Academic salaries in the UK and US.

Stevens, Philip Andrew

We examine the wages of graduates inside and outside of academe in both the UK and US. We find that in both the UK and the US an average graduate working in the HE sector would earn less over his or her lifetime than graduates working in non-academic sectors. The largest disparity occurs throughout the earlier and middle career period and so if people discount their future earnings, the difference will be even greater than these figures suggest. Academics in the UK earn less than academics in the US at all ages. This difference cannot be explained by differences in observable characteristics such as age, gender or ethnicity. This leads us to conclude that the differences in UK and US academic wages are unlikely to be due to differences in the academics themselves, but rather to differences in labour markets generally and in systems of higher education between the two countries, which suggests that there is a strong pay incentive for academics to migrate from the UK to the US.

I. Introduction

There are currently serious concerns about the recruitment and retention of higher education (HE) academics. As with other public sector jobs, the HE sector is perceived to pay lower wages than those available in the private sector (and other areas of the public sector). The concern over labour shortages in the sector has increased in recent years, particularly in areas such as business, economics and engineering. When one takes into account the present government's aims for the expansion of HE, the problem of labour shortages is likely only to get worse.

The education sector is an international one. One quarter of academic staff moves within the HE sector are to higher education institutions (HEIs) outside the UK (Stevens, 2004). Much of this movement is made up of non-UK nationals (63 per cent of academics who move to an HEI abroad are non-UK nationals). This is because non-UK academics working in UK HEIs are more than three times as likely to move to HEIs in other countries as UK nationals. Nevertheless, the US is by far the most popular destination for UK academic migrants, representing the destination of over one third of staff who move to an HEI abroad. (1)

The subject of this paper is differences in earnings. Obviously, pay is only one of many factors which influence the decision to enter and remain in academe and to move between academic institutions. It may be expected that international moves have similar determinants to intra-national moves: academics may move, for example, because they seek promotion, higher pay, a more highly ranked institution or a better quality of life. However, international moves are more problematic because they entail more unknowns, both about the employment market and about non-work aspects. At the same time, there are well-trod career paths, entailing permanent and temporary moves abroad.

We begin our study by reviewing some evidence regarding recruitment and retention difficulties in UK HE and previous international comparisons of academic salaries. The latter section is brief because previous work is scanty.

2. Evidence of recruitment and retention difficulties

Recruitment and retention difficulties for academic staff in higher education appear to be recent phenomena. Certainly, the literature pointed to few concerns in the UK about the ability of universities to recruit and retain academic staff during the 1970s and 1980s (HEFCE, 2003), (2) with the exception of concern about a potential retirement crisis. The issue became more prominent during the 1990s and evidence, based on the difficulties reported by university human resource specialists, suggests that difficulties have grown over the last decade.

With respect to retirement, concern that the age profile of academics in many disciplines was sharply skewed towards those close to retirement was causing concern in the early 1980s and the 1990s (UGC, 1984; Keep, Storey and Sisson, 1996). However, the age profile of academics is generally similar to that of the workforce as a whole (PREST, 2000), although analysis of the age structures of departments shows considerable variation, with the proportion of staff set to retire in the next five years ranging from under 10 per cent to over 25 per cent (UCEA, 2002).

In more recent years, evidence has emerged of HEIs experiencing difficulties recruiting and retaining academic staff (see, for example, Bett, 1999 and HEFCE, 2003). As might be expected, difficulties are not uniform and vary by institution, location, subject, grade and contract, and also vary by type of individual. At the end of the 1980s the only area where there were major difficulties was engineering and technology (Pearson et al., 1990). Even by 1998, there were still not widespread problems with recruitment and retention, although the number of areas in which difficulties existed had increased (Bett, 1999). Problematic areas included recruitment and retention in business subjects, IT, electronic engineering and some rarer specialisms, and in the recruitment of academics with professional experience (e.g. in law, health studies and teaching). (3) Other areas suffering problems with retention included researchers and teaching staff on fixed-term contracts (Bett, 1999). Recruitment problems were sharpest in the old universities, in the South-East (excluding London) and in the West Midlands.

However, over the next few years reported recruitment and retention problems intensified and became more widespread. Based on HEI human resource specialists reports of recruitment and retention, UCEA (2002) pointed to a considerable increase in recruitment and retention difficulties for both academic and support staff since 1998. Around one in five institutions reported experiencing difficulties filling academic positions in 2001 (18.1 per cent), compared to one in twenty in 1998 (5.8 per cent). The subject areas causing the most problems were similar to those in the Bett Report, but had expanded. They now included computing/IT, business subjects (accountancy/finance, business/ management, law and economics), engineering, science subjects (biological sciences, chemistry and physics), nursing/midwifery and professions allied to medicine and education. The increase in retention difficulties was of a similar magnitude, with 7.6 per cent reporting retention difficulties 'most of the time' or more in 2001, compared to 2.2 per cent in 1998. Institutions reported that lecturers were the most difficult to recruit, with almost 60 of all institutions reporting difficulties. The percentage reporting difficulties in recruiting lecturers was similar in new and old universities, but lower in colleges. More old universities than new found it difficult to recruit professors and research assistants.

