The best-documented watching of elections turns out to be mostly done in the open: embassies sending staff to polling stations, and hedge funds buying private polls. The covert layer, which is what Daniel is really asking about, is thinner on paper than you'd guess, and what is on paper comes from leaked tasking orders, a Senate report and a great many denials.
Good place to start, because the question has a nice shape. In any big democratic election, some people care intensely about the process rather than the winner. Who are they, what do they collect, and who pushes back?
And he wants it kept apart from influence. We covered the pushing side already, how foreign powers quietly shape elections. That was about pushing. This one is about watching.
The line blurs in a couple of places and I'll flag it when it does. But take the cleanest case I know. In February of twenty seventeen, WikiLeaks published three CIA tasking orders about the French presidential election of twenty twelve. They run from November of twenty eleven to September of twenty twelve.
Three orders. So not a one-off curiosity.
A standing request. They ask for human sources and signals intelligence on all the major parties. What are the governing party's strategic election plans, how are the campaigns funded, what do the candidates think of the United States, where do they stand on Greek debt and the euro. And the documents are marked not releasable to foreign nationals, with a note about what they call friends-on-friends sensitivities.
Polite way of saying we are spying on an ally and would prefer it stayed off the dinner table.
One caveat, and it's a real one. The agency has never confirmed those documents, and I haven't seen anyone authenticate them independently. So treat it as a leak, not a verified record. But what it shows is collection aimed at the contest itself. Not who wins. What each side intends to do, and who is paying for it.
An election is the one time a government publishes its future intentions in advance and then argues about them on television. A foreign service wants the unpublished version.
Sure. Corn, would that surprise you, an ally doing it to an ally? Or do you assume it?
I assume it. I'd be more surprised if somebody told me nobody was reading the French campaign funding. I'd want a refund.
Fair enough. The odd part is the target list. If you wanted to know how the French government would behave on the euro, you could ask the French government, or read the newspapers. The tasking goes after the plans that aren't said aloud. The funding. The fallback positions. What the campaign would do if it won.
The things a candidate tells his donors and not the voters.
Which is also the material that's valuable to a rival, and that's the uncomfortable thing about collection in this territory. The same information works for a policy briefing and for a nudge. Nothing in the document says which was intended. I'd be guessing.
You'd be guessing, and you'd enjoy it.
A little. Corn, you said you assume it. Does it make a difference to you that it's an ally, or is that just manners?
Manners, mostly. An ally still has its own interests, and a presidential election is the moment they get rewritten. I'd rather the briefing existed than not.
Hm. I'm less comfortable than you are, but I can't say why I'd rather they didn't.
Daniel's first guess, though, was people on the ground. An officer working out whether the election is really free. Is that a thing?
Plausible, and I can't point you at a document that describes it as a distinct practice. What is documented is the overt version. In Kenya in two thousand seven, about a hundred and sixty American embassy staff, in fifty-six teams, went out to watch the vote. That came out through freedom of information releases, not a leak. Colombia the same year: roughly thirty officers, deployed under accreditation from the Organization of American States.
Accredited to watch? Out loud?
Out loud. Everyone knows the embassy is counting, and the embassy knows everyone knows. It's reporting, and the reports go home.
The secret agent in the polling booth is a bit of a movie.
Maybe there's one in there too. I'd just say the record shows a very large open apparatus, and an open apparatus is a decent place to stand if you wanted attention of the other kind. That's me inferring. I can't prove it.
And does watching even work? Having people there is one thing. Knowing is another.
Iran in two thousand nine is the cautionary case. Newsweek reported, citing anonymous officials, that most American intelligence analysts initially believed Ahmadinejad had won solidly, and that the agencies were caught off guard by the street protests after it. I haven't seen a declassified assessment, so that's a press account of what officials said.
A very large machine, and it read the room wrong.
Possibly. Or the analysts were right about the count and wrong about the mood, which are different failures. I can't tell from outside.
You can't tell from outside is going to be the subtitle of this episode.
It might be the subtitle of the whole genre.
