This applies to editorial statements (opinions, interpretations); but if you're just quoting blunt facts, a primary source is preferable, no? E.g. if you're citing a statistic, it's better to cite the study that supplied the statistic, than to cite a science-journalistic review of said study. You can then cite some statement the review made to put the statistic in context; but that should be a separate, second citation, so that the statistic itself can be grounded in a primary source — especially since science-journalism does not often cite the primary sources itself, making it hard to chase the citation for the statistic otherwise.
Or, maybe to couch this in language more friendly to how a Wikipedia editor might think of the process: if you are already providing an attributed quote of a secondary source; and in the secondary source, a fact is quoted from some primary source without a true citation being provided (only a weak, implicit-in-context mention of the source); is it not better to do the "original research" of figuring out what primary source the secondary source got the data for the claim from, and then putting in a citation for the fact inside the quote, yourself, in exactly the way the secondary source's author likely would have if the format they were publishing in allowed true citation? "Repairing" the citation graph, so to speak, where the secondary source has left a gap in it.
(And this is important to the topic at hand: citing only secondary sources for "blunt facts", where those secondary sources are not themselves expected/required to provide citations to primary sources, is exactly how hoax citation graphs arise.)
>but if you're just quoting blunt facts, a primary source is preferable, no?
At least in theory, no it's not, per Wikipedia policy. Now, in practice, if you reference some government dataset for a non-controversial fact like the area of a state, only the most procedural whackjob admin is going to flag it. But, in general, you're not supposed to use primary sources.
Or, maybe to couch this in language more friendly to how a Wikipedia editor might think of the process: if you are already providing an attributed quote of a secondary source; and in the secondary source, a fact is quoted from some primary source without a true citation being provided (only a weak, implicit-in-context mention of the source); is it not better to do the "original research" of figuring out what primary source the secondary source got the data for the claim from, and then putting in a citation for the fact inside the quote, yourself, in exactly the way the secondary source's author likely would have if the format they were publishing in allowed true citation? "Repairing" the citation graph, so to speak, where the secondary source has left a gap in it.
(And this is important to the topic at hand: citing only secondary sources for "blunt facts", where those secondary sources are not themselves expected/required to provide citations to primary sources, is exactly how hoax citation graphs arise.)