FEBRUARY 2016

Knowledge Exchange in UK Universities Results from a Panel of Academics 2005 - 2015

2

Writen by: Cornelia Lawson, Alan Hughes, Ammon Salter and Michael Kitson with Anna Bullock and Robert B. Hughes

2 ••

Authors and Acknowledgements: This study was commissioned through the National Centre for Universities and Business (NCUB) and funded by the Arts and Humanities Research Council, the Department for Business, Innovation & Skills, the Economic and Social Research Council, the Engineering and Physical Sciences Research Council, the Higher Education Funding Council for England, the Medical Research Council, and the Natural Environment Research Council. The core project team consisted of Prof. Alan Hughes (Principle Investigator (PI) Imperial College Business School, London, Michael Kitson (PI), Cornelia Lawson (Research Fellow and Analytical and Editorial Work Lead), Anna Bullock (Database Manager) and Robert Hughes (Database Associate) all at the Centre for Business Research (CBR) University of Cambridge and Ammon Salter (PI) at the University of Bath. Isobel Milner provided valuable survey management support in the initial stages of the project and Robert Hughes oversaw and took part with Jessica Burgess, Caroline Druitt, David Haines, Oliver Rubinstein Baylis and Kit Westlake in the core task of hand collecting a total of over 140,000 email addresses of the academics constituting the sample frame. Michelle Osmond of Yellow Sunday Ltd developed the survey instrument and survey platform software and provided invaluable software and technical support throughout the survey process. The final steering group for the project which was responsible for oversight, review and final sign-off of the report, included Jeremy Neathey, ESRC, Ian Viney, MRC, Alex Herbert, HEFCE, Sue Smart, EPSRC, Philip Heads, NERC, Sumi David, AHRC, Dominic Rice, BIS Carolyn Reeve, BIS and Bev Sherbon of MRC. The survey team is extremely grateful to them for their comments, encouragement and support throughout the development of the survey and of this report. The core team also owes a large debt of gratitude to Rosa Fernandez of NCUB who chaired the Steering Group for her sustained and helpful project management and for many valuable comments made during the preparation of this report. The project team is, of course, hugely indebted to the tens of thousands of academics who responded to the survey.

Citation Reference: Lawson, C., Hughes, A., Salter, A., and Kitson, M. with Bullock, A. and Hughes, R.B. (2016) ‘Knowledge Exchange in UK Universities: Results from a Panel of Academics 2005 - 2015’, NCUB, London.

