Trends in secondary science teacher recruitment

24th April 2019 by Timo Hannay [link]

We are now at the busiest time of year for teacher recruitment, so this post looks at trends in school hiring, especially of science teachers¹. We are delighted to be collaborating on this with the Gatsby Foundation, publishers of a previous analysis on the same topic² and longstanding proponents of science education. We hope that these results will provide guidance for the many schools now seeking to fill positions for the start of the next academic year in September. Our findings include:

• There has been a further increase in the number of science teacher advertisements compared to the same period last year.
• The proportions of science teacher positions that require specialists in biology, chemistry or physics vary greatly by school type and location, suggesting big disparities in the provision of science education.
• Based on an analysis of repeat advertisements, there is no evidence that specialist science teacher positions are any harder to fill than general science positions.
• There are also large differences in the proportions of part-time, temporary and maternity-cover positions across various science subjects. These results are consistent with the idea that biology teachers are more likely than physics teachers to be female, with potential implications for gender disparity in subject choice at A-level.

More detail follows, so please read on. You may also wish to refer to our previous analyses of the School Workforce Census and staff development spending in schools.

Supply of teachers

Teacher recruitment and retention have long been hot topics. Last year the Education Policy Institute warned of severe shortages. Secretary of State Damian Hinds, who assumed his post in January 2018, has described it as a top priority, leading the Department for Education to issue its teacher recruitment and retention strategy at the beginning of this year. But concerns continue and the Labour Party has recently gone as far as to propose a shorter working week to attract more people into the profession.

To provide objective data, we have been tracking education recruitment activity by analysing vacant positions advertised on school websites. Summary statistics and underlying data are provided in our Jobs section, where you can see recent top-level trends by subject and role. This post takes a deeper look at teacher recruitment in the sciences, which are not only core subjects but also account for almost 20% of all teacher advertisements – considerably more than even maths or English.

Recruitment rising

Figure 1 shows the total numbers of all science teaching advertisements found between May and March (the period for which we have comparable data) during the 2017-18 and 2018-19 academic years. There was a year-on-year increase in advertisements posted of 5.3%, though this was lower than the 16% increase between the 2016 and 2017 calendar years reported in the previous Gatsby analysis. Also in contrast to previous results, this year we have so far seen a larger increase in specialist biology, chemistry and physics positions (+7.9%) than in general science positions (+4.4%). Note, however, that the present analysis does not yet cover a full 12-month period; we intend to present complete results this summer. (Hover over Figure 1 to see exact values.)

Science was not exceptional in displaying these increases. As shown in Figure 2, over the last two years it has accounted for a roughly constant proportion of just under 20% of all secondary school teacher advertisements. Though there are seasonal patterns (dips in the percentage of science teacher advertisements tend to occur in the summer, when the total number of advertisements is very low), there has so far been very little year-on-year change between the 2017-18 and 2018-19 academic years. Mathematics and English also continue to account for similar proportions to last year: respectively, just over and just under 15%. (Hover over the lines Figure 2 to see data values.)

Highs and lows

These total values hide a great deal of variability among schools. Figure 3 shows the total number of all science teaching positions (including biology, chemistry and physics) by school type, taking into account the total number of teachers employed in each group. We now say 'positions' instead of 'advertisements' because for this analysis (and those that follow) putative repeat advertisements for the same role at the same school have been removed (see Footnote 3 for details).

The types of schools showing higher levels of recruitment activity include those that are typically growing (free schools), those with low staff numbers (small schools), those teaching at high academic levels (sixth forms and grammar schools) and those that may be experiencing difficulty recruiting (schools with high deprivation). It is also interesting to note that the trends for school deprivation (based on the proportion of pupils who are eligible for free school meals) and local deprivation (based on the childhood poverty rate in the surrounding area) run in opposite directions. This may seem surprising because poorer schools tend to be located in poorer neighbourhoods, so we might expect them to correlate. This discrepancy could arise because, although deprived schools might have more difficulty in recruiting, poorer neighbourhoods are likely to have fewer alternative employment opportunities and therefore be more conducive to hiring staff. (Use the menu above Figure 3 to see these the other school groups; hover over each column to see the corresponding data value.)

Figure 4 provides a similar breakdown by region. The South East and London show the highest levels of recruitment activity while regions in the north of England tend to show the lowest levels. This presumably reflects the relative tightness of local labour markets.