Recruitment and retention problems were particularly acute in areas which had to compete with the private sector, such as law, IT and engineering (UCEA, 2002). However, the same was true in areas competing with other public sector jobs with higher pay, such as subjects allied to medicine (ibid). The result of this was that HEIs reported difficulties attracting many candidates and those that they did attract were often not of the requisite quality. Moreover, they also reported that it was difficult to recruit good young academic staff as a result of low starting salaries (ibid).

There are many similarities between the academic labour market and that for school teachers. Certainly, Dolton (1990) found that, even after accounting for education, race, gender and social class, both the level and relative wages in teaching influence the relative supply of teachers. He also found that, as with HE academics, the supply varies by subject. More recently, Chevalier, Dolton and McIntosh (2002) also found relative wages to be important. Dolton and van der Klaauw (1999) estimated a competing risks model that discriminated between movements from teaching to the non-teaching sector and those to the non-working sector. They found that higher opportunity wages increase the tendency among teachers to switch careers, but have no effect on exit to non-work. A similar analysis by reason for leaving, which discriminated between voluntary and involuntary exits, (4) also found an effect of opportunity wages on the former but not on the latter.

Nevertheless, it should be noted that there are also a number of differences between the markets for school teachers and HE academics (Johnes, 1990). Since most academics have at least one postgraduate degree, the training required to become an academic is lengthier than that to become a school teacher. Other differences include the fact that there are proportionally many fewer women in academe and higher demands placed on staff by changes in the staff/students ratio (numbers of pupils in much of primary and secondary education have remained relatively stable whereas numbers of students in HE have increased greatly).

3. Previous international studies of academic wages

There have been very few international studies of academic wages This is unusual when one considers that the university, sector labour market is one of the few that has a truly international dimension. One reason for the lack of international studies may be the difficulty in obtaining data from the HE sector The international higher education sector is a heterogeneous one, being made up of private and public sector organisations. In many countries it is a combination of the two. This has meant that the few existing studies have been based on earnings data from a small number of HEIs in different countries in order to obtain the required data on earnings. In this study we bypass this problem by using national survey data.

The Association of Commonwealth Universities (ACU) has undertaken a number of surveys of academic staff salaries and benefits in commonwealth countries. Provan (2001) compared academics in universities in six countries: Australia, Canada, New Zealand, the UK, South Africa and Singapore. The real salaries (converted using World Bank purchasing power parity series) of UK academics were found to compare poorly with Australia, New Zealand, South Africa and Singapore (but not Canada). More recently, Maxwell and Murphy (2003) expanded the set of countries to include Malaysia, with different results. Real UK academic salaries were found to be more similar to those in Australia, Canada, New Zealand and South Africa, but still lower than those in Singapore. However, the problem with using Provan (2001) as a guide to policy on the influences of international disparities in academic salaries on recruitment and retention in UK higher education is that it excludes two major destinations of UK academics: the US and mainland Europe (Stevens, 2004). Moreover, the method used by Maxwell and Murphy (2003) to account for differences in purchasing power in the different countries--the Economist's 'Big Mac Index'--is problematic. (5) This index--created by comparing the prices of McDonalds' hamburgers in different countries--is a long way short of an appropriate measure of purchasing power. Indeed, its use appears to be a step back from the association's earlier work (Provan, 2001), which used indices of purchasing power based on more representative baskets of commodities to calculate. Thus the difference in their results with those of Provan (2001) may merely be due to the change in the conversion rates, as the authors acknowledge.

Stevens (2004) used data from labour force surveys to compare a number of indicators of earnings across nine countries. These countries were selected to include the main English-speaking countries to which UK academics move (the US and Australia), together with other English-speaking nations (New Zealand and Canada) and three European countries (Denmark, France and Sweden). The results of this study were that, overall, UK academics earn less in real terms than those in the US and Canada; they earn an amount similar to those in Denmark and France, and more than those in Australia, New Zealand, Japan and Sweden. The UK academic earnings distribution was found to lie somewhat in the middle of the sample of countries, being more dispersed than those in Nordic countries, but less so than in the US. It is the comparison with the US that is perhaps the most important for this study. It was found that it is not merely the higher real salaries that set academic staff in the US apart; the distribution is much more dispersed. Although academics at the lower end of the scale in the UK and US earn similar amounts, the median worker in the US earns more than three-quarters of UK academics. The upper reaches of US academics earn far more than their UK counterparts. From this it was concluded that in the US there is a premier league of academic high flyers that does not exist in the UK.