03 ••

Contents

EXECUTIVE SUMMARY

05

Section A: Introduction

07

Section B: The CBR Academic Surveys 2008/9 and 2015

08



Coverage, Sampling and Response

08



Data Processing and Non-response Analysis

09



Panel Description

13

Section C: Introduction to External Engagement

14

Section D: Persistence in External Engagement

18



Engagement Transition Between 2008/9 and 2015

18



Motivations for and Attitudes to Engagement

22

Section E: Summary

26

Section F: Policy implications

27

Appendix A: Econometric Analysis

29



A.1 Propensity to Engage

29



A.2 Propensity to Exit and Enter

34

Appendix B: Supplementary Tables

35

Bibliography 37

04 ••

Executive Summary

• T  his report draws upon a unique panel dataset to understand the extent, nature and persistence of academic engagement with external organisations in the UK. The data relate to 4,059 UK-based academics who responded to each of two national surveys covering all academics in all disciplines in all UK Higher Education Institutions. The two surveys were carried out in 2008/9 and 2015. They generated data relating to the three year period prior to each survey. The data thus covers the period 2005 to 2015. It provides a comprehensive and unprecedented opportunity to examine changes in the scale and modes of engagement connecting UK academics with external organisations. • The survey data identifies 25 separate modes of interaction with external organisations. • T  hese modes of interaction can be combined into 5 groups of correlated activities. The first group labelled training relationships encompasses 4 modes. These are employee training, student placements, joint curriculum development and enterprise education. The second group labelled meetings, consulting and advice consists of 7 modes which do not require new original research. It includes attending conferences, standard setting forums, network participation, sitting on advisory boards, consultancy services, invited lectures and informal advice. Group three labelled joint research includes 6 modes which are joint research agreements, hosting of personnel, secondment of personnel, contract research, research consortia and joint publications. The fourth group labelled commercial activities and services includes 5 modes which are licensing research outputs, patenting and prototyping for external organisations, as well as the creation of new spin-out companies and setting up new physical facilities. Finally group five labelled public engagement includes 3 modes which are engagement through school projects, public lectures and public exhibitions. • M  ost academics engaged in a very wide range of the 25 individual modes of interaction and persisted in doing so over the two periods captured in the surveys. The results also show that those academics that did not report engagement with external actors in the first survey were also unlikely to report engagement in the second survey. So both engagement and non-engagement are persistent patterns of behaviour. • In the majority of the 25 individual modes of interaction, activity declined slightly or stayed roughly the same between the two survey periods. However, joint publication, hosting of personnel, sitting on advisory boards and lectures for the community each of which are highly frequently cited modes of interaction in both surveys were all higher in the second survey. The same was true of the much lower frequency activity of enterprise education. • T  aken as a whole these results suggest that for the majority of academics engagement is a recurrent persistent activity, firmly rooted in academic practice across the UK and across a full range of disciplines. • W  hen the data is analysed using the 5 groups of interaction modes a number of findings emerge. First there was a significant decrease for the meetings, consulting and advice group both in the mean number of modes used and in the share of respondents engaging in them. In contrast the public engagement group shows a significant increase in users and in the number of modes used between the survey periods.

05 ••

• If we focus on each of the 5 groups of modes of interaction separately we find considerable persistence within each. This suggests that developing engagement in a group of correlated modes is a learned activity that academics sustain over time. The research also shows that some groups of engagement modes contain more persistent engagement than others. Activities in the meetings, consulting and advice, joint research and public engagement groups all show highly persistent involvement. Equally non-engagement also persists. Thus for the commercial activities and services group only 11% of non-active academics in 2008/9 reported active involvement in 2015. • T  he results for the individual groups also show that reduced engagement within a group by some academics is counter-balanced by increased engagement efforts by other academics. The result is that although the level of engagement of individual academics may vary over time, the total level of engagement within each group is fairly constant. • T  he multivariate analysis of academic persistence taking all 5 groups together rather than individually shows that the decision to sustain engagement levels is largely explained by past engagement efforts and the research orientation of the individual. Prior experience in engagement and a research orientation towards applied or user oriented basic research are positively related to future engagement. These forces are particularly associated with persistence in the commercial activities and services group of modes of interaction. • W  omen have a lower propensity to engage through commercial activities and services than men, but a higher propensity to engage through meetings, consulting and advice and public engagement. • A  cademic motivations for engagement are related to five areas: learning (informing and testing own research), access to resources (funding for research, specialist equipment and materials, expertise of external organisations), teaching (student placements and teaching content), financial (personal income and business opportunities), and to foster the university’s outreach mission. The survey reveals that learning is the most important motivation for academics, while financial benefits were ranked lowest. • T  hose academics motivated in their engagement activities by opportunities to learn and gain access to resources are more likely to be persistent in engagement efforts. In contrast, those motivated by promoting outreach are more persistent in their public engagement activities. This suggests that engagement persistence is partly driven by the underlying motivations of academic for engagement and that such motivations may shape the efforts of academics over time to sustain different modes of engagement. • A  cademics’ decisions to enter and exit engagement with external actors are shaped by overall attitudes to engagement and their perception of the availability of external partners. The perceived lack of external partners increases the likelihood of non-persistence in training, joint research and public engagement groups while a negative engagement attitude increases the propensity to not persist with modes of interaction in the meetings, consulting and advice group.

