The extended project qualification in England: does it provide good preparation for higher education?

ABSTRACT

The Extended Project Qualification (EPQ) is taken by students in England aged 16–18, usually alongside A levels. It consists of an in-depth project in an area of students’ choosing. Students are required to plan the project, research and analyse sources of information, write up their analysis, draw conclusions and produce an evaluation of the processes involved. As such, it is promoted by exam boards as providing the skills required for university study or for work. Universities also seem to value the qualification, with many reducing their offers to students who have achieved a high grade in the EPQ. The main aim of this research was to use national data from England to investigate whether students taking the EPQ were better prepared for higher education (HE) than students not taking it. The results of logistic regression analyses showed that students taking EPQ were more likely to progress to HE, were less likely to drop out, and were more likely to achieve a good degree, than non-EPQ students. This was after controlling for a large number of covariates, including attainment. This provides further evidence that the skills learnt in undertaking an EPQ can be useful for preparing students for HE study.

KEYWORDS:

• Extended project qualification
• progression
• drop-out
• degree classification
• project-based learning

Introduction

The Extended Project Qualification (EPQ) is available for students aged 16 to 18 in England as they undertake sixth form study. It is usually taken alongside other qualifications, such as Advanced (‘A’) levels (the most common qualification taken by students ofthis age). The EPQ is an academic, rather than a vocational, qualification but, unlike most other academic qualifications, it is not examined. Instead, it involves students undertaking a substantial project in an area of their own choosing.

The EPQ comprises a taught element and an independent study element. During the taught element, students learn skills such as project management, research techniques, reflective learning, presentation skills, and effective writing (Oxford, Cambridge and RSA [OCR], 2023). These skills then help to support them whilst undertaking the independent study part of the qualification. For this, they are required to plan the project, research and analyse sources of information, write up their analysis, draw conclusions and produce an evaluation of the processes involved. During this stage, they also receive support from teachers in the form of supervision and verbal feedback. The project can have one of several different final outcomes, such as a report with findings from an investigation, a dissertation, an artefact, or a performance.

As the qualification includes the teaching of project management and research skills and then the application of those skills, it is thought to be excellent preparation for university study. It is promoted as such by examination boards who offer the qualification (e.g. Assessment and Qualifications Alliance [AQA], 2021; OCR, 2023). Universities also value the qualification, with some reducing the grades required (in A levels, for example) to be accepted on a particular course for students who have achieved a high grade in the EPQ (see https://qips.ucas.com/qip/extended-project-qualification-epq).

There already exists some empirical evidence of the benefits to students of taking an EPQ. For example, taking an EPQ was shown to be associated with improved performance in A levels taken at the same time (e.g. Gill, 2017; Jones, 2016). The conclusion from both these articles was that the research skills learned by students in EPQ may have been transferable to their other subjects. Jones investigated the effect on different A level subjects and found that, whilst most subjects showed a significant positive association between taking EPQ and A level performance, there was no significant effect for either mathematics or language subjects. Gill found that the size of the positive effect associated with taking EPQ was small, equivalent to an improvement of a quarter of a grade per A level.

Regarding the effects of the EPQ on higher education (HE) performance, Gill (2018) found that taking an EPQ on top of A levels was associated with an increased probability of achieving at least an upper second-class degree.¹ Similarly, Dilnot et al. (2022) found that taking an EPQ was associated with lower probability of achieving below an upper second, as well as lower probability of dropping out, or of repeating a year. In both these articles, regression models were undertaken which took into account the prior attainment of students (at age 16) and demographic characteristics (e.g. gender, socio-economic status) and other contextual factors such as school type.

Other studies have found qualitative evidence of the benefits of the EPQ in terms of preparing students for university study. Stephenson and Isaacs (2019) surveyed undergraduate students who had taken an EPQ and teachers with experience of supervising the qualification. The students reported that the EPQ stretched them academically and having completed it they felt a sense of achievement and a boost to their confidence as learners. It also gave them some insight into their preferred ways of learning. Being able to focus on an area of particular interest was also motivating for the students. Teachers reported seeing similar improvements to self-confidence in their students. They also believed that the need to be self-sufficient was likely to be beneficial to their learning. These factors are all likely to help students with the transition to higher education.

Williamson and Vitello (2018) surveyed heads of departments (HoDs) in schools and colleges in 2017 and 2018. Their focus was on the impact of planned changes to A levels, but they included some questions about the EPQ. They received responses from 188 HoDs across a range of school types (broadly representative of the distribution of school types in England). The responses showed that the HoDs were positive about the impact of EPQ. Over 95% agreed with the statement ‘EPQ helps to enhance general skills’ and 86% agreed that ‘EPQ is good preparation for university’. A higher proportion thought that EPQ was more useful for high attaining students (over 80%) than thought it was useful for students across the range of achievement (below 70%).

Cripps et al. (2018) reported on a partnership between the University of Bristol and local schools which aimed to support students in developing independent research skills whilst completing their EPQ. As part of the evaluation of the project, former students who had participated in the programme were asked about its long-term impact. They reported that the EPQ worked as a bridge between A levels and their degree study, because of the similarity between what was required in an EPQ and in researching and writing essays at university.

