Is UK Vice Chancellor Pay Justified by U

R AY B ACHAN † and B ARRY R EILLY ‡

†University of Brighton (R.Bachan@brighton.ac.uk) ‡University of Sussex (b.m.reilly@sussex.ac.uk)

Abstract

The pay of university managers (vice chancellors, VCs) in the UK has risen sharply in recent years. This paper considers the extent to which VC pay awards are related to university performance measures based around institution mission and financial probity using an individual fixed effects strategy based on data covering academic years 1998–99 to 2008–09. In contrast to earlier studies on this theme, we find evidence that VCs are rewarded for observable mission-based performance measures. Specifically, our results suggest that success in widening participation for students from comprehensive schools and those from areas with low university participation exerts a positive effect on VCs’ pay. Securing income flows from university funding council grants also impacts positively on their remuneration. However, even after controlling for a rich array of observable and unobservable factors, there have been sizeable increases in real pay in recent years that cannot be readily explained.

* Submitted September 2012. The authors acknowledge the constructive comments and helpful advice provided by the journal’s

editor and an anonymous referee that have enhanced this paper. They would also like to thank Mike Barrow, Geraint Johnes, Maura Sheehan and participants at the 2008 annual conference of the European Association of Labour Economists (EALE) in Amsterdam and of the Work, Pensions and Labour Economics Study Group (WPEG) in Sheffield for constructive comments on earlier drafts of this paper. However, the usual disclaimer applies.

Keywords: vice chancellors, CEO pay–performance, public sector, higher education. JEL classification numbers: J45, M5, M12.

© 2015 Institute for Fiscal Studies

52 Fiscal Studies

Policy points

• The remuneration packages awarded to top executives from both the

public and private sector have risen sharply in recent years. Amongst those who have benefited from such increases are university managers (vice chancellors, VCs) in the UK. This paper considers the extent to which VC pay awards are related to university performance measures, including those based around institution mission and financial probity.

• The evidence reveals that VC pay is linked to widely-published performance indicators, including those related to widening participation, suggesting that there may be scope in setting VC pay to

introduce incentive pay schemes aligned to these indicators. • However, even after controlling for a rich array of observable and

unobservable factors, there have been sizeable increases in real VC pay in recent years that cannot be readily explained.

I. Introduction

The size of and increase in the remuneration packages awarded to top executives have often encountered adverse public reaction in both the UK and the US. Critics of such high pay awards often perceive them as either undeserved or unjust. For example, in his address to the Association of British Insurers in June 2011, the Business Secretary, Vince Cable, stated: ‘I want to see progress ... in bringing excessive, unjustified, executive pay under control. ... Ridiculous levels of remuneration are going unchallenged as the norm, when there is no clear evidence of a correlation with

performance’. 1 Such concern is often focused on the pay of heads of large private sector organisations, but concern has also been expressed in the UK

regarding the pay awards granted to senior executives in major public sector organisations. For instance, in July 2003, the then deputy prime minister, John Prescott, criticised the £200,000 annual salary offered by Bradford Metropolitan Borough Council to fill its chief executive officer (CEO) position, pointing out that it exceeded the salary of the then Prime Minister. 2

More recently, the Communities and Local Government Secretary, Eric Pickles, criticised the ‘bloated salaries’ paid to senior town hall officials, emphasising that some pay packages would ‘make a football manager

blush’. 3

The growth in public sector executive pay led to the recommendation that the pay of senior and the most highly remunerated public officials should be

1 Cable’s address is available at https://www.gov.uk/government/speeches/association-of-british- insurers-conference.

2 ‘Prescott lashes out at £200,000 council job’, Yorkshire Post, 5 July 2003. 3 ‘Eric Pickles pledges to cut “bloated” council pay’, The Telegraph, 10 July 2010.

Is UK vice chancellor pay justified by university performance?

publicly disclosed. 4 In 2011, the government published a list containing the names of 258 public servants (including the heads of quangos) earning in

excess of £150,000 per annum, seven of whom earned in excess of £250,000, with the highest-paid receiving an annual salary of £379,999 plus

allowances. 5

The pay awards of UK vice chancellors (VCs) 6 has also been the subject of public scrutiny since 1994 when UK higher education institutions (HEIs)

were required to publicly disclose the annual pay of the head of their institution. It was reported that, in 2010–11, 12 VCs received annual pay in excess of £300,000 excluding pension contributions and 33 earned over

£250,000. 7 The rate of increase in VC pay has also been of particular concern. The Business Secretary, Vince Cable, was reported as being ‘taken

aback’ by the 10 per cent rise in VC pay in the academic year 2008–09 and urged ‘restraint’ in the pay awards granted to top university officers. 8

Figure 1 presents the plot of VC pay between 1997 and 2009, showing that it increased by 59 per cent in real terms over this period. In contrast,

FIGURE 1 Average annual real pay at 1998 prices:

all VCs, CEOs/directors, senior executives and university lecturers

2006 2008 2010 All VCs

CEOs/Directors

Senior Managers

Note: The VC pay used here includes all HEIs. Source: Times Higher Education Supplement; Annual Survey of Hours and Earnings, various years.

