The birthday effect is a statistical anomaly that shows higher likelihood of death on or near your birthday. For the Pudding, Russell Samora analyzes mortality data to see if this effect is for real.

The differing results across studies highlight how methodological choices—from sample selection to statistical analysis—can dramatically shape our findings. When studying something as basic as birthdays and death, how we approach the problem shapes what answers we find.

Warning: this is a statistics lesson in disguise so you might learn something.

There are plans to reduce staffing at the Social Security Administration by 7,000 and perhaps to rebuild the entire codebase, which means check delays could be a real thing. For the Urban Institute, Richard W. Johnson and Jonathan Schwabish ran the numbers for how many people could be affected if checks are just one, two, or three months late.

We find 11 percent of current Social Security beneficiaries, or 7.4 million people, do not have enough savings to replace their benefits if their Social Security checks were delayed for one month. The share with inadequate savings to replace their Social Security increases to 13 percent (8.3 million people) if checks were delayed for two months and to 14 percent (9.2 million people) if checks were delayed for three months.

For reference, SSA has never missed a benefit payment since its inception eight decades ago.

The NOAA database, dubbed Billion-Dollar Weather and Climate Disasters, tracked events that exceeded one billion dollars of damage. It’s kaput. For CNN, Andrew Freedman reports:

The disasters database, which will be archived but no longer updated beyond 2024, has allowed taxpayers, media and researchers to track the cost of natural disasters — spanning extreme events from hurricanes to hailstorms — since 1980. Its discontinuation is another Trump-administration blow to the public’s view into how fossil fuel pollution is changing the world around them and making extreme weather more costly.

Add this to the growing list of discontinued NOAA products, because they are related to climate, even just a tiny bit.

RJ Andrews held an exhibition of data graphics at 41 Cooper in New York. On physical data that exists off the screen:

After my presentation, I was asked about tools and gave my stock answer extolling paper and pencil as the most important creative weapon. This quip surprised me with the biggest applause of the night.

Beautiful information graphics off-screen, in the real world, are a real treat. In some sense the show did not feel special to me as much as it felt natural. The best way to consume data might not be alone on a pocket screen—but together, in conversation, with human-sized artifacts.

Andrews speaks of the satisfaction of such things but also of the challenges of handling physical data.

I’ve felt this with my own physical projects. It’s nice to hold the data. It’s real. It’s tangible.

It takes exponentially more time and effort.

But if we want visualization to extend beyond analytical tools and dashboards, then that’s what we will need to do.

The U.S. and China put a semi-hold on tariffs for the next 90 days. The New York Times provides the timeline so far.

Tariffs on Chinese goods are down to 30% from 145% and tariffs on U.S. goods are down to 10% from the retaliatory 125%. Has anyone explained how countries choose these numbers? They seem arbitrary.

In any case, I’m guessing these step charts will need to update a few times over the next few months. Step charts, which show constant values and then sudden changes over time, might be the chart type of the next few years.

It’s a temporary suspension, for now. For the Washington Post, Maxine Joselow reports:

The program collects data on levels of harmful air pollutants, including ozone and particulate matter, at the 63 national parks in the United States. Federal officials consult this data when deciding whether to grant permits to nearby industrial facilities, such as power plants or oil refineries.

Breathing in these pollutants is linked to a range of adverse health effects. For instance, long- and short-term exposure to particulate matter is associated with heart attacks, strokes, asthma attacks and premature death, according to the Environmental Protection Agency.

For CNN, Andrew Freedman reports:

Of the hundreds of vacancies, the 30 open meteorologist-in-charge roles are particularly worrying, current and former NWS meteorologists told CNN.

Meteorologists-in-charge serve as the captain of a team of forecasters and other specialized staff members. Their decades of experience often comes into play during high-impact weather situations, an active-duty NWS meteorologist told CNN. They also requested anonymity for fear of reprisal.

In addition to headless offices, fewer weather balloons are launching, less data is collected, and the forecasts become less accurate. Not ideal.

Researchers from the Ocean Discovery League estimated how little we know about the deep seafloor:

We then used two independent methods to estimate the amount of deep seafloor observed globally over the past seven decades (see Methods). Using the dive-based method, we estimate a maximum visual seabed coverage of 2130 km2. Using the time-based method, we estimate visual seabed coverage of 3823 km2.

Of Earth’s total surface area (510 million km2), the seafloor makes up 360 million km2 (approximately 71%). Of this ocean area, approximately 93% is deep seafloor (≥200 m), yielding 66% of the Earth’s total surface area (~335.7 million km2). In the framework of this global context, our exploration coverage estimates show that deep-sea visual tools have only observed 0.0006 to 0.001% of the deep seafloor since 1958.

We’re never going to find Atlantis at this rate.

Seth Godin on the importance of judgment:

AI is a tool, and judgment, for the foreseeable future, remains our job. It doesn’t matter how cool your hammer is, it’s still on you to decide which nails need hammering. And to be responsible for what happens when you use it.

There’s been a shift in wording brought on by the AI wave. People “ask” the AI to do things versus “use” our tools to solve problems, as if we should pass responsibility of thought to statistical models. This seems not good.

When talking about AI, I mentally substitute “computer” or “tool” in its place. I don’t ask my computer to analyze data. I use my computer like someone might use a hammer while building a planter box.

Benj Edwards reporting for Ars Techinca:

In “Large Language Models, Small Labor Market Effects,” economists Anders Humlum and Emilie Vestergaard focused specifically on the impact of AI chatbots across 11 occupations often considered vulnerable to automation, including accountants, software developers, and customer support specialists. Their analysis covered data from 25,000 workers and 7,000 workplaces in Denmark.

Despite finding widespread and often employer-encouraged adoption of these tools, the study concluded that “AI chatbots have had no significant impact on earnings or recorded hours in any occupation” during the period studied. The confidence intervals in their statistical analysis ruled out average effects larger than 1 percent.

See the full study by Anders Humlum and Emilie Vestergaard here.

This seems to make sense. Even if our time is freed up in one area, wouldn’t employers expect that time to be spent elsewhere?