4. Estimation and data

Our analysis is based on the UK Labour Force Survey Spring 2001 to Winter 2001 and the US Current Population Survey 2001. In order to remove any potential bias to our results, we focus our analysis on the full-time workers in the graduate sector of the economy (i.e. on those individuals with at least a bachelors degree) (6) since this sector of the labour market reflects the choices and opportunities open to potential and actual academics.

In order to account for the differences in academic and non-academic earnings, we model the log of weekly wages (w) as dependent upon whether the individual is an academic in addition to a quadratic in age, whether the individual only has a first degree (i.e. the baseline is a higher degree) (degree) and a vector of control variables (C) to account for differences between men and women, ethnic groups etc, (7) that is: (8)

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

It may be the case that there are important differences between the structure and profile of lifetime earnings of men and women, or academics and non-academics. Therefore, we include interaction terms in our initial specification to account for this. In the results presented below we have excluded those additional interaction terms that were not found to be statistically significant. We also included double interactions to see if there were more complex interactions, but none were found to be significant. The full set of variables used in the analysis is set out in table 1.

One issue of potential concern is the fact that the academic variable may be endogenous. The endogeneity of occupational choice has been the subject of much research. Individuals may make the choice of becoming an academic dependent upon their expectation of future earnings. School teachers, for example, tend to have lower grades on average than other graduates (Chevalier et al., 2002; Nickell and Quintini, 2002). This is less likely to be the case for academics in higher education than school teachers, but must be borne in mind when one considers the results. Indeed, one suspects that academics will not have a lower unmeasured ability than all other graduates; certainly they are on average likely to have a higher class of degree. (9) If this is the case, the estimate of [[alpha].sub.1] will be downward-biased. A related issue is the degree variable. As can be seen from table 1, academics are more likely than other graduates to have a higher degree. Since we find the return to graduates without a higher degree to be lower, ceteris paribus, one must bear this in mind when considering the size of the 'academic wage gap' suggested by [[alpha].sub.1]. In our consideration of the results, we take this into account by considering the average academic and non-academic (i.e. taking the mean values of the other variables for the relevant population) when comparing earnings. (10)

Because of the potential for our estimates to be biased because of participation decisions, we employ a 'Heckman correction/selection' methodology in estimating wages (Heckman, 1976). In this model (1) is modified to account for the fact that the dependent variable in the earnings regression is only observed if a secondary inequality is satisfied (the 'selection equation'). That is, the dependent variable in equation (1) for individual i is observed only if

(2) [gamma][z.sub.i] + [v.sub.i] > 0

where z is a set of variables that affect participation (11) and [gamma] is a set of parameters to be estimated, the error term v ~ N(0,1) and corr([epsilon], v) = [rho].

When [rho] [not equal to] 0, a standard regression of equation (1) will yield biased results. The Heckman selection model provides consistent, asymptotically efficient estimates for all the parameters in such models. In the results, in addition to reporting [rho], we report the Wald test of independence of the selection and earnings equations, i.e. the likelihood ratio test that [rho] = 0.

We estimate (1) using both OLS and a Heckman-selection method with data from the UK Labour Force Survey Spring 2001 to Winter 2001 and the US Current Population Survey 2001. The results of the selection equations are included in the appendix.

The primary data issue is how to identify higher education academics. This was done using information on occupation and industry. Individuals were considered to be higher education academics if they were in an appropriate occupation (e.g. lecturer or researcher) and working in the higher education sector (although in the US it is not possible to separate teachers and researchers at universities from those at state colleges, who also conduct teaching undertaken in the FE sector in the UK). One issue to note is that it is possible that pro-vice chancellors and deans may be excluded from our analysis if they were coded as managerial occupations. This will only take place if the majority of their time is spent in managerial tasks. Moreover, the number of such staff is very small.