06 ••

Section A Introduction

Universities are frequently viewed as key drivers of economic growth. There has been a continued debate about the extent to which collaboration between the academic and nonacademic sectors is a fundamental enabler of this role. Surveying academics about their external interactions contributes to a better understanding of these processes and how policy can intervene to further foster knowledge exchange and disciplinary research. Between September 2008 and June 2009 the largest ever survey of UK academic engagement with external organisations was undertaken by the Centre for Business Research covering the three year period prior to the survey (2005-2008). This original web-based survey attracted over 21,000 responses (Abreu et al., 2009). In 2015 a new survey was launched that covered the period 2012-2015 and attracted over 18,000 responses (Hughes et al. 2016). These are the two largest research and knowledge exchange surveys ever completed of a national Higher Education System. Both surveys highlighted the multifaceted role of the university and provided evidence of a “third mission” that is inclusive of all publics (public, civic and business) and research areas (humanities, social and natural sciences). The large and representative sample sizes permit the identification of a large panel of academics that responded to both surveys r and form the basis for a robust longitudinal analysis of change within the UK higher education institution (HEI) sector. The themes covered in both surveys relate to work roles and their recognition by the university, the balance between basic and applied research, the range and breadth of external knowledge exchange interactions and how they are initiated, and the motivations and constraints experienced by academics when engaging in knowledge exchange activities. Hughes et al. (2016) showed, using a matched sample approach, that engagement levels through most types of activities declined slightly or stayed the same between the two survey periods. The matched sample approach compares samples of respondents from each survey matched by their characteristics at the respective survey dates such as age and seniority alongside discipline and a wide range of other characteristics. This is useful and important for gauging average changes between the survey periods controlling for overall sample changes in those characteristics. This report in contrast focuses on the persistence in engagement behaviour between the two surveys using the panel of academics who responded to both surveys. Some of their characteristics (e.g. age) will have changed over time for all of the panel members but others may have changed for some but not others (e.g. seniority). This report specifically focuses on these individuals’ changes in external engagement. We do this taking into account the institutional context of each panel member, their individual motivations and their perceived constraints. These are key elements for understanding behavioural change and are important for explaining changes in the engagement behaviour of individuals. These considerations are, in particular, directly pertinent to questions of the development of policy incentives designed to alter the patterns of engagement activity at the level of the individual academic; issues that cannot be properly addressed without the kind of panel data analysis presented in this report. To the best of our knowledge this represents the first longitudinal survey data evidence on academics’ engagement with external organisations.

07 ••

Section B The CBR Academic Surveys 2008/9 and 2015

This section discusses methodological issues related to the two CBR academic surveys. This includes information on sampling and non-response bias.

Coverage, Sampling and Response The CBR academic surveys of 2008/9 and 2015 were not originally designed to produce a panel survey but rather as a stand-alone survey of the full population of academics in the UK. The sampling frame in each case included all academics active in teaching and/or research in each sample period in all disciplines in all UK HEIs. For each survey lists of academics in all departments and faculties were manually collected from the websites of UK HEI. This email directory provided the sampling frame to which a web-based questionnaire was addressed in each case. Both surveys were conducted online. The 2008/9 survey frame identified 125,900 valid email addresses and received 21,598 valid responses (response rate of 17%). The 2015 survey frame identified 131,088 valid email addresses and received 18,177 responses (response rate of 14%). Exhibit 1 gives an overview over survey responses. Comprehensive non-response analyses were performed on both surveys. A small bias towards older and slightly more engaged academics was found.

Exhibit 1

Academic survey response 2008/9 and 2015 surveys 2008/9

Total Sample*

%

2015

126,120

140,312

220

8,422

 

802

220

9,224

125,900

131,088

867

162

21,598

18,177

%

Less: No Longer at Institution/Undeliverable Not Eligible

Total Surveyed Sample Not Eligible Responses** Complete Responses Response Rate [complete responses]

17.3

13.9

* The sample consists of all HEIs in England, Scotland, Wales and Northern Ireland. ** Responses were excluded because the survey indicated inactivity through e.g. retirement or the respondents were students or administrative staff and as such not eligible to participate.