4 Hutton, 2011. 5 Data.Gov.UK. 6 The abbreviation ‘VC’ will be used in this paper as a generic term to describe all heads of UK higher

education institutions and covers the various titles of vice chancellor, principal, rector, director, provost etc.

7 See accompanying tables to ‘Executive overdrive’, Times Higher Education (THE), 10 May 2012. 8 The Telegraph, 26 May 2010.

54 Fiscal Studies

full-time lecturers in higher education received a pay increase of about 16 per cent over the same period. The data suggest that, on average, VCs received pay awards that were close to four times those of lecturers over this period and a widening of this differential over time is also detected. These differences have been a major focus of criticism vented in the media and by lecturer unions, with assertions that the pay awards granted to VCs have

been ‘outrageous’, ‘shameless’, ‘despicable’ and ‘unjust’. 9 It is argued that the leadership and managerial skills needed to lead and

administer large complex HEIs are similar to the executive skills required to manage comparably large private sector companies. 10 In this sense, the pay

awards granted to VCs in the UK higher education sector may be seen as merited and not excessive given that we might think remuneration should be broadly comparable to that received by private sector CEOs with similar executive responsibilities. As depicted in Figure 1, the average level of pay received by VCs was well below the pay of CEOs/directors of large private

sector firms. 11 However, UK HEIs differ markedly in terms of their history, size,

organisational structure and the markets for which they cater, 12 which might

be expected to affect VC pay. Figure 2 illustrates the trends in real pay of VCs who headed different types of HEIs between 1997 and 2009. 13 It is

clear that VCs of the pre-1992 or ‘old’ universities received higher average pay in all years. Their real average pay increased by 66 per cent over the period, compared with 53 per cent for VCs in the former polytechnics or ‘new’ universities and 43 per cent for those who headed art, music and drama colleges. It is also evident that the pay of VCs of both ‘old’ and ‘new’ universities is significantly more than the pay received by those who head other HEIs.

Despite a high level of public interest in the pay of VCs, only a few studies have been conducted on the pay determination process. The primary aim of this paper is to examine the relationship between VC pay and performance and, in so doing, shed light on whether or not VCs’ headline pay awards are justified. We examine VC performance across two

9 See, for example, the following articles published in THE: ‘The annual pay review: are VCs worth every penny?’ (28 March 2013), ‘Identity check: vice-chancellors’ education and pay revealed’ (24

March 2011) and ‘It was fun while it lasted’ (1 April 2010); and in the Times Higher Education Supplement: ‘Large rises at top as pay dispute grips sector’ (10 March 2006), ‘25% wage hike for V-Cs’ (10 March 2006), ‘Heads enjoy 100% rise in pay over ten years’ (25 February 2005), ‘V-C pay survey:

thriving V-Cs net 6.1% rise in wages’ (20 February 2004), ‘Disparity is a disgrace’ (1 March 2002), ‘Union fury at “shameless” V-C pay rises’ (26 January 2001) and ‘Pay rises for university chiefs more than double increases given to their staff’ (26 January 2001).

10 Farnham and Jones, 1998; Smith et al., 1999; Shattock, 1999; Bargh et al., 2000; Whitchurch, 2006. 11 These are defined here as organisations employing 500+ workers in a single site or on multiple sites. 12 Dolton and Makepeace, 1982; Parry, 2006. 13 Postgraduate institutions and medical schools are excluded.

Is UK vice chancellor pay justified by university performance?

FIGURE 2 VCs’ average annual real pay by institution type at 1998 prices

Pre-1992 Universities Former Polytechnics Other HE Institutions and University Colleges Art, Music and Drama Colleges

Note: The econometric analysis reported in this paper only uses data from 1999 to 2009 and does not include data drawn from art, music and drama colleges and from other HEIs (see text). Source: Times Higher Education Supplement; institutions’ annual financial statements, various years.

dimensions: first, in terms of the financial performance of the university and, second, in terms of furthering the institution’s ‘mission’ or strategic plan.

The latter typically includes statements on ‘widening participation’, 14 excellence in teaching and research, and growth. In this particular respect,

our research makes a novel contribution to the existing literature. We also consider whether remuneration committees set VC pay with equity considerations in mind. They may not want to grant huge pay awards that may appear unjustified, in order to mitigate public outrage. On the other hand, the committee may determine to award a pay increase that would be commensurate with the earnings of other VCs, who run comparable institutions in terms of their size or nature of student intake. We explore this issue using a suitably-defined external pay ‘benchmark’, which remuneration committees are assumed to use when setting VC pay. 15 Further, we also

consider whether internal pay structures influence the pay-determining process and infer from this whether remuneration committees seek to preserve pay differentials within their institutions. These considerations have

14 See Department for Education & Skills (2003). 15 See Committee of University Chairs (2009, p. 27).

56 Fiscal Studies

been explored before but exploited different methodologies and used more restricted data than the current study.

