Politics on Big Muddy

Adjusting for recalled past vote in political polling

Sun, 12 Apr 2026 22:55:00 -0400

The founder of Abacus Data, a Canadian polling firm, dropped kind of an interesting URL yesterday: abacus-weighting.com. It’s a advertisement in the form of a case study on why Abacus weights their political polls on past vote. It fits perfectly with the theme of yesterday’s post on how pollster’s get different results from the same data (the answer is they weight the raw data differently).

If you follow Nate Silver (or American political polling in general), you probably know that pollsters undercounted Trump support in all three elections where he was on the ballot. What I learned from this post is that support for the Conservative Party of Canada has been underestimated in their firm’s polling data in every polling wave for every election since 2011:

In every single wave, across every single election cycle, Conservative voters are underrepresented in our demographically weighted sample relative to their actual share of the vote. Not in most waves. Not in some elections. In every case we can observe.

Weighting for recalled past vote improves the estimate in every case, sometimes dramatically so:

In every election, past vote weighting moved our Conservative estimates upward and our Liberal estimates downward — consistently in the direction of the actual result. The 2021 election shows the most dramatic correction: a 7-point improvement in our Conservative estimate.

How do pollsters get different results from the same data?

Sat, 11 Apr 2026 22:36:00 -0400

Nate Silver linked to this throwback article from 2016 in The New York Times in his recent article on fake AI polls, which I also wrote about a few days ago. The article, entitled “We Gave Four Good Pollsters the Same Raw Data. They Had Four Different Results.” is a good reminder that modern polling diverges very far from the theoretical ideal of a simple random sample. Even after deciding on a methodology to sample participants and collecting the data, a lot of work goes into interpreting raw poll responses to give us top-line polling numbers. Every pollster needs to figure out how to weight the responses they get, since poll response rates are abysmal and variable across different demographic groups. As in the example given in this piece, these choices can result in large differences in those top-line numbers: from +4 Clinton to +1 Trump, all from the same raw data!

For an interesting follow-up: “Polling is becoming more of an art than a science”, also on Nate Silver’s Substack.

What is a public opinion poll without the public?

Tue, 07 Apr 2026 18:27:00 -0400

A few days ago, two professors (Leif Weatherby and Benjamin Recht) published an opinion piece in the New York Times calling attention to Axios publishing a story on maternal health using invented polling results:

A recent Axios story on maternal health policy referred to “findings” that a majority of people trusted their doctors and nurses. On the surface, there’s nothing unusual about that. What wasn’t originally mentioned, however, was that these findings were made up.

Clicking through the links revealed (as did a subsequent editor’s note and clarification by Axios) that the public opinion poll was a computer simulation run by the artificial intelligence start-up Aaru. No people were involved in the creation of these opinions.

The piece goes on to argue that this so-called “silicon sampling” is seductive because good public opinion polling is expensive, hard to do, and still prone to bias. But this shortcut magnifies the the problem of bias rather than solving it.

I’ve read a little bit about this strategy of using LLM-generated survey participants in the context of social science research in a series of posts (mostly from Prof. Jessica Hullman) over on Andrew Gelman’s blog:

Validating language models as study participants: How it’s being done, why it fails, and what works instead (2025-12-19)
Survey Statistics: Thomas Lumley writes about Interviewing your Laptop (2025-08-26)
When does it make sense to talk about LLMs having beliefs? (2025-08-15)
Better and worse ways to mix human and LLM responses in behavioral research (but you still have to figure what you’re measuring) (2025-06-12)
LLMs as behavioral study participants (2025-05-29)

Silicon sampling seems moderately interesting from a research perspective, but I can’t help but agree with the New York Times opinion piece authors that this will be ruinous for the already waning trust in public opinion polling. If you didn’t bother to ask the public, then why should the public care what you “find”? I think there is probably a lot of utility in using LLM samples to aid in designing and validating surveys, though.

One important fact about for-profit plasma donation

Mon, 30 Mar 2026 21:51:00 -0400

For-profit plasma donation is in the news today in Canada. Two people recently died after giving plasma at Grifols for-profit plasma clinics in Winnipeg, Manitoba, although Health Canada has yet to find a link between the plasma collections and the deaths. Today, it was reported that a Grifols clinic in Calgary, Alberta was found non-compliant during an inspection in December 2025:

The inspection found the Calgary centre didn’t accurately assess whether donors were suitable, didn’t collect blood according to its Health Canada authorization, didn’t thoroughly investigate errors and accidents, and didn’t carry out sufficient corrective and preventative actions.