08 ••

Exhibit 2

Sample characterstics 2008/9 and 2015 surveys (% of respondents)

Social Sciences

Subject

Health Sciences Physics, Mathematics Biology, Chemistry, Veterinary Science Arts and Humanities Engineering, Materials Science Female Professor

Position

Senior Lecturer, Reader Lecturer, Teaching Fellow Research Fellow, Associate Research/Teaching Assistant Emeritus

Age

50 and Over 40 - 49 30 - 39 Under 30 0

5

10

2008/9 Full Sample

15

20

25

30

35

40

45

2015 Full Sample

Exhibit 2 shows the distribution of academics in both surveys broken down by disciplinary area, gender, seniority and age. The seniority and subject distribution in 2015 is very similar to 2008/9, however, the share of women has increased by a small amount from 40% in 2008/9 to 42% in 2015. The 2015 sample is also slightly older with a higher share of respondents over the age of 50.

Data Processing and Non-response Analysis In order to build a panel of academics we had to identify those that answered to both surveys whilst maintaining the confidentiality under which each survey was conducted. This was done in two steps (see Exhibit 3). First, all email addresses collected during the two surveys were standardised1 and compared. To ensure that email addresses had not been reassigned we also checked for name matches. In total, 11,294 academics that had responded to the 2008/9 survey were contacted in the course of the 2015 survey using the same email address; 3,391 of these replied to the 2015 survey round. Second, to try and deal with problems arising from email address changes caused for example by changing universities we compared the 2008/9 respondents and 2015 respondents based on name and subject area and conducted some manual cross-checking. This process utilised no other data from the two surveys the data from which is kept in separate anonymized files from the contact address data. A new anonymized panel dataset was then created separately from the files containing panel email addresses and names. In this new anonymised dataset we compared the age in both surveys and discarded those matches where the age in 2015 was lower than in 2008/9. Following this process we were able to identify an additional 668 academics that had replied to both surveys. We thus have a total of 4,059 academics in our panel.

1  Standardisation was necessary where universities have more than one email domain (e.g. le.ac.uk and leicester.ac.uk), use prefixes (e.g. elec.strath.ac.uk), or changed name or merged (e.g. leedsmet.ac.uk became leedsbeckett.ac.uk).

09 ••

Exhibit 3

Matching procedure 2008/9 and 2015 surveys Surveyed Sample 2015

2008/9 Respondents 21,563

11,294

10,269

Respondents 18,177

Non-respondents 112,749

3,391

7,903

Automated Email Check

668

Semi-automated Name and Subject Check

Unmatched 14,118

The characteristics of the panel respondents in 2008/9 compared to the full 2008/9 respondent sample are provided in Exhibit 4. They show that the panel sample was more likely to be drawn from senior academics and men. These are the groups that are known to be more likely to interact with external organisations. This sample composition means that we may have a bias towards more external engagement within the panel.

Exhibit 4

Sample characterstics 2008/9 full sample and panel (% of respondents)

Social Sciences

Subject

Health Sciences Physics, Mathematics Biology, Chemistry, Veterinary Science Arts and Humanities Engineering, Materials Science Female Professor

Position

Senior Lecturer, Reader Lecturer, Teaching Fellow Research Fellow, Associate Research/Teaching Assistant Emeritus

