Four months ago on this blog, I posed the question: How many women start on statutory paid maternity leave in the UK each year?

You’d think this would be a well-known and widely used official statistic, freely available on some government website. But you’d be wrong.

As I noted back in September, there is a figure – about 650,000 – that is widely used (or at least relied upon) by academic researchers, journalists, supposedly expert campaign groups such as Pregnant Then Screwed, employment lawyers and even the TUC. But that widely used figure is wrong.

We have known for some years that the figure must be wrong, because there are two sets of official data on recipients of Statutory Maternity Pay (SMP): one unpublished but provided by HMRC in response to numerous Freedom of Information requests in recent years, which is where the figure of ‘about 650,000’ comes from, and one published by the Department for Work & Pensions (DWP). And, as the following chart shows, the figures in those two data sets are not just different – they are very different:

This matters. Because, without knowing how many new mothers start on statutory paid maternity leave each year, we cannot, for example, estimate what proportion of such new mothers use the chronically failing Shared Parental Leave scheme to transfer some of their paid leave to the child’s father (a more meaningful measure of the success or otherwise of the SPL scheme than the rate of take-up among the limited pool of eligible fathers).

Furthermore, as I noted in my September blog post about take-up of statutory paid paternity leave, the DWP’s data set is published under the terms of a Memorandum of Understanding with the Office for Budget Responsibility (OBR), and reflects the Government’s financial delivery plans. So, if it’s wrong, the Chancellor’s annual Budget is also a pile of pants.

Accordingly, late last summer, I set out to get to the bottom of the glaring discrepancy between the two sets of data.

Soon after my September blog post, HMRC confirmed to me (in response to my Freedom of Information request FOI2021/20932), that the figures in their data set are inflated by double counting. In short, where a spell of SMP extends across the boundary between two financial years, the recipient is counted twice (once in each financial year). And, with the average new mother taking nine months of SMP (see below), there are a lot of spells of SMP that extend across the boundary between two financial years. So, there is a lot of double counting in the HMRC data set.

Then, in late October, the DWP confirmed to me (in response to my Freedom of Information request FOI2021/79684) that their data on SMP caseload understates the number of women who start on SMP each year, as the data is average caseload at any point in time, not total caseload over the year.

Fortunately, armed with this information, it is a relatively simple task to adjust the two data sets accordingly. Assuming the average duration of a spell on statutory maternity leave to be nine months (as is suggested by BEIS research, by a statement by BEIS minister Paul Scully to MPs in June 2020, and by the DWP’s quarterly statistics on Maternity Allowance grants), we can adjust the HMRC figures down by a factor of 4/7, and adjust the DWP figures up by a factor of 4/3, to generate a new chart:

We can then (a) average out the remaining small difference between the two sets of data, and (b) add the some 60,000 Maternity Allowance starts each year (from the DWP’s quarterly statistics), to give us figures for the number of women who start on statutory paid maternity leave each year:

You’re welcome.

(Incidentally, I did ask policy officials in the BEIS family rights team – the team that works on e.g. maternity leave and SPL policy – to assist with my reconciliation of the HMRC and DWP data sets, and later to comment on my adjusted figures, but they repeatedly declined to do so. Make of that what you will.)