ISI KANDUNGAN

Assalamualaikum Warahmatullahi Wabarakatuh dan Salam Sejahtera

YBhg. Datuk/ Datin / Prof/ Tuan/ Puan

SDI@PTAR adalah salah satu perkhidmatan penyebaran maklumat terpilih yang

disediakan oleh Perpustakaan Tun Abdul Razak, UiTM Shah Alam untuk ahli Mesyuarat

Senat UiTM. Perkhidmatan ini bertujuan untuk menyalurkan maklumat terbaharu mengenai

isu-isu semasa di dalam dan luar negara yang memberi nilai tambah serta impak kepada

pengajaran, pembelajaran dan penyelidikan UiTM ke arah menjadi Universiti Terkemuka

Dunia.

Untuk keluaran kali ini, SDI@PTAR menampilkan artikel teks penuh mengenai Cost Saving

(Higher Education/University Financial Management) Diharapkan maklumat ini memberi

manfaat kepada YBhg. Datuk/ Datin/ Prof/ Tuan/ Puan.

Sebarang cadangan dan maklumbalas mengenai perkhidmatan ini boleh disalurkan kepada

En. Mohd Ismail bin Abidin, Timbalan Ketua Pustakawan (e-mel ismailabidin@uitm.edu.my)

dan Puan Nik Zatihulwani binti Jamaludin, Pustakawan, (e-mel

nikzatihulwani@uitm.edu.my), Bahagian Penyelidikan, Pembelajaran dan Rujukan,

Perpustakaan Tun Abdul Razak, UiTM Shah Alam.

Sekian. Terima kasih.

Bahagian Penyelidikan, Pembelajaran & Rujukan

Jabatan Perkhidmatan Perpustakaan

Perpustakaan Tun Abdul Razak Utama UiTM Shah Alam

Bil 3/2021

ISI KANDUNGAN

BIL. TAJUK SUMBER h-INDEX PENULIS

1 Assessing the cost-effectiveness of university

academic recruitment and promotion policies

European

Journal of

Operational

Research

36 Thanassoulis, E., Sotiros,

D., Koronakos, G.,

Despotis, D.

2 Costs, efficiency, and economies of scale and

scope in the English higher education sector

Oxford

Review of

Economic

Policy

22 Johnes, G., Johnes, J.

3 Does trust play a role when it comes to

donations? A comparison of Italian and US higher

education institutions

Higher

Education

9 Francioni, B., Curina, I.,

Dennis, C., (...),

Bourlakis, M., Hegner,

S.M.

Bil 3/2021

Assalamualaikum Warahmatullahi Wabarakatuh dan Salam Sejahtera

YBhg. Datuk/ Datin / Prof/ Tuan/ Puan

ASSESSING THE COST-EFFECTIVENESS OF

UNIVERSITY ACADEMIC RECRUITMENT AND

PROMOTION POLICIES

Bil 3/2021

European Journal of Operational Research 264 (2018) 742–755

Contents lists available at ScienceDirect

European Journal of Operational Research

journal homepage: www.elsevier.com/locate/ejor

Innovative Applications of O.R.

Assessing the cost-effectiveness of university academic recruitment

and promotion policies

, � , ✰✰

E. Thanassoulis a , ∗, D. Sotiros b , G. Koronakos b , D. Despotis b

a Operations and Information Management Group, Aston Business School, Aston University, Birmingham B4 7ET, UK b Department of Informatics, University of Piraeus, 80 Karaoli and Dimitriou, 18534 Piraeus, Greece

a r t i c l e i n f o

Article history:

Received 9 December 2016

Accepted 20 June 2017

Available online 27 June 2017

Keywords:

Data Envelopment Analysis

Academic promotions

Academic recruitment

Cost efficiency

a b s t r a c t

This paper develops an approach for higher education institutions to assess the economic efficiency of

their recruitment and promotion practices concerning academic staff. Research output potential is a key

criterion in most academic appointments. Generally, there is a long lead time between the conduct of

research and its ultimate value in the form of disseminated knowledge. This means higher education

institutions usually reward financially staff on the prospect of research output, albeit on the basis of

research outputs achieved up to the point of recruitment or discretionary salary rise (e.g. through pro-

motion). We propose a Data Envelopment Analysis (DEA) model which can be used retrospectively to set

salary costs against corresponding research outputs achieved as a measure of the financial efficacy of past

recruitment and promotion practices. The analysis can identify potential issues with those practices and

lead to improvements for the future.

© 2017 Elsevier B.V. All rights reserved.

1. Introduction

University academic staff is typically expected to make contri-

butions in four broad areas: Teaching, Research, administration and

‘outreach’. The first three are self-explanatory. Outreach is exter-

nally facing in the form of engagement in local or national gover-

nance, consultancy or service to the community, or a combination

thereof. The remuneration of academic staff typically covers all but

outreach activities and it represents a cost for their University. In

general, there is considerable lead time between the conduct of

research and corresponding measurable outcomes being observed.

For academic research, ultimate value is gained via its dissemina-

tion and use. Yet it can take considerable time between the gener-

ation and the dissemination and/or use of the research outcomes.

Firstly, the research itself may take months or even years before

it becomes a self-contained, publishable in principle, unit ready

for submission to be assessed. The assessment, normally by peer

� This research has been co-financed by the European Union (European Social

Fund – ESF) and Greek national funds through the Operational Program "Educa-

tion and Lifelong Learning" of the National Strategic Reference Framework (NSRF)

(Grant no. E2458-2012SE24580287) – Research Funding Program: THALES. Invest-

ing in knowledge society through the European Social Fund. ✰✰ The views expressed in this paper are those of the authors and no representa-

tion is made that they are shared by any funding body. ∗ Corresponding author.

E-mail address: e.thanassoulis@aston.ac.uk (E. Thanassoulis).

review, could lead to one or more iterations for revision and re-

assessment before the research is ultimately accepted and later

put in the public domain. Use of that research can then take still

further time, possibly years after the research has been published.

Thus, in practice there is normally considerable time lag between

the research work of an academic and the corresponding outcomes

being observed. Therefore, when it comes to research, institutions

pay an academic in advance , on the prospect of research output,

albeit the prospect itself would normally be based on prior re-

search carried out by an academic. This is especially so when an

academic is recruited to a post and to that date he/she has not

contributed to research within the recruiting institution. Payment

on the prospect of research output also occurs when an academic

is promoted primarily on his/her research outputs to the point of

promotion.

It is true that promotions in institutions can also be made on

the basis of excellence in teaching, e.g. for teaching innovation and

continual upgrade of content which can have long term effects for

students. Thus, in a sense institutions can pay in advance for the

prospect of teaching as for research excellence. However, there are

still crucial differences in the role played by research versus teach-

ing in most academic salary levels. Recruitment to a post is rarely

on the basis of teaching excellence, though it can be for exper-

tise in a teaching area, but provided the prospect of research is

met. Thus given that promotion on teaching excellence is relatively

recent and more rare than for research, and given recruitment

http://dx.doi.org/10.1016/j.ejor.2017.06.046

0377-2217/© 2017 Elsevier B.V. All rights reserved.

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 743

on teaching alone is even less prevalent, this paper focuses on

assessing recruitment and promotion practices with reference to

research output by those recruited and/or promoted.

There is considerable literature on assessing research effective-

ness at person, Department or University level. For a review of

the use of DEA in assessing education services, including uni-

versities the interested reader is referred to Thanassoulis et al.

(2016) . Assessments addressing various aspects of University ef-

fectiveness can be found in De Witte and Rogge (2010) . Assess-

ments at University level can be found in Abbott and Doucouliagos

(2003), Athanassopoulos and Shale (1997), Avkiran (2001), Coelli

(1996) and Katharaki and Katharakis (2010) . Assessments at Uni-

versity Department level can be found in Johnes (1995), Johnes

and Johnes (1993) and Doyle et al. (1996) . Assessments of Greek

Universities or University Departments can be found in Anastasiou,

Kounetas, Mitropoulos, and Mitropoulos (2007), Kounetas, Anas-

tasiou, Mitropoulos, and Mitropoulos (2011) and Karagiannis and

Paschalidou (2017) . However, none of these assessments looks at

the recruitment and subsequent career payments of an individual

relative to their research outputs. Such a view, when suitably ag-

gregated across individuals, could identify any cost deficiencies in

the policy of an institution on appointments and promotions of

academics. The aim of this paper is to contribute an approach to

addressing the issue of the cost-efficiency of the policy of recruit-

ment and promotions at higher education institutions.

In order to assess the cost-effectiveness of the recruitment and

promotion policies of a tertiary education institution we propose

the use of Data Envelopment Analysis (DEA) to set the career cost

for research by each academic of that institution against their re-

search output as it has been revealed prior to and post decision

points on recruitment and promotion. This is done at academic

person level ensuring the academics are comparable on research

by way of academic discipline and contractual obligations de facto

and de jure. While at person level there could be ad hoc events

that could affect the findings on relative costs of research output,

when the results are aggregated across suitably defined sets of in-

dividuals they can be informative about the cost-efficiency of the

policy on recruitment and promotions at constituent unit of an in-

stitution or even at institution level.

We illustrate our approach using realistic estimates of data for

comparable by discipline academic staff from a Greek University.

We set their estimated career remuneration for the conduct of

research against measures of their research output. We identify

alternative scopes for savings that might have been made under

alternative scenarios. Our headline finding is that the cost of ag-

gregate research output has limited scope to be lowered, about

7.5% of total salary bill, by simply optimising the decisions on pay

level without improving the pace of research output. If, however,

apart from making salary levels more in line with the pace of

research output, that pace itself had been faster then the time

saved would have reflected savings of 17.5% of the salary bill. It

is noted that these findings are only illustrative as they are based

on benchmarking on only a limited number of some 38 persons

and on other subjective value judgements made in the modelling,

notably on the relative values of research published pre- versus

post-appointment, and on the worth of publications in differently

ranked journals.

The paper is structured as follows. Section 2 presents the as-

sessment model used in terms of the underlying conceptualisation,

including the measures of salary cost and corresponding research

outputs used. This is followed by a mathematical formulation of

the model conceptualised. Section 4 presents an illustrative appli-

cation of the mathematical model, using realistic estimates of data

relating to 38 academic persons. Section 5 discusses policy im-

plications that could be drawn from the results on the estimated

data. Section 6 concludes the paper.

2. Conceptualising the assessment model

We wish to set the salary component of an academic over the

period the academic has been in post against the research out-

comes attributable to that academic. Ideally therefore one would

require the component of the academic’s salary that relates to their

research. Measures of the research contributions (e.g. academic pa-

pers, books, etc.) can be obtained at academic person level. How-

ever, the salary of an academic typically covers all their activities

and the component relating to their research output is not known.

One way to proceed therefore would be to use the entire salary

cost of the academic on one side and all activities (teaching, re-

search, administration, etc.) on the other. Conceptually this is a

correct approach, especially if the cost efficiency of the academic

person is the object of our assessment. However, our target here is

the recruitment and promotion policies of an institution and these

have hitherto, as noted earlier, largely hinged on research, and es-

pecially so where recruitment of staff in the first place is con-

cerned. Thus, another way to proceed is to use the whole salary

but data only from academic staff where both the contractual and

the de facto expectation is the same in terms of contributions to

teaching, research and administration. This cohort, to which we

shall refer as “Standard Teaching, Research, and Administration” or

STRA academic staff would generally be core service delivery staff

constituting a majority of the academics in an institution. STRA

staff would exclude academics where for any reason the balance

in expectation of teaching, research and administration does not

apply (e.g. by reason of senior management role within the insti-

tution, secondment outside academia, ill health, etc.). In contrast,

given that our aim is to assess policies rather than individuals, data

from past STRA academics who have left an institution could be

used, provided they remained sufficiently long at the institution

for their output in research corresponding to salary payments to

register. Our approach is then predicated on using only data from

STRA academic staff.

For our illustrative application, we have used data from STRA

staff only and so we can proceed by using the same proportion of

the total salary cost of each academic person as if it reflected com-

pensation for expected research output. Using the same proportion

of salary across staff means the same expectation of research out-

put but not necessarily the same delivery of research output. In

fact, the model aims to show where delivery of research has not

been in tandem with salary cost. Where research output by a STRA

academic is higher relative to other STRA staff, the model will find

higher cost efficiency relative to other STRA academics who have

similar salary. The opposite will be the finding when research out-

put by a STRA academic is lower relative to other STRA staff on

similar salary. Thus so long as expectation in research relative to

other duties is the same across all staff whose data is used in the

model there will be no bias in the findings for using the same pro-

portion of total salary as compensation for research. Where the to-

tal salary leads to an input level in the model for research compen-

sation not in line with research output by the person the model

will appropriately show this as high or low cost efficiency as the

case may be.

We wish to set up a model to estimate the lowest career-

aggregate cost that could have secured the career-aggregate re-

search output of an academic, if he/she had been as productive in

research quantity and quality as found in others within the sam-

ple of 38 academics we are using. The cost of a given set of re-

search outcomes can be minimised either through producing the

research in less time, or by paying a lower salary to the aca-

demic producing the research or both. The career salary profile in

turn of an academic reflects the initial salary they were appointed

at and any salary increases they may have received, normally

through the promotions process. Our model will reflect both the

744 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

Fig. 1. Savings on salary research component as % of total salary cost.

appointment and the annual mean career salary of an academic.

To make salaries across time comparable we have used the mean

salaries of a given year (2014) for each level of post (Lecturer, Asso-

ciate Professor, etc.) and applied them to the duration at each post

level by each STRA academic within our sample. This also controls

for appointment vs promotion salary inflation as could arise where

the former is susceptible to the balance of supply and demand

for academics which can change over a prolonged time period. In

Greece base academic salaries at each level are set by the state and

there is little market pressure on salaries as there is generally ex-

cess staff supply and staff mobility is limited. What the monetary

values used for salary in the model will then reflect is the level

of post at appointment and frequency of promotion. These will be

judged by the model relative to research output by the academic

concerned.

Turning to the research outputs of an academic there is no clear

consensus in the literature as to how their quality should be re-

flected. However, judging from accepted practice, e.g. the United

Kingdom periodic Research Assessment Exercises ( http://www.ref.

ac.uk/ ), papers published in peer reviewed academic journals and

some measure of ‘impact’ of research are taken as reflective of

quality of research. The quality of research output should in prin-

ciple be assessed using judgment. However, the judgments on the

quality of academic papers do tend to match in considerable mea-

sure the ranking of the journal where the papers have appeared.

For example, see Pidd and Broadbent (2015) on the strong associ-

ation of the grading of research outputs by the Business and Man-

agement panel ( Research Excellence Framework, 2014 ) and the As-

sociation of Business Schools (ABS) ranks of the journals where the

papers had appeared. This was despite the fact that the panel had

used judgment rather than journal ranks when grading research

outputs on quality. So, one could effectively use the ranking of the

journal where a paper is published as a proxy for the quality of

that paper. See also Mingers, Watson, and Scaparra (2012) on this

point.

For the illustration of our approach, to reflect the research out-

puts of an academic we have used the ranks of the journals where

the papers have been published and the citations their papers

have received, excluding self-citations. Specifically, we have used

peer-reviewed journal papers indexed in Scopus and their citations.

For quality of the papers we have used the Australian ERA2010

ranking of journals as we found this had a wider list of journals

than did the Association of Business Schools ranking of journals,

the latter being widely used for ranking journals in the UK. Ac-

cording to ERA2010, the journals are ranked from C to A + , the lat-

ter reflecting top quality under ERA2010. We have added a cate-

gory D for papers found in Scopus journals that are not ranked in

ERA2010. We have used the ranking of journals at the time of writ-

ing this paper (ERA2010) rather than at the time the papers in our

data were each published. This level of refinement of the illustra-

tive assessment was not deemed necessary though it can be used

in a real application of our approach by an institution. We have

used Scopus citations, excluding self-citations, to capture the de-

gree of dissemination of the research of an academic person. The

citations are not shared across co-authors nor allowance is made

for other refinements such as allowing for the time since publica-

tion of a paper, or the ranking of the journal where the citation

has occurred. These refinements are not deemed necessary for our

illustrative application but can be undertaken in principle in real

life applications.

Table 1 shows the specific input-output variables which we

have used within our DEA model and which can serve as a general

base for cost-efficiency assessments of this type of the research

output at academic person level. For simplicity of expression in the

remainder of this paper reference to A + publications will include A

publications converted to A + equivalent and reference to B publi-

cations will include C and D publications converted to equivalent B

publications, as converted in Table 1 . In principle, there is no need

to aggregate publications in different ranked journals. For example,

the outputs could be by journal rank A + , A, B etc. The aggrega-

tion here was to retain degrees of freedom due to the relatively

low number of only 38 observations. It is clear that the journal

rank reflects but a proxy of the quality of a paper. Some A + pa-

pers may be no better than say an A paper and an A paper could

have been worthy of publication in an A + journal. Thus, there will

always be potential stochastic error in reflecting the true quality

of research output, whether journal ranking or some other method

such as judgement is used for the assessment. However, when the

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 745

Table 1

Input–output variables for assessing the cost efficiency of academic research.

INPUTS

X 1 : First year in post;

X 2 : Duration in post after the first year.

OUTPUTS

Y AB : Publications in A + journals before recruitment, including A journals using the conversion rate A = 0 .5A + (SAE a )

Y BB : Publications in B journals before recruitment, including C and D journals using the conversion rate (B = 2C = 6D) (SAE)

Y AA : Publications after recruitment in A + journals including A journals using the conversion rate A = 0.5A + (SAE)

Y BA : Publications after recruitment in B journals including C and D journals using the conversion rate (B = 2C = 6D) (SAE)

Y C : Citations of the articles included in the foregoing four categories of journal papers

a SAE = Single Author Equivalent, each joint author credited the same fraction of the paper.

findings are ultimately aggregated at institution level the impact

of this stochasticity will to an extent be ameliorated as biases in

opposite directions cancel out to an extent across papers or other

research output units. This will be more so the more the research

output units (papers, etc.) within the assessment.

The rationale for each variable in Table 1 is as follows.

(i) First year in post, X 1 = 1 for all persons: This variable reflects

the first year in post. It enables the model to assess the efficacy

of the initial salary offered to the academic. The initial salary in

most cases is fundamental in shaping the salary profile in post for

an academic as it typically provides the basis for salary increments

going forward. Thus, the variable is instrumental in assessing the

recruitment policies of an institution in terms of salary reward for

promise of research as assessed at the time of recruitment.

(ii) Time in post – X 2 : This variable reflects the duration in post

after the first year. It enables the model to set in context the re-

search outputs of the academic concerned. The academic’s time is

a key ingredient converted to research outcomes.

(iii) Publications Y AB to Y BA : We use the peer-reviewed publi-

cations to capture the research output of an academic. In partic-

ular, we take into account only journal papers indexed in Sco-

pus. Conference papers, book chapters etc., even though indexed

in Scopus, are not taken into account falling largely in line with

practice in research assessments conducted periodically by funding

agencies (e.g. the REF in the UK in 2014 ( http://results.ref.ac.uk/ )).

Where publications are authored by more than one person we

have shared the credit equally between all authors. For example,

a publication having 2 authors will count as 0.5 publications for

each joint author denoted 0.5 SAE (single-author equivalent) above.

For alternative ways of crediting authors of multi-authored pa-

pers see Karagiannis and Paschalidou (2017) or Section 12.3.3 in

Thanassoulis et al. (2016) .

The publications are broken down into four categories. Two of

the categories are based on timing of the publication before or af-

ter the person was recruited to their current post. This subdivision

is felt necessary because while the sum total of one’s publications

constitutes his/her research profile, those prior to recruitment play

a role in whether or not a person is recruited and on the salary

they are recruited at. Those post recruitments play a role in the

retention and promotions of the person concerned. These distinct

roles of pre- and post-appointment publications will enable the

model to reveal the effect on the cost efficiency of the recruitment

as distinct from that of the promotions process.

Each one of the pre- and post-recruitment publications is fur-

ther subdivided into two categories by quality of outlet as reflected

in the Australian ERA2010 ranking of journals. One category repre-

sents journals ranked as A + or A and the other journals ranked B,

C and D. Following information on how certain Australian Univer-

sities compensate staff for publishing in journals ranked A + to C

we have adopted the following assumptions: one publication ap-

pearing in a journal ranked A counts as 0.5 publications appearing

in a journal ranked A + ; each B ranked publication was deemed

worth 2 and 6 publications appearing in C and D ranked journals

respectively. As noted earlier, these conversions using subjective

equivalences of papers are not necessary in principle as papers can

be used by each rank of journal. However, it would be advisable to

use weights restrictions, as we do below, to signal to the model at

least the rank order of journal rating (e.g. on average an A paper

would be of better quality than a B paper etc.).

(iv) Citations – Y C : We deemed citations to constitute a surro-

gate for the reach of the publications. Only citations of the papers

included in the four categories of outputs Y AB to Y BA were consid-

ered. As noted above we are aware of issues in using raw citation

counts and the arguments that citations do not necessarily reflect

quality of publication. However, as we already use four surrogate

variables for quality of publications, the citations are intended to

simply reflect the degree to which the work of an author has been

noted by others.

Clearly our choice of input, and especially output, variables is

based on subjective judgments albeit made on the extant refer-

ences and ’accepted wisdom’ as to how quality of research is to

be judged. Our assumptions are in our view sufficient to illustrate

the approach being developed here. Any institution adopting our

approach could replace our output variables with others more in

line with its own judgement as to how research output quality and

indeed quantity is to be captured (e.g. including PhD student com-

pletions or research grant income generated as reflecting research

output). The generic input remains time devoted to research and

the generic output is research quantity and quality.