There are other qualifications offered internationally to 16- to 18-year-olds that also claim to be particularly beneficial in terms of preparation for HE study. These include the Cambridge International Global Perspectives & Research AS & A level (https://www.cambridgeinternational.org/programmes-and-qualifications/cambridge-international-as-and-a-level-global-perspectives-and-research-9239/) and the Extended Essay component in the International Baccalaureate (IB) (https://www.ibo.org/programmes/diploma-programme/curriculum/dp-core/extended-essay/). These have similarities to the EPQ in that they require students to undertake a substantial piece of independent research, produce a final written report and reflect on the process. There is limited empirical evidence of the effectiveness of these qualifications in preparing students for HE and this is only for the IB. For example, Davies and Guppy (2022) found that IB students in Canada were less likely than non-IB students to drop out of HE and more likely to achieve at least an upper second-class degree (after accounting for other factors including prior attainment). Duxbury et al. (2021) found similar results when comparing IB students in England with A level students. All IB students are required to take the Extended Essay as part of their diploma, which may explain these findings. As outlined in Hopfenbeck et al. (2020), the IB (and particularly the Extended Essay component) can help with developing critical thinking skills. In that research, they found that IB students (in Australia, England, and Norway) developed better critical thinking skills whilst studying for their diploma than non-IB students. They also reported that teachers believed that the Extended Essay provided good preparation for HE study.

The research cited reveals some evidence (both quantitative and qualitative) that taking EPQ can benefit students, both in their other subjects taken at the same time and in their HE study. However, there is a shortage of studies investigating the specific impact of taking EPQ using data from a whole cohort of students and following them through from the end of schooling into HE. Investigating this is of particular importance currently, as more HE providers are reducing their offers to students with a good grade in EPQ, in the belief that it provides good preparation for HE study.

Thus, the main purpose of the research presented here was to use national data in England to investigate further whether students taking EPQ were better prepared for HE than students not taking it. The following questions were addressed:

1. Are EPQ students more likely than non-EPQ students to progress to HE (after accounting for other factors likely to affect this, such as prior attainment)?

2. Are EPQ students less likely than non-EPQ students to drop out of HE in their first year (after accounting for other factors)?

3. Is taking an EPQ associated with better degree performance (after accounting for other factors)?

Materials and methods

The main source of data for this project was a dataset linking students’ school records in England (from the National Pupil Database) with higher education outcomes (from Higher Education Statistics Agency data). The National Pupil Database (NPD) is administered by the Department for Education (DfE) and includes examination results for all students in all qualifications and subjects in schools and colleges in England, as well as student and school background characteristics such as gender, ethnicity, level of income-related deprivation and school type. The Higher Education Statistics Agency (HESA) data has information on the students who attend universities in the UK. It includes details of the institution attended, the course subject and level, the degree classification obtained (where applicable) and some additional background characteristics, such as socio-economic status and level of parental education.

We used the Key Stage 5² (KS5) extract of the NPD for 2015/16 linked to HESA data in 2016/17, 2017/18 and 2018/19. These were the most recent data available to us at the time the research started. This enabled us to investigate the relationship between taking an EPQ and the probability of progression to HE, the probability of dropping out of HE and the probability of achieving a ‘good’ degree (first class or upper second class). For all the analyses, we restricted the NPD data to students who took at least one qualification equivalent in size to an A level and who were aged 17 or 18 at the start of the academic year. This meant we could focus on students with qualifications which would allow them to progress to HE should they wish to.

For the analysis of progression to HE, we used the NPD data for 2015/16, matched to the HESA data for 2016/17, 2017/18 and 2018/19. This meant we were able to include students who had one or two years of deferment before progressing to HE. Students who were in the NPD data, but not in the HESA data for any of the three years were assumed not to have progressed to HE. It is possible that some of these students progressed in later years, or went to an HE institution in another country, but this is likely to be a small minority.

For the analysis of drop-out from HE, students who were present in the HESA data in one year (e.g. 2016/17) but were not present in the next year (e.g. 2017/18) were counted as having dropped out of HE. This is not a perfect measure, as some of these students may have transferred to a university in a different country or taken a year out, but we assumed that there would only be a very small number of such students. We were able to investigate drop-outs for two different cohorts of students: those who started HE in 2016/17, but were not in the data for 2017/18, and those who started in 2017/18 (i.e. deferred a year), but were not in the data for 2018/19.

For the analysis of degree class achieved, we used the NPD data for 2015/16 matched to the HESA data for 2018/19. This means that this analysis was limited to students who started HE immediately after finishing school and completed their degree in three years. This will include most students, but we acknowledge that some students will not be included in this analysis.

For each of the research questions, descriptive analyses showing patterns of progression to and achievement in HE were undertaken. Then, we carried out regression analyses to fully account for the students’ backgrounds when looking at how well EPQ prepared students for HE.

Regression analysis

To answer each of the research questions, a series of different regression models were fitted.

The first of these was a set of logistic regression models predicting the probability of students who completed their KS5 studies in 2015/16 progressing to HE sometime within the next three years. We used a multilevel model, as this accounted for the clustering of students within schools. For a more detailed description of multilevel logistic regressions see Goldstein (2011). The general form of the model was as follows:

(1) $log\left(\frac{{\mathit{p}}_{\mathit{ij}}}{1-p_{\mathit{ij}}}\right)={\beta }_0+{\beta }_1{x}_{1\mathit{ij}}+{\beta }_2{x}_{2\mathit{ij}}+\cdots +{\beta }_l{x}_{\mathit{lij}}+u_j+{\epsilon }_{\mathit{ij}}$(1)

where $p_{\mathit{ij}}$ is the probability of student $\mathit{i}$ from school $\mathit{j}$ progressing to HE, $x_{1\mathit{ij}}$ to $x_{\mathit{lij}}$ are the independent variables (gender, prior attainment etc.), ${\beta }_0$ to ${\beta }_l$ are the regression coefficients, $u_j$ is a random variable at school level, and ${\epsilon }_{\mathit{ij}}$ is the residual term.