Second guess. States running secret polls so they know the result before the exit polls come out.
I looked for that and I can't find a case. A state service quietly polling a foreign election to beat the exit poll... no. What exists is the open version. The State Department has an Office of Opinion Research inside its intelligence bureau, described as the government's main source of foreign public opinion research. It descends from the old United States Information Agency, which used to slip its own questions into commercial national polls abroad. A nineteen seventy-seven memo describes the practice.
Riders on somebody else's poll. Thrifty.
Very. But it's opinion research that anyone can read, not a secret read on election day.
Who does beat the exit poll?
Money. His third guess, and that one is documented. In the weeks before the Brexit referendum, the financial press reported that hedge funds and banks were commissioning their own polls, with updates on the day. Then in twenty eighteen Bloomberg Businessweek ran a long piece, The Brexit Short, saying funds had bought private polling and exit data from at least six firms, some of it arriving while people were still voting.
Legal?
Legal, as long as nothing was published before the polls closed. A pollster sells a client a number, the client keeps it, the client trades. The pound was above a dollar fifty the evening before the vote and fell to around one thirty-two.
Somebody had a number at ten at night that the rest of us learned at four in the morning.
An informed estimate, which isn't the same as knowing. But yes. And there was a row afterwards. Nicky Morgan, who chaired the Treasury Committee in parliament, wrote to Sir John Curtice, the polling professor who headed the British Polling Council. She worried about pollsters cross-subsidising their public work with private sales, and about market-sensitive data going to clients ahead of publication.
Curtice said what?
That the disclosure rule was effectively unenforceable, and told her to substantiate or withdraw the claim about perverse incentives. Months later the head of the Financial Conduct Authority said they were looking at guidance on inside information, not new rules. Whether that guidance ever appeared, I don't know.
The clean example of beating the exit poll is a private one. Not a state, a trading desk.
Hedge funds with a very narrow question. Which side wins, by how much, by when. Nothing about the process itself. Corn, should that be banned, or is it just information?
Just information, mostly. A pollster who keeps his data in a drawer until ten at night hasn't hurt anyone. The thing I'd worry about is the pollster whose public work gets cheaper because the private clients are paying for it. That's the Morgan point, and it's a fair one.
I'd go the other way slightly. If a market-moving number exists at nine in the evening and some people can buy it, that's a two-tier market in what the public is about to learn. Not illegal. Not nice either.
You've been wanting to say that for ten minutes.
Since the pound.
Hm. Here's what I can't settle. If you're watching an election, do you count only the people who vote, or the ones who stay home too?
Funny you'd ask. And now let's hear from the voicemail line, where Agnes has left us a message.
Agnes: Hello there, it is Agnes here, calling from the back porch of my wee house up past Dingwall, where I am listening with the wireless set turned up loud because the kettle drowns out the quiet bits. I wanted to say that your programme today, about folk watching other folk's elections, put me in mind of the polling day in two thousand and one when a young man stood at the end of the village hall with a clipboard and counted every single person going in and coming out, and my late husband said he was surely just the most thorough census taker we had ever had, and I have wondered ever since whether that man was counting us or something else entirely. So my question is this, and I hope it is not too daft: when these people are collecting up all the numbers of who voted, do they ever stop to count the ones who stayed home, or is it only the ones who go in that matter?
Not daft at all, Agnes. Turnout is voters divided by everyone registered, so the people who stayed home are in the sum by subtraction, and anyone serious about an election cares as much about who won't show as who will. Volunteers with clipboards outside polling stations are a familiar sight in Britain, and that's usually what they're tracking.
Who turns up is also why the next case matters. The best-documented example of a campaign's inside numbers going somewhere they shouldn't is Manafort and Kilimnik, and I'll go slowly, because the record stops short and people overshoot it.
Overshooting is your hobby.
It is not. Mostly. In the summer of twenty sixteen, Paul Manafort, then running the Trump campaign, had his deputy Rick Gates send Konstantin Kilimnik internal polling data. More than once. Robert Mueller's report says so, and says investigators couldn't determine what Kilimnik did with it.