3. Constructing the assessment model

As noted at the outset, our overarching aim is to assess the

efficacy of an institution in salary payments it has made on the

prospect of research output by its academic staff. Portela and

Thanassoulis (2014) have proposed a general-purpose DEA model

for estimating potential cost savings at a production unit when the

input levels it uses and the prices it pays per unit of input are op-

timised simultaneously, to minimise the aggregate cost of inputs,

controlling for output levels. Their model provides a useful start-

ing point in addressing the issue of efficacy of the recruitment and

promotion policies of an institution, where research by academic

staff is concerned. The input in our case would be the time taken

to conduct the research, while the outputs are the research outputs

of an academic. The input prices are the annual salaries of the aca-

demic. We differentiate between salary in the first-year post ap-

pointment (henceforth year 1) and annual salary in the subsequent

years of an academic’s duration in post at the institution. Salary in

year 1 reflects publications prior to appointment and enables the

analysis to assess the financial efficiency of recruitment policies. In

contrast, annual salaries after year 1 reflect the financial efficiency

of salary increments, notably through promotion. Certain conven-

tions, if not legal requirements, such as no reduction in salary over

time, can also be explored regarding their cost implications.

746 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

Drawing on the model by Portela and Thanassoulis (2014) we

propose Model 1 for estimating the potential cost savings in

achieving the research output of academic person o .

Model 1:

min C1 = γ1 p 1 o θ1 x 1 o + γ2 p 2 o θ2 x 2 o s.t. ∑ n

j=1 λ j x 1 j ≤ θ1 x 1 o C1 . 1 ( v 1 ) ∑ n j=1 λ j x 2 j ≤ θ2 x 2 o C1 . 2 ( v 2 ) ∑ n j=1 λ j y AB, j ≥ y AB,o C1 . 3 ( u AB ) ∑ n j=1 λ j y BB, j ≥ y BB,o C1 . 4 ( u BB ) ∑ n j=1 λ j y AA, j ≥ y AA,o C1 . 5 ( u AA ) ∑ n j=1 λ j y BA, j ≥ y BA,o C1 . 6 ( u BA ) ∑ n j=1 λ j y C, j ≥ y C,o C1 . 7 ( u C ) ∑ n j=1 z 1 j p 1 j ≤ γ1 p 1 o C1 . 8 ( u 8 ) ∑ n j=1 z 2 j p 2 j ≤ γ2 p 2 o C1 . 9 ( u 9 ) ∑ n j=1 z 1 j = 1 C1 . 10 ( u 10 ) ∑ n j=1 z 2 j = 1 C1 . 11 ( u 11 )

2 λ j − z 1 j − z 2 j ≤ 0 C1 . 12 ( u 12 ) ∑ n j=1 λ j = 1 C1 . 13 ( u o )

θ1 = 1 C1 . 14 ( δ1 ) 0 ≤ θ2 ≤ 1 C1 . 15 ( δ2 ) −γ1 p 1 o + γ2 p 2 o ≥ 0 C1 . 16 ( μ1 ) γ1 p 1 o ≥ a C1 . 17 ( μ2 ) γi ≥ 0 , i = 1 , 2 C1 . 18

λ j ≥ 0 , j = 1 , . . . , n C1 . 19

In Model 1 x ij is the level of the i th input of academic j , where

x 1 j = 1 for all j , is the first year in post. x 2 j is the career duration (in

years) of academic j , from year 2 on. y ABj is the number of publi-

cations in A

+ rated journals by academic j before recruitment. y AAj

is the number of publications in A

+ rated journals by academic j

after recruitment . y BBj , and y BAj are defined in an analogous man-

ner. y Cj is the number of citations of academic j . p 1 j is the salary

in year 1 (i.e. on recruitment) of academic j , while p 2 j is the mean

annual salary of academic j from year 2 in post on.

The variables in the model are γ , θ and λ. The constraints C1.1–

C1.7 inclusive define in conjunction with C1.13 a convex feasible set

of input–output levels in the traditional context of DEA under vari-

able returns to scale. The constraints in C1.8–C1.11 inclusive define

a feasible set of initial and mean subsequent annual salaries, as-

suming a convex combination of observed salaries is feasible even

if not observed.

In setting up an instance of the generic version of Model 1

that would reflect both the legal framework in which an institu-

tion functions and its own internal norms, the issue arises of how

to link benchmarks on research outputs and those on salary. We

would expect that if an individual person (P1) has his/her dura-

tion in post and research outputs used as benchmarks for individ-

ual o then so should be P1’s salary levels too. Otherwise a bench-

mark with low research level and low salary can be used to set the

target salary of an individual with a high research output which

would be inappropriate. Clearly it would be difficult to argue for

complete disconnect between benchmarks on research output and

salaries. The only question is what should be the nature of the

link. It is recalled that at the optimal solution to Model 1 posi-

tive λ values will identify benchmarks on research outputs while

positive z values will identify benchmarks on salaries. At its sim-

plest, the link could be set so that benchmark (i.e. referent) aca-

demics are identical both in terms of who they are and in what

proportion they each contribute to targets for the academic be-

ing assessed, both where research output levels and where salaries

are concerned. This approach would mean discarding the z vari-

ables in Model 1 and simply using the same λ variables to form

convex combinations of salaries, years in post and research out-

put measures. This in our view, while possible, would be overly

restrictive. We have opted instead for the more flexible constraint

2 λ j − z 1 j − z 2 j ≤ 0 , j = 1 . . . n in C1.12.

The coefficient of 2 for λ j in C1.12 reflects the fact that we have

two inputs in this assessment. The constraint ensures that if an

academic j is a benchmark on research outputs and duration in

post then that academic must also be a benchmark for at least one

of the salaries, (year 1 or career mean after year 1). However, the

proportions that benchmark academic contributes to target salaries

need only average to the λj value (i.e. proportion the benchmark

contributes to research output and duration in post targets). Clearly

the nature of the link between benchmarks on salaries ( z values)

and research output ( λ values) is subjective. For example, a user

may opt for a coefficient k in C1.12 different from the number of

inputs (2 in our case). The higher the value of k relative to the

number of inputs the weaker the link between benchmarks on

salary versus research outputs and duration.

The objective function of the model is minimising the career

salary cost of academic o by optimising simultaneously the val-

ues of γ and θ so that for academic o duration in post θi x io and

salary levels γi p io are optimal. Note that in constraint C1.14 we fix

θ1 = 1 as we do not permit part-time appointments in year 1 and

assume each person appointed will normally remain at least one

year in post, given the legal framework. This makes the constraint

C1.1 redundant. However, we have kept it in for the completeness

of the model as in the general case one may not wish to fix at the

outset θ1 = 1 and/or x 1 j = 1 for all j . In C1.15, θ2 ≤ 1 allows for the

fact that a person’s research output could have been achieved at a

shorter space of time than the person took. Constraints C1.16 and

C1.17 reflect the legal framework of appointments. Constraint C1.16

ensures that the benchmark mean annual salary from year 2 on

for academic o does not fall below the benchmark annual salary at

which that academic could have been appointed. This is compati-

ble with the assumption that salary reduction below the appoint-

ment salary is not permissible or at least rarely found in practice in

academic institutions. Constraint C1.17 does not permit the bench-

mark salary on appointment to be below the legal minimum an-

nual salary at which an academic could have been appointed, de-

noted in the model a . This type of constraint would apply in most

countries. Constraints C1.16 and C1.17 would need to reflect the le-

gal framework of the country of application of the model.

Institutions would normally have views as to the relative value

of publications appearing in differently ranked journals and also

the worth of citations versus publications. These views are subjec-

tive and possibly institution specific. Without claiming generality,

and for illustrative purposes only, we have incorporated in this as-

sessment the following value judgements:

– An A + publication is worth at least 2 publications in B ranked

journals; (we explore later how robust are the findings to these

trade-offs).

– 5 publications after appointment in a given rank journal are

worth 6 pre-appointment publications in a journal of the same

rank; This reflects the uncontroversial judgment that all being

equal a publication is worth more to an institution in terms of

earning it visibility and prestige if it is made by a member of

its staff when he/she was an employee at the institution rather

than as an employee elsewhere.

– 10 citations are worth more than a publication appearing before

appointment in a B ranked journal but less than a publication

appearing post appointment in a B ranked journal. It would be

uncontroversial that a citation is worth less than a publication

but it is subjective as to how many citations would make up for

one publication. Our choice of 10 in relation to a publication

of quality between A + and B is subjective and for illustrative

purposes only.

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 747

Fig. 2. Potential proportional change in recruitment salary ( γ 1 ), in mean post year 1 annual salary ( γ 2 ), and in duration in post since year 2 ( θ 2 ).

In order to illustrate the incorporation of value judgements,

such as the foregoing, we formulate Model 2 as the dual to Model

1. The dual variable associated with each constraint is shown in

brackets next to the labels C1.1–C1.17 of Model 1.

Model 2 (dual to Model 1)

max D = ( u AB y AB,o ) + ( u BB y BB,o ) + ( u AA y AA,o ) + ( u BA y BA,o ) + ( u C y C,o ) + u o + u 1 o + u 11 + μ2 α + δ1 − δ2

s.t. Dual variable

v 1 x 1 o + δ1 ≤ γ1 p 1o x 1o θ1

v 2 x 2 o − δ2 ≤ γ2 p 2 o x 2 o θ2

u 8 p 1 o − μ1 p 1 o + μ2 p 1 o ≤ p 1o θ1 x 1o γ1

u 9 p 2 o + μ1 p 2 o ≤ p 2o θ2 x 2o γ2

−v 1 x 1 j − v 2 x 2 j +

(u AB y AB, j

)+

(u BB y BB, j

)

+

(u AA y AA, j

)+

(u BA y BA, j

)+

(u C y C , j

)+ u o − 2 u 12 ≤ 0

λ j

−u 8 p 1 j + u 1 o + u 12 ≤ 0 z 1 j −u 9 p 2 j + u 11 + u 12 ≤ 0 z 2 j v i , u 8 , u 9 , δ2 , u 12 , μi ≥ 0 , i = 1 , 2

u AB , u BB , u AA , u BA , u C ≥ 0

u o , u 10 , u 11 , δ1 f ree

We can add now the following constraints to Model 2 to cap-

ture the value judgments expressed above.

Constraints on weights Dual variable

u AB ≥ 2 u BB t 1 u AA ≥ 2 u BA t 2 u AA ≥ 1 . 2 u AB t 3 u BA ≥ 1 . 2 u BB t 4 u C ≤ u BA t 5 u C ≥ u BB t 6

Thus, for example, the constraint u AB ≥ 2 u BB ensures that one

publication in an A + ranked journal is worth at least 2 publica-

tions in B ranked journals, both prior to appointment. The remain-

ing restrictions above associated with the dual t variables are in-

terpreted in a similar manner. Note that citations are in units of 10.

Using now the additional constraints and their dual variables t 1 to

t 6 as shown above we can revert back to Model 1 which modifies

to Model 3, capturing the foregoing value judgements.

Model 3:

min C3 = γ1 p 1 o θ1 x 1 o + γ2 p 2 o θ2 x 2 o s.t. ∑ n

j=1 λ j x 1 j ≤ θ1 x 1 o C3 . 1 ∑ n j=1 λ j x 2 j ≤ θ2 x 2 o C3 . 2 ∑ n j=1 λ j y AB , j − t 1 + 1 . 2 t 3 ≥ y AB C3 . 3 ∑ n j=1 λ j y BB , j + 2 t 1 + 1 . 2 t 4 + t 6 ≥ y BB C3 . 4 ∑ n j=1 λ j y AA , j − t 2 − t 3 ≥ y AA C3 . 5 ∑ n j=1 λ j y BA, j + 2 t 2 − t 4 − t 5 ≥ y BA C3 . 6 ∑ n j=1 λ j y C, j + t 5 − t 6 ≥ y C C3 . 7

n ∑

j=1

λ j = 1 C3 . 8

n ∑

j=1

z 1 j p 1 j ≤ γ1 p 1 o C3 . 9

n ∑

j=1

z 2 j p 2 j ≤ γ2 p 2 o C3 . 10

n ∑

j=1

z 1 j = 1 C3 . 11

n ∑

j=1

z 2 j = 1 C3 . 12

2 λ j − z 1 j − z 2 j j = 1 . . . n C3 . 13

θ1 = 1 C3 . 14

0 ≤ θ2 ≤ 1 C3 . 15

−γ1 p 1 o + γ2 p 2 o ≥ 0 C3 . 16

γ1 p 1 o ≥ a C3 . 17

γi ≥ 0 , i = 1 , 2 C3 . 18

λ j ≥ 0 , j = 1 , . . . , n C3 . 19

t r ≥ 0 , r = 1 , . . . , 6 C3 . 20

z i j ≥ 0 , i = 1 , 2 ; j = 1 , . . . , n C3 . 21

Notation in Model 3 is as in Model 1, while the t r , r = 1 , . . . , 6

are dual variables associated with the additional constraints re-

flecting value judgements, as detailed above.

Model 3 is a generic one. It can be modified to investigate

the impact of the recruitment and promotion policies of an in-

stitution within the degree of flexibility it has. For example, by

setting θ1 = θ2 = 1 while γ1 and γ2 are permitted to vary the insti-

tution can investigate its recruitment and promotion policies un-

der the prism that only salaries are under its control once a per-

son has been appointed because firing is only permitted by law

748 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

under exceptional circumstances. (In Greece for example Univer-

sity academics after a certain grade have tenure.) More generally,

by varying the combination of the parameters θ1 , θ2 , γ1 and γ2

that can vary one can reflect in the assessment those aspects of

duration in post, recruitment and promotion salary that are seen

as falling within the gift of the institution and legitimate factors

against which to assess the efficacy of its recruitment and promo-

tion policies.

4. An illustrative use of the foregoing model

Model 3 was applied to estimated data on 38 STRA academics

of a Department of a Greek University. Their academic disciplines

are similar and only those who had had at least 3 years in post

were included. The three-year threshold was used to ensure each

academic had had sufficient time in post to deliver research out-

put and be affected by appointment and possibly promotion poli-

cies both of which are the issue of the analysis. The input and the

output data are estimated and so our findings do not necessarily

reflect the true state of affairs at the institution concerned. How-

ever, our aim is not to assess individuals or the institution per se.

Rather it is to illustrate the approach we have developed in this

paper with realistic data and then to discuss the implications of

the findings for the recruitment and promotions policy of any in-

stitution in which findings of the type our estimated data yield

were to be the case.

4.1. Assessing the scope for savings through optimising research

output rate and recruitment and promotion practices

We begin with an overview of the findings estimating the grand

total savings that might have been possible where the component

of research expenditure is concerned. For this scenario, Model 3

has been solved in the form stated above. As it has a non-linear

objective function it was solved parametrically by numerical ap-

proximation varying θ2 in the range from 0.001 to 1 in steps of

0.001. For each value of θ2 the model becomes linear and can be

readily solved. We have assumed that 40% of an academic’s salary

is the component attributable to their research. In the DEA ap-

proach we are using, the level of this fraction does not affect the

comparative ranking for the individual persons. However, the use

of a uniform percentage of salary as input means, as argued earlier,

that we need to be using in our sample only data from past and

current STRA staff, for whom there is a uniform expectation (not

necessarily delivery) of research, teaching and administration, and

their salary would reflect this expectation. Moreover, any promo-

tions should in principle have been on the basis of research, again

in expectation rather than in practice. Where then delivery say of

research exceeds expectation that could lead to salary rise through

promotion but the model would see that as stable cost efficiency

as input (salary) would be rising in response to output (research).

The model would then capture as inefficiency where salary cost is

not in keeping with the best ‘ratio’ of research output to salary,

e.g. because promotion was on criteria other than research. Where

promotion might have been on excellence in teaching or outreach

to community (relatively rare to date) that data would not be suit-

able to use in our model. Where, however promotion might have

been on what officially the institution or state would not sanction

(e.g. recruitment or promotion despite insufficient research output)

then the model would reflect that as cost inefficiency of recruit-

ment and promotion practices.

Fig. 1 shows the proportion of the overall salary of each aca-

demic that would have been saved if their initial and subsequent

salary levels were compatible with the duration and the quan-

tity and quality of the research outputs of the benchmark indi-

viduals. The latter are the 12 individuals in Fig. 1 for whom no

scope is found for reducing the salary expenditure of their research

outputs. For the median person, some 12% of their salary might

have been saved if their research performance had matched that

of the benchmark individuals. In aggregate the potential savings

are 17.45% of the salary bill. It is worth noting that in this par-

ticular case the results were not very sensitive to the subjective

choice that a paper in an A ranked journal is worth 2 papers in

B ranked journals. When the model was rerun assuming one A

paper (before or after appointment) is worth 10 corresponding B

papers) the aggregate savings rose from 17.45% to 19.11% of total

salary and two previously benchmark individuals were no longer

so. Thus, though the assessment now valued so much more ex-

tremely papers in journals ranked A compared to B, the findings

for at institution level and therefore the derived views on the re-

cruitment and promotion practices did not alter substantially. The

effects would have been even less pronounced with a more median

trade-off between papers in A and B ranked journals.

From the institutional perspective, it is worth investigating if

there are patterns in the qualifications benchmark individuals of-

fered on recruitment and in their subsequent promotions pattern

and contrast them with those of 12 individuals at the other end of

the spectrum, showing the largest proportions of potential salary

savings. This second set of 12 individuals is estimated on average

to be able to save about 26.66% of their salary if they matched

the benchmark individuals on rate, quantity and quality of publi-

cations. Table 2 contrasts the two sets of individuals by showing

the data for the 12 non-benchmark individuals as percentage of

the corresponding data for the benchmark individuals.

Though the sample of 38 individuals is too small from which to

draw firm conclusions, and the assessment has rested on a num-

ber of subjective assumptions, to the extent our data are realistic

Table 2 can be used to illustrate how findings of assessments of

the type developed in this paper can be used by an institution. For

example, one notable difference between the two sets of individu-

als is the duration in post. Benchmark individuals have on average

been half as long in post as the non-benchmark individuals with

the highest scope for savings. However, research output prior to

appointment of the non-benchmark individuals relative to that of

the benchmark individuals is only 29.32% and 24.6% in A

+ and B

ranked journals respectively. So, on the face of it there has been

a considerable improvement over time in terms of publications of-

fered by those being recruited and this seems to have led to more

productive individuals in terms of research as the lack of scope to

save on research component of salary shows.

In terms of A

+ and B publications after appointment the 12

non-benchmarks have a mean which is 58.41% and almost 92% re-

spectively of that of the benchmark individuals before adjusting

for duration in post. After adjusting for 191.16% duration in post

of non-benchmark individuals A

+ and B ranked publications fall to

30.55% and 48.12% respectively of those of the 12 benchmark in-

dividuals. Finally, citations of benchmark individuals are about 3

times those of the non-benchmark individuals under discussion.

This difference is even more in favour of the benchmark individ-

uals when we reflect on the fact the non-benchmark individuals

have had a longer period to build up the citations compared to the

benchmarks.

Mean salary on appointment is marginally lower for non-

benchmark individuals at 94.41% of that of the benchmark individ-

uals. This is in the right direction given the lower research profile

of non-benchmark individuals at the time of appointment. How-

ever, the differential is not significantly large given the much wider

difference in research publications offered pre-appointment by the

two groups.

Here is the first policy issue that comes into sharp relief: the

salary on appointment does NOT on the face of it discriminate

sufficiently on promise of research. The individuals in the two

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 749

Table 2

Mean values of 12 non-benchmark individuals expressed as % of those for 12 benchmark individuals.

Years in post

Salary on

appointment

Salary from

year 2 on A + before B before A + after B after Citations

12 non-benchmark mean as % of 12

benchmark mean

191 .16 94 .41 113 .21 29 .32 24 .60 58.41 (adj

30.55)

91.99 (adj

48.12)

35 .72

Groups differ by about 6% in favour of benchmark individuals who,

however, offer substantially better research profile on appointment

compared to the non-benchmark individuals. Thus, on the basis of

this admittedly limited sample, the Institution is offering in an in-

consistent manner ‘advance payment’ on the promise of research

rewarding the same way those with less and those with more

promise of research output.

The next question is the promotions policy and whether it is

compatible with research output by individuals. The promotions of

individuals are reflected in their mean annual salary, post year 1.

Here we have a relatively large advantage in favour of the 12 non-

benchmarks. Their mean annual salary post year 1 is 13% higher

than that of the benchmarks. Given the superiority of the bench-

marks on research both pre- and post-appointment the salary dif-

ferential in favour of non-benchmarks is not justified, at least

where research is concerned. If an institution did have findings

of the type depicted in Table 2 it would be incumbent upon it to

check whether its promotion policy is in tune with research out-

put rate, as the indications would be that it is not. For example,

Table 2 suggests that promotions are a reward for longevity in post

rather than for achieved research output. It is true that research

may be seen as less important in career terms by those already es-

tablished in post for considerable time, or after they have achieved

the promotion they sought. However, while this phenomenon may

be understandable from a practical perspective, it is not compatible

with promotion on merit, where merit is earned through research

output. The model identifies where salary increases are not justi-

fiable on research terms, at least not relative to benchmark indi-

viduals. It would be for the institution to decide whether it wishes

to condone ‘slackening’ on research output by individuals and thus

not deem it as an unintended consequence of recruitment and pro-

motion practices where those individuals are concerned. The anal-

ysis brings to the fore the need for the institution to clarify its cri-

teria for promotion when research does not appear to be the main

one in practice.

A further additional consideration from the institutional per-

spective is whether recruitment and promotion decisions have

been deficient in identifying potential for rate of research going

forward or have been too generous in salary terms, given research

output. Fig. 2 shows the optimal values of θ2 , γ1 and γ2 in Model

3. Individuals are ordered by the value of θ2 .

It is clear from Fig. 2 that the biggest range of optimal values

is shown by θ2 , ranging from under 0.2 to 1. This means there is

great variability in rate of research output as the lower the value

of θ2 the shorter the time justified by the person’s research out-

put and so their rate of output should have been higher. The me-

dian value of θ2 is 0.736, which in effect means the median person

has spent about 26% of their time for research unproductively. As

we can see, however, in Fig. 2 there is great variability at individ-

ual person level. For the benchmark individuals, setting the stan-

dard for salary and output profiles, we have θ2 = γ1 = γ2 = 1. In

contrast at the other extreme some individuals have wasted about

80% of their time for research (we have optimal θ2 = 0 . 2) and even

then, they would only justify about 70% of their mean career salary

( γ 2 = 0.7). This can be seen clearly at the person level bars, on the

left end of Fig. 2 .