The second set of logistic regression models predicted the probability of a student dropping out of HE in their first year. There were two separate hierarchies within this data, with students clustered in schools and students clustered in HE institutions. This was accounted for by using a cross-classified multilevel model. The general form of the model was:

(2) $log\left(\frac{{\mathit{p}}_{\mathit{ijk}}}{1-p_{\mathit{ijk}}}\right)={\beta }_0+{\beta }_1{x}_{1\mathit{ijk}}+{\beta }_2{x}_{2\mathit{ijk}}+\cdots +{\beta }_l{x}_{\mathit{lijk}}+u_j+u_k+{\epsilon }_{\mathit{ijk}}$(2)

where $p_{\mathit{ijk}}$ is the probability of student $\mathit{i}$ from school $\mathit{j}$ and attending HE institution $\mathit{k}$ dropping out of HE, $x_{1\mathit{ijk}}$ to $x_{\mathit{lijk}}$ are the independent variables, ${\beta }_0$ to ${\beta }_l$ are the regression coefficients, $u_j$ is a random variable at school level, $u_k$ is a random variable at the HE institution level, and ${\epsilon }_{\mathit{ijk}}$ is the residual term.

The third set of models predicted the probability of achieving a first-class degree (and separately the probability of achieving at least an upper-second class degree). A cross-classified multilevel model was employed with students nested in schools and in HE institutions. The general form of the model was as in equation (2)(2) $log\left(\frac{{\mathit{p}}_{\mathit{ijk}}}{1-p_{\mathit{ijk}}}\right)={\beta }_0+{\beta }_1{x}_{1\mathit{ijk}}+{\beta }_2{x}_{2\mathit{ijk}}+\cdots +{\beta }_l{x}_{\mathit{lijk}}+u_j+u_k+{\epsilon }_{\mathit{ijk}}$(2) , but with $p_{\mathit{ijk}}$ being the probability of student $\mathit{i}$ from school $\mathit{j}$ achieving a first (or, separately, at least an upper second) in HE institution $\mathit{k}$.

In each of the three sets of models the main variable of interest was whether or not a student took the EPQ. A statistically significant coefficient for this variable would indicate that taking an EPQ had a significant association with the outcome variable. We also ran additional models with the grade achieved in EPQ as a predictor variable. The purpose of these models was to investigate whether achieving a higher grade in the EPQ had a significant association (over and above just taking the qualification). It is important to note that finding a significant association does not mean that taking an EPQ ‘causes’ better outcomes. There may be other unobserved variables (for example motivation) which are related to both the probability of taking an EPQ and HE outcomes.

The regression models were fitted using the glmer function from the lme4 package (Bates et al., 2015) in the R programming language.

For each regression model, contextual variables which were likely to have had an impact on the outcome variable were included. These were: gender, concurrent attainment, deprivation, ethnic group, first language, special educational needs (SEN) status, school type, school gender, and school mean KS5 attainment. For the models predicting the probability of drop-out or the probability of achieving a good degree, we added students’ socioeconomic classification, their parents’ level of education, and the degree subject (all of which were in the HESA data).

Whilst the relationship between these students’ characteristics and the outcomes of interest is out of scope in this work, it was important that the contextual information was included in the models because it meant we could be more confident that any significant association between taking EPQ and the outcome variable was genuine and not down to differences in the other factors. They were all characteristics which previous research (e.g. Chowdry et al., 2013; Gill, 2018; Vidal Rodeiro, 2019) found to be significant factors in determining the likelihood of progression to HE, drop-out, and degree class achieved.

The measure of concurrent attainment used was the students’ average KS5 points score. This variable was available in the NPD data and was calculated by assigning a points score to each achieved grade³ and averaging this across all KS5 qualifications (at least equivalent in size to an A level⁴) taken by a student. The measure, therefore, excluded the grade achieved in EPQ (for those students who took it), as this is equivalent in size to half an A level.

Student deprivation was measured by the Income Deprivation Affecting Children Index (IDACI), which indicated the proportion of children in a very small geographical area (Lower Layer Super Output Area or LSOA) living in low-income families (Smith et al., 2015). This variable was available in the NPD and varied between 0 and 1. However, this measure only indicated how income deprived the area was that a student lived in, not how income deprived the student was.

We used the ethnicity categories in the NPD to group students into seven groups: Asian, black, Chinese, mixed, white, other, and unclassified. Chinese students were in a category of their own (rather than grouped with other Asian students) due to a well-known tendency to perform very well in exams in England in comparison to other Asian students. Students were also grouped by their first language (English or other).

For grouping students according to SEN, we used the categories in the NPD. These were ‘No SEN’, ‘SEN, no statement’, and ‘SEN, with statement’, with the last of these indicating children requiring the most support.⁵

These four variables (deprivation, ethnicity, first language, and SEN) were reported by schools to the DfE as part of the school census returns. However, some types of schools (independent schools and colleges) were not required to provide this information, leading to large amounts of missing data. This meant that, for each of the three sets of analysis (progress to HE, drop-out from HE, degree performance) approximately 50% of students had missing data on these four variables. Students with missing data for any of these variables were excluded from the regression analyses.

For the analysis by school type, schools were grouped into five categories: comprehensive schools, colleges (further education/tertiary/sixth form), independent schools, selective schools, and other schools. Most students in England attend comprehensive schools or colleges which are free and usually non-selective. A minority of students attend selective schools, which only admit high-ability students, or fee-paying independent schools. Due to the missing school census data outlined above, the main regression analyses did not include any students attending independent schools or many of the students attending colleges. As an illustration of the effect of the missing data, Table 1 shows the number of students attending each school type, split by whether they took EPQ and whether they were included in the regression analysis (of progression to HE).

Table 1. Number of students in KS5 cohort, split by school type, EPQ, and inclusion in regression models or not.