Kilimnik is?
A business associate whom the Senate Intelligence Committee, in August of twenty twenty, called a Russian intelligence officer. It called Manafort sharing campaign information with him a grave counterintelligence threat. It also said it couldn't establish where the data went next, and that the lack of records wasn't dispositive.
That's a lawyer's way of saying we can't prove it and we wouldn't bet against it.
Then in April of twenty twenty-one the Treasury Department went further, in a sanctions announcement. It said Kilimnik had given Russian intelligence sensitive information on polling and campaign strategy. What evidence sits behind that isn't public.
Why is a poll worth stealing, though? I'd have said it's the least secret thing in a campaign. Everyone's polling.
A voter roll tells you who exists. A poll tells you where the campaign thinks it's weak, which states it believes it can win, where it's about to spend. It's the campaign's own map of itself. And a poll that models who actually turns out is worth far more than one that only records who people like.
The stay-at-homes are in the data.
Where the value is. If I wanted to know where to push, that's what I'd want: not the secret, the map. Which is also why a stolen poll and an influence campaign sit so close together. The collection is the first half of the push.
So Daniel's line between watching and pushing is a line on a staircase.
Mostly. It's still worth drawing, because the evidence for the two is very different. For the pushing we have indictments and takedowns. For the watching we have a record that stops at a name. Kilimnik got the data. After that the documents run out, and anything I said about what he did next would be me improvising.
Improvising from a donkey who has read the whole Senate report. Wouldn't be the first time.
Not the first time you've said it, either.
Now the blurry category. Collecting a poll is watching. Paying a politician to abstain is pushing. You said the line wobbles.
The nineteen seventy-two Brandt vote. A no-confidence motion against Chancellor Willy Brandt in the West German parliament needed two hundred and forty-nine votes and got two hundred and forty-seven. One reason it fell short: a Christian Democrat member, Julius Steiner, took fifty thousand marks from the East German Stasi to abstain.
Fifty thousand marks to move the Cold War by two votes.
That's influence, plain. Other names get attached to the story, and I'd treat those as allegations, not findings. Same with Taiwan. In late twenty twenty-three, two people were detained there and later indicted, accused of publishing fabricated presidential polls under direction from Chinese officials. The link was a trip to Xiamen paid for by a mainland newspaper, plus some messages prosecutors couldn't authenticate. I don't know how the trial ended. Taiwan's National Security Bureau also said in late twenty twenty-three that it was tracking money flowing from polling firms it suspected of mainland ties.
A fake poll isn't watching. It's a forged weather forecast.
It gets filed under election intelligence anyway, because the same offices tend to handle both. I keep it separate, but I understand why nobody else does.
Who pushes back on the watchers? That was the other half of what Daniel asked.
In the United States the answer got thinner in twenty twenty-five. The attorney general disbanded the FBI's Foreign Influence Task Force that February, and the FBI confirmed it. About a hundred and thirty people were laid off at the Cybersecurity and Infrastructure Security Agency, including roughly ten regional election-security specialists, and funding was halted that March for a center that shared election threat information with local officials.
To be fair to the people who did it, the argument as I understand it was that some of that work had drifted into policing speech, which is a fair complaint about parts of it.
That was the argument, yes. Critics say the cost was fewer people left watching. I haven't checked where any of it stands now, so call that the early twenty twenty-five picture. Corn, can a country do counterintelligence on its own election without it turning partisan? What's your instinct?
Slowly, and with someone watching. Whoever runs the election also runs the people who defend it, so the defence is always partly a political choice. I'd want it boring and independent. It never is.
Israel built an answer into the structure. During an election the Shin Bet operates under the authority of the chair of the Central Elections Committee, who is a Supreme Court justice, not a politician. In January of twenty nineteen its chief told a closed forum that a foreign state would try to interfere in the April vote.
Which state?
The name was gagged. Russia denied it was them anyway. Then in March, Channel Twelve reported that the Shin Bet suspected Iran of hacking the phone of Benny Gantz, the main challenger, and had told him about five weeks earlier. Gantz denied it, Iran denied it, and it was never officially confirmed.