It is also interesting to look at the proportions of vacant positions that are for specialist science teachers – ie, in biology, chemistry or physics. These show even bigger variations, as illustrated in Figure 5.

In sixth forms and grammar schools, the overwhelming majority of science teacher posts are specialist positions. Among single-sex schools and Ofsted 'Outstanding' schools they represented about half. In contrast, secondary schools without a sixth form and those with high levels of deprivation or Ofsted 'Inadequate' ratings showed much lower proportions – around 10% or less⁴. (Use the menu above Figure 5 to view these the other school groups.)

Differences by region are less dramatic, but substantial all the same. As shown in Figure 6, specialist science teacher positions are most common in London and the North East (interestingly, the two regions with the highest levels of deprivation). They are least common in Yorkshire and the Humber.

Thus the probability that a pupil will be taught by a specialist in biology, chemistry or physics seems to vary hugely – perhaps as much as tenfold – depending on the type and location of the school they attend.

Repeat, repeat

Of course, it is not enough for a school simply to advertise a vacant position, they must also fill it. There is no foolproof way to determine whether or not a given vacancy has been successfully filled, but one indicative approach is to identify putative repeat advertisements for the same position at the same school³. Overall, we classified about 25% of science teacher advertisements as repeats⁵, though this varied by school type and (in particular) degree of specialisation, as shown in Figure 7.

In general, schools with high levels of deprivation or located in urban areas showed higher levels of repeat advertising, consistent with the idea that such schools find it harder to attract suitable candidates. Interestingly, so did schools with high Ofsted ratings, which seems inconsistent. Perhaps this latter group is more likely to readvertise because it is more discerning in its recruitment.

Importantly, the overall repeat rates for specialist positions (red columns) are invariably lower, not higher, than for general science positions (blue columns)⁶. This supports the conclusions of the previous Gatsby study and contradicts the common assumption that specialist positions are harder to fill. It is true that this does not compare like with like because, as we saw in Figure 5, schools that are more likely to advertise specialist positions may also be in a stronger position to attract suitable candidates. But it is striking that general science teaching positions always seem to show higher repeat rates than specialist positions whichever way we divide up schools, whether by school deprivation, Ofsted rating or academy status. (Use the menu provided above Figure 7 to explore these and other groupings.)

Figure 8 shows the same analysis by region. The South East and London showed the highest levels of repeat advertising while regions in the north and south west showed the lowest levels. Similar to the results we saw in Figure 4, this is consistent with the idea that the labour market in the latter areas may be more accommodating for schools.

Once again, in every region the putative repeat rates for general science teaching positions (blue columns) was higher than for specialist positions (red columns).

These results suggests that schools of any kind should not avoid trying to recruit specialist science teachers simply out of fear of not being able to fill the position. On the contrary, it may even make the task easier. This is not as counterintuitive as it may seem: science teachers might be more, not less, attracted to roles that are specific to their own area of expertise. Consistent with this, a recent policy paper published by the Gatsby Foundation pointed out that teachers working outside their specialisms are more likely to leave their current school.

Permanent marker

Finally, we analysed the incidence of part-time, temporary and maternity positions⁷. As shown in Figure 9, these accounted for small proportions (3.5%-4.5%) of all science teacher positions found. It is striking that biology positions are much more likely to be maternity, part-time or temporary positions than are physics positions, with chemistry coming somewhere in between. One obvious explanation is that biology teachers are much more likely to be women, which could at least partly account for the large gender disparities seen when students choose their A-level subjects. (Compare, for example, these SEEdash charts showing GCSE and A-level subject choice by boys versus girls.)

Headhunting season

That concludes our analysis of teacher recruiting activity up to the end of March 2019. We plan to provide an update just before the summer holidays to show what has happened in practice over the whole of this academic year. Until then, we welcome your thoughts: [email protected].