School type	EPQ?	Students	Included in regression model	Not included in regression model	% in model	% not in model
Colleges	No	166956	113	166843	<0.1	>99.9
	Yes	10254	<10	SUPP	<0.1	>99.9
Comp./Academy	No	136876	133361	3515	97.4	2.6
	Yes	18066	17768	298	98.4	1.6
Independent	No	33366	<10	SUPP	<0.1	>99.9
	Yes	4081	<10	SUPP	<0.1	>99.9
Other	No	9627	8697	930	90.3	9.7
	Yes	878	788	90	89.7	10.3
Selective	No	18398	18162	236	98.7	1.3
	Yes	5125	5079	46	99.1	0.9

Schools were also categorised by their ‘gender’ (i.e. boys only, girls only, or mixed), which we derived from the percentage of girls in each school. If this was greater than 95% then the school was categorised as a girls’ school, if it was less than 5% it was categorised as a boys’ school. Otherwise, it was categorised as a mixed school.

For the school KS5 attainment measure, we calculated the average KS5 points score (as described above) amongst students at the end of KS5 in each school.

The socioeconomic classification variable in the HESA data indicated the classification of the student if they were 21 years old or over or the classification of their parents if under 21. The categories used are standard categories used in the UK census, which run from 1 (‘Higher managerial & professional occupations’) to 8 (‘Never worked & long-term unemployed’), with 9 indicating ‘not classified’ (including students). For a full list of the different categories, see the HESA website (https://www.hesa.ac.uk/collection/student/datafutures/a/entryprofile_sec).

The HESA data also included a variable indicating whether students’ parents had a HE qualification.

Finally, the degree subject group was included in some of the regression models. HESA used a system called JACS (Joint Academic Coding System) to classify subjects into one of 18 different subject groups (https://www.hesa.ac.uk/support/documentation/jacs/jacs3-detailed). For students taking combinations of subjects in different subject groups we applied the following rule: if the percentage of the course within one subject group was greater than 50%, then assign the student to that group; otherwise assign the student to an additional group called ‘Combined’.

Students with missing data for any of these variables were excluded from the regression models. However, all students were included in tables of descriptive data.

Results

Are EPQ students more likely than non-EPQ students to progress to HE?

Descriptive statistics show that 88.5 per cent of the EPQ students (amongst those who finished KS5 in 2016/17) went on to HE within the next three academic years, compared with 66.8 per cent of students who did not take an EPQ (see Table 2). However, a higher proportion of non-EPQ students (17.1%) than EPQ students (14.1%) progressed to HE after deferring for one or two years.

Table 2. EPQ and non-EPQ students progressing to HE.

Year started HE	EPQ students	EPQ %	Non-EPQ students	Non-EPQ %
2017	28,602	74.5	182,397	49.7
2018	4,757	12.4	51,586	14.1
2019	649	1.7	10,956	3.0
All years	34,008	88.5	244,939	66.8
Did not progress	4,410	11.5	121,939	33.2

Table 3 presents the number and the percentage of students progressing to HE by their EPQ grade. This shows higher progression rates to HE amongst students with the highest EPQ grades (except for those achieving a grade X⁶).

Table 3. Students progressing to HE, by EPQ grade.

EPQ grade	No. of students	% of these students progressing
A*	6,897	95.4
A	9,176	94.0
B	8,232	90.3
C	6,629	84.6
D	3,793	80.2
E	1,890	74.8
U	832	68.4
X	969	76.3
Not taken	366,927	66.8

However, the EPQ grade had a substantial correlation with the A level mean (Pearson correlation coefficient = 0.462, n = 38,418) which may partly explain why progression was higher amongst students achieving higher grades.

The results of the regression analyses looking at the probability of progressing to HE are presented in Table 4. This shows the parameter estimates with their standard errors in brackets. Statistical significance (at the 5% level) is indicated by an asterisk. We had two main models. Model 1 includes an indication of whether EPQ was taken or not, the contextual variables that we needed to control for, and any significant interaction effects between EPQ and the contextual variables. Model 2 includes the grade achieved at EPQ instead of the binary indicator, along with the contextual variables, but no interaction effects.

Table 4. Results of regression models, predicting the probability of progressing to HE.

Variable		Parameter estimate (std. error in brackets) Model 1 (n = 183,979)	Parameter estimate (std. error in brackets) Model 2 (n = 183,979)
Intercept		0.949 (0.019)*	0.953 (0.019)*
Taken EPQ	No
	Yes	0.997 (0.029)*
EPQ grade	No EPQ
	A*		1.462 (0.088)*
	A		1.356 (0.061)*
	B		1.137 (0.052)*
	C		0.798 (0.047)*
	D		0.660 (0.056)*
	E		0.444 (0.076)*
	U		0.185 (0.112)
	X		0.414 (0.124)*
Mean KS5 points score		0.017 (0.000)*	0.017 (0.000)*
Gender	Female
	Male	−0.170 (0.013)*	−0.166 (0.013)*
IDACI score		−0.696 (0.060)*	−0.691 (0.060)*
Ethnic group	White
	Asian	0.889 (0.031)*	0.865 (0.031)*
	Black	1.262 (0.037)*	1.243 (0.037)*
	Chinese	0.934 (0.112)*	0.933 (0.112)*
	Mixed	0.468 (0.033)*	0.458 (0.032)*
	Other	0.868 (0.064)*	0.841 (0.062)*
	Unclassified	0.355 (0.062)*	0.363 (0.060)*
First language	English
	Other	0.398 (0.026)*	0.398 (0.026)*
	Unclassified	0.008 (0.108)	0.012 (0.108)
SEN status	None
	SEN, no statement	−0.187 (0.029)*	−0.183 (0.029)*
	SEN, with statement	−0.148 (0.069)*	−0.146 (0.070)*
School type	Comp./College
	Selective	0.859 (0.065)*	0.819 (0.064) *
School gender	Mixed
	Boys	−0.057 (0.091)	−0.052 (0.090)
	Girls	0.132 (0.072)	0.118 (0.071)
School mean KS5 points score		0.005 (0.001)*	0.005 (0.001)*
Taken EPQ*ethnic group	White
	Asian	−0.362 (0.095)*
	Black	−0.358 (0.141)*
	Chinese	−0.017 (0.439)
	Mixed	−0.136 (0.132)
	Other	−0.418 (0.230)
	Unclassified	0.123 (0.263)
Taken EPQ*school type	Comp./College
	Selective	−0.229 (0.085)*

For both these regression models, it was necessary to combine some of the school type categories. This was because the models failed to converge when all categories were used. We combined comprehensive schools, colleges, and other schools into one category and kept selective schools as a separate category.