5. Results

5.1 UK

The results of the estimation of our model in the UK are presented in table 2. We can see that wages increase over an individual's lifetime, although the effect is non-linear--they do so at a decreasing rate. This can be seen more clearly when we consider the predicted lifetime earnings profiles below. Academics have a different lifetime earnings profile to non-academic graduates, with their earnings starting from a lower point, but increasing faster over their working life, as evidenced by the positive coefficient on academicxage. The negative coefficient on femalexage suggests that women's wage profile, on the other hand, is flatter than men's, increasing at a slower rate. The negative coefficient on degree suggests that graduates without a higher degree earn less than those with them once one accounts for whether the individual is an academic. This suggests that the result that is often found of a negative return to higher degrees (e.g. O'Mahony and Stevens, 2004) is likely to be due to the occupational choices made by these individuals, rather than negative within-occupation returns indicative of individuals with lower unmeasured ability undertaking graduate study or of a negative amount of human capital being imparted by graduate instruction.

Once we account for age, qualifications and ethnicity, UK academics earn less than non-academics, ceteris paribus. However, one has to be careful to interpret the significant negative coefficient on academic in the wage equation as evidence of a ceteris paribus earnings gap on its own because of the non-linear nature of the equation (i.e. because of the inclusion of the academicxage term) and possibly because of the endogeneity of the academic term. Once we account for non-linearity, we see that the wage an individual can earn as an academic will be below that which they could earn elsewhere until the end of their career (chart 1). The experiences of men and women working in higher education are different. Women's wages generally fall relative to men over their working life. This may be due to the effect of periods out of the labour force on experience or because there is inequality in promotion behaviour. However, the significant positive coefficient on the femalexacademic term suggests that female academics fare well relative to their male colleagues within academe by comparison to those in non-academic employment.

[GRAPHIC OMITTED]

Elsewhere in this issue, Dolton and Chung (2004) calculate the rate of return on career choice (RRCC) of being a school teacher relative to other occupations. If we apply this method to academics and use the wages of other graduates, as the value of the alternative to (or 'cost' of) working in academe, controlling for personal characteristics, we find that the RRCC of working in academe is -0.78 per cent for men and -0.83 per cent for women. Note that these figures are based on taking the working life from 25 onwards and controlling for whether or not the individual has a doctorate. If we consider the cost of a PhD (which is increasingly seen as a prerequisite for working in academe) as an additional cost (both in terms of fees and foregone earnings), these figures will be even lower.

We can get a clearer picture of the pattern of earnings by considering the lifetime wage profiles of our average workers by age, presented in chart 1. This figure shows the annual wage of average academics and nonacademics alike. (12) Academics earn less than nonacademics at almost all ages. At the age of thirty, the average academic earns 72 per cent of the average graduate outside academe. At forty this figure is 79 per cent and at fifty it is 86 per cent. Academics achieve parity with non-academics only at the very end of their working life. It is important to note that this difference is not due to differences in personal characteristics. Although these lines represent the average academic and non-academic graduate, the effect of differences in characteristics is negligible: using our estimates, if we plotted a line to show what an average academic would earn outside of academe, it would be indistinguishable from the lifetime wage profile for non-academics. Therefore, this difference represents the pay cut a new academic would have to suffer entering from outside of the HE sector and the pay premium an academic would be expected to achieve if they left the sector. (13)

It is important to note that, not only are the lifetime earnings of academics lower than that of non-academics, but that the disparity is greater throughout the earlier and middle career period. Because of this, if people discount their future earnings, the difference between academic and non-academic earnings is greater than these figures suggest. These differences may, therefore, have a greater impact upon recruitment and retention of young staff.

5.2 US

The results of our estimation for the US are presented in table 3. The earnings profile of US academics is similar to those in the UK, although it is flatter (chart 2). The average US academic's earnings continues to increase over his/her entire working life, although the rate at which these earnings increase slows over his/her career. Unlike in the UK, US academic wages reach a plateau at the end of their career.

[GRAPHIC OMITTED]

A question of particular interest to this study is what UK academics could--or believe that they could--earn in the US. Academics in the US earn more at all ages (table 4 and chart 3). However, it is important to note that the US data include individuals working in the US state colleges, who undertake teaching similar to UK FE colleges and so this comparison may actually understate the differences in wages.

[GRAPHIC OMITTED]

If the US figure does indeed represent what they could earn in the US, this will provide an incentive to UK academics. However, we must first discount the possibility that the difference in earnings represents differences in the personal characteristics of UK and US academics.

When we account for differences in personal characteristics and see what the average UK academic would earn in the US, (14) we see that the lifetime wage profiles of the average UK and US academic are almost identical. This leads us to conclude that the differences in UK and US academic wages highlighted in table 4 are unlikely to be due to differences in the academics themselves, but rather to labour markets generally and to systems of higher education in the two countries. This suggests that there is a strong pay incentive for academics to migrate from the UK to the US.