In contrast to θ2 the values of γ 1 (appointment salary) show

little variation around 1. Thus, not much can be achieved where

savings are concerned by varying appointment salary. This of

course in large measure reflects the legal minimum salary that

must be offered. Nevertheless, some optimal γ 1 values are above 1

suggesting certain individuals should have been offered better ini-

tial salaries albeit at the expense normally of their subsequent an-

nual salary. In one case however, the model suggests that savings

can be achieved through raising both the appointment and subse-

quent salary of an individual if they could produce their research

outputs faster. Finally, the values of γ 2 range from about 0.6 to 1.1

with a median value of 0.936. As can be seen in Fig. 2 for most

individuals the optimal value of γ 2 is below 1. Had this been the

finding for an institution it would be suggesting there is scope for

downward salary adjustment, which in effect means slower pace

of promotion.

The policy implications of the findings above can be sum-

marised as follows:

◦ There is little scope for savings in terms of salary offered at

appointment. Indeed, in a number of cases they should have

been offered higher initial salaries than materialised.

◦ There is scope for savings by making promotions more compat-

ible with research output. There appears significant correlation

between duration in post and career mean annual salary and

yet not matched by research output.

◦ The largest scope for savings is available through a faster pace

of research output. The median research output profile appears

to justify only about 74% of the time they have actually taken

to deliver the research and in some cases a lot less.

It should be noted that in interpreting the results depicted in

Figs. 1 and 2 in terms of the institution’s recruitment and promo-

tion policies we have not gone further to see whether the optimal

values of θ2 , γ1 and γ2 in Model 3 are associated in some way

with year of appointment and/or year of promotion. If such an as-

sociation were to be found, e.g. on average lower θ2 optimal values

for those appointed say 15 years ago compared to those appointed

5 years ago, it would suggest policy in terms of recruiting those

with a better rate of research output had improved over time. As

we wish findings to be comparable across time we can run the as-

sessment using all time periods but we can then examine optimal

values by time period as desired, e.g. to assess the impact of some

policy change that was implemented at some given point in time.

We next consider the particular changes an institution might adopt

to make salaries more in line with research output.

4.2. Assessing the scope for savings through optimising only

recruitment and promotion practices

The preceding section identified potential savings in aggregate

through both salary adjustments and optimising the rate of re-

search output. In this section, we isolate the scope an institu-

tion would have had to control the cost of its research when the

sole instrument at its disposal is the salary it offers its staff. Ar-

guably an institution has more immediate control of this instru-

ment rather than influencing the pace at which its academic staff

deliver research outputs.

We can isolate the scope for savings through salary adjustments

alone by solving a modified version of Model 3 where we set

θ1 = θ2 = 1. Model 3 remains otherwise as stated above so that γ 1

750 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

Fig. 3. Potential saving as per cent of career salary solely through salary adjustments.

Table 3

Mean values by quartile of salary at appointment, standardised on quartile 1 = 100.

Quartiles on

recruitment salary

Salary at

appointment Years in post

Mean annual

salary A + (before) B (before) A + (after) B (after) Citations

Savings

case 2 as %

of full

salary

Q1 100 100 100 100 100 100 100 100 100

Q2 103 .27 113 .67 111 .02 140 .43 154 .37 130.56 (adj

114.86)

289.12 (adj

254.35)

161 .99 147 .76

Q3 114 .60 124 .13 123 .05 191 .49 193 .20 104.17 (adj

83.92)

208.84 (adj

168.24)

145 .46 244 .40

Q4 128 .74 87 .37 127 .50 278 .72 180 .58 61.11 (adj

69.94)

144.90 (adj

165.85)

135 .11 271 .27

and γ 2 are permitted to vary. Thus, the model seeks to minimise

the cost of the research outputs of an academic person solely by

assessing whether his/her salary at appointment and/or the career

mean annual salary could have been different but not below the

legal minimum. Fig. 3 shows the findings at person level.

The median saving is just under 3% while the aggregate savings

are 7.63% of the career total pay of the individuals. These figures

contrast with median potential savings of 12% and aggregate sav-

ings of 17.45% of the total salary bill when rate of research output

is also optimised. Clearly therefore the bulk of the savings possi-

ble, as already established through Fig. 2 , is if rate of research out-

put is optimised. Put another way, the cost per unit of research

output can primarily be reduced through increasing the rate at

which research outputs are produced rather than by manipulating

the salaries of the academics. Indeed, behaviourally rewarding with

higher salaries may be more effective in raising the rate of research

output rather than the other way around.

Fig. 3 suggests that there is significant scope for savings only

in the case of a minority of individuals, perhaps those where es-

timated savings are above 10% of career salary. Such individuals

constitute about 25% of the sample of individuals. The aggregate

potential saving of 7.63% can be taken as an index of the effective-

ness of the recruitment and promotion practices of the institution

when we accept that duration in post, once a person has been ap-

pointed, is in large measure outside the control of the institution

(e.g. because of the legal framework on hiring and firing staff). The

practices relate to decisions made at the time an individual is re-

cruited and at the time or times that individual is promoted.

Taking first the issue of salary at appointment we can inves-

tigate how its levels relate to publications prior to appointment

and performance in research post appointment. Table 3 shows the

mean values of publications before and after appointment and the

scope for savings of full salary through salary adjustments alone.

The data is arranged by quartile on recruitment salary level. Publi-

cations after appointment also show adjusted normalised data for

duration in post.

We would expect that salary at appointment would have a

strong correlation with publications prior to appointment as one

key criterion on which academic appointments are normally made.

However, while the relationship revealed in Table 3 is in the right

direction it is perhaps weaker than we would expect. This is espe-

cially so in going from quartile 1 to quartile 2 where mean ap-

pointment salary rises by only 3.37% yet A

+ publications before

appointment rise by 40.43% and B publications by 54.37%. A more

pronounced lack of proportionality between salary at appointment

and publications before appointment exists between quartile 1 and

quartile 3. Salary rises by about 14.6% while publications before

appointment are nearly double both for A

+ and B ranked journals.

It is true of course that we would not expect a constant returns-to-

scale type proportionality between salary at appointment and level

of publications prior to appointment but here there is relatively lit-

tle reward at appointment time for publications offered. Another

way to view the situation is that it has become tougher to enter

the profession and ever higher level of publication is required for

being offered a post while salaries are low and differ little between

different grades. Appointments do not appear to be made at senior

level unless of course salary differentials between senior and entry

level grades are no more than about 30% of starting salary.

Turning to the relationship between appointment salary and

pre-appointment research profile on the one hand and post ap-

pointment research performance on the other, the picture is mixed.

Publications post appointment in B ranked or equivalent journals

rise on the whole both with appointment salary and with pre-

appointment publications. The rise is not so pronounced however

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 751

where publications in A

+ journals are concerned and in fact ad-

justed for duration in post publications fall. More significantly, as

salary at appointment rises so does scope for savings as can be

seen in the right most column of Table 3 . For example, those ap-

pointed with a mean annual salary in the fourth quartile have

nearly 3 times (271.27%) the scope of savings of those appointed

with a salary in the first quartile. This means post appointment

there is a slackening of research output and a drift towards B from

A journals and the higher the salary at appointment the more pro-

nounced the slackening and the drift. As noted earlier, such slack-

ening in output post appointment and/or promotion may be un-

derstandable at a practical level. However, the model identifies it

as it may be seen as out of line with the institution’s own formal

expectations of STRA academics and thus a potential defect in its

recruitment and promotion criteria.

In summary, had the findings in Table 3 been based on accu-

rate real data they would call into question the relationship be-

tween the salary structure and research output post appointment.

The emerging picture is that salary at appointment up to and

including quartile 3 shows little variation in absolute values yet

publications offered at the time of appointment are significantly

higher. Salary in quartile 4 is somewhat better in line with pub-

lications pre-appointment. Post appointment publications tend to

veer towards B journals but keep to some extent pace with pub-

lications offered at appointment, at least for quartiles 2 and 3 on

appointment salary. The quantity and quality overall however, of

publications and citations post appointment is not commensurate

with expectation relative to benchmark individuals as can be de-

duced from the significant scope for savings (right most column

of Table 3 ). Table 3 shows that the higher the appointment salary

the higher the mean career salary, as we might expect. In fact, up

to and including quartile 3 the mean annual salaries move almost

in unison with mean number of years in post. Only for those ap-

pointed at the higher quartile 4 salaries does the duration in post

not run in line with mean annual salary, suggesting they might

have been appointed more recently to senior posts. However, in

all cases as mean annual salary rises so does the scope for savings.

This suggests promotions and initial appointment salary while they

lead to improvement in number of publications, those are not in

as highly ranked journals and in any case not commensurate with

those of benchmark individuals and hence the significant scope for

savings.

4.3. Assessing the scope for savings through optimising only

promotion practices

We conclude this section by exploring potential savings on the

cost of research outputs through better alignment of the promo-

tions practices, given the salary at which each person has been

appointed and their duration in post. This is achieved by solving

Model 3 in a modified form in which θ1 = θ2 = 1 and a in con-

straint C1.17 is set equal to the actual salary at appointment of the

individual, denoted p 10 . Thus, the model can only now minimise

the cost of research output through varying the mean career salary

given duration in post is fixed and appointment salary could not

have been lower than what transpired at the point of recruitment.

Fig. 4 shows the potential savings in ascending order. Percentages

are of full salary cost.

The median saving when promotions policy is in effect the sole

instrument for minimising the cost of research outputs is 2.35%.

This compares to just under 3% and 12% when ability to reduce ap-

pointment salary and ability to ‘control’ the pace of research out-

put is added respectively. The aggregate savings are 5.83% com-

pared to 7.63% and 17.45% of the career total pay of the individuals

when ability to reduce appointment salary and to ‘control’ the pace

of research output is added respectively. Clearly once we discount

the ability to control the pace of research output, once a person

has been appointed at a certain salary then realistically the sole

instrument for the institution to minimise the cost of research out-

puts is the promotions policy as it influences the person’s mean

annual salary. This, as seen above, offers only a modest scope of

saving only about 6% of the total salary bill.

In order to investigate any link between promotions and perfor-

mance in research before and after appointment we have divided

mean career salary by salary at appointment, to which we shall

refer as the ‘ promotions ratio ’. The higher the promotions ratio the

more likely it is the individual has had one or more promotions.

Table 4 shows mean values of the variables listed, including esti-

mated potential for saving by quartile of promotions ratio. The val-

ues as usual are standardised on quartile 1 = 100 and publications

after appointment are adjusted for duration in post.

We should expect promotions ratio and years in post to move

in the same direction if we assume a person’s research output is

uniform in rate over their career and that promotion is largely

based on research outputs. This expected relationship does not

hold in Table 4 . As we move from quartile 1 to quartile 2 we have

a promotion ratio rise of about 6% while in round figures publi-

cations in B journals post appointment rise by 58% (adjusted for

duration in post) but those in A

+ journals fall by 17%. As we move

further to quartiles 3 and 4 by promotion ratio, publications in A

+

and B ranked journals post appointment fall and only citations ex-

ceed significantly those prior to promotion.

The rise in mean annual salary being out of synch with research

output is strongly reflected in the significant rise in potential to

save in salary, captured in the right most column of Table 4 . Here

we see that while quartile 4 individuals have a mean salary 41%

higher than at the time of their appointment their savings feasi-

ble are 25 times those available for individuals in quartile 1, who

in general have not yet had a promotion. This high factor is an

artefact of the division by a very small number as those who have

not yet had a promotion (quartile 1) have not got a high scope

for savings. It is interesting that there is a general agreement be-

tween promotions ratio and publications in B journals prior to ap-

pointment. It is as if individuals are getting repeat credit by way

of promotion for publications prior to appointment and those have

tended to be in B ranked journals.

5. Policy implications of the findings

A key aim of the approach outlined in this paper is to help

institutions improve their recruitment and promotions policies of

academic faculty from the perspective of the payments they make

for research output. A key notion is that an institution pays for re-

search in prospect rather than in retrospect in the sense that salary

level is set on the research output prospect, albeit the prospect is

largely based on actual research outcomes to date. Once an ap-

pointment or promotion has been made, an expenditure stream is

set to follow going forward, and the duration of this can be in-

dependent of research outcomes for a considerable period of time,

depending on managerial styles, institutional conventions and legal

frameworks. The main aim of the approach developed in this paper

is to aid institutions in taking timely retrospective review of poli-

cies and practices on recruitment and promotion with a view to

making them more effective from the cost perspective going for-

ward. A secondary possibility, whether or not an aim, would be

to aid institutions in managing the research performance of aca-

demics who are still within the institution. This becomes possible

as the review of the effectiveness of past recruitment and promo-

tion policies regarding cost effectiveness of research is through a

review of individual academics from the cost effectiveness of re-

search at person level. The individuals whose data is analysed need

not all be current members of faculty. The data of any faculty that

752 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

Fig. 4. Potential saving as per cent of career salary solely through mean annual salary adjustments.

Table 4

Mean values by quartile of promotions ratio, standardised on quartile 1 = 100.

Quartile on

Promotion

Promotions

ratio Years in post Initial Sal Career Mean A + (before) B (before) A + (after) B (after) Citations

Savings

case 2 as %

of full

salary

Q1 100 100 100 100 100 100 100 100 100 100

Q2 106 .05 78 .02 109 .50 103 .3 85 .19 403 .48 64.88 (adj

83.16)

122.48 (adj

158)

188 .39 538.01

Q3 117 .88 99 .54 101 .49 100 .43 27 .74 178 .01 32.14 (adj

32.29)

78.89 (adj

79.25)

119 .34 971.40

Q4 141 91 .34 105 .20 102 .27 80 .91 356 .00 49.14 (adj

53.80)

76.22 (adj

83.45)

149 .96 2515.33

have left the institution within the period of time analysed are still

legitimate to analyse in terms of conveying evidence on the effi-

cacy of the recruitment and promotions policies of the institution.

From the institutional perspective at the level of policy review

the following issues can be investigated through the analysis car-

ried out in this paper:

a) Have research profiles at recruitment proven good predictors of

performance on research in post?

b) Are there research profiles at recruitment which are associated

with good post recruitment performance in research?

We shall investigate these questions using the basic Model 3 so

we capture the full potential flexibility of action on the part of an

institution, including affecting the rate of research output.

5.1. Did research profile at recruitment prove a good predictor of

ultimate performance in research?

In order to capture the research profile of an individual we ag-

gregate their pre-recruitment A + and B papers using A

+ = 2B . We

are unable to allocate citations to pre- or post-appointment papers

but it is safe to assume most citations will have occurred post ap-

pointment given the generally long periods in post of the individu-

als whose estimated data is being analysed. Table 5 shows the data

by quartile on publications before appointment, standardised on

Q1 = 100. Publications after appointment are also shown adjusted

for duration in post.

There is an interesting non-uniform association between pub-

lications pre- and post- appointment. As we move from quartile

1 to 2 on pre-appointment publications, pre-appointment publi-

cations more than triple reflecting a very low level of publica-

tions by the bottom quartile. Looking at adjusted for duration in

post publications post appointment they generally rise, especially

so for publications in B ranked journals. The rise is not, however

by the same factor by quartile as is for pre-appointment publica-

tions. Citations rise too, as we might expect. There is a hint from

A + publications after appointment that there is some threshold of

publications prior to appointment so that only those offering pre-

appointment publications above the threshold do proceed to of-

fer a level of research output post appointment which is some-

what consistent with their research output pre-appointment. The

threshold appears to be at the median level of publications pre-

appointment. The more interesting finding appearing in Table 5 is

that in the right most column. Potential savings in cost of research

output drop consistently as publications before appointment rise.

This would be in line with intuition, i.e. the stronger the record in

research at appointment the better the subsequent research out-

put, not leaving much scope for savings in cost per unit research

output. One very prominent feature in Table 5 is that those in

Quartile 1 who offer the lowest level of publications at recruit-

ment, have been the longest in post. They have been on average

twice as long in post as the remaining 75% of academics. This sug-

gests as also noted earlier that in general over time the institu-

tion started to recruit academics with better qualifications. This,

did lead to better cost efficiency of research overall.

So, one clear implication, had these findings been based on real

rather than realistic data, is that the institution has been moving

in the right direction in terms of lowering the cost per unit of

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 753

Table 5

Publications and citations after appointment by quartile on papers prior to appointment.

Quartile on Pre-appointment

publications A + Before Years in post A + after B after Citations

Potential

savings as % of

total salary cost

Q1 100 100 100 100 100 100

Q2 328.56 46.46 43.33 (adj. 93.26) 58.97 (adj.

128.86)

57.94 67.90

Q3 643.16 56.64 83.05 (adj. 146.63) 94.86 (adj.

167.47)

122.57 38.88

Q4 1798.62 58.14 66.27 (adj. 113.98) 84.76 (adj.

147.78)

168.30 27.42

Table 6

Benchmark individuals and their performance in research pre- and post-appointment.

Benchmark index

Res output rate and

quality Initial salary

Mean annual

salary A + Equiv before

A + Equiv after

(adj) Citations

Initial Salary

(scaled)

Mean annual

salary (scaled)

Person 1 11 .99 7 .49 16 .48 1 .23 0 .7 13 .10 3 .57 3 .57

Person 2 8 .38 10 .79 5 .97 1 .80 3 .8 40 .40 3 .57 3 .83

Person 3 3 .38 5 .12 1 .64 6 .67 4 .35 95 .10 4 .09 5 .81

Person 4 2 .91 3 .46 2 .35 1 .71 3 .76 26 .30 4 .09 5 .33

Person 5 2 .37 3 .05 1 .70 2 .89 0 .11 13 .70 4 .09 4 .39

Person 6 2 .33 1 .02 3 .63 1 .84 2 23 .70 4 .09 4 .09

research output produced. This has been dropping as the institu-

tion has been recruiting persons with better research output pro-

files over time. However, the institution could do even better by

improving still further the pace of research output, as we saw in

Table 1 .

5.2. Identifying benchmark individuals as case studies where

recruitment and promotions have led to good cost efficiency of

research in post

One additional use of the analysis undertaken would be to iden-

tify appointments which have turned out well in terms of cost of

research in order to identify decision points and relevant decisions

that were made at the time which could inform future policy on

recruitment and promotions.

Table 6 shows the key 6 benchmark individuals identified.

These are individuals who each were used at least once as a com-

parator for setting targets for someone else. The three columns

listed under ‘Benchmark Index’ show the sum of the optimal λor z values for each benchmark individual as Model 3 was solved

over the 38 academics in the assessment sample. Thus, the in-

teger part of the index (plus 1 for any decimal part) shows re-

spectively at least the number of times the person concerned was

used as a benchmark for research output quality and rate, for ini-

tial salary and for mean annual salary. For all these individuals,

we identify no scope to lower their cost per unit of research out-

put. The table also shows the publications before and after recruit-

ment of the individual, having converted them to A + equivalent

using 2 B publications as equivalent to 1 A

+ publication. Salary

data though estimated within the analysis have here been further

scaled for reasons of confidentiality. The same scaling constant was

used for initial and annual mean salary to enable comparison of

the two. Citations are in units of 10. The adjusted publications

post-appointment have been normalised on a confidential duration

in post.

Person 1, used most frequently (12 times on research output

and even more on mean annual salary) is perhaps not an obvi-

ous benchmark. His/her publications record declines after appoint-

ment. However, his/her use as a benchmark may be the result of

having one of the lowest pre- and post-recruitment salaries and a

fair profile if pre- and post-appointment research is set against a

short duration in post. Person 2 has closer to a profile we would

expect for a benchmark. In post, he/she has improved their re-

search output while their salary rise has been only 7.3%. Person

3 has a stable pre- and post- recruitment research output. He/she

is a good example of output and initial salary but not so for pro-

motion (used only about twice as benchmark on post appointment

salary). This is because his/her salary has risen by 42% post ap-

pointment. The person offers by far the best citations record. Per-

son 4 has a significant improvement in their research post appoint-

ment. Their mean annual salary shows a corresponding significant

increase. The last 2 persons show a deterioration in research out-

put over time but their duration in post is relatively short. They

are needed it would appear as benchmarks for initial and mean

annual salary respectively.

In summary, the six individuals with the highest benchmark in-

dices, are used either to benchmark the research output of oth-

ers or to benchmark their salary. The method trades off research

output with salary to minimise the cost of research. In a real-life

context, internal review of the benchmarks would be much more

meaningful without the constraint of confidentiality. The institu-

tion would be able to review the cases to identify what were good

and what not so good about the decisions made at the time of

their appointment so far as financial considerations are concerned.

For example, person 6 might have come in at a higher than need

have been the case salary while person 3 might have been pro-

moted too soon and hence their mean annual salary is not attrac-

tive as a benchmark for others.

6. Conclusion

This paper has developed an approach which universities can

use to assess retrospectively the efficacy of their recruitment and

promotion practices for those where the key criterion for recruit-

ment and promotion is anticipated performance in research. There

is generally a long lead time between the conduct of research

and its appearance in the form of published output. Thus, aca-

demic institutions at the recruitment stage offer salary levels on

the prospect of quantity and quality of research output they per-

ceive the academic concerned will deliver in post. The economic

efficiency of the judgments made can only be ascertained in ret-

rospect after those recruited have had a period of time in post to

deliver the research expected of them. This paper has developed

an instrument for this purpose and illustrates its use on realistic

data (for confidentiality reasons) from a University in Greece.

754 E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755

The DEA model developed is applied at person level. Ideally it

should set the salary component that relates to the research out-

puts of the person concerned against the quantity and quality of

that research. However, as research specific components of salary

are typically not available for academics, a uniform proportion of

the salary of each academic can be used without detriment to

the findings of the model provided contractually the persons being

compared are required to deliver on research in the same man-

ner across their full portfolio of duties. Research outputs by the

individual are treated differently depending on whether they had

appeared prior to recruitment at the institution or after. The lat-

ter are deemed more valuable in terms of promoting the prestige

of the home institution. Further, research outputs are divided for

quality. There is no consensus on how to judge quality of research.