The results of Model 1 show a significant positive association (0.997) between taking EPQ and the likelihood of progressing to HE. Interpretation of parameter estimates in logistic regression models is not straightforward, as they represent the log of the odds of achieving the outcome variable (i.e. progressing to HE). However, we can make comparisons by converting the results to predicted probabilities of progressing for specific groups of students. These students will be referred to in the rest of the article as ‘typical’ students, who were in the base/reference category for all categorical variables, and with a value of the continuous variables (mean KS5 points score, IDACI score and centre mean KS5 point score) equal to the mean amongst all students.

Figure 1 presents the probabilities for a typical student, with different mean KS5 points scores, and whether they took EPQ or not.

Figure 1. Predicted probabilities of progressing to HE by EPQ, and KS5 mean points score.

This illustrates the size of the difference in probability of progression between EPQ and non-EPQ students. For example, for students with a mean KS5 points score equal to the mean amongst all students (220, equivalent to one B grade and two C grades at A level), the probability was 0.87 for EPQ students and 0.72 for non-EPQ students.

We also ran a model (not presented here) which omitted the census variables (IDACI score, ethnic group, first language, and SEN status) because, as mentioned earlier, these had a lot of missing data and this led to the exclusion of a high number of students from the regression analyses. This additional model, therefore, had a much larger number of students (n = 403,725). It allowed us to check whether excluding the students with missing data changed the results in a meaningful way. Reassuringly, there was very little difference in the parameter estimate for the EPQ variable (0.981, compared with 0.997).

The results of Model 2 showed that higher EPQ grades were associated with greater likelihood of progressing to HE. Compared with students not taking EPQ, every grade had a significant and positive impact on the probability of progressing (apart from grade U). This effect was over and above any effect of KS5 attainment, as this was accounted for in the model.

Figure 2 shows the probabilities for typical students achieving each EPQ grade. There was very little difference in the probabilities for grades A* to B, which suggests that there was not much additional effect of achieving above a grade B. There was a significant positive effect of only achieving a grade X, compared with not taking EPQ. This may be because even starting (but not finishing) an EPQ was beneficial.

Figure 2. Predicted probabilities of progressing to HE by EPQ grade.

There were two interaction effects in Model 1 which were statistically significant. The first of these was between ethnic group and taking EPQ. The negative parameter estimates for Asian (−0.362) and for black (−0.358) students mean that the positive association between taking EPQ and progressing to HE was less for these students than for the reference group (white students).

The second significant interaction was between school type and taking EPQ, with the negative estimate for selective schools (−0.229) indicating that the effect of taking EPQ was less for students in these schools.

Are EPQ students more likely than non-EPQ students to drop out in year 1?

Table 5 presents the number and percentage of EPQ and non-EPQ students dropping out from HE in year 1. The results show that non-EPQ students were more likely to drop out in their first year (5%) than EPQ students (2.3%).

Table 5. EPQ and non-EPQ students dropping out, by year started HE.

Year started HE	EPQ students dropping out	EPQ %	Non-EPQ students dropping out	Non-EPQ %
2017	646	2.3	9,002	4.9
2018	128	2.7	2,715	5.3
All	774	2.3	11,717	5.0
Did not drop out	32,585	97.7	222,266	95.0

Table 6, which presents the number and the percentage of students dropping out by their EPQ grade, shows that students with higher EPQ grades were less likely to drop out, and this pattern held across all grades. However, as mentioned earlier, the EPQ grade correlated substantially with the A level mean which may partly explain why drop-out rates were lower amongst students achieving higher grades.

Table 6. Drop-out rate by EPQ grade (year 1).

EPQ grade	No. of students	% of these students dropping out
A*	6,520	1.2
A	8,498	1.6
B	7,289	2.0
C	5,478	3.0
D	2,949	4.0
E	1,366	4.7
U	548	4.9
X	711	5.9
Not taken	233,998	5.0

The results of the regression analyses looking at the probability of dropping out from HE are presented in Table 7. These models included extra variables (degree subject, socioeconomic classification, and parental education), which were available in the HESA data. In Table 7 (and all subsequent tables), we do not report the parameter estimates for the degree subject or the socio-economic classification due to the large number of different categories in each (19 and 8 respectively).

Table 7. Results of regression models, predicting the probability of drop-out from HE in year 1.