Once we adjust for the minor differences in personal characteristics, we find that the RRCC of working in UK academe compared to the US, is -0.23 per cent for men and -0.63 per cent for women. This figure is smaller than the difference between academic and nonacademic graduates in the UK noted above, implying that UK academics would be better off choosing nonacademic careers in the UK than academic careers in the US, ceteris paribus. The RRCC for a potential UK academic working in the US academic sector relative to working in the UK non-graduate market is -0.54 per cent for men and -0.21 per cent for women. For men, this negative return is two-thirds of that found when one compares UK academe to the UK non-academic sector; for women this deficit is one-quarter. If the non-pecuniary benefits of HE are the same for men and women, these figures suggest that the incentive to leave academe altogether are higher for men than women, ceteris paribus. If the non-pecuniary advantages of academe are even higher for women--because it is easier to combine work with family commitments, for example--this difference in incentives would be even greater. However, the differences in the non-pecuniary rewards of working in the US and UK HE sectors are likely to be marginal by comparison. (15)

6. Conclusion

In both the UK and the US an average graduate working in the HE sector would earn less over his or her lifetime than graduates working in non-academic sectors. The largest disparity occurs throughout the earlier and middle career period. Thus, if people discount their future earnings, the difference between academic and non-academic earnings is even greater than charts 1 and 2 suggest. It is likely therefore that these differences will have a greater impact upon the recruitment and retention of young staff. However, it is unlikely that the balance between the pecuniary and non-pecuniary aspects will remain constant over an academic's working life. Young academics may be willing to accept lower earnings as academics because of the other attractions of academic life, such as the ability to conduct research, flexible working hours, leave entitlement and the ability to attend foreign conferences. In later life, family commitments may increase the weight of salary relative to non-pecuniary benefits in the individual's utility function. Moreover, if the more negative aspects of academic life--bureaucracy and administrative tasks, for example--increase over their career, a higher wage will be required to compensate (Rosen, 1986).

Academics in the UK earn less than academics in the US at all ages. The difference between the wages of UK and US academics is particularly pronounced at the later stages of the academic's working life. This may reflect the fact that HEIs in the US are less constrained in their pay scales and are able to reward academic high-flyers in order to attract and retain them. It is certainly the case that there is much higher variation in US academic salaries than those in the UK, both across grades and across subjects (Machin and Oswald, 2000; Stevens, 2004). The difference between the wages earned by UK and US academics cannot be explained by differences in observable characteristics such as age, gender or ethnicity. This leads us to conclude that the differences in UK and US academic wages are unlikely to be due to differences in the academics themselves, but rather to labour markets generally and to systems of higher education between the two countries.

Data Appendix

UK

The UK data come from the four quarters of the 2001 Labour Force Survey (LFS). Academics are defined using the three-digit 1992 Standard Industrial Classification (SIC) and the three-digit 2000 Standard Occupational Classification (SOC). Academics are defined as being in SOC codes 2311, 2319, 2321, 2322 and 2329. The main SOC unit group for academic staff is 2311, which refers to 'higher education teaching professionals'. Other teaching staff from SOC unit group 2319 'teaching professionals not elsewhere specified' if they were also coded as working in SIC class 80.3 'higher education'. Also included as academics were individuals who were in SOC minor group 232 "research professionals, i.e. unit groups 2321 'scientific researchers', 2322 'social science researchers' and 2329 'researchers not elsewhere specified' if they were also coded as working in SIC class 80.3 'higher education'. The two measures of salaries used are gross weekly wages (variable grsswk).

US

The US data come from the Merged Outgoing Rotation Group of the 2001 Current Population Survey. The industry an individual worked in was determined using the three-digit 1980 Census of Population Industry Classification (CPIC). The appropriate CPIC code was 822 'Colleges and universities'. The individual's occupation was determined using the four-digit 1980 Census of Population Occupation Classification (CPOC). The majority of academics were identified as individuals whose occupation was CPOC code 221 'teachers, postsecondary' and who's CPIC industry code was 822 'Colleges and universities'. Unfortunately the CPOC does not have an occupational group for researchers per se. Therefore, these individuals were identified using a number of scientific occupations (both in the social and natural sciences) in combination with the CPIC code. Unfortunately it is impossible to differentiate between teachers and researchers teaching in the HE sector and the state colleges. The wage data used are 'earnings per week' (earnwke).