Footnotes:

1. This analysis is based on data gathered by indexing school websites every weekday in order to search for any new vacancies advertised there. This process does not capture all vacant positions because: (a) not all positions are advertised on school websites, (b) even when they are, they are not necessarily presented in a way that can be automatically indexed, and (c) websites are sometimes unresponsive or otherwise unavailable. For this reason, the data presented should be thought of as being based not on a comprehensive list of all vacancies but on a subset. However, positions were detected for well over 90% of schools and the representation of different subgroups was generally consistent, so there are no obvious reasons to suspect systematic biases. This means that any variations, such as differences between subjects or changes over time, are likely to represent genuine underlying trends. All data presented in this analysis are from mainstream state secondary schools, sixth-form colleges and FE colleges in England. Primary schools, special schools, independent schools and establishments outside England have been omitted.
2. Note that this analysis is based on a different data set to the one used in the previous Gatsby analysis, so the results of these two studies are not directly comparable.
3. Repeat advertisements are defined as new advertisements that overlap with the subject area of an earlier advertisement for the same school which appeared at least 21 days but no more than 90 days previously. This method almost certainly creates both false positives and false negatives, but manual inspection suggests that the results are in line with classifications made by human experts.
4. Interestingly, we see a similar pattern among technician vacancies, where science positions account for about 40% of the total. Grammar schools, single-sex schools and Ofsted 'Outstanding' schools are much more likely to advertise for technicians in biology, chemistry or physics, while schools with higher levels of deprivation or lower Ofsted ratings are much more likely to recruit for general science technicians. However, absolute numbers for technician posts are much lower than for teachers (roughly fivefold smaller), making them less statistically reliable and they are not analysed in detail here.
5. The corresponding figure in the previous Gatsby study was 34%, but the methods used differed somewhat so these two figures are not directly comparable.
6. It is important to qualify this finding. Because general science positions are more numerous than biology, chemistry and physics positions, it is more likely that we will overestimate the repeat rate for the former. This is because it is more likely for a given school to have two science vacancies in quick succession than (say) two biology vacancies. In either case, if they are sufficiently close together we may misidentify the second advertisement as a repeat of the first. There is no reliable way to correct for this short of speaking with each school about each position. It is therefore better to think of these results on their own as being consistent with the idea that specialist biology, chemistry and physics positions are no harder to fill than general science positions rather than providing definitive evidence that they are easier to fill.
7. These were defined as positions for which the part-time, temporary or maternity status was explicitly stated within or close to the job title. It does not include positions where this was specified elsewhere or not included at all in the online information.

A new map of English education

12th April 2019 by Timo Hannay [link]

The previous posts in this series have looked at the English education system through the lenses of poverty, urbanisation, immigration and politics. All of these aspects show a great deal of variation across the country. This final instalment will try to pull together the various threads and consider the bigger picture. In particular, we will ask what all this means for the way in which we should think about education across England?

Bird's-eye view

As shown in Map 1 below, England is traditionally divided into nine administrative regions – running from the North East to the South West and covering everywhere in between. These are in turn divided into 152 local authorities. (Use the menu to highlight specific regions; hover over the map to see local authority names.)

Such regional groups make intuitive sense and can be useful in understanding disparities based on national geography. But it's also important to realise that they hide a lot of intra-regional variation. Just because local authority areas are physically close, it doesn't necessarily follow that they are similar. So while it's tempting to generalise about each region (as we frequently do on this blog), such an approach rarely addresses the most significant underlying differences across the country, which usually manifest themselves within regions as well as between them.

To illustrate this, Figure 1 shows the distributions of values across local authorities for a selection of social, demographic and educational indicators¹. Of these, ethnicity shows the greatest variation, with proportions of white British pupils ranging from low single-digit percentages to well over 90 percent. One or two regions – notably the North East – are composed of local authorities that fall within a relatively narrow range. But most do not. For example, the West Midlands and the South East display wide ranges, which suggests that (at least on this measure) their constituent local authorities share relatively little in common.

A similar picture emerges for urbanisation, where regions like the North West, Yorkshire and The Humber and the West Midlands show almost as much variation as the whole of England.

Poverty shows a narrower national range but a similar regional story. Local authorities in the North East and East of England cluster fairly closely, but others such as those in the East Midlands and London show wide variations.

Phonics ability at age 7 shows a very narrow range, but by age 11, attainment shows more variation and this increases again by the time pupils take GCSEs at age 16. Here, too, we often see large intra-regional variation: for example in the North West, East of England and South East. (To explore the data further, use the menu above Figure 1 to select an indicator and click on the figure legend to turn individual regions on or off.)

An important consequence is that social or education policies that focus on regions are unlikely to be helpful because there's too much variation within each region for any single approach to apply. Yet addressing each local authority separately seems like overkill. If only there were a way to group together local authorities that are similar even if they are in different parts of the country. Fortunately there is: the clustering algorithm.