The data set we use covers 1999 to 2009, a period of considerable change in the funding and nature of the UK higher education sector. It should be noted that it is not our intention to explore the relationship between the personal characteristics of VCs and their pay given that most of these attributes are immutable across time. Our major focus is on the pay– performance relationship. Thus, at a more general level, this paper also contributes to the relatively small body of research on the determinants of CEO pay in the public sector with particular reference to this relationship.

The paper is arranged as follows. Section II provides a brief review of the relevant literature. Section III describes the methodology employed and Section IV presents the key features of the data. The results are presented in Section V and a summary of the conclusions is contained in Section VI.

II. Literature review

The relationship between executive pay and factors assumed, a priori, to influence such rewards has received substantial attention in the academic literature and this interest has grown considerably over the last two decades. This research is, however, almost entirely concerned with executive pay in

16 the private sector. 17 In addition to human capital theory, three general approaches to explain the relatively high levels of pay for CEOs can be

identified in the literature: the principal–agent approach, tournament theory and the theory of managerial power.

In certain respects, VC pay in the UK can be viewed as a principal–agent problem where it is necessary to motivate the agent to act in the principal’s best interest by designing ‘optimal contracts’ where executive effort is unobserved. 18 This theoretical framework points to a link between executive pay and the ‘size’ and ‘performance’ of the firm. 19 In particular, within the

public sector, agents may have vaguely-defined tasks and face multiple principals who may have to compete for the agent’s effort. These characteristics of public sector organisations result in agents facing several incentive structures that may offset one another, making the overall

16 For comprehensive reviews of the literature, see Murphy (1999), Prendergast (1999) and Jensen, Murphy and Wruck (2004).

17 Mincer, 1974; Becker, 1993. 18 See Mirrlees (1976) and Holmstrom (1979). 19 See Jensen and Zimmerman (1985), Murphy (1985 and 1986), Jensen and Murphy (1990), Gregg,

Machin and Szymanski (1993), Conyon, Gregg and Machin (1995), Conyon and Murphy (2000), Conyon et al. (2000), Girma, Thompson and Wright (2007), Gabaix and Landier (2008), Frydman and Jenter (2010), Gregg, Jewell and Tonks (2010) and Bell and Van Reenen (2011).

Is UK vice chancellor pay justified by university performance?

incentive weak. 20 Marsden (2010) discusses the problems of introducing performance-related schemes in the UK public sector.

A second strand uses tournament theory to explain the high level of executive pay. 21 The basic framework assumes that promotion lotteries

provide the necessary incentives for high-ranking executives where executive effort is again unobservable. Firms and organisations are prepared to spend large sums of money on CEO salaries and benefits to reward capability because it also serves to motivate workers at all levels in the firm to work hard for promotion. Workers are ranked according to their relative performance and winners secure the ‘prize’ (known in advance) in terms of higher pay and the opportunity to participate in subsequent promotional tournaments. The ultimate prize is the promotion to the rank of CEO. Employees are assumed to exert effort to increase the likelihood of securing the ‘prize’, and the effort expended depends on the differential in pay between a high-ranked position and a lower-ranked one, the number of

competitors in the lottery and the likelihood of winning. 22 It is possible that the pay package awarded to high-ranking university officers (for example, a

VC) is influenced by the pay packages received by those in lower-ranked positions (for example, a pro-VC) or simply by the presence of highly-paid academics and/or administrative staff. Moreover, we may expect that a vacant VC position in a prestigious institution, noted for its quality in research and teaching, will attract more candidates for the post than a similar vacancy in a lower-ranked institution on the basis of these attributes. Thus, the likelihood of winning the ‘tournament’ and securing the post in a highly- ranked institution is lower in comparison, and pay will be higher in these institutions than in a lower-ranked one.

A third approach – the theory of managerial power – focuses on the influence CEOs exert over their own pay package. The strength of this influence will depend on their relationship with the board of directors and/or

the remuneration committee. 23

The determination of VC pay in UK higher education

Research into the determination of VC pay in UK higher education is limited, not least because it is often difficult to determine what exactly

20 See Tirole (1994) and Dixit (2002). 21 See Lazear and Rosen (1981) and Rosen (1986). 22 See O’Reilly, Main and Crystal (1988), Main, O’Reilly and Wade (1993), Knoeber and Thurman

(1994), Bognanno (2001), Audas, Barmby and Treble (2004), DeVaro (2006), Clark, Kristensen and Westergard-Nielsen (2009) and Chen, Ezzamel and Cai (2011).

23 Main and Johnston, 1993; Blanchard, Lopez-de-Silanes and Shleifer, 1994; Main, O’Reilly and Wade, 1995; Conyon, 1997 and 2006; Newman and Mozes, 1999; Bertrand and Mullainathan, 2001;

Bebchuk and Fried, 2003 and 2006; Acharya and Volpin, 2010; Frydman and Jenter, 2010; Morse, Nanda and Seru, 2011.