The paper adopts the practice of using the ranking of the journals

where research has appeared as a proxy for its quality. However, an

institution can use its own measures of quantity and quality of re-

search outputs within the broader type of model developed here.

The DEA model constructed is non-linear minimising the cost in

the form of aggregate product of salary and time taken to deliver

the research.

The use of the model is illustrated using data on 38 academics

who had spent a minimum of 3 years in an academic department

of a Greek University. The data for reasons of confidentiality are re-

alistic rather than actual. The paper illustrates how the institution

can estimate retrospectively the scope for savings of the attained

research outputs and thereby draw lessons from the recruitment

and promotions policies practiced hitherto. Such lessons are not so

much for making economies going forward. Rather, the financial

information is a proxy for the mismatch, if any, between expec-

tations of research output and research actually delivered, conse-

quent on the recruitment and promotion policies practiced. Such

information can be used to streamline recruitment and promotion

policies going forward better with the institution’s objectives on

research output by its academic staff.

In the case of the illustrative data analysed some of the key

findings are:

– The institution is recruiting progressively better qualified staff

in terms of research offered at appointment and subsequent re-

search delivered. However, salary offered at appointment does

not reflect the significant differential in research between the

better qualified staff at recruitment who go on to deliver bet-

ter research in post. This could be demotivating for staff able to

deliver good research outputs.

– Promotions appear to match longevity in post better than they

do research output. As in the case of salary at appointment

there is clear mismatch between research delivered in post and

promotions as reflected in mean annual salary post appoint-

ment. This again can prove counter-productive for the research

culture of the institution.

– The key reason a mismatch is found between research outputs

and financial rewards is not the level of salary at appointment

or indeed at promotion. Rather it is the rate of research output.

This has implications for recruitment and for promotion poli-

cies in terms of motivating staff to a higher rate of research

output.

– Publications prior to appointment do tend to be associated

with a better publications record post appointment but only

where pre-appointment publications pass a certain threshold.

The identification of the threshold could help make more pro-

ductive appointments where research is concerned.

The approach developed in the paper can be applied by an aca-

demic institution where research is a key criterion for recruitment,

to monitor the effectiveness of its recruitment and promotion poli-

cies as they evolve. The model developed is a generic one. It can

be adapted to the legal framework where the institution is located

(e.g. restricting the γ values so as not to permit salary below the

legal minimum for each level of post, or by not permitting salary

reduction overtime), or constraining the θ values so as to allow

for a minimum duration, consistent for example for any proba-

tionary period for new staff. Moreover, the institution may wish

to judge its staff against absolute attainment levels rather than

relative ones. For example, it can include within the observations

‘ideal’ staff who would have produced research output of quality

and quantity and within a time frame it would deem ideal. Where

these ideal targets exceed observed data, they would be used as

benchmark within the framework of the DEA model and the find-

ings will then reflect how effective the institution’s recruitment

and promotion policies are relative to the ideal targets. This of

course means the ideal targets would need to be carefully crafted

to be realistic and attainable and within the legal context of the

institution.

Another refinement of the generic model would be to set the

standard below that of the very top performers in research when

assessing recruitment and promotion practices. For example, the

12 benchmark individuals depicted in Fig. 1 can each be assessed

in turn relative to the rest of the sample but without permit-

ting the individual being assessed to also be a potential bench-

mark. This is the spirit of the so-called “super-efficiency” model of

Andersen and Petersen (1993) adapted to the framework of Model

3. The model would need to allow for θ2 to exceed 1 in C3.15 so

that the cost of the academic concerned could exceed their ob-

served salary cost in post, reflecting the potential for such person

to earn a salary in excess of what they were in fact paid. Subjec-

tively then a proportion (e.g. 5%) of the cohort showing the high-

est potential salary in this manner could be excluded from being

benchmark so that potential savings for all the rest are estimated

relative to the remaining STRA staff in the sample. (For an illus-

tration of using super-efficiency models to lower targets for school

pupils in this manner see Thanassoulis, 1999 .) We have not gone

down this avenue but it is clearly an option in the implementa-

tion of Model 3. This avenue could be pursued for example if it

is deemed some exceptional performers in research, perhaps who

were not paid as much as they might have deserved, should not

be used to set standards by which to judge the financial efficacy of

institution’s recruitment and promotion practices.

Once the regal and institutional framework has been integrated

along with any ideal targets within the model, it can be run using

data on existing and past academic staff in order to explore the

implications of past recruitment and promotion policies and draw

lessons for the future.

References

Abbott, M. , & Doucouliagos, C. (2003). The efficiency of Australian universities: A

data envelopment analysis. Economics of Education Review, 22 (1), 89–97 . Anastasiou, A. , Kounetas, K. , Mitropoulos, P. , & Mitropoulos, Y. (2007). Evaluating

efficiency of university academic departments. Journal of Interdisciplinary Math- ematics, 10 (3), 409–431 .

Andersen, P. , & Petersen, N. C. (1993). A procedure for ranking efficient units in data envelopment analysis. Management Science, 39 (10), 1261–1264 .

Athanassopoulos, A. D. , & Shale, E. (1997). Assessing the comparative efficiency of

higher education institutions in the UK by the means of data envelopment anal- ysis. Education Economics, 5 (2), 117–134 .

Avkiran, N. K. (2001). Investigating technical and scale efficiencies of Australian uni- versities through data envelopment analysis. Socio-Economic Planning Sciences,

35 (1), 57–80 . Coelli, T. (1996). Assessing the performance of Australian universities using data en-

velopment analysis . Centre for Efficiency and Productivity Analysis, University of New England, NSW .

De Witte, K. , & Rogge, N. (2010). To publish or not to publish? On the aggregation

and drivers of research performance. Scientometrics, 85 (3), 657–680 . Doyle, J. R. , Arthurs, A. J. , Green, R. H. , Mcaulay, L. , Pitt, M. R. , Bottomley, P. A. ,

et al. (1996). The judge, the model of the judge, and the model of the judged as judge: Analyses of the UK 1992 research assessment exercise data for business

and management studies. Omega, 24 (1), 13–28 .

E. Thanassoulis et al. / European Journal of Operational Research 264 (2018) 742–755 755

Higher Education Funding Council for England (HEFCE), Research Excellence Frame- work. http://results.ref.ac.uk/ .

Johnes, G. , & Johnes, J. (1993). Measuring the research performance of UK Economics departments: An application of data envelopment analysis. Oxford Economic Pa-

pers, 45 (2), 332–347 . Johnes, G. (1995). Scale and technical efficiency in the production of economic re-

search. Applied Economics Letters, 2 (1), 7–11 . Karagiannis, G. , & Paschalidou, G. (2017). Assessing research effectiveness: A com-

parison of alternative non-parametric models. Journal of the Operational Research

Society, 68 (4), 456–468 . Katharaki, M. , & Katharakis, M. G. (2010). A comparative assessment of Greek uni-

versities’ efficiency using quantitative analysis. International Journal of Educa- tional Research, 49 (4-5), 115–128 .

Kounetas, K. , Anastasiou, A. , Mitropoulos, P. , & Mitropoulos, I. (2011). Departmental efficiency differences within a Greek university: An application of a DEA and

Tobit analysis. International Transactions in Operational Research, 18 (5), 545–559 .

Mingers, J. , Watson, K. , & Scaparra, M. P. (2012). Estimating business and manage- ment journal quality from the 2008 research assessment exercise in the UK.

Information Processing & Management, 48 (6), 1078–1093 . Pidd, M. , & Broadbent, J. (2015). Business and management studies in the 2014 re-

search excellence framework. British Journal of Management, 26 (4), 569–581 . Portela, M. C. S. , & Thanassoulis, E. (2014). Economic efficiency when prices are not

fixed: Disentangling quantity and price efficiency. Omega, 47 , 36–44 . Thanassoulis, E. (1999). Setting achievement targets for school children. Education

Economics, 7 (2), 101–119 .

Thanassoulis, E. , De Witte, K. , Johnes, J. , Johnes, G. , Karagiannis, G. , & Portela, M. C. S. (2016). Applications of data envelopment analysis in educa-

tion. Data envelopment analysis. In J. Zhu (Ed.), A handbook of empirical studies and applications . International series in operations research : Vol. 238. New York,

US: Springer .

Assalamualaikum Warahmatullahi Wabarakatuh dan Salam Sejahtera

YBhg. Datuk/ Datin / Prof/ Tuan/ Puan

COSTS, EFFICIENCY, AND ECONOMIES OF SCALE

AND SCOPE IN THE ENGLISH HIGHER EDUCATION

SECTOR

Bil 3/2021

1

COSTS, EFFICIENCY AND ECONOMIES OF SCALE AND SCOPE IN THE ENGLISH HIGHER EDUCATION SECTOR

Jill Johnes

Huddersfield University Business School University of Huddersfield

Queensgate Huddersfield HD1 3DH

Email: J.Johnes@hud.ac.uk Tel: +44 1484 472231

Geraint Johnes

Department of Economics Lancaster University LA1 4YX Email: G.Johnes@lancs.ac.uk

Tel: +44 1524 594215

February 2016

ABSTRACT

Following recent changes to the funding mechanism for higher education, students in England face high ticket prices on tuition. The taxpayer continues to subsidise this education heavily, however, both through direct contributions and throughwriting off unpaid loans. Government therefore retains an interest in the efficiency with which higher education is delivered. We review the literature, and, using data for 2013-14, apply the appropriate frontier methods to model the structure of costs in this diverse sector. In doing so, we uncover information about the returns to scale and scope within the higher education sector, and identify differences in efficiency across institutions.

JEL Classification: I23

Keywords: costs, efficiency, higher education

2

1. Introduction

Under the current funding mechanism for higher education in England, many students will not pay off the whole of the debt that they accrue while studying. The Resource Accounting and Budgeting (RAB) cost – the proportion of the value of student loans that, owing to the write-off of debt after 30 years, will not be repaid – is currently estimated to amount to around 45 per cent1. The tuition fee charged by providers is not necessarily the same, therefore, as the amount paid by customers. In addition, the fact that students do not make up-front payments might make the demand for higher education less price sensitive than would otherwise be the case. These factors have the potential to produce a market failure such that the usual competitive pressures fail to incentivise providers to become more efficient. Moreover, the government continues to subsidise both teaching (through covering the cost of student debt that remains unrecovered after 30 years and also through subsidy of some of the costliest subjects) and research, and therefore has an interest in the efficient operation of all aspects of higher education. An analysis of the cost structure and efficiency of higher education institutions (HEIs) is therefore of on-going interest and importance.

Extensive work has been undertaken on evaluating efficiency in the higher education sectors of various countries. Work in the United Kingdom (UK) is of particular relevance here (see, for example, Johnes 1990; Johnes and Taylor 1990; Johnes, J 1996; Johnes et al. 2005; Johnes 2008; Thanassoulis et al. 2011). Much of the literature on efficiency measurement has emphasised the statistical evaluation of costs (Cohn et al. 1989), since efficiency concerns how a given output can be produced at as low a cost as possible. Statistical and econometric techniques have been developed which allow efficiency to be evaluated for each institution. These statistical methods do not drill down into the detail of how institutions do what they do2; rather they offer the analyst both an understanding of how costs are determined in higher education institutions as a whole, and a measure of the extent to which different institutions manage to produce their outputs efficiently. They thereby allow an assessment to be made of the extent to which institutions differ in terms of their efficiency. The methods provide, at a higher level of abstraction, much the same input into benchmarking exercises as do more detailed qualitative exercises, but offer the advantage of a focus on the front-end activities of teaching and research. A number of studies exist which have adopted this general approach for UK higher education3 (Glass et al. 1995a; 1995b; Johnes, G 1996; Johnes 1997; 1998; Izadi et al. 2002; Johnes et al. 2005; Stevens 2005; Johnes et al. 2008b; Johnes and Johnes 2009; Thanassoulis et al. 2011).

The purpose of this paper is to undertake an empirical study of costs and efficiency in English higher education using data on the most recent available year, namely 2013-14. The paper is in 5 sections of which this is the first. A review of empirical studies of costs in UK higher education is presented in section 2. Section 3 discusses the methodological issues associated with estimating cost functions, and examines how estimates of efficiency can be derived from the cost function. The results of the empirical analysis are presented in section 4. Conclusions are drawn in section 5.

1http://www.publications.parliament.uk/pa/cm201314/cmhansrd/cm140320/text/140320w0002.htm. 2 Unlike, for example, the Transparent Approach to Costing (TRAC). 3 Note that there are also notable studies of cost structures of higher education systems of other countries such as Japan, Italy, Spain, Portugal, the USA and Germany, respectively (Hashimoto and Cohn 1997; Agasisti and Salerno 2007; Johnes and Salas Velasco 2007; Johnes et al. 2008a; Agasisti and Johnes 2010; Johnes and Schwarzenberger 2011).

3

2. Review of the literature on costs in higher education in the UK

As in many developed countries, higher education funding in the UK poses challenges for governments under pressure to reduce public budget deficits and for HEIs which face continuous competitive pressure to do more with less. A thorough understanding of universities’ costs and economies of scale and scope is crucial in determining how universities should be organised to make the best use of their resources.

There is now a considerable literature concerning the cost structure and efficiency of systems of higher education. While the earliest work on university cost functions for the UK (Verry and Layard 1975; Verry and Davies 1976) acknowledges that universities are multi-product firms (producing teaching and research), both the estimation method and specification of the cost function are restrictive since they allow for only limited economies of scale and preclude altogether the possibility of economies of scope. Indeed, the complexities of the operation of multiproduct organisations identified by Baumol et al. (1982) were first recognised in the higher education context in the seminal work of Cohn et al. (1989). Subsequent studies have exploited developments in frontier estimation methods (Aigner et al. 1977; Charnes et al. 1978; 1979) in order to combine the estimation of multi-product cost functions with estimation of efficiency (in the UK context see, for example, Johnes, G 1996; Johnes 1998; Izadi et al. 2002; Johnes et al. 2005; Stevens 2005; Thanassoulis et al. 2011).

The adoption of frontier estimation techniques to estimate cost functions and efficiency leads to analysts facing a choice of methods of analysis. While all empirical cost efficiency evaluations are theoretically rooted in the work of Farrell (1957), they have generally employed one of two main methodological approaches. The non-parametric approach of data envelopment analysis (DEA) (Charnes et al. 1978; 1979) allows, in the evaluation of a production or cost technology, input and output weightss to differ across institutions;the method consequently has an advantage when applied to a sector comprised of a highly diverse set of institutions as it allows universities to pursue their own specific missions without penalising their estimated efficiency. By way of contrast, the parametric approach of stochastic frontier analysis (SFA) (Aigner et al. 1977) is less flexible in its basic form, estimating a cost function with – in its simplest variant at least – identical parameters for all units in the sector, but it has the advantage of permitting statistical inference and calculation of economies of scale and scope.

Institutions within the English higher education sector are highly diverse in terms of, for example, age, size, subject mix, research intensity and external engagement. There is no reason to expect that cost structures and efficiency should be the same across all these HEIs, and parametric estimation methods have therefore needed to be adapted in order to accommodate this heterogeneity.

One way of addressing this is to focus attention on particular (pre-defined) groups of universities, and to estimate a separate cost function for each one. The English higher education sector can be categorised into three broad groups. Traditional universities, which had university status prior to 1992, offer degree programmes across the spectrum of academic subjects and have a well-developed research mission. Indeed these institutions often describe themselves as being ‘research led’. A large cohort of other institutions received university status in 1992; these institutions have a

4

balanced portfolio offering degree programmes across a range of academic and vocational subjects, and undertake research. Finally, since 2003 a third wave of institutions, previously colleges of higher education, have been awarded university status. This last group comprises a diverse set of institutions; some are small and specialist, and many lack a strong research mission.

Early cost studies focus solely on traditional (pre-1992) universities (Glass et al. 1995a; 1995b; Johnes, G 1996; Johnes 1998) where scale economies appear to be significant and unexhausted for the typical university; evidence on economies of scope is mixed and varies according to how the outputs are aggregated. In essence, the more highly aggregated the teaching outputs across subjects in the cost function specification, the less likely we are to observe economies of scope.

Later studies use data across both pre- and post-1992 HEIs to estimate cost functions (Johnes 1997; Izadi et al. 2002), and more recent work has included English universities from across all three groupings (Johnes et al. 2005; Johnes et al. 2008b; Thanassoulis et al. 2011). These studies have taken advantage of the increasing availability of appropriate data which permit the use of panel estimation methods as a means of dealing with unobserved heterogeneity. But changes over time in the funding regime of the English higher education sector mean that obtaining a panel which is comparable over time is difficult, and this has led to modelling problems in the panel context, especially for longitudinal data over a relatively long time period (Johnes and Johnes 2013). These studies find that scale economies are close to constant or decreasing for the typical university (Johnes 1997; Izadi et al. 2002; Johnes et al. 2005; Johnes et al. 2008b; Johnes and Johnes 2009) while global diseconomies of scope are a consistent finding (Johnes 1997; Izadi et al. 2002; Johnes et al. 2005; Johnes et al. 2008b; Johnes and Johnes 2009).

There is also clear evidence that efficiency estimates vary by mission group: former colleges of higher education appear to be least efficient, followed by post-1992 and then pre-1992 HEIs (Johnes et al. 2005; Johnes et al. 2008b; Johnes and Johnes 2009; Thanassoulis et al. 2011). There is, across the full range of institutions, a considerable range in efficiency scores; for example, Thanassoulis et al. (2011) find, using SFA, that while mean efficiency is 0.75, it varies from 0.06 to 0.99. This vast range is likely a consequence of the diversity of the HEIs in the sample. Institutions at the lower end of the distribution of efficiencies tend to have characteristics (such as quality, size or specialisation), inadequately captured in the ‘one size fits all’ specification of the cost function, that ‘explain’ their relatively high costs. The efficiency scores attached to these institutions therefore need to be treated with considerable caution.

One way of addressing the issue of diversity adopted in a number of studies is to add a set of exogenous control variables which might affect costs into the estimated cost function. Such factors include ‘quality’ of students, input prices (as reflected by geographical location dummies), real estate costs, success in strategies to widen student participation in higher education, and measures of third mission (knowledge transfer) activity. While one study finds that the proportion of students achieving first and upper second class degrees has a positive influence on both costs and on efficiency (Stevens 2005), student quality has generally not been found to be a significant determinant of costs (Verry and Davies 1976; Johnes et al. 2005; Johnes et al. 2008b; Johnes and Johnes 2009). Accounting for third mission activity has proved to be very difficult in practice because of the paucity of data. Variables used to reflect third mission include income from other services rendered (Johnes et al. 2005), income from intellectual property (including contract research,

5

consultancy and income from engagement with business and the community), and staff time directed at and attendees at public events (Johnes and Johnes 2013). Data around public events seem particularly unreliable; income from intellectual property has the expected positive relationship with costs (Johnes and Johnes 2013). More generally, identifying control variables that might influence costs, over and above the standard outputs of teaching and research, has not proved particularly successful in studies to date.

A small number of recent cost studies has therefore experimented with random parameter stochastic frontier models (Tsionas 2002; Greene 2005). These models do not require data on time-invariant influences on costs, but – as an advance on fixed effects models – allow all parameters of the cost equation to vary across institutions. Much like DEA, the random parameter stochastic frontier estimation method thus effectively estimates a separate technology for each unit of observation. Applications of this method are limited with only one applied to higher education in England (Johnes and Johnes 2009) and the others to universities in Italy, Germany and the USA (Agasisti and Johnes 2010; Johnes and Schwarzenberger 2011; Agasisti and Johnes 2015). A common theme of these studies is that, for the typical university, economies of scale are generally exhausted (although this is not the case for typical universities in Germany) and opportunities for savings arising from global economies of scope are limited. The disadvantage of this approach is that the model is very demanding of the data and can, in consequence, be difficult to estimate. In addition, by allowing each HEI to have its own mission and be judged in isolation, the random parameters approach might be considered to be too tolerant of high-cost practices.

An alternative to random parameter frontier estimation involves estimating separate frontiers for two or more groups of institutions. This is less permissive than the random parameters method in that it allows variation in the parameters between institutions in different classes, but no variation within each class. The latent class approach is particularly attractive because it allows the membership of each class to be determined by the data without need for the analyst to prescribe which institutions belong in which class. Unlike the random parameter approach, the latent class model can be estimated using a cross-section of data; in a policy environment that is rapidly changing, this is an important advantage. To date this method has not received much attention in the higher education literature (exceptions include Johnes and Johnes 2013; Agasisti and Johnes 2015). This is a gap which we intend to fill with this study.

3. Methodological approach

The review of the literature indicates that the functional form of the cost function, the modelling of economies of scale and scope, and the choice of estimation technique are all important issues to address in undertaking an analysis of university costs and efficiency.

3.1 Functional form

A cost function relates costs to the set of outputs produced, given prices of inputs. For institution 𝑘𝑘 this is written in the general form

𝐶𝐶𝑘𝑘 = 𝑓𝑓(𝑦𝑦𝑖𝑖𝑘𝑘 ,𝑤𝑤𝑙𝑙𝑘𝑘) (1)

Where 𝐶𝐶𝑘𝑘 represents costs for university 𝑘𝑘, 𝑦𝑦𝑖𝑖𝑘𝑘 is quantity of output 𝑖𝑖 for university 𝑘𝑘 and 𝑤𝑤𝑙𝑙𝑘𝑘 is the price of input 𝑙𝑙 for university 𝑘𝑘. We estimate a quadratic cost function; this satisfies the desiderata

6

identified by Baumol et al. (1982) – including the requirement that the equation should produce sensible estimates of costs when there is zero production of one or more of the outputs. The specification is therefore:

𝐶𝐶𝑘𝑘 = 𝛼𝛼0 +∑ 𝛽𝛽𝑖𝑖𝑦𝑦𝑖𝑖𝑘𝑘𝑖𝑖 + 12∑ ∑ 𝛾𝛾𝑖𝑖𝑖𝑖𝑦𝑦𝑖𝑖𝑘𝑘𝑦𝑦𝑖𝑖𝑘𝑘𝑖𝑖𝑖𝑖 + ∑ 𝛿𝛿𝑙𝑙𝑤𝑤𝑙𝑙𝑘𝑘𝑙𝑙 + 𝜀𝜀𝑘𝑘 (2)

where 𝜀𝜀𝑘𝑘 is an institution-specific residual, and 𝛼𝛼0, 𝛾𝛾𝑖𝑖𝑖𝑖 and 𝛿𝛿𝑙𝑙 are parameters to be estimated.