Variable		Parameter estimate (std. error in brackets) Model 1 (n = 135,362)	Parameter estimate (std. error in brackets) Model 2 (n = 135,362)
Intercept		−4.336 (0.270)*	−4.340 (0.270)
Taken EPQ	No
	Yes	−0.524 (0.055)*
EPQ grade	No EPQ
	A*		−0.557 (0.156)*
	A		−0.689 (0.122)*
	B		−0.692 (0.112)*
	C		−0.509 (0.107)*
	D		−0.304 (0.123)*
	E		−0.068 (0.164)
	U		−0.233 (0.278)
	X		0.335 (0.222)
Mean KS5 points score		−0.004 (0.001)*	−0.004 (0.000)*
Gender	Female
	Male	0.250 (0.032)*	0.247 (0.032)*
IDACI score		1.162 (0.123)*	1.160 (0.123)*
Ethnic group	White
	Asian	−0.464 (0.059)*	−0.464 (0.059)*
	Black	−0.412 (0.066)*	−0.416 (0.066)*
	Chinese	−0.434 (0.205)*	−0.434 (0.205)*
	Mixed	−0.171 (0.071)*	−0.172 (0.071)*
	Other	−0.470 (0.119)*	−0.473 (0.119)*
	Unclassified	0.029 (0.125)	0.021 (0.125)
First language	English
	Other	−0.188 (0.051)*	−0.187 (0.051)*
	Unclassified	−0.348 (0.257)	−0.347 (0.257)
SEN status	None
	SEN, no statement	0.019 (0.070)	0.019 (0.070)
	SEN + statement	0.177 (0.157)	0.174 (0.157)
Degree subject	**
Socioeconomic class.	**
Parent educated to HE?	Yes
	No	0.258 (0.035)*	0.257 (0.035)*
	Unknown/unavailable	0.368 (0.048)*	0.366 (0.048)*
School type	Comp
	College	0.325 (0.612)	0.346 (0.612)
	Selective	−0.441 (0.072)*	−0.440 (0.072)*
	Other	0.113 (0.076)	0.111 (0.076)
School gender	Mixed
	Boys	−0.115 (0.111)	−0.114 (0.111)
	Girls	−0.165 (0.085)	−0.166 (0.085)
School mean KS5 points score		0.004 (0.002)*	−0.002 (0.001)
Taken EPQ*mean KS5 points score		−0.004 (0.001)*

As with the analysis of progression, we fitted two different regression models, one including a binary indicator of taking EPQ or not and one including the EPQ grade. In these models, a negative parameter estimate indicates a lower probability of dropping out.

In Model 1 the parameter estimate for the EPQ indicator was −0.524. This means that taking EPQ was associated with a lower probability of dropping out. However, the model also shows that there was a significant interaction between taking an EPQ and the students’ concurrent attainment, measured by the KS5 mean points score. To better understand the effect of this interaction, Figure 3 presents the predicted probabilities of dropping out (for a typical student as defined before⁷), by KS5 mean points score and whether EPQ was taken. This shows that the (negative) association between taking EPQ and dropping out was less for students of lower ability. In fact, for students with very low mean KS5 points score (below 90), the likelihood of dropping out was higher for EPQ students. However, there were very few students in the data with such low points scores (275 students, of which 25 dropped out).

Figure 3. Predicted probabilities of drop-out in year 1 by EPQ, and KS5 mean points score.

Overall, the difference in the probability of drop-out between EPQ and non-EPQ students was very small, despite it being statistically significant. For an EPQ student with a KS5 mean of 230 (equal to the mean amongst all students progressing to HE) the probability of dropping out was 0.020, compared with 0.035 for non-EPQ students.

As before, we ran an additional model which excluded census variables (n = 261,075). However, as the result of interest in this model was very similar to the result in Model 1 (EPQ parameter estimate = −0.461), we were satisfied that the exclusion of students with missing data on the census variables had little effect on the outcomes.

Figure 4 presents the predicted probability by EPQ grade (derived from Model 2). This was for typical students with a mean KS5 points score of 230 (equivalent to two B grades and one C grade at A level). This shows that the differences in drop-out rates were very small, although they were significant for grades A* to D, compared with not taking EPQ.

Figure 4. Predicted probabilities of drop-out in year 1 by EPQ grade.

Is taking an EPQ associated with better degree performance?

For the final set of analyses, we looked at the probability of achieving a good degree (first-class or upper second-class), by whether an EPQ was taken. Table 8, which presents the breakdown of degree class for EPQ and non-EPQ students, shows that EPQ students were more likely to achieve a first or an upper second than students without an EPQ.

Table 8. Degree class distribution for EPQ and non-EPQ students.

Degree class	EPQ students	EPQ %	Non-EPQ students	Non-EPQ %
First-class	4,843	30.97	22,437	24.55
Upper second-class	8,869	56.72	50,343	55.09
Lower second-class	1,776	11.36	16,748	18.33
Third-class	136	0.87	1,740	1.90
Unclassified	13	0.08	108	0.12
All	15,637	100.00	91,376	100.00

Regarding HE performance and EPQ grade, there was a fairly consistent pattern with higher grades indicating a higher percentage of students achieving a good degree (Table 9). As before, the correlation between EPQ grade and A level performance might partially explain these results.

Table 9. Achieving a first, by EPQ grade.

EPQ grade	No. of students	% of these students achieving first-class	% of these students achieving at least upper second-class
A*	2,928	45.8	96.0
A	4,014	33.9	91.7
B	3,647	28.4	88.1
C	2,634	24.3	81.8
D	1,364	21.3	79.2
E	558	14.7	71.9
U	218	17.4	78.0
X	274	19.3	73.7
Not taken	91,378	24.6	79.7

Table 10 presents the results of the regression models predicting the probability of achieving a first or at least an upper second (also known as a ‘two-one’, 2:1, 2.i or 2(i)). For both dependent variables, we ran two separate models: Model 1 with the EPQ indicator, and Model 2 with the EPQ grade. The same contextual variables as in the models described in the previous section (Table 7) were included in all the regression models.

Table 10. Results of regressions models, predicting the probability of achieving a first and at least an upper second-class degree.