58 Fiscal Studies

constitutes ‘performance’ in higher education (for example, income generation, research/teaching quality, student enrolment), and indeed what elements of performance are legitimately ascribed to VC effort alone. 24 An

early study by Baimbridge and Simpson (1996), using a cross-section of 64 VCs for the academic year 1993–94, found some evidence that pay is influenced by university income largely based on research grants and fees. Dolton and Ma (2003), using a similar specification to Baimbridge and Simpson, estimated the earnings relationship for VCs in UK HEIs using panel data covering the nine academic years from 1993–94 to 2001–02. They found some evidence that institutional characteristics such as type,

size, internal pay structures, income and research performance 25 exerted a positive and significant effect on VC pay.

In a more recent study, Tarbert, Tee and Watson (2008), using VC salary data for the period 1997 to 2002, modelled the relationship between pay and university performance employing dynamic pay change models. They also included controls for internal and external pay benchmarks. In contrast to Dolton and Ma (2003), they found no evidence that VC pay is influenced by observable performance measures such as university income and student recruitment. This may be unsurprising given the poor explanatory power of first-difference pay change models regardless of the definition of pay and the

performance measures employed. 26 However, they did find a positive and significant relationship between VC pay and the presence of ‘highly’-paid

staff in the institution 27 and between VC pay and the pay of VCs leading comparable HEIs. Moreover, when their sample was split between pre-1992

(‘old’) universities and post-1992 (‘new’) universities, they found evidence that changes in VC pay were related to changes in both research income and the number of postgraduate students for pre-1992 universities and to changes in the total number of students for post-1992 universities.

The foregoing review suggests that VC remuneration may be partly related to performance and not simply bureaucratically determined as predicted by agency theory. In the current paper, we investigate this

24 There has been some empirical research on the determinants of the pay of college presidents in the US. For example, Ehrenberg, Cheslock and Epifantseva (2001) found a weak link between pay and

institution performance (i.e. the president’s success in securing private donations). However, they did find some evidence linking the president’s tenure and experience to pay and further evidence linking pay to institution size, type and income, though these effects are not robust across the specifications reported. Cornell (2004) found that elite US universities do not find it difficult to recruit suitable candidates for the position of president even though they are paid significantly less than and have similar skills and abilities to top corporate CEOs in the US.

25 They used the scores from the Research Assessment Exercise (RAE), which inform the allocation of government research funds to UK universities.

26 See Conyon, Gregg and Machin (1995). 27 ‘Highly paid’ was defined as earning over £50,000. Whether this is indeed ‘high pay’ is debatable, as

mean academic pay over the sample period covered was about £33,000 (Annual Survey of Hours and Earnings, various years).

Is UK vice chancellor pay justified by university performance?

relationship using various measures of university performance including the income a university attracts from UK funding councils and other non- governmental sources and variables that can capture VC success in furthering the university’s mission or strategic plan. Therefore, we include in our analysis variables that explicitly capture certain dimensions of the university’s mission statement such as those relating to ‘widening participation’ and institution expansion. There is also evidence that internal pay structures influence VC pay, as predicted by tournament theory. Thus, as with previous studies, we include the proportion of highly-paid staff within the institution to investigate this issue. We also assess whether remuneration committees rely on external pay benchmarks in determining VC pay, using a suitably-constructed benchmark for each VC based on VC pay in similar institutions.

III. Methodology

The primary objective of this paper is to investigate the relationship between VC pay and an array of observable performance indicators, including variables that capture mission-based and financial-based performance. We do not report results for pay change models, due to their poor explanatory

power, 28,29 and instead model the relationship of interest by employing a VC fixed effects methodology. 30 The regression model is expressed as

1, 2,..., T where: w th

i == 1, 2,..., N ; j = 1, 2,..., ; Jt

ijt is the natural log of real annual pay for the i VC in institution j at time t and X ijt is a vector of time-varying VC- and institution-specific pay- determining variables excluding those that are mission and financial

performance relevant. A location-specific variable for the j th institution is

28 See Tarbert, Tee and Watson (2008). 29 Pay change models were also estimated. The annual difference in the log of CEO pay was used as the

regressand and the annual difference in the performance measures were used as regressors. The specifications were very poorly determined and no inference could be drawn from the estimated

parameters. Conyon, Gregg and Machin (1995) discuss the problems associated with models of this kind.

30 We favour the use of individual VC rather than institution fixed effects in this application because we believe the remuneration package offered is more strongly influenced by the attributes the individual

brings to the job than by the characteristics of the institution. In addition, we believe the assumption of time-invariant individual attributes is more persuasive in the current context than the assumption of time- invariant institutional characteristics, given the rapidly changing nature of the university landscape in the UK over the period covered by our analysis. However, the robustness of our results is investigated using an alternative model containing institutional fixed effects also (see Table A2 of the online appendix, available at http://www.ifs.org.uk/docs/fsmar15_bachan&reilly_appendix.pdf).

60 Fiscal Studies

also included in

X to capture local economic conditions that may influence VC pay (here captured using the log of average county-level house prices).