3.2 Economics of scale and scope

Measures of the returns to scale and scope suggested by the estimated cost function are evaluated following Baumol et al. (1982). These measures are defined in Table 1. In the case of the returns to scale, the measures all draw on the idea, familiar from the literature on single-product firms, that average costs are higher than marginal costs over the range of output where the former are decreasing. Where the measure of ray or product-specific returns to scale exceeds unity, there are increasing returns to scale; where the measure is below unity, returns to scale are decreasing. The evaluation of global economies of scope involves examining the cost of producing all the outputs of the typical university together and comparing that to the sum of the costs of producing each output (at the same level) in separate production units. Product-specific economies of scope refer to the cost savings (or otherwise) of producing one specific output along with all the others. Economies of scope (global or product-specific) are observed when the corresponding measure is positive.

<Table 1 here>

3.3 Estimation method

A review of the literature identifies problems of estimating higher education cost functions for a diverse higher education sector such as that observed in England. Panel data estimation with a random parameters specification might offer a way forward, but recent changes in the student funding mechanism – notably the increase in undergraduate tuition fees to £9000 – limit the extent to which data from different years are comparable. In this paper, we therefore apply a latent class stochastic frontier estimation approach to cross-section data. Specifically the latent class stochastic frontier model for each class 𝑚𝑚 is

𝐶𝐶𝑘𝑘,𝑚𝑚 = 𝛼𝛼0,𝑚𝑚 + ∑ 𝛽𝛽𝑖𝑖,𝑚𝑚𝑦𝑦𝑖𝑖𝑘𝑘𝑖𝑖 + 12∑ ∑ 𝛾𝛾𝑖𝑖𝑖𝑖,𝑚𝑚𝑦𝑦𝑖𝑖𝑘𝑘𝑦𝑦𝑖𝑖𝑘𝑘𝑖𝑖𝑖𝑖 +∑ 𝛿𝛿𝑙𝑙,𝑚𝑚𝑤𝑤𝑙𝑙𝑘𝑘𝑙𝑙 + 𝑣𝑣𝑘𝑘,𝑚𝑚 + 𝑢𝑢𝑘𝑘,𝑚𝑚 (3)

is estimated, with the analyst prescribing how many latent classes exist, but with the membership of each latent class being determined alongside the parameter values and the residual terms by maximum likelihood.

To summarise, it is possible to combine the stochastic frontier and latent class approaches so that (i) cost frontiers (or envelopes) are estimated (ii) yielding measures of the efficiency of each organisation in the data set and (iii) establishing which organisations belong in each of the latent classes or groups. It is useful to illustrate this method graphically. Consider Figure 1. This shows a scatter plot of points, each of which describes the costs and output levels of a single observation. Each observation might represent a decision-making unit or organisation – in our case a higher education institution. A straightforward latent class analysis of these data might involve the analyst in specifying that there are two different types of institution in the data set. The latent class model

7

therefore fits two lines to the data. These are shown by the two dashed lines. In fitting these two lines, the model also determines which observations belong to which of the two latent classes – thus the model classifies some of the cost-output pairings into class X and some into class Y. These letters are shown as the data points on the diagram, but it should be emphasised that the observations are placed in these classes by the maximum likelihood algorithm used in the latent class estimation itself; the observations are not placed within one class or the other by the analyst.

<Figure 1 here>

Now the two dashed lines represent the best fit that is associated with the observations (given that there are two latent classes), but they do not represent the cost envelope faced by organisations within each of these two classes. To find these cost envelopes, the latent class method must be used alongside a stochastic frontier model. Doing this moves the lines down (though this is not necessarily a parallel shift). The resultant cost envelopes are represented by the solid lines. Note that, within each latent class, some observations lie below the cost frontier (because of the stochastic error component), but most lie above. The preponderance of observations above the frontiers represents inefficiency. The technique allows the efficiency of each observation to be evaluated by reference to its position relative to the frontier for the latent class to which the observation belongs.4

4. Empirical analysis

Data on costs and outputs in higher education are published by the Higher Education Statistics Agency (HESA) in a series of annual publications, including Students in Higher Education and Finances of Higher Education Providers. In the analysis that follows, we draw on these data to estimate a frontier model of costs as a function of several outputs and input prices, recognising the multi-product nature of higher education.

The costs variable includes current expenditures excluding ‘hotel’ costs associated with residences and catering. Student numbers – classified, in the case of undergraduates, into the sciences and other subjects – are expressed as full-time equivalents. We eschew the option to employ a finer disaggregation of the student body owing to problems of multicollinearity. Following the precedent set by earlier studies, we use research income as the measure of research activity. This measure has the virtue of providing a market value for research, hence appropriately weighting quantity and quality. While it is a prospective measure, and may be criticised for being an input rather than an output, it is typically highly correlated with measures (such as publications or citations) that are more unambiguously considered to be research outputs, but which are more retrospective in nature (Johnes and Johnes 2013).

Previous analyses of university costs have, with few exceptions, failed to control for variations in costs due to the impact of labour market conditions differentially affecting institutions. An important exception – albeit one that predates the use of frontier methods – is that of Cohn et al. (1989) where wage is found to be a significant variable in the cost function. In this paper, we use as a control a measure of hedonic costs in the labour market derived as the residual from a regression of institutions’ salary costs against a vector of variables describing the numbers of staff in each of ten age groups. Definitions of all variables can be found in Table 2.

4 This is done using a method developed by Jondrow et al. (1982).

8

<Table 2 here>

The sample used for estimation comprises 103 higher education institutions in England.5 A latent class stochastic frontier model (with 2 classes) is applied to these data, employing a quadratic specification to capture returns to scale and scope, the latter being due to synergies across the various teaching activities and research. Results are shown in Table 3.6

<Table 3 here>

These results are not amenable to straightforward interpretation, and we defer to later a consideration of what they imply about cost structures in the higher education sector.7 First, we examine the composition of the two latent classes. Recall that the classification of institution into one class or the other is determined (alongside the coefficient estimates and the one-sided residual that captures inefficiency) by the criterion that this should optimise the fit of the model to the data. In essence, the latent classes each contain institutions that are, in some sense not directly observed in the variables, similar to one another but are distinct from those in the other class. Using the latent class approach thus accommodates a degree (albeit limited) of unobserved heterogeneity across institutions, and a look at the composition of each class should teach us something about which institutions are alike and which are not.

Table 4 reports descriptive statistics for the key variables for the institutions belonging to each class, and detail of class membership appear in Table 5. On most measures, the means of the variables are quite similar across classes, though the extent of research activity is on average somewhat higher in the second latent class than in the first. More detailed investigation shows some more pronounced differences between the two classes, however. Latent class 2 contains the largest universities, and also includes many of the smaller institutions that have gained university status since the turn of the millennium (see Table 5 for details). The standard deviation attached to all variables is correspondingly greater for this latent class than for the first.8

<Tables 4 and 5 here>

In Table 6, we report the levels of average incremental costs for each output. The calculations are provided for the institution producing, respectively, an average output vector, twice and half the average output vector, in each of the latent classes. It is worth noting that an average institution (or

5 We exclude from the sample a number of institutions that are, for one reason or another, idiosyncratic. These are: the ancient universities of Oxford and Cambridge, whose costs are affected by their internal structures and teaching methods; small and specialist institutions with costs below £25m per year; the University of Arts, London, for which, owing to an unusual employment structure, we were unable to obtain hedonic salary cost; Buckingham, which is fully private; Open University, which specialises in distance learning; and the London University (Institutes and Activities), which comprises a number of highly specialised research centres. Data for University Campus, Suffolk are divided equally between the Universities of Essex and East Anglia; figures for Liverpool Tropical Medicine are added to Liverpool University. 6 The Akaike Information Criterion rejects the 1-class model (a conventional model applied across the whole sample) in favour of the 2-class latent class model at a conventional significance level. 7 Relatively few coefficients differ significantly from zero. This is usual in this type of analysis, given the highly nonlinear nature of the functional form. 8 The mixed character of the institutions in the second latent class begs the question of whether the number of classes ought to be extended. We have tried to do this, but statistical considerations (specifically a singular variance matrix) prevent estimation.

9

one that is double or half the size of the typical institution) is a hypothetical construct; in reality, HEIs may specialise to a greater or lesser extent in a particular output.

In class 1, the costs associated with undergraduate education for the typical university in that class average between £6000 and £7000 for both the sciences and other subject areas. This is low in comparison with tuition fees, which in most cases amount to the maximum permitted £9000 for domestic students. Postgraduate tuition involves institutions in greater expenditure per student. This is unsurprising in view of the facts that classes for taught programmes are often relatively small, and that research requires one-on-one supervision. The average incremental cost of research is high, at around 2.5 – indicating that each pound of research income is associated with more than twice as much expenditure. This confirms the conventional wisdom that income from teaching is used to cross-subsidise research. The main distinction between classes 1 and 2 is that postgraduate costs are considerably higher in class 2 than class 1. The general pattern of average costs across output types in both classes is similar for the scenarios in which output is greater than or less than the average.

<Table 6 here>

In Table 7, we report estimates of product-specific and ray returns to scale for institutions within each latent class for the HEI of average, twice and half the average output vector. Ray economies of scale are unambiguously exhausted across both classes at large sizes; but there are potential economies for the smallest HEIs (in both classes) and also for typical universities in class 1. Note that the typical HEI in class 2, for which ray economies of scale are already exhausted, is larger in some dimensions than its counterpart in class 1.

<Table 7 here>

Typically product-specific economies of scale are exhausted, but there are two exceptions – unexhausted economies are observed for postgraduate education and for research in institutions in the second latent class. This finding is consistent with earlier studies (Johnes et al. 2008b; Johnes and Johnes 2009), and suggests that there is some scope for increased concentration of postgraduate and research activity amongst these institutions.

Estimates of economics of scope are reported in Table 8. We find that global economies of scope are exhausted (to varying degrees) in both classes. Undergraduate teaching in subjects other than science, however, exhibits economies of scope across both classes and all sizes of HEI. The opposite is observed for undergraduate teaching in science subjects and research.

<Table 8 here>

The one-sided regression residuals that emerge from the stochastic frontier model can be compared across institutions by calculating the ratio of predicted costs to the sum of predicted costs and the residual. This gives a measure, bounded from above by unity, of the extent to which each institution is efficient. As may be observed in Figure 2, the degree of efficiency in our sample is very high: all institutions are better than 90% efficient, and most are very close to full efficiency9. The institution

9 Note that the efficiency component for each of the latent class models is not significantly different from zero, adding further evidence that efficiency is high.

10

achieving the lowest efficiency score, at 90.6%, is the University of Arts, Bournemouth – a small and specialist institution whose costs are only marginally above the lower limit of £25m used as a cutoff for the purposes of this study.

<Figure 2 here>

5. Conclusions

The statistical approach to evaluating efficiency offers an evidence base on which to begin a more refined consideration. Most studies of efficiency in the higher education have demonstrated that the sector appears to be reasonably efficient – though it should be borne in mind that efficiency is defined by reference to best observed practice. The results do not, therefore, support the notion that substantial sector-wide gains could be made by using efficiency scores as a criterion for resource allocation.

That said, efficiency is a slippery concept. A user of the results of a statistical analysis may deem some characteristics of institutions, but not others, to be legitimate explanations of cost variations. This issue is further complicated by the fact that some of the characteristics that influence costs can be measured whereas others cannot – though, by using latent class modelling, both observable and (some) unobservable characteristics can be allowed for in the calculation of an efficiency score. In sum, statistical analysis can take us some of the way towards understanding efficiency, but a full understanding requires more qualitative research.

11

Figure 1: Illustration of the latent class approach

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Y

Y Y

Y

Y

Y

Y

Y

Y

Y

costs

output

Key:

Line of best fit

Stochastic frontier

12

Figure 2: Distribution of efficiency scores

Freq

uenc

y

EFF

0

18

36

54

72

. 900 . 925 . 950 . 975 1. 000

13

Table 1: Evaluating economies of scale and scope

Ray/global Product-specific Economies of scale 𝑆𝑆𝑅𝑅 =

𝐶𝐶(𝑦𝑦)∑ 𝑦𝑦𝑖𝑖𝐶𝐶𝑖𝑖(𝑦𝑦)𝑖𝑖

𝑆𝑆𝑖𝑖(𝑦𝑦) = 𝐴𝐴𝐴𝐴𝐶𝐶(𝑦𝑦𝑖𝑖)/𝐶𝐶𝑖𝑖(𝑦𝑦)

Implication If 𝑆𝑆𝑅𝑅 > 1 (𝑆𝑆𝑅𝑅 < 1) then we have ray economies (diseconomies) of scale.

If 𝑆𝑆𝑖𝑖(𝑦𝑦) > 0 (𝑆𝑆𝑖𝑖(𝑦𝑦) < 0) then there are product-specific economies (diseconomies) of scale for output 𝑖𝑖.

Economies of scope 𝑆𝑆𝐺𝐺 = ��𝐶𝐶(𝑦𝑦𝑖𝑖)

𝑖𝑖

− 𝐶𝐶(𝑦𝑦)� /𝐶𝐶(𝑦𝑦)

𝑆𝑆𝐶𝐶𝑖𝑖 = ��𝐶𝐶(𝑦𝑦𝑖𝑖)𝑖𝑖

+ 𝐶𝐶(𝑦𝑦𝑛𝑛−𝑖𝑖) − 𝐶𝐶(𝑦𝑦)� 𝐶𝐶(𝑦𝑦)�

Implication If 𝑆𝑆𝐺𝐺 > 0 (𝑆𝑆𝐺𝐺 < 0) then we observe global economies (diseconomies) of scope from producing all the outputs together rather than each one in a separate firm.

If 𝑆𝑆𝐶𝐶𝑖𝑖 > 0 (𝑆𝑆𝐶𝐶𝑖𝑖 < 0) then there are product-specific economies (diseconomies) of scope for output 𝑖𝑖 suggesting that there are cost savings (dissavings) from producing this output with all the others.

Where 𝐶𝐶(𝑦𝑦) is the total cost of producing the output vector 𝑦𝑦; 𝐶𝐶𝑖𝑖(𝑦𝑦) is the marginal cost of producing the 𝑖𝑖th output; the average incremental cost of output 𝑖𝑖 is 𝐴𝐴𝐴𝐴𝐶𝐶(𝑦𝑦𝑖𝑖) = [𝐶𝐶(𝑦𝑦𝑛𝑛) − 𝐶𝐶(𝑦𝑦𝑛𝑛−𝑖𝑖)]/𝑦𝑦𝑖𝑖; 𝐶𝐶(𝑦𝑦𝑛𝑛) is the total cost of producing all the outputs at the levels in vector 𝑦𝑦; 𝐶𝐶(𝑦𝑦𝑛𝑛−𝑖𝑖) is the total cost of producing all outputs at the levels in vector 𝑦𝑦 except for output 𝑖𝑖 which is set to zero; 𝐶𝐶(𝑦𝑦𝑖𝑖) is the cost of producing output 𝑖𝑖 in a separate firm at the same level as in the output vector 𝑦𝑦.

14

Table 2: Data Definitions

Variable name Definition Units Dependent variable COST Total expenditure minus expenditure on residences and catering

operations £000s

Undergraduate teaching UGS UGA

Undergraduate students (first degree and other) in sciences (medicine and dentistry, subjects allied to medicine, biological sciences, veterinary science, agriculture and related subjects, physical sciences, mathematical sciences, computer science, engineering and technology, and architecture, building and planning) Undergraduate students (first degree and other) in all other subjects (social studies, law, business and administrative studies, mass communications and documentation, languages, historical and philosophical studies, creative arts and design, and education)

FTEs FTEs

Postgraduate teaching PG Postgraduate students in all subjects FTEs Research RES HEFCE R plus income from research grants and contracts £000s Input prices WAGE The residual from a hedonic wage function i.e. a regression of institutions’

salary costs against a vector of variables describing the numbers of staff in each of ten age groups

£

15

Table 3: Latent class stochastic frontier quadratic cost functions

Latent class 1 Latent class 2 Constant 36.179 5.526 (573x105) (5.36) Undergraduates: non-science (UGA) 1.990 4.553 (4.93) (2.45) Undergraduates: science (UGS) 5.065 4.067 (6.05) (2.59) Postgraduates (PG) 16.450 19.118 (18.03) (7.38) Research (RES) 2.365 1.664 (1.09) (0.21) UGA2 0.217 0.023 (0.76) (0.31) UGS2 0.003 0.498 (1.14) (0.49) PG2 2.592 -3.653 (4.32) (3.55) RES2 0.004 -0.008 (0.01) (0.00) UGA*UGS 0.014 -0.220 (1.51) (0.57) UGA*PG -1.199 0.622 (4.28) (1.93) UGA*RES 0.176 -0.061 (0.22) (0.07) UGS*PG 0.843 0.893 (3.29) (1.52) UGS*RES -0.025 0.005 (0.17) (0.08) PG*RES -0.284 0.398 (0.43) (0.22) Hedonic wage costs 0.797 0.222 (0.48) (0.31) Log likelihood -407.88 Note: standard errors in parentheses

16

Table 4: Descriptive statistics of variables, by latent class

Latent class 1 Latent class 2 mean SD mean SD Cost 193.443 123.661 184.298 205.650 Undergraduates, science (thou)

4.938 2.648 5.078 3.997

Undergraduates, other (thou)

6.029 2.955 5.819 3.530

Postgraduates (thou) 2.579 1.410 2.536 2.465 Research (mill) 23.045 43.774 28.784 58.878 Number in class 54 49

17

Table 5: Latent class membership, ranked within each class from highest to lowest total cost Latent Class 1

>£200K £100K-£200K <£100K Imperial College Lancaster Northampton

Liverpool City Southampton Solent Southampton Surrey St George’s Hospital

Bristol Nottingham Trent SOAS Warwick Sussex West London

Queen Mary College Kent Royal Veterinary College Exeter Bath University for the Creative Arts York Portsmouth Falmouth

Durham Anglia Ruskin Leicester Salford Reading Middlesex

Sheffield Hallam Brunel London School of Economics Hull

East Anglia Brighton Northumbria Westminster Hertfordshire De Montfort

Wolverhampton Cranfield East London Oxford Brookes Bradford South Bank Sunderland Derby Royal Holloway and Bedford Huddersfield London Metropolitan Keele London Business School Lincoln Latent Class 2

>£200K £100K-£200K <£100K University College London Central Lancashire Birkbeck

Manchester Kingston Edge Hill King’s College Greenwich Institute of Cancer Research Nottingham Liverpool John Moores Chester

Leeds Leeds Beckett Goldsmiths Sheffield Birmingham City Roehampton

Birmingham Essex Institute of Education Newcastle-upon-Tyne London Sch Hygiene & Trop Med Worcester

Plymouth Bournemouth Gloucestershire Manchester Metropolitan Bedfordshire Cumbria

Coventry Staffordshire Buckinghamshire New Loughborough Teesside Bath Spa

West of England Aston Winchester Canterbury Christ Church Liverpool Hope York St John Chichester Bolton Royal College of Art University College Birmingham St Mary’s Twickenham Harper Adams Arts University Bournemouth

18

Table 6: Estimates of Average Incremental Costs (AICs) at various levels of output by latent classes

Latent class 1 Latent class 2 Mean 2*Mean 0.5*Mean Mean 2*Mean 0.5*Mean Undergraduate sciences 6763 8461 5914 7726 11386 5896 Undergraduate other 4337 6684 3164 3401 2250 3977 Postgraduate 13533 10616 14992 29474 39830 24296 Research 2.67 2.97 2.52 2.58 3.46 1.96

Table 7: Estimates of returns to scale at various levels of output by latent classes

Latent class 1 Latent class 2 Mean 2*Mean 0.5*Mean Mean 2*Mean 0.5*Mean Undergraduate sciences 1.00 1.00 1.00 0.75 0.69 0.82 Undergraduate other 0.77 0.72 0.83 0.96 0.89 0.98 Postgraduate 0.67 0.44 0.82 1.46 1.87 1.24 Research 0.97 0.94 0.98 1.37 1.66 1.11 Ray returns 1.06 0.86 1.38 0.94 0.85 1.02

19

Table 8: Estimates of economies of scope at various levels of output by latent classes

Latent class 1 Latent class 2 Mean 2*Mean 0.5*Mean Mean 2*Mean 0.5*Mean Undergraduate sciences -0.04 -0.08 -0.02 -0.03 -0.06 -0.02 Undergraduate other 0.15 0.03 0.32 0.08 0.10 0.09 Postgraduate 0.13 0.24 0.06 -0.30 -0.55 -0.15 Research -0.03 -0.05 -0.01 -0.20 -0.36 -0.07 Global returns 0.01 0.02 0.00 -0.20 -0.37 -0.10

20

References Agasisti, T. and G. Johnes (2010). 'Heterogeneity and the evaluation of efficiency: The case of Italian

universities.' Applied Economics 42(11): 1365-1376. Agasisti, T. and G. Johnes (2015). 'Efficiency, costs, rankings and heterogeneity: The case of US

higher education.' Studies in Higher Education 40(1): 60-82. Agasisti, T. and C. Salerno (2007). 'Assessing the cost efficiency of Italian universities.' Education

Economics 15(4): 455-471. Aigner, D., C. A. K. Lovell and P. Schmidt (1977). 'Formulation and estimation of stochastic frontier

production models.' Journal of Econometrics 6: 21-37. Baumol, W. J., J. C. Panzar and R. D. Willig (1982). Contestable Markets and the Theory of Industry

Structure. London, Harcourt Brace Jovanovich. Charnes, A., W. W. Cooper and E. Rhodes (1978). 'Measuring the efficiency of decision making units.'

