是如此。这些假民调似乎有利于民主党人，尤其是偏左的民主党人：它们声称共和党在密西西比州参议员选举中仅以微弱优势领先，而预测市场认为共和党获胜的概率为90%。Wedgewood的民调塑造了关于中期选举、捐赠者行为、选民认知乃至投票的叙事。而这一切损害都来自一家今年夏天才起步的民调公司。
Not only politicians are affected. The New York Times used Wedgewood numbers in its poll aggregates, as did the influential polling analyst Nate Silver. The fake polls seemed to favor Democrats, and particularly left-leaning ones: They claimed the Republicans had a very narrow lead in the Mississippi Senate election, which prediction markets score as 90% likely to be a GOP win. Wedgewood’s polls have shaped narratives about the midterm elections, donor behaviors, and voter perceptions and perhaps even votes. All this damage from a polling company only started over the summer.
这类把戏并不新鲜：2006年，两名被告在被控欺诈雇用他们进行内部民调的竞选团队后，对联邦欺诈指控认罪。
Such shenanigans are nothing new: In 2006 two defendants pleaded guilty to federal fraud charges after being charged with defrauding campaigns that hired them to do internal polling.
与预测市场不同，假民调机构有腐蚀选举的确凿证据。他们成功骗过了一家重要报纸和一位顶尖民调专家，使其使用了他们的假数字。问题可能会变得更糟，因为恶意行为者看到了利用民调扭曲选举的轻而易举和巨大威力。
Fake pollsters, unlike prediction markets, have a smoking gun of corrupting elections. They have managed to trick an important newspaper and a leading polling expert into using their phony numbers. The problem could get worse, as malicious actors see the ease and power of using polls to distort elections.
民调行业自身无力解决这些问题，因此立法者应考虑对民调施加监管。要求对其数据接受审计。确保它们与既得利益集团之间没有未披露的关联。对欺骗性做法处以经济处罚。许多政界人士和倡导者曾对预测市场提出这类要求，如今他们应当把注意力转向唯一被证明会损害选举公正性的预测——民调。这是保护我们民主制度最起码该做的事。
The polling industry has been unable to address these issues on its own, so lawmakers should consider imposing regulations on polls. Require their numbers to be audited. Make sure they have no undisclosed ties to vested interests. Impose financial penalties for deceptive practices. The many politicians and advocates who have made such demands from prediction markets should turn their attention to the only forecasts that have proved to damage election integrity. It is the least we can do to protect our democracy.
克兰先生是罗格斯大学统计学教授。斯特伦普夫先生是维克森林大学经济学教授。
Mr. Crane is a professor of statistics at Rutgers University. Mr. Strumpf is a professor of economics at Wake Forest University.