Table A. 1 Selection equations

 UK US

academic -0.541 * -0.777 ***
 (0.299) (0.256)
female x academic -0.085 -0.169
 (0.136) (0.118)
age 0.068 *** 0.1111 ***
 (0.010) (0.005)
age (2)/100 -0.001 *** -0.002 ***
 (0.000) (0.000)
female x age 0.005 *** 0.008 ***
 (0.001) (0.001)
academic x age 0.024 *** 0.032 ***
 (0.006) (0.005)
nonwhite -0.555 *** -0.043 *
 (0.048) (0.026)
female x nonwhite 0.028 0.063 *
 (0.073) (0.035)
academic x nonwhite -0.145 0.254
 (0.225) (0.199)
degree -0.120 *** -0.003
 (0.030) (0.017)
female x degree 0.000 -0.162 ***
 (0.046) (0.024)
academic x degree -0.027 -0.299 ***
 (0.144) (0.151)
married 0.169 *** 0.188 ***
 (0.035) (0.022)
female x married -0.111 ** -0.465 ***
 (0.051) (0.029)
children -0.008 -0.023 **
 (0.015) (0.009)
female x children -0.174 *** -0.218 ***
 (0.024) (0.012)
q1 -0.154 ***
 (0.030)
q2 -0.018
 (0.030)
q3 0.004
 (0.030)
Constant -0.595 *** -1.162 ***
 (0.191) (0.106)

Uncensored Obs. 14,785 50,795
Censored Obs. 9,164 36,145

Notes: Selection equation for whether individual is included in wage
equation. Standard errors in parenthesis. * significant at 10 per
cent; ** significant at 5 per cent; *** significant at I per cent.
See table 1 for variable definitions.

Table 1. Variables used in empirical analysis

 Non-
 academics

Variable name UK US

Ln_Wage Log gross weekly earnings 6.32 6.82
academic HE academic 0 0
female Female 0.40 0.45
degree Only has first degree (a) 0.73 0.67
nonwhite Non-white 0.07 0.14
married Married 0.58 0.68
children No. of dependent children 0.69 0.75
age Age 38.8 41.6
femaleXvariable Female interacted with variable
academicXvariable Academic interacted with variable

UK-only variables

ql Spring 2001 survey 0.23
q2 Summer 2001 survey 0.25
q3 Autumn 2001 survey 0.26

 Academics Total

Variable name UK US UK US

Ln_Wage 6.32 6.87 6.32 6.83
academic 1 1 0.04 0.03
female 0.39 0.38 0.40 0.45
degree 0.27 0.13 0.71 0.66
nonwhite 0.05 0.13 0.07 0.14
married 0.59 0.71 0.58 0.68
children 0.66 0.61 0.69 0.74
age 43.0 45.6 39.0 41.7
femaleXvariable
academicXvariable

UK-only variables

ql 0.23 0.23
q2 0.23 0.24
q3 0.27 0.26

Notes: (a) Note sample does not include individuals with only
a 2-year college degree. Data relate to full-time staff only.

Table 2. UK results

 OLS Heckman

Academic -0.625 *** -0.574 ***
 (0.112) (0.114)
femalexacademic 0.135 *** 0.141 ***
 (0.050) (0.051)
Female 0.034 0.036
 (0.045) (0.044)
age 0.096 *** 0.090 ***
 (0.004) (0.005)
[age.sup.2]/100 -0.001 *** -0.001 ***
 (0.000) (0.000)
femalexage -0.006 *** -0.006 ***
 (0.001) (0.001)
academicxage 0.011 *** 0.009 ***
 (0.002) (0.002)
Nonwhite -0.019 0.030
 (0.025) (0.028)
femalexnonwhite -0.020 -0.020
 (0.038) (0.038)
academicxnonwhite -0.049 -0.046
 (0.104) (0.106)
Degree -0.075 *** -0.064 ***
 (0.014) (0.014)
femalexdegree 0.006 0.006
 (0.021) (0.022)
academicxdegree 0.067 0.069
 (0.054) (0.055)
q1 -0.042 *** -0.029 **
 (0.013) (0.014)
q2 -0.023 * -0.021
 (0.013) (0.013)
q3 -0.008 -0.008
 (0.013) (0.013)
Constant 4.306 *** 4.461 ***
 (0.090) (0.096)

Observations 9,164 14,785
Adjusted [R.sup.2] 0.150
F/[chi] 101.9 *** 1,597.9 ***
[rho] -0.321
Wald 13.32 ***

Notes: Standard errors in parenthesis. * significant
at 10 per cent; ** significant at 5 per cent;
*** significant at 1 per cent. See table 1
for variable definitions.