European Journal of Operational Research 2(4): 429-444. Charnes, A., W. W. Cooper and E. Rhodes (1979). 'Measuring the efficiency of decision making units:

a short communication.' European Journal of Operational Research 3(4): 339. Cohn, E., S. L. W. Rhine and M. C. Santos (1989). 'Institutions of higher education as multi-product

firms: economies of scale and scope.' Review of Economics and Statistics 71(2): 284-290. Glass, J. C., D. G. McKillop and N. S. Hyndman (1995a). 'The achievement of scale efficiency in UK

universities: a multiple-input multiple-output analysis ' Education Economics 3(3): 249-263. Glass, J. C., D. G. McKillop and N. S. Hyndman (1995b). 'Efficiency in the provision of university

teaching and research: an empirical analysis of UK universities.' Journal of Applied Econometrics 10(1): 61-72.

Greene, W. (2005). 'Reconsidering heterogeneity in panel data estimators of the stochastic frontier model.' Journal of Econometrics 126: 269-303.

Hashimoto, K. and E. Cohn (1997). 'Economies of scale and scope in Japanese private universities.' Education Economics 5(2): 107-115.

Izadi, H., G. Johnes, R. Oskrochi and R. Crouchley (2002). 'Stochastic frontier estimation of a CES cost function: the case of higher education in Britain.' Economics of Education Review 21(1): 63-71.

Johnes, G. (1996). 'Multi-product cost functions and the funding of tuition in UK universities.' Applied Economics Letters 3: 557-561.

Johnes, G. (1997). 'Costs and industrial structure in contemporary British higher education.' Economic Journal 107: 727-737.

Johnes, G. (1998). 'The costs of multi-product organisations and the heuristic evaluation of industrial structure.' Socio-Economic Planning Sciences 32(3): 199-209.

Johnes, G., A. S. Camanho and M. C. A. S. Portela (2008a). 'Assessing Efficiency Of Portuguese Universities Through Parametric And Non-Parametric Methods.' Portuguese Journal of Management Studies 13(1): 39-66.

Johnes, G. and J. Johnes (2009). 'Higher education institutions’ costs and efficiency: taking the decomposition a further step.' Economics of Education Review 28(1): 107-113.

Johnes, G., J. Johnes and E. Thanassoulis (2008b). 'An analysis of costs in institutions of higher education in England.' Studies in Higher Education 33(5): 527-549.

Johnes, G., J. Johnes, E. Thanassoulis, P. Lenton and A. Emrouznejad (2005). An Exploratory Analysis of the Cost Structure of Higher Education in England. London, Department for Education and Skills.

Johnes, G. and M. Salas Velasco (2007). 'The determinants of costs and efficiencies where producers are heterogeneous: the case of Spanish universities.' Economics Bulletin 4(15): 1-9.

Johnes, G. and A. Schwarzenberger (2011). 'Differences in cost structure and the evaluation of efficiency: The case of German universities.' Education Economics 19(5): 487-499.

21

Johnes, J. (1990). 'Unit costs: some explanations of the differences between UK universities.' Applied Economics 22(7): 853-862.

Johnes, J. (1996). 'Performance assessment in higher education in Britain.' European Journal of Operational Research 89: 18-33.

Johnes, J. (2008). 'Efficiency and productivity change in the English higher education sector from 1996/97 to 2004/05.' The Manchester School 76(6): 653-674.

Johnes, J. and G. Johnes (2013). Effiicency in the Higher Education Sector: A Technical Exploration. London, Department for Business Innovation and Skills.

Johnes, J. and J. Taylor (1990). Performance Indicators in Higher Education. Buckingham, Society for Rresearch into Higher Education and Open University Press.

Jondrow, J., C. A. K. Lovell, I. S. Materov and P. Schmidt (1982). 'On the estimation of technical inefficiency in the stochastic frontier production function model.' Journal of Econometrics 19(2-3): 233-238.

Stevens, P. A. (2005). 'A stochastic frontier analysis of English and Welsh universities.' Education Economics 13(4): 355-374.

Thanassoulis, E., M. Kortelainen, G. Johnes and J. Johnes (2011). 'Costs and efficiency of higher education institutions in England: A DEA analysis.' Journal of the Operational Research Society 62(7): 1282-1297.

Tsionas, E. G. (2002). 'Stochastic frontier models with random coefficients.' Journal of Applied Econometrics 17: 127-147.

Verry, D. W. and B. Davies (1976). University Costs and Outputs. Amsterdam, Elsevier. Verry, D. W. and P. R. G. Layard (1975). 'Cost functions for university teaching and research.'

Economic Journal 85: 55-74.

Assalamualaikum Warahmatullahi Wabarakatuh dan Salam Sejahtera

YBhg. Datuk/ Datin / Prof/ Tuan/ Puan

DOES TRUST PLAY A ROLE WHEN IT COMES TO

DONATIONS? A COMPARISON OF ITALIAN AND

US HIGHER EDUCATION INSTITUTIONS

Bil 3/2021

Vol.:(0123456789)

Higher Educationhttps://doi.org/10.1007/s10734-020-00623-1

1 3

Does trust play a role when it comes to donations? A comparison of Italian and US higher education institutions

Barbara Francioni1  · Ilaria Curina1 · Charles Dennis2 · Savvas Papagiannidis3 · Eleftherios Alamanos3 · Michael Bourlakis4 · Sabrina M. Hegner5

Accepted: 9 September 2020 © Springer Nature B.V. 2020

AbstractHigher education institutions (HEIs) have experienced severe cutbacks in funding over the past few years, with universities examining options for alternative funding streams, such as alumni funding. Identifying the factors influencing their alumni’s intentions to invest in their alma mater can be of significant importance when establishing a sustainable revenue stream. Within this context, empirical research on the potential role of trust is scarce. This paper aims to deepen the analysis of the relationship between alumni trust and engagement as well as three outcomes, namely support, commitment, and attitude toward donation. A structural equation model was tested on two samples of US (n = 318) and Italian (n = 314) alumni. Although both countries are affluent and developed countries, the USA has an established tradition of alumni donations, which is not such a developed practice in Italy. For both countries, results confirm that engagement is an antecedent of trust, which in turn leads to the three investigated outcomes (support, commitment, and attitude toward donations). In contrast, the effect of commitment on attitude toward donations is significant only for the USA universities. The paper has interesting theoretical and managerial implications. From a theoretical point of view, the study aims to address a gap concerning the role of trust in the HE context. Managerially, the study has significant implications for universities that want to change alumni attitude toward donations.

Keywords Higher education institutions · Engagement · Trust · Commitment · Support · Attitude toward donations

Introduction

Over the past few years, higher education institutions (HEIs) have experienced severe cutbacks in funding from government sources (Stephenson and Yerger 2014). Such austerity measures, as part of public funding reviews coupled with increasing international competition and new entrants, as well as events exogenous to HE, such as the COVID-19 pandemic, have had an impact on traditional income sources. As a result, universities depend ever more on private

* Barbara Francioni barbara.francioni@uniurb.it

Extended author information available on the last page of the article

Higher Education

1 3

donors, who have become an important part of the financial mix and the well-being of HEIs (Tsao and Coll, 2005; Weerts and Ronca 2009; Iskhakova et al. 2020). Even though alumni contributions have always represented a significant source of university funding (Baruch and Sang 2012; Durango-Cohen and Balasubramanian 2015; Stephenson and Bell 2014), encouraging such a practice can make a positive impact on the stability and longevity of an HE institution.

Given the above, it has become crucial for HEIs to identify the factors influencing alumni’s intentions to invest in their alma mater, especially in terms of financial support (Baruch and Sang 2012). For this reason, HEIs are increasingly exploring relational practices (Hemsley-Brown and Goonawardana 2007; Pinar et al. 2011) that are able to strengthen the interactions with their students and, especially, with their alumni (Stephenson and Yerger 2014). From a theoretical perspective, among the different antecedents leading to more positive attitudes toward donations, the concept of trust is a fairly new construct (Yousaf et al. 2018). Ghosh et  al. (2001) were among the first to analyse trust in the higher education (HE) context, describing it as a significant long-term solution that universities should adopt in order to face the fierce competition in the sector. However, despite the relevance of the concept of trust in HE, only a handful of studies have considered it and, as such, further research in the HE context is required (e.g. as argued by Carvalho and Oliveira Mota 2010; Yousaf et al. 2018). More recent studies have conceptualised the role of trust by identifying some of its possible antecedents (Dennis et  al. 2016; Jillapalli and Jillapalli 2014) and outcomes (Carvalho and Oliveira Mota 2010; Dennis et al. 2016; Jillapalli and Jillapalli 2014).

The present work aims to enhance this stream of research by further investigating and testing trust’s antecedents and outcomes, especially when it comes to donations. More specifically, the research objectives of the study are twofold: firstly, to deepen the investigation of the alumni’s trust by identifying and testing a possible antecedent, namely engagement, and three outcomes, which are support, commitment, and attitude toward donations, and secondly, to analyse the possible relationships between commitment and support and attitude toward donations. By addressing these two objectives, this paper seeks to make important contributions to the existing HE literature with regard to the underlying factors that encourage donations, which in turn could have significant implications for practice. To this end, six hypotheses have been developed and tested through a structural equation model, with data coming from two countries. The USA has been selected because alumni donations are widespread within a more market-oriented HE system. Hence, it acts as a relative backdrop on which to establish a comparison. On the other hand, even though donations to universities exist in Italy too, such a practice is not widespread and HEIs are primarily funded by the state (Baruch and Sang 2012; Sung and Yang 2009).

The remainder of the article is structured as follows: the next section presents the hypothesis development by discussing the relevant literature. This is followed by the methodology and results. Finally, the last two sections discuss the findings, the theoretical and managerial implications, the limitations of the work, and the potential directions for future research.

Background and hypothesis development

Engagement and trust in the HE context

Engagement has been analysed within multiple disciplines, such as marketing, sociology, political science, and educational psychology (Brodie et al. 2011). By specifically focusing

Higher Education

1 3

on the HE context, different studies have investigated the influence of students’ engagement on their behaviour and attitude towards universities, both during their studies (Baruch and Sang 2012; Koranteng et al. 2019; Sung and Yang 2009; Weerts and Ronca 2008; Snijders et al. 2020) and after their graduation (Snijders et al. 2019). For instance, Weerts and Ronca (2008) analysed the impact of students’ engagement with a college by highlighting the importance of the experiences that they had during their university course. Conceptually, the main aspects characterising the engagement construct concern the level of students’ aca-demic and social involvement experienced with faculty and staff, the interactions with the alma mater, the exposure to diverse points of view, and the high quality of the academic pro-grams. In this respect, engagement can be portrayed by the interactions experienced by stu-dents during their degree course (Weerts and Ronca 2008). As such, the quality of students’ experiences during their course (e.g. academic programs, relationships with the academic staff, and extra-curricular activities) becomes a key factor characterising their engagement with the university (Baruch and Sang 2012), both in relation to face to face interactions and via remote channels such as social media, with various levels of perceived quality of interac-tion depending on the channel (Koranteng et al. 2019). Hence, interactions that facilitate a high quality of communication can enhance the engagement of students with the university where a student studies (Sung and Yang 2009). This engagement can, in turn, result in a long-term relationship and trust (Heffernan et al. 2018). Rojas-Méndez et al. (2009) concep-tualised the personal experiences built with the institution as a fundamental antecedent of students’ trust alongside other positive outcomes, such as loyalty (Snijders et al. 2020). This loyalty can lead to long-lasting relationships after the graduations of students, translated into monetary and non-monetary contributions to the university (Snijders et al. 2019).

The definition of trust has evolved over time and multiple conceptualisations have been proposed in the context of marketing (Wong and Ho 2011; Schlesinger et al. 2017). Some authors defined trust as a kind of customers’ belief or confidence (Moorman et al. 1992; Mor-gan and Hunt 1994) that any future interaction with organisations and firms will be identical and positive (Sultan and Wong 2012). Other researchers (Dowell et  al. 2015; Morgan and Hunt 1994) conceptualised trust as a consumer’s expectation that firms/organisations will not adopt opportunistic behaviours, and they will offer products and services at a level of quality expected by the consumer. Overall, trust is perceived as a complex construct characterised by two elements: a cognitive one, which is based on the consumer’s knowledge of the organi-sation and its capacities, and an affective one, concerning the emotional connection that an individual develops over time with the organisation (Heffernan et al. 2018).

In the context of HE, Ghosh and colleagues (2001, p. 325) were the first to examine the role of trust. For them, trust is “the degree to which a student is willing to rely on the insti-tute to take appropriate steps that benefit him and help him achieve his learning and career objectives”. Hence, the personal learning and student experience someone has while study-ing can lay the foundations for the future relationship with their university and positively influence their behaviour towards them (Hennig-Thurau et al. 2001). To this end, reliability and integrity are key dimensions of a trusted relationship as captured by Morgan and Hunt (1994, p. 23), who defined trust as “when one party has confidence in an exchange part-ner’s reliability and integrity”. This definition has been adopted by a number of studies in the educational context which examined the role of trust in relationship commitment. Specifically, students’ trust in the university is influenced by the following: the common and shared values between them and the organisation (Wong and Ho 2011), their emotional connection with the organisation (Komljenovic 2019), the perceived quality of teaching and other services that students receive (Hennig-Thurau et al. 2001), as well as their over-all satisfaction. Trust, in turn, can have a significant effect on the loyalty that students feel

Higher Education

1 3

towards the institution (Rojas-Méndez et al., 2009). Trust can also be built after graduation by encouraging interaction between the graduates and the university by promoting respect, responsibility, and reciprocity (Harrison 2018).

Given the above, it is expected that the higher the level of students’ engagement with the university, the higher their trust toward it is. We thus formulate the first hypothesis as follows:

H1. Prior engagement with the university positively influences trust towards it.

Attitude toward donations

Faced with the extremely competitive marketplace characterising the HE context, institutions and universities are increasingly searching for new strategies to connect with their graduates and seeking ways in which they can contribute to the institution (Johnson et al. 2010; Weerts and Ronca, 2008). Such methods may involve tools that are typically used in the commercial sphere, such as crowdfunding, and go beyond the typical scope of donations (Cho et al. 2019). Irrespective of the mechanism or channel by which money is raised, it also worth noting that the widespread credit crunch of recent years has led to a reduction in the level of donations for all universities (Gallo, 2018; Baruch and Sang 2012). In this respect, additional income sources other than tuition fees are of relatively higher importance for the HEIs as they help diversify income streams and avoid over-reliance on tuition fees. In particular, alumni donations have become a fundamental part of these sources (Baruch and Sang, 2012; Bastedo et al. 2014; Dennis et al. 2016). For this reason, the interest of scholars in alumni attitude toward donations has been growing increasingly, especially in recent years (Stephenson and Yerger 2014). Indeed, over the past two decades, several authors have attempted to identify the main factors leading alumni to donate to their alma mater (Weerts and Ronca 2008). Other studies investigated socio-economic variables, such as income and education, past giving, sector of employment, type of financial aid received, as well as demographic indicators (Gunsalus 2005; Newman and Petrosko 2011; Wunnava and Lauze 2001), such as age, ethnicity, income, gender, residence, and marital status (Clotfelter 2003; Drezner 2018; Holmes 2009; Hueston 1992; Okunade and Berl 1997; Weerts and Ronca, 2008). Other researchers focused their attention on behavioural factors, such as volunteering for the college, membership in alumni chapters, and reunion attendance (Durango-Cohen and Balasubramanian 2015). Furthermore, characteristics related to institutions and universities have been investigated, such as size, type of institution, or endowment value. For instance, McAlexander and Koenig (2010) found that alumni who belong to smaller institutions feel more integrated and inclined to support their university community than do graduates of larger institutions. Similarly, past work has examined how institutional reputation can potentially impact attitude and, in turn, support and donations (Shaari et  al. 2019). Finally, work has specifically investigated the social exchange factors, such as the experiences undergone by alumni during their university course; the quality of education, career gains, and social connections; satisfaction with student affairs; and campus resources (Leslie and Ramey 1988; Stephenson and Yerger 2015; Taylor and Martin 1995). Notably, social exchange theory proposes that alumni attitude toward donations is in part influenced by their perceptions of the quality of their past experiences with the university (Clotfelter 2003; Weerts and Ronca 2008).

Starting from these assumptions and from the need to identify additional factors that may explain the alumni intention of donating to their alma mater (Baruch and Sang 2012),

Higher Education

1 3

in the present study, the trust construct has been investigated as a possible predictor of alumni attitude toward donations. In this way, the work aims to enrich the literature focused on the identification of additional social exchange factors leading alumni to donate to their university. Sargeant and Lee (2004) found how trust plays a key role in facilitating the stimulation of monetary donations, while Melendez (2001, p. 121) suggested that “donors do not contribute to organisations they do not trust and about which they do not feel confi-dent”. Based on these findings, we hypothesise that trust can lead to more positive attitudes toward donations, also in the HE context. In particular, as stated above, trust within this sector can be perceived as a feeling that grows as students undergo positive experiences with their university (Heffernan et al. 2018). By also considering that the alumni attitude toward donations is influenced by the quality of past experiences they had during their uni-versity course (Weerts and Ronca 2008), it could be assumed that the better the experi-ences undergone by alumni during their university course, the higher the level of alumni’s trust toward their university, the more positive their attitude toward donation is. This argu-ment leads to the second hypothesis:

H2. Trust towards the university positively influences attitude toward donations.

Support

In the HE context, the concept of support refers to the active participation of alumni who decide to support their university beyond graduation (Weerts and Ronca 2008). In addition to welcoming philanthropic donations, HE institutions often consider alumni as a key enabler to contribute to the mission and vision of the university via various non-monetary forms of support, e.g. their time (Gallo 2018), which can be invested in a number of different activities or by promoting the university to potential students (Sung and Yang 2009). Alumni can support their alma mater in multiple ways, e.g. by attending alumni events, assuming the role of mentor to students, collaborating with the university’s academic staff, participating in research projects, or becoming a volunteer for the university (Mael and Ashforth 1992; McDearmon 2013; Sung and Yang 2009). As stated by Weerts and Ronca (2008), alumni volunteers represent valuable assets to their institutions. Through their professional and social networks, they can lend their experience, help formulate strategic directions for the institution, and act as mentors, recruiters, and club leaders.

Based on the relevance of alumni as a critical source of support for HEIs (Mael and Ashforth 1992), different studies have attempted to identify the main factors leading to the creation of supportive alumni–university long-term relationships (Weerts and Ronca 2008). In particular, emotional attachment, the high quality of educational experiences (Weerts and Ronca 2008), and loyalty (Hennig-Thurau et al. 2001) are some of the main antecedents of alumni support identified in the literature. Such student experience can be of a very personal nature and even be the result of unconscious motives (Drezner and Garvey 2016).

By underlining the key role of trust in building and maintaining long-term relations between alumni and their alma mater, Ghosh and colleagues (2001) confirmed the strategic implication, for universities, of having alumni who trust and consequently support them. Therefore, it could be hypothesised that the more the alumni are confident about their alma mater, the more they will support it. Hence:

H3. Trust towards the university positively influences support.

Higher Education

1 3

In addition, graduates of a university can make monetary contributions to their univer-sity. Previous literature suggests that undergraduates who have a close relationship with their university while studying are more likely to become donors after graduation (Sung and Yang 2009). However, strong relationships between students and their university can continue after graduation by enhancing reciprocity and trust (Harrison 2018). Hence, we hypothesise that the stronger the supportive behaviour in the form of non-monetary con-tributions, the more likely are positive attitudes towards donating to the university. Alumni who actively support the university’s activities are expected to be more philanthropic and consequently more inclined to sustain their alma mater from an economic perspective (Newman and Petrosko 2011). Similarly, Taylor and Martin (1995) have identified how alumni are generally more involved in being active donors when they are active in organis-ing the university’s activities, like for instance sporting events (Diaz Vidal and Pittz 2019).

Based on this, the fourth hypothesis is formulated:H4. Support for the university positively influences attitude toward donations.

Commitment

Previous literature (Johnson et al. 2010; Poole 2017) has identified three different forms of commitment that a person can experience with an organisation: (1) affective commitment, related to the emotive connection towards the organisation; (2) continuance commitment, namely a connection based on the fact that the costs of maintaining the relationship are less than the costs of ending it; and (3) normative commitment, which represents a sense of obligation to maintain the relationship. However, most studies consider affective com-mitment as the form that is most likely to occur in consumer-organisation relationships (Sung and Yang 2009). Affective commitment represents a connection that makes it possi-ble to increase loyalty, positive word-of-mouth, consumers’ participation, and volunteering (Poole 2017).

In the HE context, affective commitment represents a key factor (Schlesinger et  al. 2017; Weerts and Ronca 2008), since it is viewed as a student’s belief that a continuous relationship with their alma mater is so significant that it justifies maximum effort in order to maintain it (Dennis et al. 2016; Johnson et al. 2010; Jillapalli and Jillapalli 2014; Pedro et al. 2020).

By focusing on commitment’s possible antecedents, some authors (Dennis et al. 2016; Frasquet et al. 2012; Hennig-Thurau et al. 2001; Jillapalli and Jillapalli 2014; Rojas-Mén-dez et al., 2009; Schlesinger et al. 2017) have analysed the role of trust in reducing anxie-ties and dissonances in the HE relationships. When students develop trust in their institu-tion during their degree course, it will be easier for them to build committed relationships with it (Dass et al. 2020; Jillapalli and Jillapalli 2014; Pinar et al. 2020; Yousaf et al. 2020), which may extend beyond their graduation. Moreover, Dennis et al. (2016) found that trust can enhance the efficiency of a relationship with a consequent positive effect on satisfac-tion and commitment.

Based on these findings, it could be hypothesised that the more a university is trusted, the higher is the level of alumni commitment towards it.

