Actors by genre by year: list the credits, not a tally

Annual counts by genre look like the most solid numbers in any index. They are the least solid, because every one of them is the product of three decisions the reader cannot see: which label, whose label, and which year.

This page is about those decisions and about the practical answer for an actors directory: present a performer's work by genre and year as a dated, sourced list of credits rather than as a tally.

What to take away

The three hidden decisions

Which label. Most productions can be described by more than one term, and most sources give more than one. Counting the first one listed is a decision. Counting all of them is a different one. Neither is wrong, and a number produced by one is not comparable with a number produced by the other.

Whose label. A distributor, a broadcaster, a classification body and a library catalog will not agree, because they are labeling for different purposes. A count that pools them is adding different measurements, which the notes on whose vocabulary a genre label came from work through in detail.

Catalog scope shows the problem plainly. The AFI Catalog of Feature Films, a filmographic database started in 1968 as a research tool for film historians, holds entries on more than 60,000 feature films and 17,000 short films produced from 1893 to 2011. A count drawn from it describes that catalog's scope and its labels, and nothing after 2011.

Library catalogs label with a controlled vocabulary instead. The Library of Congress publishes its Genre/Form Terms for Library and Archival Materials (LCGFT), with a manual of guidelines for applying the terms, and the OLAC Cataloging Policy Committee has written best practices for using LCGFT for moving images. A directory that mixes those terms with the looser labels a distributor uses for marketing is mixing two measuring systems, so state which one each label on a credit comes from.

Which year. Production year and release year differ, sometimes by a lot, and release year differs by territory. Choose one, label it, and store the others, for the reasons set out in the three years a production has.

Counting productions against counting assignments

The third row is the one to avoid. Splitting a production across three labels as a third each looks precise and encodes a judgment nobody made. Where a single total is needed, prefer the first method with the rule stated. Where accuracy matters more than a tidy total, use the second and say that the total counts assignments.

What the annual series is actually showing

Even done carefully, a genre count by year is a measurement of a catalog rather than of an industry.

Coverage is uneven by country, by language, by production size and by period, so a rising line frequently records improving documentation rather than increasing output. That is ordinary selection bias, and it does not go away with a larger sample drawn from the same sources.

Retroactive labeling adds a second distortion, opposite in direction. Older productions accumulate labels over decades as writers reach back for them. A category can then appear to grow in a period long after that period ended.

Why that happens, and how to date a label separately from a production, is the subject of how a genre label changes over time.

Publishing a number responsibly

That last point separates a statistic from a claim. Keep raw values beside derived ones. A changed rule then means re-derivation, not a re-count. That is the ordinary discipline behind data cleansing done well.

The general argument against publishing derived scores as though they were observations is in why a ranking of live work measures exposure.

Performers are not countable this way

A tally of how many productions of a given genre a person appeared in is three unstable things multiplied together: an unstable label, an unstable year, and an unstable credit definition.

It answers a question nobody should ask: what someone worked in records what was being made and what they could be hired for.

The pillar on how a genre label attaches explains why the label belongs to the production. A credit resists a simple count in the fields a television credit needs.

What a directory can show instead

An actors directory can still answer the reader who wants to see a performer's work by genre and year. It does so by listing, not tallying:

Here is what that looks like for three of Clint Eastwood's films, with the labels and years as Wikipedia's articles give them:

Three credits carry three different labels from one source. A tally would report one action film, one western and one drama and tell the reader almost nothing. The list shows the run of work in order and where each label came from, and when a second catalog files the same films differently, the source column shows why.

Common questions

Why does the same year give different totals on two sites?

Because they used different label sources, different year definitions, or different counting units. Check all three before assuming either is wrong.

Should a production with four labels count four times?

It can, provided you say that the unit is a label assignment and publish the production count alongside it. What it should not do is count four times inside a total presented as a production count.

Is a percentage safer than a count?

No, it just hides the denominator. A percentage of an unstated base is less checkable than a raw number, not more.

Can these counts show a trend?

Only within one catalog, one vocabulary and one year definition, held constant across the whole series. Compare a catalog with itself. Comparing two catalogs over time measures the catalogs.