Age

50 and Over 40 - 49 30 - 39 Under 30 0

5

10

2008/9 Full Sample

10 ••

15

20

2008/9 Panel

25

30

35

40

45

Changes in the sampling frame populations between the two years meant that not all academics could be surveyed twice. Some had retired or left for other reasons since 2008/9, while some had entered newly into academia. This compounds the potential problems arising from the fact that not all of those contacted in both surveys replied both times. We therefore estimate a selection model to see which 2008/9 respondents are more likely to reply, taking into account whether or not they were included in the second sampling frame.2 This will help to establish whether there is a non-response bias in our panel. Demographic characteristics should explain the propensity to be contacted while prior engagement experience should increase the propensity to respond. Results are reported in Exhibit 5 and confirm these assumptions.3 The results show that those employed in non-tenure track roles (research fellow, assistant or teaching fellow contracts) in 2008/9 are less likely to have been surveyed again compared to professors, senior lecturers or lecturers (75% less likely in the case of research fellows). Also, those that had already retired in 2008/9 are less likely to still be in academia in 2015 and therefore less likely to be in the sample frame for that year. The oldest and youngest groups of academics are less likely to be resurveyed compared to those that were between the ages of 30 to 49 in 2008/9. Older academics may have retired and younger academics may have been in casual employment and left. Exhibit 6 plots the propensity to be resurveyed for different age groups. It shows that those between the ages of 40 and 49 have an almost 60% propensity to be surveyed again. Other characteristics are also significant. Those that had management responsibility at the time of the first survey or held research council grants4 are more likely to be contacted again in the second survey. The latter, in particular, is a very strong predictor for having remained in an academic post. In contrast, those that had professional experience outside academia prior to the 2008/9 survey are slightly less likely to still be in the 2015 sample frame population. There are also some minor differences by subject area. Academics in the arts and humanities are more likely to be in both sample frame populations, as are those in physics and mathematics. In contrast, those in health sciences are less likely to be included in both. Taking into account the probability of being in the second survey frame population, the Stage 2 column in Exhibit 5 looks at the characteristics in 2008/9 influencing whether a response is received in 2015 from those academics surveyed. The 2008/9 survey asked about external engagement in the previous three years through a variety of channels which were grouped into four categories: commercial, people-based, problem-solving and community-based activities (see Hughes et al., 2016). We find that those involved in more community-based or problem-solving interactions in the three years prior to 2008/9 are more likely to respond to the 2015 survey, however, the effects are very small. For example, those engaged through a high number of different problem-solving activities are about 9% more likely to respond than those with a medium number of different activities. Neither involvement in teaching nor research orientation affect response propensity. We do, however, find that women are slightly less likely than men to respond again. It is notable that the seniority and age effects which influence inclusion in the second survey have little or no effect with regard to their decision to respond to that survey. They are therefore not included in the second stage. Taken as a whole these results suggest that we do not need be too concerned about non-response bias relating to the characteristics examined.

2  Strictly speaking this test is not comprehensive since we cannot analyse individuals who were non-respondents in the second survey but who had changed their institution and hence email address compared to the first survey. 3  Most demographic criteria, such as seniority, age and discipline serve as exclusion restrictions in the first stage as they are expected to determine whether academics have remained in academia and are contactable but not whether they respond again. The exclusion restrictions are supported as none of these variables significantly influence response if included in the second stage. 4 

Research council funding data was provided by the research councils.

11 ••

Exhibit 5

Propensity of 2008/9 respondents to reply to 2015 survey, conditional on being in the 2015 sampling frame population

Gender

Female

Stage 1

Stage 2

Surveyed 2015

Responded 2015

Coef.

S.E.

Coef.

S.E.

-0.01

0.02

-0.06

0.03

*

Basic

base

Research Orientation09

User-inspired

-0.02

0.04

Applied

-0.05

0.04

None

-0.04

0.08

External Engagement05-08

Commercialisation

-0.03

0.03

Community-based Intensity

0.05

0.02

**

Problem-solving Intensity

0.09

0.02

**

People-based Intensity

0.00

0.02

0.01

0.05

 

 

 

-0.71

0.09

**

Activity09

Teaching

Rootedness in Academia

Management Responsibility09

0.08

0.02

**

Research Council Funding05-08

0.30

0.03

**

Professional Experience Outside HEI09

-0.09

0.02

**

Position09

Age09

Discipline09

Institution Type09

Professor

base

Senior Lecturer/Reader

-0.01

0.03

Lecturer

-0.06

0.03

Emeritus/Honorary (retired)

-0.23

0.09

*

Research Fellow

-0.75

0.04

**

Research Assistant

-1.05

0.06

**

Teaching Fellow

-0.50

0.07

**

>50

base

40-49

0.38

0.02

**

30-39

0.21

0.03

**

=50 [base] Age 15