The vector X ijt also includes an external benchmark pay variable reflecting the pay awarded to VCs of comparable universities. This variable is constructed for each year in the data set according to the type of university (‘old’ or ‘new’) and the institution size determined by the total number of full-time-equivalent (FTE) students (undergraduate and postgraduate). By university type, each university is arranged in ascending order according to the total number of FTE students. The universities are then grouped into fives, starting from the smallest institution. Within each group and for each VC, the average group pay is calculated excluding the pay of the VC who is subjected to the benchmark. This variable is calculated for all VCs and its natural log is used to define the external benchmark pay that remuneration

committees potentially use to inform their pay awards. 31 It enters the regression model (1) lagged by one year. We expect VC pay to be positively

related to this variable, indicating that as the pay of comparable VCs increases, so too does the pay of the VC under consideration.

Z ijt–1 is a vector of lagged-one-year mission-based performance measures and V ijt–1 is a vector of lagged-one-year financial performance variables. A

set of year-specific dummies, D t , is also included, to capture exogenous events that are taken to commonly affect all VCs over time. Finally, the unknown parameters α i (the intercepts) are the VC fixed effects assumed to capture differences across VCs in terms of their unobservable characteristics.

IV . Data

The data employed in this paper were collected from a variety of sources listed in the online data appendix. 32 The data include information on the 193

VCs who led 95 UK universities between 1998–99 and 2008–09 (inclusive). Once allowance is made for missing values, these data represent an unbalanced panel comprising 1,045 observations. 33 As described in the

previous section, we use information on VC pay, individual characteristics and variables that capture mission-based and financial-based performance. It should be noted that the mission-based performance data used in this study are only available for all institutions from 1997–98. The pay and all other

31 This was confirmed by interviews with the VCs’ personal assistants (PAs) in a number of universities. It would appear that in setting VC pay, one important source of information available to

remuneration committees is the pay of VCs (assumed to be the previous year’s pay) in ‘comparable’ institutions. In the interviews, PAs also indicated that university type and size of institution by number of students were important considerations in framing VC pay.

32 http://www.ifs.org.uk/docs/fsmar15_bachan&reilly_appendix.pdf. 33 There were 125 exits in the data, including 90 retirements and 13 others appointed VC at universities

not included in our sample.

Is UK vice chancellor pay justified by university performance?

TABLE 1 Summary statistics at vice-chancellor level

All VCs

VCs of

VCs of z-score /

‘new’ t-statistic a universities universities

‘old’

VC pay Real average pay (£)

(19,021) Real average benchmark pay (£)

(17,904) VC characteristics Mean age (years)

(4.08) Age ≤55 years

0.316 –1.24 Age 56–60 years

0.429 –0.05 Age >60 years

0.255 2.23 Tenure (years)

(4.152) University characteristics Pre-1992 university

0.010 8.02 than £70,000 p.a.

Proportion of staff earning more 0.025

(0.007) Real house price (£)

(54,756) Mission-based performance Merger (number)

17 14 3 Hit benchmark for

0.912 –5.32 comprehensive school students

0.551 –3.05 low-participation areas

Hit benchmark for students from

Financial-based performance Real funding council grants

(12,086) Real tuition fees (£000s)

(10,358) Real research council grants

Number of observations

429 Number of VCs

74 a z-scores are used to test differences in proportions between the characteristics of VCs of ‘old’ and ‘new’

universities; t-statistics are used to test differences in means. The critical value at the 0.05 level using a two-tailed test is ±1.96. The null of no association between university type and the VC age categories is

upheld by the data using a χ 2 test with two degrees of freedom. The test value is 3.172 and the corresponding p-value is 0.205. Note: Standard deviations are reported in parentheses beneath the means for continuous variables only.

62 Fiscal Studies

financial data used are expressed in real terms and adjusted to reflect 1998 prices. 34 The associated summary statistics computed at the level of the VC

are reported in Table 1 and are the focus of the discussion below. 35

We exclude from our analysis postgraduate institutions, medical schools, university colleges and other HEIs including art, drama and music colleges, as relevant performance data are unavailable for these institutions for our time period. Moreover, these institutions differ from ‘old’ and ‘new’ universities in terms of their size and the nature of the programmes of study offered. We thus focus our analysis on the pay–performance relationship for VCs of ‘new’ and ‘old’ universities, which we argue represent a more homogeneous group and for which performance data are readily available for the period under review here. Furthermore, it is the pay of these VCs that attracts most public attention and criticism. The set of ‘old’ universities accounts for about 59 per cent of the sample (see Table A1 of the online appendix). The institutions that are included in the analysis are detailed in

Table A3 of the appendix. 36 The annual data on VC pay were obtained from the Times Higher

Education Supplement (THES), and university financial accounts for the period covering 1998–99 to 2008–09 were also used. The pay data used include any performance-related pay and an estimated value of benefits in kind but exclude pension contributions made by the institution. It is not possible to distinguish between the elements that comprise the final pay from information provided. The mean pay over the sample period is £141,537 at 1998 prices. The sample data also reveal that VCs of ‘old’ universities were paid, on average, close to 8 per cent more than their counterparts in ‘new’ universities.