This leads to the following hypothesis:H5. Trust towards the university positively influences commitment to the university.Previous literature (Balaji et al. 2016; Jillapalli and Jillapalli 2014) attempted to iden-

tify the main outcomes resulting from the development of committed relationships between

Higher Education

1 3

alumni and their university. Notably, past studies provided evidence as to how commitment can create a strong sense of alumni identification with their university, thus motivating the alumni to maintain the relationship with their alma mater and to continue their postgradu-ate studies at the same institution (Perin et  al. 2012). In addition, commitment has been found to increase loyalty, relationships with the institution, participation, positive word-of-mouth, and volunteering (Jillapalli and Jillapalli 2014). Moreover, Poole (2017) detected how passionate and very committed graduates are generally more active supporters in dif-ferent ways. In particular, by specifically focusing on the alumni intention to make dona-tions, Baruch and Sang (2012) suggested that alumni commitment can contribute to this specific attitude. Starting from these findings, we expect that a higher level of commitment leads to a more positive attitude toward donations. Therefore:

H6. Commitment to the university positively influences attitude toward donations.Figure 1 presents the conceptual model and the related hypotheses.

Research method

Data collection

The study was carried out in two countries, namely the USA and Italy. Collecting data from more than one country to test a model is not necessary, but, nevertheless, it helps to confirm the stability of the model (Cadogan 2010). The USA and Italy were chosen in order to take advantage of the contrast in the HE contexts between the two countries (Clark and Cullen 2016). The USA was selected as HEIs, in this country, have strong global brands, high levels of engagement, and very positive attitudes towards donation (McDon-ald 2014; Drezner 2019). In particular, as regards the last aspect, the donation functions in the USA are among the leading ones in the world and universities usually pursue activities aimed at encouraging donations, like for instance the organisation of fundraising events, commonly named “days of giving”.1 As such, the USA offers a benchmark on which to

Fig. 1 Proposed conceptual framework

1 See for instance: https ://www.umass .edu/givin g/; https ://dayof givin g.uni.edu/; https ://louis ianat echgi vingd ay.org/givin g-day/12228 ; https ://givin gday.uconn .edu.

Higher Education

1 3

compare and contrast findings from other countries, such as Italy (Baruch and Sang 2012; Sung and Yang 2009), in which alumni donations are not such a popular practice as in the USA. Although we do not express our hypotheses in terms of cultural dimensions, we also note that the two countries in our sample vary substantially on Hofstede’s (2003) cultural constructs, which might be relevant to our research topic. For example, Italy is higher on uncertainty avoidance and long-term orientation but lower on indulgence, so Italians may consider higher education more as a long-term investment compared with US counterparts. In addition, the Italian HE is more centrally controlled compared with the USA and other G7 countries, such as the UK, that could have been chosen (De Feo and Pitzalis 2017). G7 countries, i.e. the seven most advanced economies in the world (Canada, France, Germany, Italy, Japan, the UK, and the USA), offer a sufficiently homogenous group with regard to the size of their economies and the development of their HE systems.

The study employed an online survey. To prevent potential biases, different recommendations of MacKenzie and Podsakoff (2012) have been adopted. First, in order to reduce social desirability bias, the complete anonymity and confidentiality of responses have been assured through a statement included in the introductory part of the questionnaire. Second, we assured respondents that there are no right or wrong answers and that they could have different opinions about the issues examined. Finally, in the “Introduction” section, we anticipated that some of the questions were personal, and, in order to be completed, they consequently required the absence of possible factors of disturbance.

In the USA, a market research company recruited participants in order to control quotas of gender, age, and area of residence. After sending the survey link, the market research company obtained 318 valid responses. In Italy, a pilot test was conducted and a preliminary version of the translated questionnaire into Italian was administered to five volunteer participants. The questionnaire was adapted and adjusted based on the comments of the respondents related to language, the order of the questions, and the understanding of the concepts. In order to ensure the consistency of the Italian and English versions of the questionnaire, translation and back-translation procedures (Behling and Law 2000; Brislin 1970) were adopted. When it came to recruiting participants, two of the authors and two external collaborators handled this phase. The data was collected via an online survey, obtaining 314 valid responses. Table 1 presents the profile of the two samples.

Construct measurement

For all constructs, we used the same question (“Please select the option that applies to each of the statements below”) and a seven-point scale has been adopted (ranging from 1 = “Not likely at all” to 7 = “Extremely likely” for support, from 1 = “Strongly disagree” to 7  =  “Strongly agree” for engagement, trust, and commitment and from 1  =  “Do not agree” to 7  =  “Completely agree” for attitude/intention toward donations). More specifically, the statements concerning engagement were formulated through a revision of the scale proposed by Banahene (2017), while those related to trust were developed and validated by Jillapalli and Jillapalli (2014). The alumni support items were extracted and adapted starting from the study of Bellezza and Keinan (2014), while the scale proposed by Jillapalli and Jillapalli (2014) has been adapted to measure the concept of commitment. Finally, attitude/inclination toward donations was measured by adapting the items extracted from the study of Johnson et al. (2010).

Higher Education

1 3

Common method bias was assessed by employing Harman’s single-factor test, which suggests that common method bias has affected the results of a study when a single factor containing all items included in a questionnaire explains over 50% of the variance (Podsa-koff et al. 2003). In this study, the single factor explains 42.2% of the variance; therefore, common method bias is not an issue.

The constructs, in this study, are reflective measures as specified by Jarvis et al. (2003). The authors develop a set of criteria for determining reflective models, which are fulfilled by the construct measurement in this study. The direction of causality is from construct to items and indicators are manifestations of the construct. Further, indicators share a com-mon theme and covary with each other. Additionally, the nomological net for the indicators does not differ.

Data analysis

As constructs are measured reflectively with several items and the sample size is large enough, the authors apply covariance-based structural equation techniques (Hair et  al. 2018). Testing the proposed model requires an operationalisation of the hypothesised latent

Table 1 Respondents’ profile

Characteristic FrequencyUSA

% USA FrequencyItaly

% Italy

Gender  Male 141 44.3 103 32.8  Female 177 55.7 211 67.2

Age (years)  20–29 126 39.6 250 79.6  30–39 185 58.2 39 12.4  40 or over 7 2.2 25 8.0

Area of residence  Urbanised area 159 50 86 27.4  Urban cluster 120 37.7 184 58.6  Rural 39 12.3 44 14.0

Income  $0–$24,999 29 9.1 114 36.3  $25,000–$49,999 82 25.9 117 37.3  $50,000–$74,999 73 23.0 45 14.3  $75,000–$99,999 70 22.1 31 9.9  More than $100,000 63 19.9 7 2.2

Education attainment  Current university student 8 2.5 3 1.0  University graduate (e.g. bachelor) 225 70.8 203 64.6  Graduate degree (e.g. master) 74 23.3 100 31.8  Doctorate 11 3.5 8 2.5

Donating behaviour  Donation since graduation 74 23.3 12 3.8  No donation since graduation 244 76.7 302 96.2

Higher Education

1 3

constructs and associated indicators, which is only possible with SEM (Bagozzi and Yi 2012). When covariance-based structural equation modelling is performed, the error terms are modelled for each indicator and loadings of the specific indicator are obtained; thus, the quality of the latent constructs modelled can be adapted and improved. Confirmatory factor analysis in SEM allows all latent constructs to covary and thereby permits the evaluation of both convergent and discriminant validity for each construct (Bagozzi and Yi 2012; Hair et al. 2018). The fit between the observed and estimated models can be obtained and evaluated. Further, relationships including mediating variables can be measured in one model, which is a major improvement over multiple regression (Bagozzi and Yi 2012; Fabrigar et al.f 2010; Schreiber 2008). AMOS (version 22.0) was used as it represents a user-friendly software package to model covariance-based SEMs (Byrne 2016).

The following steps for analysing the data in order to assess the relationships among the underlying constructs were taken: an exploratory factor analysis (EFA) was conducted to identify the underlying relationships between the measured variables. This was followed by confirmatory factor analysis (CFA), which was used to assess the validity of the constructs. In turn, a multi-group structural equation model (SEM) was performed to test the hypothesised relationships between the constructs and to assess differences between the countries. Finally, an additional model was tested, including possible covarying variables.

Results

Measurement reliability and validity

With 318 respondents in the USA and 314 in Italy, we were above the rule of 200 (Kline 2011) and the sample-to-item ratio was about 12.5, which is higher than the acceptable ratio of 5:1 (Gorsuch 1983), and this leads us to conclude that we have an adequate sample size. We calculated the Kaiser–Meyer–Olkin (KMO) as well as Bartlett’s Test of Sphericity to measure sampling adequacy (Hutcheson and Sofroniou 1999). The KMO is 0.957 (>  0.5), and Bartlett’s Test of Sphericity is significant (p  <  0.001); therefore, the data are suitable for factor analysis. We used principle component analysis with varimax rotation. As hypothesised, all five constructs had eigenvalues >  1, explaining 79.3 per cent cumulative variance. Of the initial 25 items, no item had significant cross-loadings (> 0.50) and all loaded on the original constructs.

Prior to testing the structural model with data for both countries, the requirements of instrument validity and reliability must be met. Confirmatory factor analysis (CFA) using AMOS 22.0 was performed to determine the discriminant and the convergent validity of the scales. Hair et al. (2018) recommend a factor loading (FL) value higher than 0.50 for an item to be significant. Table 2 presents the factor loading values for the individual items. Additionally, at construct level, Hair et al. (2018) proposed the calculation of composite reliability (CR) and average variance extracted (AVE) instead of Cronbach’s alpha when using structural equation modelling (SEM).

Convergent validity was examined by calculating the average variance extracted (AVE) and the construct reliability (CR). AVE needs to be >  0.50 (Fornell and Larcker 1981), while the CR should be > 0.60 (Bagozzi and Yi 1988). All our AVE and CR values are above the recommended thresholds. To test for discriminant validity, all AVE values need to be higher than the squared inter-construct correlation estimates (SIC). Details for AVE, CR, and SIC values are provided in Table 3.

Higher Education

1 3

Tabl

e 2

Fac

tor l

oadi

ng v

alue

s fro

m th

e C

FA fo

r the

diff

eren

t ite

ms m

easu

ring

the

rese

arch

con

struc

ts

Con

struc

tsIte

ms

Fact

or lo

adin

gs

Prio

r eng

agem

ent w

ith th

e un

iver

sity

I rea

lly li

ke to

talk

abo

ut th

is u

nive

rsity

with

oth

ers

0.86

2I a

m a

lway

s int

eres

ted

in le

arni

ng m

ore

abou

t thi

s uni

vers

ity0.

869

I wou

ld b

e in

tere

sted

in m

erch

andi

se w

ith th

is u

nive

rsity

’s b

rand

nam

e on

it0.

787

I am

pro

ud to

hav

e ot

hers

kno

w I

study

in th

is u

nive

rsity

0.83

5C

omm

itmen

tI a

m v

ery

com

mitt

ed to

this

uni

vers

ity0.

893

This

uni

vers

ity is

ver

y im

porta

nt to

me

0.92

0I r

eally

car

e ab

out t

his u

nive

rsity

0.87

3I b

elie

ve th

at th

is u

nive

rsity

des

erve

s my

effor

t to

mai

ntai

n a

rela

tions

hip

0.84

1I f

eel a

nee

d to

ens

ure

the

succ

ess o

f thi

s uni

vers

ity0.

768

Supp

ort

How

like

ly a

re y

ou to

atte

nd a

lum

ni e

vent

s?0.

812

How

like

ly a

re y

ou to

bec

ome

a m

ento

r for

stud

ents

in th

is u

nive

rsity

?0.

871

How

like

ly a

re y

ou to

giv

e a

gues

t lec

ture

r in

this

uni

vers

ity?

0.84

7H

ow li

kely

are

you

to v

olun

teer

for t

his u

nive

rsity

?0.

867

How

like

ly a

re y

ou to

hos

t a st

uden

t pro

ject

for t

his u

nive

rsity

?0.

846

How

like

ly a

re y

ou to

eng

age

with

aca

dem

ic st

aff in

this

uni

vers

ity?

0.84

0H

ow li

kely

are

you

to ta

ke p

art i

n a

rese

arch

pro

ject

in th

is u

nive

rsity

0.79

8H

ow li

kely

are

you

to a

ttend

a p

ublic

lect

ure

in th

is u

nive

rsity

0.82

4Tr

ust

This

uni

vers

ity c

an b

e tru

sted

0.86

9Th

is u

nive

rsity

can

be

coun

ted

on to

do

wha

t is r

ight

0.85

8Th

is u

nive

rsity

has

hig

h in

tegr

ity0.

899

This

uni

vers

ity k

eeps

its p

rom

ises

0.87

2A

ttitu

de to

war

ds d

onat

ion

Don

atin

g m

oney

to th

is u

nive

rsity

is a

goo

d id

ea0.

917

I am

favo

urab

le to

war

ds d

onat

ing

mon

ey to

this

uni

vers

ity0.

931

Don

atin

g m

oney

to th

is u

nive

rsity

is a

wis

e id

ea0.

959

I am

pos

itive

abo

ut d

onat

ing

mon

ey to

this

uni

vers

ity0.

951

Higher Education

1 3

Fit of the measurement model

A multi-group structural equation model was conducted to assess the relationships among the underlying constructs. The measurement model was tested to determine its fit to the research data. An acceptable model fit was achieved with χ2 = 777.15; df = 249; p  <  0.01; χ2/df  =  3.12; IFI  =  0.97, TLI  =  0.96, CFI  =  0.97; and RMSEA  =  0.06. For a meaningful comparison of the model for both countries, the instrument measuring the various constructs has to possess cross-country equivalence. To meet the requirement of equivalence, configural and, at least partial, metric or scalar invariance has to be confirmed to compare the findings for the two groups of consumers (Hair et al. 2018; Steenkamp and Baumgartner 1998; Vandenberg and Lance 2000).

Metric invariance was tested by means of nested multiple-group CFA. We found a significant difference between the free and the restricted model (i.e. factor loadings restricted to being equal across countries) (Δχ2  =  59.41, df  =  20, p  <  0.01). However, a partial metric invariance model, in which two factor loadings of the construct support were constrained to be equal, leads to a non-significant difference between the constrained and the unconstrained models (Δχ2  =  20.59, df  =  18, p  =  0.08) compared with the unconstrained model. Hence, the assumption of partial metric invariance has been met (Thøgersen et al. 2015).

Test of the structural model

We conducted a structural equation model by using SPSS AMOS to assess the relationships among the underlying constructs. The results suggest an acceptable model fit (χ2 = 1191.02; df = 253; χ2/df = 4.70; IFI = 0.94, TLI = 0.93, CFI = 0.94; RMSEA = 0.08). By examining the equality of structural weights, the significance of the overall difference in the factors influencing attitude towards donations of both Italian and US alumni was determined. The path coefficients, as well as the critical ratios for significant differences on the individual paths between the samples, are reported in Table 4.

In the overall model, all hypothesised relationships are confirmed. The multi-group analysis reveals that for both samples, the influence of engagement on trust is statistically significant. The same holds true for the influence of trust on commitment. Trust signifi-cantly influences attitude towards donations both directly and indirectly in both coun-tries. One significant difference between the Italian sample and the US sample is found in

Table 3 Validity results for the model

All correlations are significant, p  < 0.01

SIC

AVE CR Engagement Commitment Support Trust Attitude

Engagement 0.70 0.91 1.00Commitment 0.74 0.93 0.60 1.00Support 0.70 0.95 0.40 0.39 1.00Trust 0.77 0.93 0.39 0.49 0.18 1.00Attitude 0.88 0.97 0.41 0.32 0.30 0.28 1.00

Higher Education

1 3

relation to the influence of trust on support. This influence is stronger for US respondents (β = 0.54, p < 0.001) compared with Italian respondents (β = 0.33, p < 0.001). A second significant difference between the two countries refers to the path between commitment and attitude towards donation. Again, the influence of commitment on attitude towards donations is stronger for US respondents (β = 0.26, p < 0.001) than for Italian respondents (β = 0.13, p = 0.12). No significant differences between Italian and US respondents were detected for the relationship between trust and attitude towards donations or for support and attitude towards donations. An additional model was tested including gender, age, and years since the graduation as covariates. In the Italian sample none of these variables is significant. In the US sample, only years since graduation (β = − 0.15, p = 0.003) are sig-nificant. Figure 2 and 3 present the findings from the structural model.

Discussion and conclusions

Discussion of the results

The empirical results support H1 since, in both countries, a positive relationship was confirmed between engagement and trust. This finding confirms how, in both HE contexts, prior engagement with the university represents a significant antecedent of trust toward it. By confirming this relationship, the results corroborate the relevance

Table 4 Standardised parameter estimates of the structural model

ns not significant*p < 0.05; **p < 0.01; ***p < 0.001

Parameters Overall samplen = 632

Italian alumnin = 314

US alumnin = 318

Critical ratio

Engagement trust 0.68*** 0.62*** 0.74*** 0.15 nsTrust support 0.48*** 0.33*** 0.54*** 3.66**Trust commitment 0.74*** 0.75*** 0.75*** 0.61 nsTrust attitude 0.30*** 0.26** 0.30*** 0.12 nsCommitment attitude 0.16** 0.13 ns 0.26*** 2.37*Support attitude 0.31*** 0.16** 0.35*** 1.48 ns

Fig. 2 Structural model (ITA)

Higher Education

1 3

of the engagement construct as a key factor in the building of students’ trust toward the alma mater. Overall, this outcome strengths prior studies (Hennig-Thurau et  al. 2001; Komljenovic 2019; Weerts and Ronca 2008; Wong and Ho 2011) that claim that students’ trust in the university is also influenced by their emotional connection with it.

With respect to H2, data confirmed a positive relationship between trust and attitude toward donations, for both countries. Such a finding is in line with recent findings suggesting that alumni trust is a predictor of self-reported giving and attitudes, even when controlling for socio-demographic characteristics (Drezner et  al. 2020). In this way, it has been possible to corroborate how significant this tie is, not only in the charity and non-profit sectors (Melendez 2001; Sargeant and Lee 2004), but also in the HE context. Notably, the confirmation of this positive relationship has made it possible to detect the role of trust as a key factor leading alumni to donate to their alma mater (Baruch and Sang 2012), thus enriching the literature focused on the identification of social exchange factors able to stimulate alumni attitude toward donation.

The positive relationship between trust and commitment (H5) confirms, for both countries, how commitment represents an outcome of trust, thus corroborating previous studies (Dass et al. 2020; Dennis et al. 2016; Helen and Ho 2011; Hennig-Thurau et al. 2001; Jillapalli and Jillapalli 2014; Pedro et al. 2020; Pinar et al. 2020; Rojas-Mendez et al. 2009; Schlesinger et al. 2017; Yousaf et al. 2020). In detail, the development of students’ trust toward their institution can lead to the building of committed student-university relationships, which may continue beyond graduation.

Conversely, a significant difference between the two samples emerges concerning the influence of trust on support (H3) and commitment on attitude toward donations (H6), which are stronger for US respondents. In other terms, in the American context, not only does the trust construct represent a more significant factor leading to the establishment of supportive alumni–university long-term relations, but commitment also plays a more active role in the formation of alumni attitude toward donations. Therefore, these results allow us to corroborate and enrich the extant studies analysing the role of trust as an antecedent of student support (Ghosh et al. 2001) as well as those examining the role of commitment as a factor leading to the formation of alumni attitude toward donations (Baruch and Sang, 2012; Jillapalli and Jillapalli 2014; Poole 2017).

Although much less significant, a difference in favour of the American sample also emerges for the relationship between support and attitude toward donation (H4).

Fig. 3 Structural model (USA)

Higher Education

1 3

More specifically, this result underlines how supportive alumni–university relationships represent, for the American HEIs, a more significant predictor in the development of alumni inclination to donate to their alma mater, thus confirming previous studies analysing this connection (Diaz Vidal and Pittz 2019; Newman and Petrosko 2011).

Overall, these results underline how the American universities are more able to trigger a virtuous cycle in the relations with their students, since when students enter the American HEIs, they become involved in the university’s life to the point of transforming this engagement status into trust and then into support and affective commitment, which in turn leads to positively influencing attitudes toward donation, once studies are over. In contrast, Italian universities appear to be less able to activate this process, failing to transform their alumni commitment into a more marked inclination to support them with donations. In regard to the insignificance of the relationship between commitment and attitude toward donations in the Italian context, this result may appear at first sight to be in contrast to what might be expected in light of Italy’s higher scores for uncertainty avoidance and long-term orientation (Hofstede 2003), which might have been expected to lead to Italians investing in higher education for the future. On the other hand, we speculate that perhaps alumni spending on donations might be considered as indulgent and thus in line with the USA’s higher scores on indulgence (Hofstede 2003). As noted earlier, collecting data from more than one country was intended primarily to help confirm the stability of the model (Cadogan 2010). In this respect, it appears that Italy and the USA have similar models of engagement through trust to commitment and also in the direct and indirect influence of trust on attitude towards donations. Nonetheless, there are differences in the “support” arm of the model, particularly in turning commitment into donations, which may be down to cultural differences and possibly due to the fact that Italians may consider higher education as a long-term investment, which is not the case for the Americans. The influence of differences in sampling cannot be eliminated but gender and age are non-significant as covariates.

