“The U.S. election was a stunning triumph for democracy at a moment when it was sorely needed.” Professor Clifford Orwin, The Globe and Mail

If political polling is an effort to use the understanding of preferences to predict voting patterns, it is the voting itself that is a cornerstone of democracy. As long as polling is merely a matter of recording and possibly influencing voting, it is not contrary to democracy. In fact, as Ira Basin informed me, George Gallup and Saul Rae (Bob Rae’s father) wrote a book in 1940 called The Pulse of Democracy in which they argued that, “public opinion polling would be the saviour of democracy because it would allow leaders, for the first time, to truly know what ordinary citizens were thinking, as opposed to the rich and powerful who’s voices could always be heard.” Polling can enhance rational choice. Polls can also be deliberately distorted to influence a vote. If so, that is a serious step in undermining democracy. But it may not be illegal. It may only be an undermining of democratic norms.

Was polling misused to distort voting? Can polling provide an answer and explain the results of the vote in one swing state, Michigan? By a margin of 47.2% Democratic to 40.1% Republican, Michigan voters seemed to prefer the Democratic Party to control the U.S. Senate since, in 2018, Debbie Stabenow (Democrat) was re-elected by that margin. In the polling for the Michigan 2020 Senate race between Democrat incumbent Gary Peters and Republican challenger John James, in the WDIV/Detroit News poll leading up to the Michigan senate race, Gary Peters lead John James by a margin of 48.4%-38.8%; 8.7% of voters were undecided. It was very similar to the lead shown for Joe Biden and Kamala Harris. But Peters seemed to be running behind Trump by 2.8% of voters. However, the final result of the 2020 Michigan senate race was razor thin, 49.8% for Gary Peters (2,721,207 votes) and 48.3% for John James (2,636,892). Why were the polls so wrong in the Senate as well as the presidential race?

Look at one key county, Macomb in eastern Michigan and part of the more prosperous northern metro Detroit. Obama carried the county. So did Governor Gretchen Whitmer, though by a smaller margin. But Hillary Clinton lost it in 2016 by 12 points and in 2020 Joe Biden lost it by 8, though both got more votes the closer they got to the city of Detroit. On the other hand, Trump, though he got 40,000 more votes from the county in 2020 compared to 2016, received a smaller percentage. Even in a county he lost, Biden improved his vote among suburban voters, particularly female suburban voters.

Go back one election for a clearer example of intermixing polling manipulation and competition for votes. In 2018 when Democratic Senator Debbie Stabenow stood for re-election, a poll ascribed to Delphi Analytics published in The Washington Free Beacon showed that rapper Kid Rock, then purportedly considering the Republican nomination, if put up as a competitor to Senator Stabenow, would win 54% of the vote compared to Stabenow’s 46%. This was Kid Rock who had sung, “If I was president of the good ol’ USA, you know I’d turn our churches into strip clubs and watch the whole world pray.” The poll went viral.

The poll was credited to Delphi Analytics. Was that organization a legitimate political pollster? It would seem not from its own website.

Delphi Analytics is a business centered around data, around analytics and around value. We are focused on problem solving for global customers with big-data and hard-to-manage problems. Although we began in the United States in the early 1990’s with an emphasis on the financial markets – we’ve now established a 25-year history in Europe developing markets with both financial and trade-related solutions.

Or, in another version of itself:

Delphi Analytics is a technology company that utilizes a variety of data analytics, models and machine learning to generate algorithms to increase predictability and outcomes in trading and corporate financial risk management. Delphi constructs proprietary algorithms and indices (termed AiCi’s = Artificial Intelligence Confidence Indices) as well as develops AiCi’s for specific/custom use.

Whatever its record in large scale data analysis, the company seemed to have no background in political polling. So why was it cited in a Michigan non-race for a Senate in the summer of 2018 for a candidate who never ran? In the early race for the Senate seat in Michigan in 2020, John James (Republican) was running against the Democratic incumbent, Greg Peters. John James was Black; Greg Peters was White. The Republicans had a plan to wean Democrat-held Senate (and house) seats by running women and minority candidates in Democratic strongholds.