Human capital variables such as age, tenure and experience are expected to enhance VC pay. Thus, we include the time-varying covariates of age and tenure of the VC in our analysis. Age enters the analysis as a set of mutually exclusive dummy variables reflecting the VC’s age at time t. These data reveal that the average age is just over 58 years and there is no significant difference in mean age between those VCs heading the two types of institution. We note that ‘old’ universities employ proportionately more VCs who are sexagenarians (34 per cent versus 25 per cent). The average length of tenure (years in post) is just over five years, but VCs in ‘new’ universities, on average, have been in post for just over a year longer than their counterparts in ‘old’ universities (5.8 years versus 4.6 years). Of course, it is

34 We use the RPI (all items) available at http://www.ons.gov.uk/ons/rel/cpi/consumer-price-indices/ december-2013/cpi-time-series-data.html (accessed June 2010).

35 Summary statistics at the university level, which are broadly similar to those reported in Table 1, are reported in Table A1 of the online appendix for comparison.

36 http://www.ifs.org.uk/docs/fsmar15_bachan&reilly_appendix.pdf.

Is UK vice chancellor pay justified by university performance?

readily acknowledged that these comparisons may be misleading as they are based on incomplete tenure durations in some cases.

As noted in Section II, tournament theory suggests that the presence of highly-paid staff in an institution drives up the pay of CEOs or other senior directors. In the context of a university, such highly-paid staff generally comprise senior executives and professors. Data on the pay of senior executives and professors are not publicly available but we do have information on the number and proportion of academic and non-academic

staff earning in excess of £70,000 for each year. 37 On average, about 2.5 per cent of university staff earn over £70,000, but the proportion is higher for the

‘old’ universities.

There has been considerable growth in student enrolments in UK higher education over the 11-year period spanning these data. Institutions have also invested heavily in developing and expanding their teaching and learning environments. Such expansion can reflect investment in new facilities both in the UK and abroad or can be brought about through merging with other institutions. These developments can be seen as part of the university’s mission and successful expansion through a merger or acquisition may be attributed to successful leadership and good strategic management. We would expect VC pay to reflect such outcomes. In these data, we note that there have been 17 episodes of major mergers over the period. It is clear that in most cases VCs in ‘old’ universities were more active in this regard.

Many universities have missions to ‘widen participation’ by attracting students from socially disadvantaged groups and from areas where university participation is low. We use two measures to capture institutional performance along this dimension. These are the percentage of students that universities enrol from areas where traditionally there has been low participation in higher education (i.e. areas where participation rates are less than two-thirds of the national figure) and the percentage of young students enrolled from state comprehensive schools. Each university is given a ‘benchmark’ intake for students from state comprehensive schools and under-represented areas, which is expressed as a percentage of its first-year intake number. The ‘benchmark’ used is calculated as a sector average, which is adjusted for the university’s subject mix, student qualifications and age on entry. The data reveal that just under 69 per cent of all VCs achieve their benchmark figure in attracting students from comprehensive schools, but they do less well in attracting students from traditionally low-

37 Although universities reported the number of staff earning over £100,000 over the sample period, many did not or reported a zero figure in their annual financial statements. This was generally the case

with many ‘new’ universities. The number/proportion of staff earning over £100,000 was very small – in many cases less than 1 per cent – except for a few large civic universities. The estimated effect for the

variable proved insignificant in the regressions undertaken and the variable was omitted from subsequent analysis. Most universities registered some staff earning over £70,000 over the period and this variable was thus used in the analysis reported.

64 Fiscal Studies

participation areas, with less than half achieving the benchmark figure. Moreover, it appears, perhaps unsurprisingly, that VCs in ‘new’ universities have been more successful in widening participation for these particular groups of students.

We expect VC pay to be positively related to university income, as a reward for sound financial management and leadership. We employ three income variables to test this proposition: income in the form of UK funding council block grants for general teaching and research and capital expenditure; tuition fees and educational grants; and income from externally- sponsored sources (i.e. research grants and contracts from UK research councils and other UK and non-UK funding sources). The sample mean incomes from these sources are about £48.5 million, £31.0 million and

£23.7 million respectively. It is clear that the VCs of ‘old’ universities attracted significantly more income from these sources than those in ‘new’ universities.

We do not, however, include other putative measures of institution performance, such as the outcomes from various rounds of the Research Assessment Exercise (RAE; a measure of university research performance used to allocate government research funds) and the Teaching Quality Assessment (TQA; an assessment of teaching quality across a selection of subjects). This is because there have been major changes over time in the methods used to calculate the RAE scores, rendering the published figures difficult to compare and interpret. Also, in many cases, we are unable to attribute RAE scores to the effort of a single VC. The TQA was not repeated for most subjects and, in any event, scores are not available for many

institutions. 38

V. Empirical results

The regression estimates based on equation (1) using VC fixed effects are reported in Table 2. 39 On the basis of the goodness-of-fit measure reported,

the specification fits the data well by the standards of broadly comparable fixed effects applications. 40 We also note that the estimated coefficients for

the academic year dummies are jointly well determined at a conventional level and suggest ceteris paribus a steady monotonic increase in VC pay over time. In general, the point estimates suggest substantial pay inflation since the academic year 1998–99 once we control for individual fixed effects

38 See Higher Education Funding Council for England for a description of the methods used to calculate the RAE score: http://www.hefce.ac.uk/research/ref/reform/.