Birds of a feather

Map 2 shows a different way of grouping local authorities based not on their geographical proximity but on their social, demographic and educational characteristics (see Footnote 2 for details of the method used). The other big difference from Map 1 is that this is a 'cartogram' in which each local authority area is scaled according to the number of pupils attending school there. Think of it as a map of school children rather than land. (Hover over the map to see local authority names.)

Somewhat coincidentally, the clustering process resulted in nine separate groups, mirroring the nine regions of England, but there the similarities end. Group 1 ('Metropolitan areas') are situated in large urban conurbations in London and beyond. They have high levels of poverty and low proportions of white British pupils but relatively high educational performance, especially at Key Stage 2. Group 2 ('Northern cities') is composed mainly of built-up areas in the north – though Thurrock to the east of London is included too. These tend to be very urban with high levels of poverty, large proportions of white British pupils and relatively low educational attainment. Group 3 ('Underperforming provinces') contains various provincial towns and cities with modest overall levels of urbanisation but relatively high levels of poverty and low average attainment at all stages of education. Group 4 ('High-performing urban') consists of three areas on the eastern fringes of London with moderate levels of poverty and ethnically mixed populations that perform well academically, especially at primary school. Group 5 ('Affluent shires') contains three wealthy areas to the west of London that show high academic performance at all stages of education. Group 6 ('Affluent urban') consists of two areas to the south of London that are much more urban than Group 5, but show similarly high educational attainment. Group 7 ('Underperforming minorities') contains Leicester and Luton, which both show high levels of poverty and large ethnic minority populations along with relatively low performance at every educational stage.

Group 8 ('Outliers') is a special class which includes 13 local authorities that didn't fit readily into any other group. This is inevitably something of a hotchpotch and includes some poor areas with relatively low educational attainment (Blackpool, Knowsley, Nottingham, Tower Hamlets and Newham), much wealthier areas with higher attainment (Trafford, Rutland and Richmond upon Thames), a couple of areas with medium levels of poverty by relatively poor academic results (Bedford and the Isle of Wight), and Slough, which has a high level of poverty but tends to do quite well educationally. Finally, there's Group 9 ('Other areas'), which contains nearly half of all local authorities. These are perhaps best described as unexceptional: they are broadly similar to each other and in that sense are perhaps the most typical of English local authorities, at least based on the factors analysed here.

Figure 2 shows the same analysis we saw in Figure 1, but by local authority cluster instead of region. For example, it shows that Group 1 ('Metropolitan areas') has high levels of urbanisation and poverty, low proportions of white British pupils and relatively high attainment, especially at Key Stage 2. This is why the algorithm put them in the same cluster. (To explore other indicators and groups, use the menu to select a metric and click on the figure legend to turn individual groups on or off.)

Final analysis

It's important to acknowledge that no analysis of this kind is ever completely objective because different metrics and algorithms will produce different groups. But it is at least as valid as our usual regional groupings – and arguably more useful when thinking about national disparities and policies to address them.

It is interesting that, despite the algorithm knowing nothing about England's geography, most London boroughs cluster together anyway. This suggests that the capital does indeed have special characteristics (as we've observed before). But not unique ones: certain other metropolitan areas such as Birmingham, Manchester and Bradford emerge in the same group. Conversely, some London boroughs fall into other groups – and Thurrock (next to London) clusters with northern cities such as Sheffield and Newcastle. So while these results confirm some familiar observations, they also provide a new way to think about education in England – one that goes beyond the usual distinctions of north versus south or London versus everywhere else.

Over the course of this four-part series (see Parts 1, 2 and 3), we have seen again and again that when it comes to English schooling, location matters. Perhaps the most important question is how to make this less so. Ironically, the answer may lie in making policy less consistent and more responsive to the distinct characteristics and challenges of different parts of the country. We hope that the perspective presented here will inform that debate.

We welcome your views and suggestions. Please send them to: [email protected].

Footnotes:

1. For a list of sources, see the first post in this series.
2. The clustering analysis used the DBSCAN algorithm applied to the following data fields: 'Urbanisation (%)', 'Income Deprivation Affecting Children Index (%)', 'Pupils identifying as white British (%)', 'KS1: Meets expectations in phonics (%)', 'KS2: Expected standard in reading, writing and maths (% of pupils)' and 'Pupils achieving grades 9-4 in English and maths GCSEs (%)'. As with the rest of the analysis in this series, the clustering process omitted two small local authorities, the City of London and the Isles of Scilly, because they do not generate enough data for meaningful comparisons. They were added afterwards to the 'Outliers' group.