This is obviously a problem for for-profit plasma collection in Canada, where the practice is already controversial. Paid plasma collection is illegal in Canada’s three largest provinces: Ontario, British Colombia, and Quebec, though Ontario allows a few for-profit clinics to operate through an agreement with Canadian Blood Services, Canada’s independent blood authority. British Colombia and Quebec together make up over 35% of Canada’s population; including Ontario, it’s nearly 80%. Besides Ontario, for-profit clinics exist in some other smaller provinces.

Vocal advocacy exists against paid plasma collection, leading to municipal resolutions against the practice in Ontario, even as clinics open. This advocacy is often premised on the fear that paid plasma will undermine voluntary donations. To my mind, the central fact in the for-profit plasma collection debate is that only a handful of countries are self-sufficient in plasma collection, and all of them allow for paid plasma collection. They are: the United States, Germany, Czechia, Austria, and Hungary (Egypt may have also recently joined the list). While other countries, like Canada, may have achieved self-sufficiency for plasma for direct infusion, no other country can meet its own needs for plasma-derived medical products. The world relies on a small number of self-sufficient countries, primarily the United States, to meet the demand for plasma products.

Colorado advances ban on algorothmic price and wage discrimination

Fri, 27 Mar 2026 17:48:00 -0400

The Colorado House voted today to ban the use of personal data to algorithmically set the price of a product or determine a wage. The legislation will now advance to the Colorado Senate for consideration. The summary of the bill, HB26-1210, reads:

Surveillance data is defined in the bill as data that is obtained through observation, inference, or surveillance of consumers or workers and that is related to personal characteristics, behaviors, or biometrics of an individual or group. The bill prohibits discrimination against a consumer or worker through the use of automated decision systems used to engage in:

Individualized price setting based on surveillance data regarding a consumer; or

Individualized wage setting based on surveillance data regarding a worker.

Obviously, the bill enumerates exceptions to the above rules, as it is not intended to ban, for example, charging a customer more to deliver an item a longer distance nor to prohibit schemes like discounts for students or seniors. One of the challenges of writing laws like this is to ensure they are written narrowly enough to target dystopian hyper-individualized pricing based on tracking of Internet and phone activity rather than normal business practices like pricing insurance policies according to demographic risk factors.

Colorado is one of at least a dozen American states considering similar bans. I don’t believe any of these proposed broad-based bans have been signed into law yet. I wrote about algorithmic price discrimination (surveillance pricing) last week in the context of proposed legislation in the Canadian province of Manitoba.

Make buses faster and more reliable by having fewer stops

Thu, 19 Mar 2026 07:30:00 -0400

This fascinating article by Nithin Vejendla in Works in Progress makes the case that bus networks would benefit from bus stop balancing: having fewer stops spaced further apart. This is especially true in the United States where stops tend to be only 700–800 feet (roughly 210–240 metres) apart. While having many bus stops theoretically improves access to the transit network, it also means that buses are slower (more time is spent accelerating, decelerating, and loading/unloading passengers) and less frequent, which reduces where you can actually go in a fixed amount of time, as well increasing the variability in the time it takes to get there.

The biggest problem holding back public transit in North America is that it is unreliable, and bus stop balancing is a rare policy solution that offers improved service without having to spend more. With fewer stops, the same number of buses can complete the same route faster and with greater frequency. This stops a single missed or delayed bus from ruining your plans or forcing you to build in extra time.

A research study from my city of Montreal even gets a shout out. As a big public transit user, I avoid buses where possible in favour of the metro and walking, because these modes of transportation tend to be much more reliable and less variable when it comes to the question of “how long will it take for me to get from point A to point B”. Stop balancing could go a long way toward addressing one of the main complaints about public transit: too many routes are not frequent or reliable enough to let riders stop worrying about the schedule.

Manitoba introduces bill to ban algorithmic price discrimination

Wed, 18 Mar 2026 07:30:00 -0400

The Canadian province of Manitoba has introduced a bill to ban algorithmic price discrimination (also known as surveillance pricing), i.e., the use of personal data to set prices for individual consumers:

New Democrats announced in December they would begin cracking down on what’s known as differential or predatory pricing. That is when retailers charge different amounts for the same products based on the timing of customer purchases, where they live or other personal data. […] The proposed legislation would render the use of “personalized algorithmic pricing,” both online or in store, an unfair business practice.