The public record is a leak and two denials.
That's all of it. The duller lesson came the next year. A Likud-used campaign app exposed the details of roughly six and a half million voters, because the credentials were readable in the page source. A vulnerability, not espionage, and it isn't clear anyone downloaded the data. The point is who holds the register. Parties do.
The least cinematic spy story of the year.
Most of them are. The warnings ahead of this month's Israeli vote follow the same pattern: in July the Shin Bet director, David Zini, said Iran and others were trying to interfere, and in June the State Comptroller said the country was completely unprepared for influence operations. Both are mostly about influence, bots and fake profiles, not about collection.
Pushing, in other words. We're drifting back into last time's episode.
Which is the problem with these agencies. They announce what they can say in public, and the public part is the pushing. Now the group Daniel most clearly means when he says people who care about the process: observers. International missions are meant to be open. Member states nominate people, and the OSCE's election office, ODIHR, doesn't recruit them. So whatever vetting happens, happens at the sending end. In December of twenty twenty-three, media reported that a Russian diplomat named Aleksandr Studenikin, previously expelled from the EU, was among four Russians on the long-term observer list for Serbia. A senior European official told Radio Free Europe he was Russian foreign intelligence. He denied it.
Allegation, not proof.
With a named source and a denial. And it runs the other way. In September of twenty twenty-four, Maria Zakharova of the Russian foreign ministry said the United States would use ODIHR to interfere in Georgia's October election, and called the office a tool of the collective West. No evidence offered. Russia sends its own observers on those missions, and says it does so at the request of friendly countries.
Each side says the other side's observers are spies, and each side has a clipboard.
Hold both accusations at arm's length. The structure is the real thing. A polling station in a contested election has party agents, domestic monitors, an international mission, the press, a handful of people whose business nobody quite knows, and then the ones who are just...
Lampposts. Standing by the wall with nothing in their hands. It's not a handful. Eleven a hall, in the halls I counted, and I counted for the Circle for years on top of the gas board.
Eleven lampposts. That's a lot of street for one village hall.
The Circle ruled the clock over the door counts as a lamppost. It's on the books as a member. Four pounds fifty a year, and Arthur Pelling pays it. Whether the tea urn counts has been in the minutes since before I joined. I wanted a brass tally counter. Thirty-one pounds in the catalogue. Didn't buy it. I put tick marks on my shirt cuff, the wash took them on Sundays, so I counted in fours.
A cuff is a workable ledger if you read it before laundry. The clock is the harder case, because if it counts, every hall goes up by one and any year-on-year comparison needs a footnote.
A clock never misses a meeting and never speaks. Best attendance of any member I've heard of. But the urn is on all day, and it listens. The urn has the better case.
Pelling won't pay for two.
Lampposts is a better word than observers, for what I was describing. Most of them are in the room for reasons that have nothing to do with collection. And if it's eleven a hall rather than a handful, the vetting gap I mentioned gets a lot bigger. Eleven people nobody vetted is a bigger surface than four.
The question for anyone who does collect is which of the eleven files a report, and nobody outside can count that.
Nobody inside can either, most likely.
Do you think the covert version is bigger than the open one? Bet.
Bigger, by a margin I can't put a number on, and I'd rather say that than invent one.
Here's what stays with me. Sorted plainly, some of it is documented: the French tasking, the embassy teams, the hedge funds, the Kilimnik handover. Some of it is plausible and thin: officers judging freeness, observers as cover. And some, like a state secretly beating the exit poll, nobody has shown at all.
The pattern is that the watching is documented while what the watchers do with it mostly isn't. The record keeps stopping at a name.
Which leaves me wondering who counts the watchers. Everyone tallies the voters and the people who stayed home, and nobody tallies the ones standing by the wall. I'd like to see someone try it with a pencil and a shirt cuff.
Thanks to our producer, Hilbert Flumingtop, for putting the episode together.
This has been My Weird Prompts. There's more of the archive at my weird prompts dot com.
See you tomorrow.