39 A Hausman test was conducted to test whether a random effects model was preferable to the VC fixed effects specification expression (1). The resultant χ 2 statistic with 23 degrees of freedom = 33.65 (p = 0.0704) favours the fixed over a random effects parameterisation in the current application.

40 For example, see Gregg, Jewell and Tonks (2010) and Fabbri and Marin (2012).

Is UK vice chancellor pay justified by university performance?

TABLE 2 Vice-chancellor pay and performance: VC fixed effects estimator

Coefficient Robust standard error VC and university characteristics Age ≤55 years

–0.0003 0.018 Age 56–60 years

–0.016* 0.009 Age >60 years

Ref. Tenure (years)

0.0004 0.005 (ln) Real external pay benchmark (lagged one year)

0.151** 0.068 Proportion of staff earning >£70k

0.434** 0.222 Pre-1992 university

0.011 0.053 (ln) Real average house price

–0.046* 0.025 Mission-based performance measures (lagged one year) Merger/expansion

0.055** 0.023 Hit benchmark for comprehensive school students 0.023*

0.013 Hit benchmark for students from low-participation areas 0.021**

0.010 Financial-based performance measures (lagged one year) (ln) Real funding council grants

0.106*** 0.035 (ln) Real tuition fees

0.012 0.041 (ln) Real research council grants

–0.010 0.012 Year dummies Year dummy 1999

Ref. Year dummy 2000

0.018* 0.011 Year dummy 2001

0.074*** 0.016 Year dummy 2002

0.116*** 0.020 Year dummy 2003

0.166*** 0.027 Year dummy 2004

0.205*** 0.035 Year dummy 2005

0.241*** 0.042 Year dummy 2006

0.274*** 0.047 Year dummy 2007

0.289*** 0.053 Year dummy 2008

0.352*** 0.058 Year dummy 2009

Number of observations 1,045 Number of VCs

193 R 2 – within

0.6985 Note: The robust standard errors are corrected for clustering at the level of the VC. * denotes significant

at 10 per cent, ** denotes significant at 5 per cent and *** denotes significant at 1 per cent, all using two- tailed tests. ‘Ref.’ denotes base category in estimation.

66 Fiscal Studies

and other pay-determining characteristics. 41 The estimated coefficient on the university type dummy is not significantly different from zero in our

specification, suggesting no ceteris paribus differential in pay across ‘old’ and ‘new’ universities. It is readily acknowledged that this may be explained by the fact that the VC fixed effects absorb most of its variation such that the estimated effect is identified through the small number of VCs who switch between these two types of universities, so some interpretational caution is required here.

Age enters as a set of dummy variables corresponding to the age group to which the VC belongs at time t. Those VCs in the 56–60 age group are paid less on average than those aged over 60. The point estimate suggests that a VC in the younger of these two age groups is paid, on average and ceteris paribus, about 1.6 per cent less. There also appear to be no significant differences in VC pay between those who are 55 years or under and those aged over 60. This particular result may be capturing a quality effect where the bidding for talent, through the award of higher pay, is independent of

age. In addition, we find no evidence that VC pay is influenced by tenure. 42

Typically, the remuneration committee sets pay awards for professors and other senior staff prior to determining the pay of their VC. 43 It is assumed

that such committees have information regarding the pay awards granted to senior or highly-paid staff in any given year when setting their own VC’s pay. Our results in this regard confirm findings from earlier research. There is evidence that VC pay is positively influenced by the presence of highly- paid staff in the institution (i.e. staff earning in excess of £70,000). The point estimate suggests that a 1 percentage point increase in the proportion of staff earning in excess of £70,000, on average and ceteris paribus, raises VC pay by about 0.43 per cent.

We argue that remuneration committees use an external pay benchmark to inform the annual pay of their VC. The external benchmark will typically

be the pay awarded to VCs in institutions comparable in terms of size and type. 44 We argue that it is unlikely remuneration committees will know the

current pay awarded to VCs in comparable institutions, though we assume they will have information on the previous year’s awards. To capture this effect, our measure of the external pay benchmark is lagged by one year.

41 VC pay and other financial data were also transformed to real variables using the consumer price index and the average pay index for the public sector (including bonuses) available at

http://www.ons.gov.uk/ons/ (accessed January 2014). These transformations made no material difference to the results reported in Table 2.

42 We also created a VC previous experience dummy variable capturing whether or not the incumbent had previously been appointed a VC at another university (including non-UK universities) and included

this as a regressor. The estimated coefficient was statistically insignificant (|t|=1.02 [p=0.31]). In addition, the exclusion of this variable made no material difference to the empirical results for the key performance

variables reported in Table 2. It was therefore dropped from further analysis. 43 Tarbert, Tee and Watson, 2008.