Okay, I guess there’s a lot of different names for this particular practice. Whatever we call it, I believe bills cracking down on algorithmic price discrimination will be very popular, as it constitutes a very clear example of companies using our data against us to rip us off. The most famous recent exposé of this practice is Groundwork Collaborative’s report on how grocery delivery service Instacart charges users different prices depending on who they are.

Manitoba isn’t the only jurisdiction introducing bills targeting this practice, but I don’t believe anywhere in the US or Canada has actually managed to ban it yet. However, New York has made in mandatory for companies to disclose when they use personal data to set prices.

Prediction markets incentivize bad behaviour

Tue, 17 Mar 2026 18:19:00 -0400

The Times of Israel journalist Emanuel Fabian is claiming that Polymarket gamblers (sorry, “traders”) have threatened his life over a report he released about an Iranian missile attack on Israel on March 10. According to the rules, this bet resolves as true if Iran strikes Israel using a drone, a missile, or an air strike on this date. At issue here is this specific rule:

Missiles or drones that are intercepted and surface-to-air missile strikes will not be sufficient for a “Yes” resolution, regardless of whether they land on Israeli territory or cause damage.

On March 10, Fabian reported a single missile had hit an open area outside the Israeli city of Beit Shemesh; he included in the report a video of the strike. This would resolve the bet as “Yes”. Evidently, holders of “No” shares would very much like him to change his report to say that the missile was intercepted, which would resolve the bet as “No”, according referenced above. This bet has seen more than 23 million USD in trading volume.

If you look at the vitriol in the comments of the bet on Polymarket, I have no trouble believing people would send threats to a journalist demanding him to change his story, whether out of desperation to change their fortunes or just in an attempt to be edgy.

Getting citizenship just got a lot harder for those of Italian descent

Sat, 14 Mar 2026 22:18:00 -0400

Many people in the Americas would probably be surprised to learn that, in much of the rest of the world, being born in a country does not by itself make you a citizen. In most of the Americas, citizenship is automatically granted on the basis of jus soli (“right of soil”): birth on the territory. Elsewhere, citizenship is more often based on jus sanguinis (“right of blood”): descent. This is the case in most of the EU.

Citizenship in an EU country is considered unusually desirable because of the mobility rights and powerful passport it confers. However, the rules concerning exactly what kind of descent confers citizenship varies widely among member states. Italy used to be considered among the easiest, requiring only that an applicant prove they had an Italian ancestor alive after March 17, 1861, when the Kingdom of Italy was founded. That changed last year, when the country passed a new law significantly tightening the requirements for citizenship, which was recently upheld by the country’s Constitutional Court. The new law brings requirements more in line with norm among EU member states:

Now, only individuals with at least one parent or grandparent born in Italy will automatically qualify for citizenship by descent. The amended law does not affect the 60,000 applications currently pending review. Additionally, dual nationals risk losing their Italian citizenship if they “don’t engage” by paying taxes, voting or renewing their passports.

Canada exports a lot of coal, but not for power generation

Sun, 08 Mar 2026 14:05:00 -0400

This provocatively titled piece in the The Hub (“Why the world needs even more Canadian coal”) made me realize I know very little about one of Canada’s most important exports: coal.

Coal is often villainized because it is incredibly dirty way of generating power. I vaguely recall an article from maybe 20 years ago claiming something along the lines of “if everyone in Canada replaced their incandescent bulbs with energy-efficient ones, the greenhouse gas savings would be cancelled out by a single coal plant that China is building every [some shockingly short amount of time]”. Although, China’s dependence on coal for power has been falling for the past two decades.

It turns out LLM-assisted search is fantastic for finding these half-remembered quotes. Here is the exact article and quote I was remembering, from a 2008 Macleans magazine article (I was pretty close):

Even if every household in the U.S. screwed in an energy-efficient light bulb today, the savings in greenhouse gas emissions would be wiped out by fewer than two medium-sized coal plants - the kind of plant that is being built in China at a rate of one a week.

But coal is also used to make most of the world’s steel (“metallurgical coal”), and this is the kind of coal that Canada (or specifically, British Columbia) overwhelmingly exports. The article goes on to claim that Canada’s production of metallurgical coal is among the cleanest (by greenhouse gas emissions) in the world.