Theoretical and managerial implications

Theoretically, the study aimed to address a gap concerning the role of trust in the HE context (Yousaf et  al. 2018; Carvalho and De Oliveira Mota 2010). In particular, the relationship between engagement and trust has been analysed, thus enriching the literature focused on the analysis of the antecedents of trust (Dennis et al. 2016; Jillapalli and Jillapalli 2014). The study also investigates the connections between trust and (i) support, (ii) commitment, and (iii) alumni’s attitude to donating. Concerning the first two relationships, the study corroborates the findings of previous studies (Dennis et al. 2016; Ghosh et  al. 2001; Jillapalli and Jillapalli 2014), by identifying the influence of trust on both constructs. As regards the relationship between trust and attitude to donating, the present research underlines the significance of this tie in the HE context too, thus enriching previous findings which have analysed this relationship in different sectors, such as the charity and non-profit ones (Melendez 2001; Sargeant and Lee 2004). Moreover, starting from the key role currently assumed by alumni contributions in the HE context (Durango-Cohen and Balasubramanian 2015), the paper identifies three outcomes leading alumni to become more inclined to donate to their alma mater, namely trust, support, and commitment, although with different results for the two investigated samples. Overall, the study provides relevant contributions to the extant HE literature by analysing the influence of trust on the formation of alumni’s attitudes and behaviours and by identifying a possible

Higher Education

1 3

process leading alumni to change their attitude towards donating to their universities. Testing in more than one country helped to confirm the stability of the model. The main difference between the countries is in the USA being more successful at turning commitment into donations. In this respect, the results highlight how the American universities are more able to manage this virtuous cycle since students who are engaged with them will probably become more confident and committed alumni, supporters, and donors. Such findings may need to be contextualised more, though, as both the student and the HE education contexts in which they operate can affect alumni loyalty and, in turn, support (Iskhakova et al. 2020).

Managerially, the study provides significant implications for universities that want to change alumni attitude toward donations. In particular, both Italian and American HEIs should encourage students to become more engaged in their university, by instilling a sense of affinity between students and their alma mater. Notably, Dennis et  al. (2016) recommended that universities build marketing and student recruitment by creating and nurturing relationships in novel ways with students, for example using networking events, social media (Dyson et al. 2015), campaigns, customised clothing, and regalia. Moreover, all these activities could make the university experiences, as lived by students, ever more positive, thus transforming them into more confident individuals, beyond their graduation too. For this reason, alumni’s trust should be nourished over time by universities since it is becoming fundamental in order to build committed and supportive relationships with their alumni. Therefore, by continuing beyond graduation, this sense of belonging could stimulate alumni’s attitudes toward volunteering and donations. Overall, the results confirm the key role of trust in shaping alumni attitudes toward donations. It is thus crucial for American and Italian HEIs to make every effort, for example, to organise extra-curricular activities, which can help to enhance the quality of their educational experiences. On the other hand, affective commitment represents a significant predictor of attitude toward donation only for American universities. This means that Italian universities should do more to stimulate the alumni level of commitment with their alma mater from an emotional point of view. In particular, universities could adopt different tools (e.g. university’s merchandising, social media communities, official websites) in order to stimulate the level of alumni’s commitment, thus making it possible to develop and maintain continuous and direct contacts with them.

Finally, the identification of a virtuous cycle able to increase the inclinations of alumni to donate to their alma mater represents a further implication provided by the study. The more positive the experiences undergone by students during their degree course are, the more they will become confident alumni in the future. This trust, built and nourished over time, could be translated into emotional attachment (commitment) and supportive behaviours (support), which in turn can also lead to a more marked alumni inclination to donate. Faced with this possible process, universities are recommended to organise specific activities for each phase in order to (i) allow students to have positive experiences during their degree course (not only experiences related to their academic career, but also linked to moments of leisure and interaction between each other); (ii) maintain and constantly nurture the trust of their alumni, by adopting activities aimed at reminding them that the university has not forgotten its students after their graduation (e.g. organisation of events dedicated to ex-students; submission of newsletters or online surveys with the final aim of learning more about their work path after graduation, or in order to inform them about new educational opportunities such as Masters, which could allow them to build upon their previous studies); (iii) adopt multiple tools able to strengthen the alumni’s emotional attachment (e.g. related to the university’s merchandising); and (iv) stimulate the alumni’s

Higher Education

1 3

supportive behaviours, thus transforming them into active stakeholders (e.g. management of the social media communities, forums, blogs, creation of online groups specifically targeting ex-students with the aim of encouraging them to participate in university life even after their graduation).

Limitations and future research

This paper is subject to some limitations. First, we focused our attention only on the engagement, trust, support, and commitment constructs as antecedents of attitude toward donations. In the future, it could be interesting to analyse additional constructs such as attachment strength, reputation, satisfaction, and perceived quality (Dennis et  al. 2016; Jillapalli and Jillapalli 2014). Similarly, future studies could also focus attention on additional outcome variables, such as the actual behaviours of alumni giving instead of their attitude toward donations. Second, we employed a convenience sample, which compromises the ability to generalise to the population. Future studies could use the insights emerging from this work as a basis for developing studies on the same or extended target populations by drawing on probability samples. Third, only two countries (USA and Italy) have been selected and analysed. As there are differences between these two countries, additional countries should be investigated in a cross-national study. In particular, cultural differences should be investigated, not least to explore the possible influence of indulgence. In addition, future studies could focus their attention on more countries, showing differences in terms of marketisation of HE, fees, or systems (for instance, public versus private). Samples could not be representative of each country, but rather aimed to provide a reasonable distribution among demographic characteristics. The differences that we report between the samples might be cultural differences surrounding the idea of donating to higher education, but might also arise from the uncontrolled nature of the schools represented, the degrees awarded, the ages of the participants, or something else not measured. Cross-cultural research with more controlled samples is recommended to resolve these issues. Finally, since only the point of view of alumni has been investigated in this work, future research could examine the specific actions adopted by the management of American and Italian universities, in order to identify how the former manage to create the virtuous cycle, and why the latter are not able to create it.

References

Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of the Academy of Marketing Science, 16(1), 74–94.

Bagozzi, R. P., & Yi, Y. (2012). Specification, evaluation, and interpretation of structural equation models. Journal of the Academy of Marketing Science, 40(1), 8–34.

Balaji, M. S., Roy, S. K., & Sadeque, S. (2016). Antecedents and consequences of university brand identifi-cation. Journal of Business Research, 69(8), 3023–3032.

Banahene, S. (2017). The impact of brand personality and students’ self-concept on brand engagement. International Journal of Business and Social Research, 7(8), 12–25.

Baruch, Y., & Sang, K. J. (2012). Predicting MBA graduates’ donation behaviour to their alma mater. Jour-nal of Management Development, 31(8), 808–825.

Bastedo, M. N., Samuels, E., & Kleinman, M. (2014). Do charismatic presidents influence college applica-tions and alumni donations? Organizational identity and performance in US higher education. Higher Education, 68(3), 397–415.

Higher Education

1 3

Behling, O., & Law, K. S. (2000). Translating questionnaires and other research instruments: problems and solutions (Vol. 133). London: Sage.

Bellezza, S., & Keinan, A. (2014). Brand tourists: how non–core users enhance the brand image by elic-iting pride. Journal of Consumer Research, 41(2), 397–417.

Brislin, R. W. (1970). Back-translation for cross-cultural research. Journal of cross-cultural psychology, 1(3), 185–216.

Brodie, R. J., Hollebeek, L. D., Jurić, B., & Ilić, A. (2011). Customer engagement: conceptual domain, fundamental propositions, and implications for research. Journal of service research, 14(3), 252–271.

Byrne, B. M. (2016). Structural equation modeling with AMOS: basic concepts, applications, and pro-gramming. New York: Routledge.

Cadogan, J. (2010). Comparative, cross-cultural, and cross-national research. International Marketing Review, 27(6), 601–605.

Carvalho, S. W., & De Oliveira Mota, M. (2010). The role of trust in creating value and student loyalty in relational exchanges between higher education institutions and their students. Journal of market-ing for higher education, 20(1), 145–165.

Cho, M., Lemon, L. L., Levenshus, A. B., & Childers, C. C. (2019). Current students as university donors?: determinants in college students’ intentions to donate and share information about uni-versity crowdfunding efforts. International Review on Public and Nonprofit Marketing., 16, 23–41.

Clark, K.D., & Cullen, P. (2016). Study abroad for business students: the Italian case. In P. Cullen and M.E. Montironi (Eds.), Teaching Business Culture in the Italian Context. Global and Intercultural Challenge (pp. 15-40). Frankfurt: Peter Lang.

Clotfelter, C. T. (2003). Alumni giving to elite private colleges and universities. Economics of Education review, 22(2), 109–120.

Dass, S., Popli, S., Sarkar, A., Sarkar, J. G., & Vinay, M. (2020). Empirically examining the psycho-logical mechanism of a loved and trusted business school brand. Journal of Marketing for Higher Education, Article in press.

De Feo, A., & Pitzalis, M. (2017). Service or market logic? The restructuring of the tertiary education system in Italy. Rassegna Italiana di Sociologia Fascicolo 2 aprile-giugno, 58(2), 219–250.

Dennis, C., Papagiannidis, S., Alamanos, E., & Bourlakis, M. (2016). The role of brand attachment strength in higher education. Journal of Business Research, 69(8), 3049–3057.

Diaz Vidal, D., & Pittz, T. G. (2019). Educating beyond the classroom: alumni giving and the value of campus culture. Studies in Higher Education, 44(12), 2208–2222.

Dowell, D., Morrison, M., & Heffernan, T. (2015). The changing importance of affective trust and cog-nitive trust across the relationship lifecycle: a study of business-to-business relationships. Indus-trial Marketing Management, 44, 119–130.

Drezner, N. D., & Garvey, J. C. (2016). LGBTQ alumni philanthropy: exploring (un)conscious motivations for giving related to identity and experiences. Nonprofit and Voluntary Sector Quarterly, 45(1), 52–71.

Drezner, N. D. (2018). Philanthropic mirroring: exploring identity-based fundraising in higher educa-tion. The Journal of Higher Education, 89(3), 261–293.

Drezner, N. D. (2019). The global growth of higher education philanthropy and fundraising. In Philan-thropy in education. Edward Elgar Publishing.

Drezner, N., Pizmony-Levy, O., & Anderson-Long, M. (2020). In “alma mater” we trust? Teachers Col-lege Record: Exploring attitudes toward institutions and alumni giving.

Durango-Cohen, E. J., & Balasubramanian, S. K. (2015). Effective segmentation of university alumni: mining contribution data with finite-mixture models. Research in Higher Education, 56(1), 78–104.

Dyson, B., Vickers, K., Turtle, J., Cowan, S., & Tassone, A. (2015). Evaluating the use of Facebook to increase student engagement and understanding in lecture-based classes. Higher Education, 69(2), 303–313.

Fabrigar, L. R., Porter, R. D., & Norris, M. E. (2010). Some things you should know about structural equation modeling but never thought to ask. Journal of Consumer Psychology, 20(2), 221–225.

Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of marketing research, 18(1), 39–50.

Frasquet, M., Calderón, H., & Cervera, A. (2012). University–industry collaboration from a relationship marketing perspective: an empirical analysis in a Spanish University. Higher Education, 64(1), 85–98.

Gallo, M. L. (2018). How are graduates and alumni featured in university strategic plans? Lessons from Ireland. Perspectives: Policy and Practice in Higher Education, 22(3), 92–97.

Ghosh, A. K., Whipple, T. W., & Bryan, G. A. (2001). Student trust and its antecedents in higher educa-tion. The Journal of Higher Education, 72(3), 322–340.

Gorsuch, R. (1983). Factor analysis. Hillsdale, NJ: Erlbaum.

Higher Education

1 3

Gunsalus, R. (2005). The relationship of institutional characteristics and giving participation rates of alumni. International Journal of Educational Advancement, 5(2), 162–170.

Hair, J. F., Black, W. C., Anderson, R. E., & Babin, B. J. (2018). Multivariate data analysis (8th ed.). NJ: Prentice Hall.

Harrison, V. S. (2018). Understanding the donor experience: applying stewardship theory to higher edu-cation donors. Public Relations Review, 44, 533–548.

Heffernan, T., Wilkins, S., & Butt, M. M. (2018). Transnational higher education: the importance of institutional reputation, trust and student-university identification in international partnerships. International Journal of Educational Management, 32(2), 227–240.

Helen, W. S. M., & Ho, W. K. (2011). Building relationship between education institutions and students: student loyalty in self-financed tertiary education. IBIMA Business Review, 22(2011), 1–31.

Hemsley-Brown, J., & Goonawardana, S. (2007). Brand harmonization in the international higher educa-tion market. Journal of Business Research, 60(9), 942–948.

Hennig-Thurau, T., Langer, M. F., & Hansen, U. (2001). Modeling and managing student loyalty: an approach based on the concept of relationship quality. Journal of service research, 3(4), 331–344.

Hofstede, G. (2003). Culture’s consequences: comparing values, behaviors, institutions and organiza-tions across nations. Thousand Oaks, CA: Sage Publications.

Holmes, J. (2009). Prestige, charitable deductions and other determinants of alumni giving: evidence from a highly selective liberal arts college. Economics of Education Review, 28, 18–28.

Hueston, F. R. (1992). Predicting alumni giving: a donor analysis test. Fundraising Management, 22, 19–22.

Hutcheson, G., & Sofroniou, N. (1999). The multivariate social scientist. London: Sage.Jillapalli, R. K., & Jillapalli, R. (2014). Do professors have customer-based brand equity? Journal of

Marketing for Higher Education, 24(1), 22–40.Iskhakova, L., Hilbert, A., & Joehnk P. (2020): Crosscultural research in alumni loyalty: an empirical

study among master students from German and Russian universities, Journal of Nonprofit & Public Sector Marketing.

Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in marketingand consumer research. Journal of consumer research, 30(2), 199-218.

Johnson, J. W., Thomas, V., & Peck, J. (2010). Do specialized MBA programs cultivate alumni relation-ships and donations? Journal of Marketing for Higher Education, 20(1), 119–134.

Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford.

Komljenovic, J. (2019). Making higher education markets: trust-building strategies of private companies to enter the public sector. Higher Education, 78(1), 51–66.

Koranteng, F. N., Wiafe, I., & Kuada, E. (2019). An empirical study of the relationship between social networking sites and students’ engagement in higher education. Journal of Educational Computing Research, 57(5), 1131–1159.

Leslie, L. L., & Ramey, G. (1988). Donor behavior and voluntary support for higher education institu-tions. The Journal of Higher Education, 59(2), 115–132.

Mael, F., & Ashforth, B. E. (1992). Alumni and their alma mater: a partial test of the reformulated model of organizational identification. Journal of organizational Behavior, 13(2), 103–123.

MacKenzie, S. B., & Podsakoff, P. M. (2012). Common method bias in marketing: causes, mechanisms, and procedural remedies. Journal of retailing, 88(4), 542–555.

McAlexander, J. H., & Koenig, H. F. (2010). Contextual influences: building brand community in large and small colleges. Journal of Marketing for Higher Education, 20(1), 69–84.

McDearmon, J. T. (2013). Hail to thee, our alma mater: alumni role identity and the relationship to insti-tutional support behaviors. Research in Higher Education, 54(3), 283–302.

McDonald, M. (2014). Colleges raise record $33.8 billion exceeding U.S. peak in 2009. Retrieved 26th June, 2014, from https ://www.bloom berg.com/news/artic les/2014-02-12/colle ge-donat ions-rise-to-recor d-as-stock s-gain-fuele d-givin g.

Melendez, S. E. (2001). The nonprofit sector and accountability. New Directions for Philanthropic Fun-draising, 31, 121–132.

Moorman, C., Zaltman, G., & Deshpande, R. (1992). Relationships between providers and users of market research: the dynamics of trust within and between organizations. Journal of marketing research, 29(3), 314–328.

Morgan, R. M., & Hunt, S. D. (1994). The commitment-trust theory of relationship marketing. Journal of marketing, 58(3), 20–38.

Higher Education

1 3

Newman, M. D., & Petrosko, J. M. (2011). Predictors of alumni association membership. Research in Higher Education, 52(7), 738–759.

Okunade, A. A., & Berl, R. L. (1997). Determinants of charitable giving of business school alumni. Research in higher education, 38(2), 201–214.

Pedro, I. M., Mendes, J.C., & Pereira, L.N. (2020) Understanding alumni-alma mater commitment relation-ships upstream and downstream. Journal of Marketing for Higher Education

Perin, M. G., Sampaio, C. H., Simões, C., & De Pólvora, R. P. (2012). Modelling antecedents of student loyalty in higher education. Journal of Marketing for Higher Education, 22(1), 101–116.

Pinar, M., Girard, T., & Basfirinci, C. (2020). Examining the relationship between brand equity dimensions and university brand equity: an empirical study in Turkey. International Journal of Educational Man-agement, 34(7), 1119–1141.

Pinar, M., Trapp, P., Girard, T., & Boyt, T. E. (2011). Utilizing the brand ecosystem framework in design-ing branding strategies for higher education. International Journal of Educational Management, 25(7), 724–739.

Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behav-ioral research: a critical review of the literature and recommended remedies. Journal of applied psy-chology, 88(5), 879–903.

Poole, S. M. (2017). Developing relationships with school customers: the role of market orientation. Inter-national Journal of Educational Management, 31(7), 1054–1068.

Rojas-Méndez, J. I., Vasquez-Parraga, A. Z., Kara, A. L. I., & Cerda-Urrutia, A. (2009). Determinants of student loyalty in higher education: a tested relationship approach in Latin America. Latin American Business Review, 10(1), 21–39.

Sargeant, A., & Lee, S. (2004). Trust and relationship commitment in the United Kingdom voluntary sector: determinants of donor behavior. Psychology & Marketing, 21(8), 613–635.

Schlesinger, W., Cervera, A., & Pérez-Cabañero, C. (2017). Sticking with your university: the importance of satisfaction, trust, image, and shared values. Studies in Higher Education, 42(12), 2178–2194.

Schreiber, J. B. (2008). Core reporting practices in structural equation modeling. Research in Social and Administrative Pharmacy, 4(2), 83–97.

Shaari, H., Salleh, S. M., Yong, P. L., Perumal, S., & Zainol, F. A. (2019). Assessing the effect of university brand personality and attitude towards donation on alumni donor behavioural intention: Malaysian per-spective. International Journal of Management in Education, 13(4), 377–396.

Snijders, I., Wijnia, L., Rikers, R. M., & Loyens, S. M. (2019). Alumni loyalty drivers in higher education. Social Psychology of Education, 22(3), 607–627.

Snijders, I., Wijnia, L., Rikers, R. M., & Loyens, S. M. (2020). Building bridges in higher education: stu-dent-faculty relationship quality, student engagement, and student loyalty. International Journal of Educational Research, 100, 101538.

Steenkamp, J. B. E., & Baumgartner, H. (1998). Assessing measurement invariance in cross-national con-sumer research. Journal of consumer research, 25(1), 78–90.

Stephenson, A. L., & Bell, N. (2014). Motivation for alumni donations: a social identity perspective on the role of branding in higher education. International Journal of Nonprofit and Voluntary Sector Market-ing, 19(3), 176–186.

Stephenson, A., & Yerger, D. (2014). Optimizing engagement: brand identification and alumni donation behaviors. International Journal of Educational Management, 28(6), 765–778.

Stephenson, A., & Yerger, D. (2015). The role of satisfaction in alumni perceptions and supportive behav-iors. Services Marketing Quarterly, 36(4), 299–316.

Sultan, P., & Yin Wong, H. (2012). Service quality in a higher education context: an integrated model. Asia Pacific Journal of Marketing and Logistics, 24(5), 755–784.

Sung, M., & Yang, S. U. (2009). Student–university relationships and reputation: a study of the links between key factors fostering students’ supportive behavioral intentions towards their university. Higher Education, 57(6), 787–811.

Taylor, A. L., & Martin, J. C. (1995). Characteristics of alumni donors and non-donors at a research I, pub-lic university. Research in Higher Education, 36(3), 283–302.

Thøgersen, J., de Barcellos, M. D., Perin, M. G., & Zhou, Y. (2015). Consumer buying motives and attitudes towards organic food in two emerging markets: China and Brazil. International Marketing Review, 32(3/4), 389–413.

Tsao, J. C., & Coll, G. (2005). To give or not to give: factors determining alumni intent to make donations as a PR outcome. Journalism & Mass Communication Educator, 59(4), 381–392.

Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance litera-ture: suggestions, practices, and recommendations for organizational research. Organizational research methods, 3(1), 4–70.

Higher Education

1 3

Weerts, D. J., & Ronca, J. M. (2008). Characteristics of alumni donors who volunteer at their alma mater. Research in higher education, 49(3), 274–292.

Weerts, D. J., & Ronca, J. M. (2009). Using classification trees to predict alumni giving for higher educa-tion. Education Economics, 17(1), 95–122.

Wong, S. M. H., & Ho, W. (2011). Building relationship between education institutions and students: stu-dent loyalty in self-financed tertiary education. IBIMA Business Review Journal, 2011, 1–22.

Wunnava, P. V., & Lauze, M. A. (2001). Alumni giving at a small liberal arts college: evidence from con-sistent and occasional donors. Economics of Education review, 20(6), 533–543.

Yousaf, A., Mishra, A., & Bashir, M. (2020). Brand trust, institutional commitment, and their impact on student loyalty: evidence for higher education in India. Studies in Higher Education, 45(4), 878–891.

Yousaf, A., Mishra, A., & Bashir, M. (2018). Brand trust, institutional commitment, and their impact on student loyalty: evidence for higher education in India. Studies in Higher Education, 1–14.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Affiliations

Barbara Francioni1  · Ilaria Curina1 · Charles Dennis2 · Savvas Papagiannidis3 · Eleftherios Alamanos3 · Michael Bourlakis4 · Sabrina M. Hegner5

Ilaria Curina ilaria.curina@uniurb.it

Charles Dennis c.dennis@mdx.ac.uk

Savvas Papagiannidis savvas.papagiannidis@ncl.ac.uk

Eleftherios Alamanos Eleftherios.Alamanos@newcastle.ac.uk

Michael Bourlakis m.bourlakis@cranfield.ac.uk

Sabrina M. Hegner sabrina.hegner@hs-bremen.de

1 Department of Communication Sciences, Humanities and International Studies, University of Urbino, Carlo Bo Via Saffi 15, 61029 Urbino, Italy

2 The Business School, Middlesex University, London NW4 4BT, UK3 Newcastle University Business School, Newcastle University, Newcastle upon Tyne NE1 4SE, UK4 Cranfield School of Management, Cranfield University, MK43 0AL Bedford, UK5 Bremen University of Applied Sciences, Werderstr. 73, 28199 Bremen, Germany