It was very apparent to many that the 2018 poll showing that Kid Rock would beat Stabenow was not a legitimate poll; it is even very doubtful than any poll was ever conducted. Was it a form of mock poll? Possibly. Was it used to publicize Kid Rock so his new record release would sell better? Possibly. However, it is more likely that the exercise was not intended to elicit preferences, but to see if the voting population was fluid. Kid Rock was simply a place holder with significant name recognition in the Black community, particularly in Detroit. Many who read the poll took it seriously. (Further, there were some allegations that the poll was used to make money on internet election betting sites. If so, that did not seem to be the prime intention of the poll, but an indirect benefit.)

Voting may be sacrosanct. Voting may indeed be a cornerstone of democracy. But manipulating voting is as old as the practice of voting itself. For the object is to get more votes for your side than opponents get for theirs’. There can be undemocratic voting – when there is only one permitted candidate, for example. In single party states, we often encounter “guided” voting. Or, as in Iran, there may be candidates who compete, but only if they are approved by the religious establishment. But in North America, winning is as much a transactional exercise as an expression of the will of the people.  

I was the guest of one of the candidates trying to win the Progressive Conservative nomination in a suburban riding in Toronto. There were three males competing for the nomination – the establishment WASP candidate who had run and lost twice before and two other candidates from two different ethnic groups with a significant presence in the riding, but neither was anywhere near a majority. One of the ethnic candidates had signed up virtually the whole of his large church as members of the PC party. The other ethnic candidate had signed up large numbers of ethnic Canadian citizens of the same ethnicity as himself from his riding, but they did not have a common institutional affiliation. In the first round of voting, the WASP candidate ran third.

The non-churched ethnic candidate, who had not been able to sign up as many members as the churched one, had made a deal that if he or the WASP candidate ran third (neither expected to run first on the initial ballot), that candidate would ask his supporters to vote for the other. The WASP candidate ran third and did as he had agreed to do. But the non-churched ethnic candidate saw that even with the supporters of both candidates who ran 2^nd and 3^rd, the church candidate would still win. The non-churched candidate made a deal with a candidate with a different ethnicity running in the riding next door. Each would do their best to get their ethnic compatriots to join the PC party in the respective adjacent ridings and ask those new members to vote for the ethnic candidate whose ethnicity they did not share. They had traded ethnic support with the adjacent ridings. Who said that politics was not akin to a sporting match?

There were two serious problems. The WASP candidate had little control over his supporters compared to the ethnic candidate. Second, the neighbouring ethnic candidate, who won his nomination, did not put in as much effort into getting his fellow ethnic members to sign up for the PC party in the adjacent riding. The Church candidate won over the non-church one by 800 to 600 in a constituency of about 80,000 voters.

Democratic party professionals know that this is how the system works on the ground level. So do Republican party professionals. The selection of the candidate by a relatively small group is the first key step. Then advertising, phone banking, text banking, door-knocking, debates and now the extensive use of social media kick in. The Republicans in Michigan failed to take into account the discipline and organization, largely in the hands of Black women in Detroit. James as a Black was expected to make significant inroads into the black vote; he only won 10% of the votes of Blacks in Detroit according to exit polls. That is the major reason why Peters won. That is why the Biden/Harris ticket won in Michigan.  

Did anyone breach democratic norms? No, everyone had competed according to the same norms for getting the vote. The democratic norms were not simply about the right to vote but about how best to capture that vote. Political scientists have bewailed the weakening of norms as a result of Trump’s assault on the norms of democracy (see my next blog). That determination to win at all costs posed a danger to democratic institutions. However, whatever the case for such charges and worries, generally on the ground, the election was run according to democratic norms that govern both voting and capturing the vote. The voting was very broadly conducted securely, honestly, fairly and transparently. The Republican assault and charges that those norms had been broken did not appear to have any substance. But the charges may play a roll in the exercise of manipulation for control of the Trump party between 2020 and 2024.