44 See Committee of University Chairs (2009, p. 27).

Is UK vice chancellor pay justified by university performance?

The estimated coefficient for this variable is well determined and the corresponding point estimate suggests that a 1 per cent increase in relative VC pay in comparable institutions (in the previous year) increases VC pay,

on average and ceteris paribus, by about 0.15 per cent. 45 This suggests that external benchmarking influences the framing of VC pay even after

controlling for institution performance and VC individual fixed effects.

The average county-level house price enters the regression model (in logged form) to control for regional economic conditions that may impact on local living costs. A significant but negative effect is detected. This seems counterintuitive, but could reflect the receipt of some form of payment in ‘kind’, such as subsidised housing or transport costs, provided to those VCs living in areas where the cost of living is high. This may be particularly the case for VCs at universities in major cities or the heads of prestigious universities such as Oxford and Cambridge known to provide accommodation in kind and where VC pay is sometimes relatively modest. 46

We now turn attention to the relationship between VC pay and mission and financial performance indicators. We first review the estimates for the variables assumed to capture dimensions of the university mission: growth and widening participation. One way in which an institution may grow is through an acquisition or merger, and the management of such a growth strategy appears to enhance VC pay by about 5.5 per cent per annum. There is also a significant effect detected for reaching the benchmark level for admitting students from areas where university participation has traditionally been low. The point estimate suggests that a VC in charge of a university achieving the relevant benchmark earns, on average and ceteris paribus, a

2.1 per cent pay premium. In addition, reaching the benchmark set for student intake from comprehensive schools enhances VC pay by a similar

magnitude. 47

Now we examine the role of university income variables that reflect one dimension of an institution’s financial performance. These comprise, as already noted, funding council grants, tuition fees and grants from research councils received in the previous year. Only the estimated coefficient on the

45 We also experimented with a wider external benchmark measure comparing incumbent VC pay in groups of six (rather than five) comparable institutions. The estimated coefficient remained significant but

reduced in magnitude.

46 Regressions that excluded the VCs of Oxford and Cambridge were also estimated but this had no effect on the sign and significance of the estimated house price coefficient.

47 We also estimated models including variables that capture the extent to which the two benchmarks were reached in place of the mission-based dummies. The estimated coefficients suggested that VC pay

increases by about 2.9 per cent if the proportion of students from low-participation areas increases by 1 percentage point above its benchmark figure and by about 1.3 per cent for a similar increase in the proportion of students from comprehensive schools. However, this latter effect was not statistically significant at a conventional level. The inclusion of these variables made no material difference to the significance and size of the other coefficients reported in Table 2 except that the age variables became insignificant.

68 Fiscal Studies

funding council grant achieves statistical significance at a conventional level. The point estimate suggests that a 1 per cent increase in funding council grants increases VC pay by just over 0.1 per cent, on average and ceteris paribus. Although statistically insignificant, the negative coefficient

on research council grants has been detected in previous studies. 48 The poorly-determined nature of some of the financial performance measures

may be due to the high correlations between these variables. 49 As a check of the robustness of the results reported above, we re-estimate

specification (1) using university rather than VC fixed effects. The results using institution fixed effects are reported in Table A2 of the online appendix along with the results already reported in Table 2 to permit a direct

comparison. 50 We note that the signs of the estimated coefficients are broadly similar for the two models and some of the estimated effects remain

stable across the two specifications. The effects of internal pay structure and the external pay benchmark remain significant across the two specifications, although the external pay benchmark now yields twice the estimated effect. The coefficient on the merger variable remains significant and its estimate is now numerically larger than when VC fixed effects are included. In terms of financial performance, the effect of attracting funding council grants remains robust across the two specifications. The impacts reported for the year dummies, though still significant and exhibiting a monotonic increase, are attenuated by about one-half in most cases. Further, VC pay is not found to

be affected by success in ‘widening participation’ when the set of university fixed effects are used in estimation. This may be attributable to the fact that the institutional fixed effects absorb most of the variation associated with the two measures of participation. This is perhaps to be anticipated since these social inclusion measures reflect explicit university policy. Overall, these results taken together provide for a general degree of confidence in the robustness of the results reported in Table 2, when VC fixed effects were used.

VI. Conclusion

There has been recent public interest in the determination of CEO pay in the public sector. This paper empirically examines the relationship between university vice chancellor pay and mission-based and financial performance indicators over an 11-year period for the UK. In addition, the data employed

48 See Dolton and Ma (2003) and Tarbert, Tee and Watson (2008). 49 The correlation coefficients are 0.65 (p = 0.00) between research council grants and funding council

grants, 0.56 (p = 0.00) between research council grants and tuition fees and 0.75 (p = 0.00) between funding council grants and tuition fees.

50 The correlation coefficient between the 95 current VC fixed effects and the comparable university fixed effects is 0.75, suggesting a high intercorrelation between the institutional and individual

unobservables in this application.

Is UK vice chancellor pay justified by university performance?