Or, how a labour market survey turned into a lesson on interpreting research findings.
It’s been a quiet bank holiday weekend, with little thought given to postgrad careers by your less-than-dedicated postgrad careers blogger (hey, I need a life too and that garden ain’t gonna prune itself). So, this morning I pounced on an e-mail with an interesting link to some survey results on public sector recruitment – that’s my blog post for today, I thought.
It seemed to imply that a large majority of public sector employers were planning to cut back on outsourcing and using external consultants, rather than cutting back on permanent recruitment – just the sort of information in which you might be interested if you’re dead set on working for the public good.
I tracked down the source of the report and found much more detail, including a table of results, types of public sector employers surveyed, dates of survey etc on a LinkedIn group set up by the recruitment consultancy who conducted the survey (which had now turned into “a poll” – distant sound of alarm bells started to ring).
Luckily, my love for numbers (and statistical cynicism) kicked in. The percentages quoted were unfamiliar – but oddly, “86%” and “57%” both occurred twice, in different parts of the “poll”. Extracting all the figures quoted in the report, and arranging them in order, gave the following sequence:
14%, 29%, 43%, 57%, 71%, 86%
All just a bit too neat for coincidence, I fear. So, I won’t be reassuring you with tales of impending public sector recruitment, but I will remind you that when you’re reading reports full of statistics, for your research or just for interest, if they don’t quote sample sizes, take it all with a pinch of salt.
(for anyone who hasn’t spotted the sequence yet, I’ll put it in the comments)
1/7 = 14%
2/7 = 29%
3/7 = 43%
… you get the picture.
Talking to seven employers is not a good basis for a robust survey, in my books (but it took me a while to spot the unfamiliar sequence, well hidden by using percentages and, of course, quoting the figures randomly and out of sequence throughout a long article).
Ha ha!
You never know, the sample size could be a multiple of seven.