Table 3. US results

 OLS Heckman

Academic -0.706 *** -0.604 ***
 (0.081) (0.085)
female x academic 0.052 0.059 *
 (0.034) (0.035)
Female -0.114 *** -0.069 ***
 (0.026) (0.027)
Age 0.058 *** 0.047 ***
 (0.002) (0.003)
age (2)/100 -0.001 *** -0.000 ***
 (0.000) (0.000)
female x age -0.004 *** -0.004 ***
 (0.001) (0.001)
academic x age 0.012 *** 0.009 ***
 (0.002) (0.002)
Nonwhite -0.129 *** -0.124 ***
 (0.011) (0.011)
female x nonwhite 0.086 *** 0.077 ***
 (0.015) (0.016)
academic x nonwhite 0.050 0.035
 (0.049) (0.051)
Degree -0.182 *** -0.181 ***
 (0.008) (0.008)
female x degree 0.035 *** 0.015
 (0.012) (0.012)
academic x degree 0.023 0.011
 (0.050) (0.052)
Constant 5.555 *** 5.810 ***
 (0.051) (0.061)
Observations 36,145 50,795
Adjusted [R.sup.2] 0.134
F/[CHI] 432.05 *** 5164.56 ***
[rho] -0.447
Wald 45.52 ***

Notes: Standard errors in parenthesis * significant at 10 per cent;
** significant at 5 per cent: *** significant at 1 per cent. See
table 1 for variable definitions.

Table 4. Comparative actual academic earnings,
[pounds sterling] PPP

 UK

 Men Women

Under 30 19,744 [pounds sterling] 20,386 [pounds sterling]
30-39 28,277 [pounds sterling] 26,094 [pounds sterling]
40-49 34,802 [pounds sterling] 28,455 [pounds sterling]
50-59 37,290 [pounds sterling] 31,330 [pounds sterling]
Over 60 41,433 [pounds sterling] 31,824 [pounds sterling]
Total 32,763 [pounds sterling] 26,821 [pounds sterling]

 US

 Men Women

Under 30 21,730 [pounds sterling] 18,997 [pounds sterling]
30-39 38,331 [pounds sterling] 28,891 [pounds sterling]
40-49 44,535 [pounds sterling] 35,611 [pounds sterling]
50-59 47,461 [pounds sterling] 38,354 [pounds sterling]
Over 60 52,440 [pounds sterling] 39,252 [pounds sterling]
Total 43,225 [pounds sterling] 33,580 [pounds sterling]

Note: Real salaries are converted using PPP rates from OECD Main
Economic Indicators.

NOTES

(1) One third of UK 'academic migrants' move to the US; for non-UK academics the figure is one-fifth.

(2) The story is slightly different in the US, where predictions of staff shortages from the 1970s (e.g. Freeman, 1971, 1975; Stapleton, 1989) led to an increase in interest in academic labour markets (see Ehrenberg, 2003, for a summary). In the US, however, the main thrust of the enquiry was the shortages of PhD students rather than issues pertaining to academic staff themselves (Carster, 1976; Ehrenberg, 1991, 1992)

(3) These figures come from the staff survey conducted for the Independent Review of Higher Education Pay and Conditions. This was the result of surveys sent to heads of personnel of all 178 UK HEIs. The overall response rate was 77 per cent, giving information pertaining to 246,027 of the estimated 300,780 staff in the UK, including 110,070 of the estimated 133,977 academic, clinical academic and research staff in UK HEIs.

(4) Which included exits for family reasons.

(5) This method is also used by OnE and Mitchell (2000), who include institutions from the US, with similar results (i.e. UK academics salaries are the lowest in their sample).

(6) In the US. this excludes those whose highest qualification is a two-year degree.

(7) For the UK estimation we also include a dummy variable for each of the first three quarters because the data is made up of four appended quarters of data to ensure a large enough sample size.

(8) Unfortunately we cannot account for degree subject, because there is no information on this for the US.

(9) For example, a recent survey of ten institutions of higher education conducted by NIESR found that three-quarters of academic staff had at least an upper second or equivalent.

(10) Running the regressions excluding the academic term provides almost identical coefficients for the other terms (including the degree term).

(11) These include the variables in the wage equation, plus variables to account for whether the individual is married, the number of dependent children, and a number of additional interactions. The selection equations are set out in the appendix to this paper.

(12) That is, the values of the variables in table 2 are set to their mean values for academics and non-academics and the predicted value of the wage is calculated at each age.

(13) It may also be the case that there are certain unobservable differences between academics and non-academic graduates, although it is not clear what these might be. In the returns to education literature, it is argued that the most important omitted variable is 'ability'. It seems unlikely that academics have lower ability than graduates as a whole. A more likely candidate is a taste for certain non-pecuniary aspects of academic life that are unavailable elsewhere.

(14) i.e. given the coefficients in the US earnings regression

(15) For a classic statement of the equalising differences model, see Rosen (1986).

REFERENCES

Bett, M. (1999), Independent Review of Higher Education Pay and Conditions, Norwich, Stationery Office.