Variable		Parameter estimate (std. error in brackets). First Model 1 (n = 58,414)	Parameter estimate (std. error in brackets). First Model 2 (n = 58,414)	Parameter estimate (std. error in brackets). At least 2(i) Model 1 (n = 58,414)	Parameter estimate (std. error in brackets). At least 2(i) Model 2 (n = 58,414)
Intercept		−0.830 (0.196)*	−0.864 (0.197)*	1.699 (0.295)*	1.640 (0.294)*
Taken EPQ	No
	Yes	0.168 (0.034)*		0.306 (0.037)*
EPQ grade	No EPQ
	A*		0.718 (0.055)*		1.183 (0.132)*
	A		0.356 (0.046)*		0.572 (0.076)*
	B		0.115 (0.050)*		0.346 (0.066)*
	C		0.030 (0.059)		0.013 (0.066)
	D		−0.112 (0.088)		−0.011 (0.089)
	E		−0.438 (0.148)*		−0.478 (0.124)*
	U		−0.502 (0.235)*		−0.086 (0.215)
	X		−0.386 (0.221)		−0.352 (0.201)
Mean KS5 points score		0.019 (0.000)*	0.019 (0.000)*	0.012 (0.000)*	0.012 (0.000)*
Gender	Female
	Male	−0.247 (0.025)*	−0.218 (0.023)*	−0.474 (0.026)*	−0.468 (0.026)*
IDACI score		−0.951 (0.099)*	−0.953 (0.100)*	−1.233 (0.105)*	−1.235 (0.105)*
Ethnic group	White
	Asian	−0.257 (0.040)*	−0.252 (0.040)*	−0.313 (0.042)*	−0.311 (0.043)*
	Black	−0.799 (0.059)*	−0.788 (0.059)*	−0.662 (0.049)*	−0.647 (0.049)*
	Chinese	−0.172 (0.108)	−0.163 (0.108)	−0.361 (0.117)*	−0.351 (0.117)*
	Mixed	−0.194 (0.051)*	−0.194 (0.051)*	−0.218 (0.056)*	−0.216 (0.056)*
	Other	−0.158 (0.084)	−0.147 (0.084)	−0.174 (0.085)*	−0.166 (0.085)
	Unclassified	−0.255 (0.102)*	−0.242 (0.102)*	−0.301 (0.107)*	−0.289 (0.107)*
First language	English
	Other	−0.188 (0.038)*	−0.188 (0.038)*	−0.183 (0.038)*	−0.182 (0.038)*
	Unclassified	−0.028 (0.172)	−0.024 (0.172)	−0.007 (0.185)	−0.006 (0.186)
SEN status	None
	SEN, no statement	−0.192 (0.057)*	−0.185 (0.057)*	−0.299 (0.059)*	−0.294 (0.059)*
	SEN + statement	−0.078 (0.146)	−0.073 (0.146)	−0.132 (0.158)	−0.128 (0.158)
Degree subject	**
Socioeconomic classification	**
Parent educated to HE?	Yes
	No	−0.095 (0.023)*	−0.094 (0.023)*	−0.037 (0.027)	−0.036 (0.027)
	Unknown/unavailable	−0.116 (0.036)*	−0.111 (0.036)*	−0.170 (0.039)*	−0.162 (0.039)*
School type	Comp.
	College	−0.180 (0.662)	−0.221 (0.664)	−0.515 (0.536)	−0.585 (0.537)
	Selective	0.321 (0.040)*	0.317 (0.041)*	0.428 (0.052)*	0.423 (0.052)*
	Other	−0.231 (0.060)*	−0.229 (0.060)*	−0.135 (0.064)*	−0.129 (0.064)*
School gender	Mixed
	Boys	−0.034 (0.072)	−0.026 (0.072)	0.023 (0.085)	0.025 (0.085)
	Girls	−0.040 (0.048)	−0.041 (0.048)	0.020 (0.059)	0.015 (0.059)
School mean KS5 points score		−0.009 (0.001)*	−0.009 (0.001)*	−0.006 (0.001)*	−0.006 (0.001)*
Taken EPQ*Mean KS5 points score	No
	Yes	0.002 (0.001)*		0.004 (0.001)*
Taken EPQ*gender	No – Female
	Yes – Male	0.124 (0.054)*

There was a significant positive association between taking EPQ and both outcome variables (achieving a first and achieving at least an upper second). The grade achieved in EPQ was also a significant factor for both dependent variables, with grades A*, A and B associated with an increased probability and grade E associated with a reduced probability of achieving a first (or at least an upper second) compared with not taking EPQ.

Figure 5 presents the predicted probabilities of achieving a first and at least an upper second for EPQ and non-EPQ students, by KS5 points score. These probabilities were for a typical student, and they took account of the significant interaction effect between taking an EPQ and the mean KS5 points score. For both outcome variables, the interaction effect was positive but very small and can be seen by the slightly larger gap between the lines at higher mean KS5 points scores. At very low mean KS5 points scores the effect was reversed so that taking an EPQ was associated with a lower probability. However, there were very few students (230) with such low points scores achieving at least an upper second.

Figure 5. Predicted probabilities of achieving a first or at least an upper second by EPQ, and KS5 points score.

Although significant, the size of the association between taking EPQ and achieving a first (or at least an upper second) was not large. EPQ students with a mean KS5 points score of 230 had a predicted probability of achieving a first of 0.38 and the probability of at least an upper second of 0.91, compared to 0.34 and 0.87 respectively for non-EPQ students.

The predicted probabilities by EPQ grade for a typical student with a mean KS5 points score of 230 are show in Figure 6. This shows that the differences in the probabilities of achieving a good degree (first or at least upper second) were quite small, despite some of them being statistically significant (see Table 10).

Figure 6. Predicted probabilities of achieving a 1st and at least a 2(i) by EPQ grade.

Table 10 also showed that, for the probability of achieving a first, there was a significant interaction effect between taking an EPQ and gender. This positive effect indicated that the association between taking EPQ and achieving a first was larger for male students than for females. This effect was not large; the probability for (typical) non-EPQ students was 0.43 for females and 0.37 for males. For EPQ students the probabilities were 0.47 and 0.44 respectively.

