Salary structure is a key determinant of a firm's success, and the positive relationship between average wage and firm performance is well established in the literature (e.g. Nickell and Wadhwani 1990, Nickell et al. 1994, Hildreth and Oswald 1997, Abowd et al. 1999). The pay–performance relationship holds true not just in commercial firms (Lazear 2000), but also in some organisations that lack a monetary measure of success, such as schools (Lavy 2009). What about universities? There is surprisingly little systematic evidence on this important question. In a recent paper, we address this, focusing on professors in UK universities (De Fraja et al. 2016).
Universities ‘get what they pay for’
Matching data on annual full professors' wages for 2013 with research performance measures for their departments from the 2014 Research Excellence Framework, we show that departments that pay their professors more perform better (Figure 1). This result holds in regressions where we control for a range of departmental characteristics and for university and subject fixed effects, in a model with very high explanatory power; it also holds across the full range of academic disciplines, even if appears to be weaker in the most established and research intensive universities.
Figure 1 Average professorial pay and REF Funding Score
Our data highlight the substantial dispersion of the UK full professors’ wage distribution. The highest paid professors in some of the elite institutions earn as much as seven times the national agreed minimum (the salary typically earned by newly promoted full professors at less prestigious universities).
Our analysis suggests that professorial wage inequality within a department is also associated to better performance. This finding, however, is weaker than for the average level of pay, and holds in some research areas only – the “Arts and Humanities” and even more so “Science and Engineering” – while we do not find a significant correlation in the “Social Sciences” or in “Medicine and Biology”.
What drives the relationship between wages and REF performance?
The positive correlation between how much academics are paid and how well they perform in research can have two different explanations. It could be due to the fact that some institutions have more funds at their disposal and can therefore outcompete their less resourced rivals, in the fashion of football clubs with a wider supporter base. It could, however, instead be due to the fact that some institutions just strike lucky and succeed in research, like firms hitting on popular products, and can share with their senior staff the financial rewards brought by their success.
We can distinguish between these possible explanations by exploiting an important feature of the rules of the evaluation exercise. The REF process involves peer-review assessment, by 36 subject-specific expert panels, of the ‘reach and significance’ of the research carried out by the academics which the universities submit for assessment. Panels evaluate academic departments in three areas: research outputs, environment, and impact. The three components are then averaged (with weights of 65%, 20%, and 15%, respectively) to construct an overall numerical ‘quality’ profile.
Output is assessed through the evaluation of scholarly publications (e.g. books and journal articles) attributed to the academic at the institution in which they are employed on 31 October 2013 (the REF census date), even if the output was produced while the academic was employed elsewhere. The environment component is described in a written submission highlighting the achievements of the department, together with data on research grant income and PhD completions. Similarly, impact is assessed through written submissions of ‘case studies’ (one for every eight academics submitted), accompanied by supporting evidence which shows how the research of the department has brought benefits outside of academia through, for example, influence on government policy or industry practice.
When we break down the overall quality profile into its three components, we find that the overall positive association between average salary and REF performance is driven primarily by the relationship between salary and publications and, to a lesser extent, by that between salary and assessment of the quality of the environment. There is no evidence of a positive relationship between average salary and the impact of a department's research. This is consistent with institutions being unable – under REF rules – to ‘buy in’ impact success. This in turn implies that they give more weight, when hiring or promoting academics, to high performance along the output dimension, as an academic with a good record in this ‘transferable’ dimension contributes more to the REF outcome than one who is relatively better in the ‘impact’ dimension.
Do all universities behave in the same way?
We also find evidence of difference in behaviour between universities, which we divide into four groups: (i) the most established institutions, the original ‘Russell’ group; (ii) those that formed the disbanded ‘1994’ group – younger and smaller research-intensive universities; (iii) the ‘New universities’, mostly created from locally controlled vocational institutions, typically former polytechnics; (iv) the rest (‘Other’), which historically put less emphasis on research. Our data highlight a clear hierarchy in REF outcomes for these four groups, as illustrated in Figure 2. Departments based in ‘Russell’ group universities tend to perform better, followed by those in the ‘1994’ group, then by the ‘Others’, and finally by the ‘New universities’.
Figure 2 Research funding performance across university types
The association between average pay and research performance becomes stronger as the research performance itself becomes weaker. Thus an increase of 1% in the average professorial pay is associated with a negligible increase in research performance in the ‘Russell’ group universities, a 1% increase in the ‘1994’ group, a 1.75% increase in the ‘Other’ universities, and a full 2.1% in the ‘New’ universities. This result can be clearly seen in the increase in the slope of the regression line in Figure 3 as one moves from one group to the other.
Figure 3 Correlation between mean departmental pay and funding score
Our analysis indicates that other characteristics of the department also matter. For instance, we find that the age composition of the professoriate plays a significant role: replacing an older professor with an under 50 is associated with an increase in funding of 1.6-3.1%.
Are there biases in the REF assessment?
More importantly, we uncover systematic evidence of a positive impact of having a member of the department in the panel on REF success. This effect is particularly strong in the ‘Arts and Humanities’ and in the ‘Social Sciences’ main panels, and in explaining success in those components of the REF score that are more prone to a subjective assessment (environment and impact), rather than in those more ‘objectively’ measurable (quality of the scholarly publications). This result might suggest the need to adjust the design of the assessment exercise in the future.
Abowd, J. M., F. Kramarz, and D. N. Margolis (1999), “High wage workers and high wage firms”, Econometrica 67: 251–333.
De Fraja, G., G. Facchini and J. Gathergood (2016), “How much is that star in the window? Professorial salaries and research performance in UK universities”, Nottingham School of Economics GEP Working Paper 16/13.
Hildreth, A. K. G. and A. J. Oswald (1997), “Rent-sharing and wages: Evidence from company and establishment panels”, Journal of Labor Economics 15: 318–337.
Nickell, S., J. Vainiomaki, and S. Wadhwani (1994), “Wages and product market power”, Economica 61: 457–473.
Nickell, S. and S. Wadhwani (1990), “Insider forces and wage determination”, The Economic Journal 100: 496–509.