The increasingly inevitable social media ban for kids

Fri, 20 Feb 2026 23:57:00 -0500

Jon Haidt writes on his Substack about the increasingly popular movement to ban social media for kids, following the implementation of Australia’s under-16 social media ban a few months ago.

US Medicaid data gets DOGE'd

Sat, 14 Feb 2026 10:29:00 -0500

The US Health and Human Services DOGE team (I guess DOGE still exists in some form) just released a new aggregated, provider-level Medicaid claims database covering January 2018 through December 2024. With this dataset, you can track the monthly claims for each procedure (by HCPCS Code) and provider over time.

Even if the framing around this dataset’s release is partisan—tied to allegations of Medicaid fraud in Minnesota—it is a genuine advance in transparency for the US’s third largest spending program. No doubt this accomplishment required a lot of work on the backend to harmonize countless fragmented datasets into one tidy schema. These data were difficult to access before, and now they are free for anyone to use. Journalists, policy researchers, and companies working in the US healthcare sector will benefit the most, but every taxpayer benefits from added transparency about where their tax dollars go.

I would say there is the potential for these data to be misused to spark witch hunts, but this is more or less the stated purpose for this data release. Per Elon Musk: “Medicaid data has been open sourced, so the level of fraud is easy to identify.” If you go on Twitter, you will find several people have already plugged in the dataset to Claude Code and trumpeted their ASCII tables of providers flagged for potential fraud. Inevitably, some of these providers targeted by public scrutiny for their unusual billing patterns will have perfectly innocent explanations. But if ProPublica is excited about the release of this new dataset, then so am I.

An end-to-end AI pipeline for policy evaluation papers

Thu, 12 Feb 2026 19:11:00 -0500

Prof. David Yanagizawa-Drott from the Social Catalyst Lab at the University of Zurich has launched Project APE (Autonomous Policy Evaluation), an end-to-end AI pipeline to generate policy evaluation papers. The vast majority of policies around the world are never rigorously evaluated, so it would certainly be useful if we were able to do so in an automated fashion.

Claude Code is the heart of the project, but other models are used to review the outputs and provide journal-style referee reports. All the coding is done in R (though Python is called in some scripts). Currently, judging is done by Gemini 3 Flash to compare against published research in top economics journals:

Blind comparison: An LLM judge compares two papers without knowing which is AI-generated Position swapping: Each pair is judged twice with paper order swapped to control for bias TrueSkill ratings: Papers accumulate skill ratings that update after each match

The project’s home page lists the AI’s current “win rate” at 3.5% in head-to-head matchups against human-written papers.

Prof. Yanagizawa-Drott says “Currently it requires at a minimum some initial human input for each paper,” although he does not specify exactly what. If we look at initialization.json that can be found in each paper’s directory, we see the following questions with user-provided inputs:

Policy domain: What policy area interests you?

Method: Which identification method?

Data era: Modern or historical data?

API keys: Did you configure data API keys?

External review: Include external model reviews?

Risk appetite: Exploration vs exploitation?

Other preferences: Any other preferences or constraints?

The code, reviews, manuscript, and even the results of the initial idea generation process are all available on GitHub. Their immediate goal is to generate a sample of 1,000 papers and run human evaluations on them (at time of posting, there are 264 papers in the GitHub repository).

The CIA World Factbook has been memory holed

Thu, 05 Feb 2026 16:37:00 -0500

Another staple of my childhood is gone, this time the CIA’s World Factbook. I have fond memories of consulting the World Factbook for school projects in my elementary school computer lab. But as of yesterday, the entire publication along with all of its archives have been suddenly and unceremoniously wiped from the agency’s website. At least archives of the website are still available on the Internet Archive, with complete zip files up to 2020 and Wayback Machine snapshots thereafter.

Total electoral wipeout

Sat, 31 Jan 2026 13:01:00 -0500

The 2002 Turkish general election is the canonical example of total electoral wipeout. Every party holding seats in the previous legislature was completely wiped out. Of the two parties that won seats in the 2002 election, the one that formed government didn’t even exist at the time of the previous election (current president Erdoğan’s AK Party, formed in 2001). Of note, it wasn’t a complete changing of the guard: one of the three independent members from the 1999 parliament won his seat again in 2002 (Mehmet Ağar), though it seems he took over as leader of one of the wiped-out parties shortly after the election.

_{Hat tip to kynakwado2 on Twitter.}