Take another state and another county, this time in Texas that Trump won. Democrats expected and counted on making significant inroads with the Latino population in Texas, overwhelmingly of Mexican background. This was especially true since Donald Trump had insulted Mexicans. The plight of asylum seekers in the state, and especially the images of child separation, were expected to enhance the Latino vote in Mexico for the Democrats.

But that is not what happened. Rio Grande Valley with a huge Latino population went for Trump in increasing numbers, though not generally a majority. Though Biden came within 6 points of winning the solid red state of Texas, for the first time since Lyndon Baines Johnson, Trump, a Republican candidate, won in blue Zapata County. Further, in Hidalgo County that Hillary Clinton won by 40 points, Biden’s victory was cut to 17 points. He had dramatically smaller numbers than Clinton along the 1,200-mile Rio Grande border. Thus, in spite of – or perhaps because of – Trump’s treatment of new Mexican immigrants, and given the absence of any significant federal infrastructure investment in border areas during Trump’s four years, one might expect an increased support for Biden. In fact, that vote as a percentage declined. Trump ran ahead of past GOP numbers, offsetting Biden gains in suburban Texas. Trump won more votes than George W, Bush who had specifically reached out for the Latino vote and supported immigration reform.

The proponent of the wall, the insulter of Mexican immigrants, had gained votes among rural Latinos. A combination of factors was at work. Many Latinos feared the competition for jobs from the new arrivals. It reminded me of 1979 when we were active in the private sponsorship movement of the initial wave of ethnic Chinese fleeing Vietnam. Other than the Hong Kong professional class who had immigrated to Toronto, we found that we had weak support and even opposition from the waiters in Chinatown. They explicitly feared economic competition for their jobs.

Trump had an additional appeal to Latino males versus the urban and suburban educated Latinos which increasingly supported the Democratic Party. Ethnic Mexicans worked in the oil fields of West Texas and Biden promised a switch out of fossil fuels. Ethnic Mexicans were not involved in the high-tech construction of windmills to capture energy from the winds of west Texas. Trump had the image of a regular guy; he did not read. He was macho in his treatment of women. He supported a gun culture. Many Latinos had switched from Catholicism to evangelical Christianity. They often worked as Border Patrol and U.S. Customs agents. Thus, in Hidalgo County, Biden got 127,507 votes, 8,698 over Hillary’s total, but Trump got 89,991 votes, 41,349 over his 2016 vote. Trump lost but he won. He cut into Biden’s majority so severely that Biden could not catch up in the rest of the state. Instead of the counties along the Rio Grande swinging more for Biden, Trump increased his percentage intake of the Latino vote and denied Biden a victory in Texas and a chance to turn a previously solid red state blue.

Polls did not capture the great diversity in education, skills, generations, gender, degree of assimilation and immigration history of Latinos in Texas where the ground game from county to county was critical. Republicans had concentrated on raising the profile of Hispanics in the party. Aron Peña in Hidalgo, a former Democratic campaign worker, became disaffected at his and his compatriots lack of recognition and switched to the Republicans. Where Democrats had run campaigns that stressed local engagement and status, several counties went for Bernie Sanders in the primaries. Overall, however, the pros in the Democratic Party had done too little and what they did came very late. Beto O’Rourke had failed to convince the party to change its strategy in winning the Latino vote. Neglect does not win support. Many Latinos were alienated rather than engaged and Trump attracted alienated voters.

The United States proved to be a robust democracy in spite of Donald Trump’s authoritarian propensities. More people cast ballots than ever before, even for the loser, Donald Trump. There was no significant violence. There was no significant intimidation. And there was almost no fraud. The norms of voting and of recruiting voters had by and large been followed.