As before, to check for the effects of excluding students with missing data in the census variables, we ran some additional models (n = 104,875). These showed small increases in the size of the EPQ parameter estimates (0.297 for a first, 0.383 for at least an upper second). However, they did not change the overall finding of an increased likelihood of achieving a good degree for students taking an EPQ.

Discussion

The main purpose of the research presented here was to investigate whether students taking an EPQ were better prepared for HE than similar students not taking an EPQ. No previous studies looked specifically at the impact of EPQ on HE outcomes using data from a whole cohort of students.

We found that students taking an EPQ were more likely to progress to HE (88.5% within the next 3 years) than those not taking the qualification (66.8%). This pattern persisted across different groups of students (defined by their background characteristics, such as gender or ethnic group) and even after accounting for factors that might affect the likelihood of progression.

The size of this association between EPQ and the likelihood of progressing to HE was substantial. According to the regression model, a typical student had a probability of progressing of 0.87 if they took EPQ and 0.72 if they didn’t. There was also evidence that achieving a higher EPQ grade was associated with greater likelihood of progressing, after accounting for other factors including KS5 attainment.

The association between taking EPQ and progression varied slightly by ethnicity and school type. In particular, it was larger for white students than for students from any other ethnic background and was larger for students attending college or comprehensive schools than for those attending selective schools. Both these interactions suggest that the EPQ had more of an impact on students who were less likely to progress (i.e. white students attending a college or comprehensive school). This may be because teachers were aware that EPQ could be useful in preparing students for HE and therefore encouraged some of the less able students who wanted to progress to HE to take the qualification. This finding suggests that EPQ may be a useful qualification in terms of widening participation in HE for socio-economically disadvantaged students. This has been a stated aim of government policy in the UK for many years (Chowdry et al., 2013).

The results of the analysis of drop-out from HE found that students taking an EPQ were less likely to drop out in year 1 (2.3%) than non-EPQ students (5.0%) and that achieving a higher grade in the EPQ was associated with lower probability of dropping out. These findings were confirmed by the regression analyses, which accounted for students’ backgrounds and their attainment at KS5. The differences in predicted drop-out rates between EPQ and non-EPQ typical students were small. However, if we think about the differences in relative terms then taking EPQ reduced the probability of drop-out by more than 50% compared to not taking the EPQ. These results confirm the findings from previous research (e.g. Dilnot et al., 2022).

EPQ students were more likely to achieve a good degree (31% achieved a first and 87.7% at least an upper second) than non-EPQ students (24.6% and 79.6% respectively). The association between taking the EPQ and increased likelihood of achieving a good degree persisted even after accounting for students’ background, their KS5 attainment and their degree subject area. This result supports the findings from previous research (e.g. Dilnot et al., 2022; Gill, 2018), which showed that students taking EPQ had a higher probability of achieving a first or at least an upper second than students taking A levels only. It also aligns with previous findings on the benefits of taking the IB (and particularly the Extended Essay component) on HE outcomes, with IB students more likely than non-IB students to achieve a first, or at least an upper second (Davies & Guppy, 2022; Duxbury et al., 2021).

There was also evidence that students achieving a higher grade in their EPQ were more likely to get a good degree. For example, a typical student achieving a grade A* (and with a mean KS5 score of 230) had a predicted probability of a first of 0.51, compared with just 0.25 for a student achieving a grade E.

Overall, the results presented here suggest that EPQ students were better prepared for HE than non-EPQ students, that is, they were slightly less likely to drop out and more likely to achieve a first or at least an upper second-class degree. As such, these results can help to explain the lower offers of some universities made to students taking the EPQ. Universities want to enrol students who are likely to succeed, and this research has shown that the EPQ is potentially one way of identifying such students.

However, the results of this research come with a substantial caveat. We cannot be sure of a causal relationship between taking the EPQ and increased likelihood of progression to HE or achieving better outcomes. In terms of progression, it may be that students chose to do the EPQ because they had already decided to attend HE and they believed it would mean that they were better prepared. For these students, taking the EPQ did not increase their chances of progressing. In terms of drop-out rates and achieving good degrees, it may be that academically motivated students were more likely to take an EPQ and it is this motivation which led to better outcomes at HE, rather than taking the EPQ per se. Unfortunately, motivation is something that is not easily measured and, therefore, is difficult to control for in studies of this type.

Acknowledgements

This work was produced using statistical data from ONS. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Tim Gill

Tim Gill is a research officer in Cambridge University Press and Assessment’s Research Division. His main areas of interest are the application of multilevel models in analysing assessment data, and the analysis of patterns of uptake and attainment in schools and colleges.

Notes

1. UK degree classifications are as follows: First-Class, Upper Second-Class (2(i)), Lower Second-Class (2(ii)), Third-Class.

2. Key Stage 5 is a label to describe the two years of education undertaken by students in England aged 16–18.

3. For example, the scores for A level grades were: A* = 300, A = 270, B = 240, C = 210, D = 180, E = 150, U = 0.

4. Most students in England take A levels, but some take other qualifications (e.g. International Baccalaureate, Cambridge Technicals, BTECs).

5. A ‘statement’ of special educational needs is a legal document which outlines the educational needs of the child and how they will be met by the local education authority.

6. A grade X indicates ‘no result’ and could be for several reasons, including the candidate failing to complete work for all components, failing to provide an internal assessment sample, an incorrect combination of components, or a script not being available to be marked (Oxford, Cambridge and RSA [OCR], 2016).

7. For the additional variables included in these models we need to select the base category. For degree subject group, we selected Biological Sciences (the largest group); for parental education we selected ‘Yes’ to parents having HE qualifications; for socioeconomic classification we selected group 2 – ‘Lower managerial & professional occupations’.