How to make sense of deaths by month causes

Group anything by month and the eye starts looking for a season. That instinct is useful in a lot of fields. On a memorial index it produces confident nonsense, and it is worth understanding exactly why.

No person is named here and no death is reported. Nothing on this page describes how anyone died.

What to take away

• A monthly index has no cause column because it has no cause data. The absence is not an omission waiting to be filled.
• Monthly grouping is an unusually strong generator of false patterns, because it gives twelve buckets to a small, selected set of entries.
• Questions about when and why people die are answerable, but from civil registration rather than from an archive of notices.

Where the cause would have to come from

For a cause to appear against an entry, somebody with direct knowledge would have to have said it publicly, or an official finding would have to have been published. Neither is common.

A death is certified where it occurred, under that jurisdiction's rules, by whoever that jurisdiction authorizes. What gets recorded goes onto a document whose access is restricted in most places: in the United States, the practical route to a copy runs through the state or territory that registered it, as the federal guide to obtaining a certified death certificate sets out, and eligibility is decided at that level rather than nationally. An archive compiling monthly pages is not an eligible applicant for anyone, and would have no business becoming one.

So the honest position for a monthly index is that it holds the fact of a death and the identity of the person, and nothing about how. That is the same position the parent guide takes in cause of death guides, and it does not change with the unit of time.

The pattern that the month invents

Here is the specific trap. Suppose an index holds a few hundred entries for a year. Split across twelve months, some months will hold noticeably more than others. They always do. A handful of entries either way is enough to make one month look like a spike.

A reader who then reaches for an explanation will find one, because explanations for a bad month are easy to invent: weather, a season, an anniversary, an age cohort. None of them was tested. The month with more entries has more entries because of who happened to be written about, and because of how quickly those particular confirmations came through.

Two further things make the illusion worse on this kind of page.

The first is that the underlying set is selected, not sampled. An archive holds the deaths that were publicly noticed, which correlates with profession, with country, with language, and with how long a person had been out of view. Every one of those correlates with time of year in some small way, and small effects on a small set produce large-looking swings.

The second is that entry dates cluster for reasons that have nothing to do with deaths. A desk that clears a backlog in one week puts a dozen entries into one month. That is a fact about the desk.

What a reader who wants numbers should do instead

Go to the bodies that count everyone rather than the ones that write about some people.

National statistical offices and civil registration authorities compile mortality from complete registration, code the causes to a common classification, and publish the results with the definitions and caveats attached. The international collection of those returns is the WHO Mortality Database, and it is the right starting point for any question of the form "does this happen more in some months than others". It will also show you how much work goes into making two countries' figures comparable, which is itself a useful corrective.

What you will not find there is a name. That is the trade. Statistical sources answer questions about populations and refuse questions about individuals. Memorial archives do the reverse. Trying to get a population answer out of an archive of individuals is the mistake this page exists to head off.

How a careful monthly page handles the question

• It says on its face that it records no cause, so a reader does not go looking for a column that was never there.
• It resists ranking months against each other, and it does not publish a chart of monthly totals, because a chart is an argument.
• Where a family has publicly stated something, it carries the statement as the family's, attributed, and does not extend it.
• It keeps the date of death and the date of entry as separate fields, so that anyone who wants to understand the shape of the page can see the confirmation lag for themselves.
• It points at statistical sources for statistical questions instead of trying to answer them.

Where this sits

The mechanics of building the monthly unit are in deaths by month. The same argument at annual scale, where the selection effects are easier to see, is in deaths by year. For what a shorter window can and cannot contain, see deaths this week, and for the first day after a death, deaths today.

Common questions

Do more people die in winter?

In many populations mortality does vary through the year, and statistical agencies publish that. You cannot learn it from a monthly memorial index, which is why this page sends you elsewhere for it.

A month here has twice as many entries as the month before. What happened?

Almost certainly nothing happened. On a set this size and this selective, a doubling between adjacent months is ordinary variation plus the rhythm of the desk that entered them.

Why not add a cause field and leave it blank when unknown?

Because a blank field is an invitation. Someone fills it from a report, then someone else treats the filled field as the archive's finding. A field that would be empty in almost every row and wrong in some of the rest is worse than no field.

If a family has said publicly how someone died, would that appear?

As a statement attributed to the family, where it is relevant, and not as a data field to be sorted or counted. What a family chooses to say is theirs, and it does not become a category by being said.

Is there any month-level question this index can answer?

Yes, but they are questions about the record rather than about mortality: how quickly entries were confirmed, how many were later amended, how much of a month arrived after the month closed.