Cartter, A.M. (1976), Ph.D.'s and the Academic Labor Market, New York, McGraw Hill

Chevalier, A, Dolton, P. and McIntosh, S. (2002), "Recruitment and retaining teachers in the UK: an analysis of graduate occupation choice from the 1960s to the 1990s', CEE Discussion paper no 21.

Dolton. P.J. (1990), The economics of UK teacher supply', Economic Journal, 100, S91-S104.

Dolton, P. and Chung, T.P. (2004). 'The rate of return to reaching: how does it compare to other graduate jobs?', National Institute Economic Review, 190, October.

Dolton, P. and van der Klaaw, W. (1999), 'The turnover of teachers: a competing risks explanation," Review of Economics and Statistics, 81, pp. 543-52.

Ehrenberg, R.G. (1991), 'Academic labor supply', in Clotfelter, Ehrenberg Getz and Siegfried (eds), Economic Challenges in Higher Education, Chicago, University of Chicago Press.

--(1992), 'The flow of new doctorates', Journal of Economic Literature, 30, pp. 830-75.

--(2003), 'Studying ourselves: the academic labor market', Journal of Labor Economics, 21, 2, pp. 267-87.

Freeman, R.J. (1971), The Market for College-Trained Manpower, Cambridge, Mass., Harvard University Press.

--(1975), "Supply and salary adjustments to the changing science manpower market: physics 1948-1973', American Economic Review, 65, pp. 287-311.

Heckman, J. (1976), 'The common structure of statistical models of truncation, sample selection, and limited dependent variables and a simple estimator for such models', The Annals of Economic and Social Measurement, 5, pp. 475-92.

HEFCE (2003), Appointment. Retention and Promotion of Academic Staff in Higher Education Institutions, Report to the HEFCE by the Scottish Council for Research in Education, University of Glasgow and Nottingham Trent University.

Johnes, G. (1990), The Economics of Education, Basingstoke, MacMillan.

Keep, E., Storey. J. and Sisson, K. (1996), 'Managing the employment relationship in higher education: quo vadis?', in Cuthbert, R. (ed.), Working in Higher Education, SRHE/OUP.

Machin, S. and Oswald, A.J. (2000), 'UK economics and the future supply of academic economists', The Economic Journal, 110, pp. 334-49.

Maxwell J. and Murphy, D. (2003), Academic Staff Salaries and Benefits in Seven Commonwealth Countries, 2001-2002, Association of Commonwealth Universities.

Nickell, S.J. and Quintini, B. (2002), 'The consequences of the decline in public sector pay in Britain: a little bit of evidence', Economic Journal, 112, pp. F107-18.

O'Mahony, M., Stevens, P.A. (2004), 'International comparisons of performance in the provision of public services: outcome based measures for education', presentation to NIESR conference on 'Productivity and Performance in the Provision of Public Services' at the British Academy.

Ong, L.L., and Mitchell, J.D. (2000), 'Professors and hamburgers: an international comparison of real academic salaries', Applied Economics, 32, pp. 869-76.

Pearson, R., Buchan. J. Bevan, S., Jackson, C. and Stock, J. (1990), The Recruitment and Retention of University Academic and Academic Related Staff, IMS Paper 157, Brighton.

PREST (Policy Research in Engineering, Science and Technology University of Manchester) (2000), Impact of the Research Assessment Exercise and the Future of Quality Assurance in the Light of Changes in the Research Landscape. Report to the Higher Education Funding Council for England.

Provan, D. (2001), Academic Staff Salaries and Benefits in Six Commonwealth Countries, 2000-2001, Association of Commonwealth Universities.

Rosen, S. (1986), 'The theory of equalizing differences', in Ashenfelter, O. and Layard P.R.G. (eds), Handbook of Labor Economics, I, Amsterdam, North Holland.

Stapleton, D.C. (1989), 'Cohort size and the academic labor market', Journal of Human Resources, 24, pp. 221-52.

Stevens, P.A. (2004), An International Comparison of Academic Wages, Report to the Department for Employment and Skills.

UCEA (2002), Recruitment and retention of staff in UK higher education: A survey and case studies, report commissioned by the HEFCE, SCOP, UCEA and UUK.

UGC (1984),A Strategy for Higher Education into the 1990s, University Grants Committee.

Philip Andrew Stevens *

* National Institute of Economic and Social Research (e-mail: [email protected]). The research reported here was funded by the DfES. Data were supplied by the Higher Education Statistical Agency. Thanks go to Martin Weale, Peter Dolton and participants at the NIESK/DfES/ESRC conference on 'Recent Developments in the UK Graduate Labour Market' for helpful comments, although the usual caveats apply.