Ephemeral factoids

Another threat to scholarship: ephemeral factoids. 

I was reading an article which cites a piece of information from the Office for National Statistics (ONS) website (I can already hear the loud groans). Something didn't look right about it, so I wanted to go back to the original source to check how the numbers were actually arrived at, what definitions were used etc. Quite properly the article cites a URL. When I type in the URL I am simply taken to a generic ONS page with lots of links, none of which take me to the document I am interested in. A generic search on the title of the document brings up nothing apart from a couple of other sources that all repeat the same factoid and cite the same, apparently now non-existent, document. 

I probably take an old fashioned view about this, but should we really be citing ephemera in academic papers to support our arguments? Isn't a basic principle of science that what we say should be able to pass the show me test? The article I was reading was just published in the last few months, so we are not talking about ancient broken links. Once we lose the discipline of rigorously only citing as evidence things that can be consulted by others, then we are in danger of basing our scholarship on nothing more substantial than, at worst, an enormous sack of wind and at best an appeal to: trust me, I'm a scientist, a social scientist...