Much of the previous literature on the wage experiences of the 'Third Age' (1) in UK regions focused on aggregated data thus disregarding individual level issues. This paper uses microeconometric data (2) to examine hourly net wage patterns adjusting for sex, ethnicity, qualifications, industry and region.

The authors find evidence of differences in net hourly earnings between the sexes for those in the 'Third Age' with men earning £1.70 more. Ethnicity and region prove not to be significant.

Education is significant with possession of a graduate qualification for those in the 'Third Age' resulting in an increase in net hourly earnings of over £2. The authors tentatively suggest that this may be due either to these workers receiving an high economic rent encouraging them to remain in the labour force or that such workers are employed in jobs where productivity does not decline significantly with age.

INTRODUCTION

In Britain there has been a significant decline in the level of economic activity amongst Third Age males, but not females, over the last quarter of a century. However, there remain in this age group a number who wish to continue in employment and earn a wage as an alternative, or an addition, to non-earned benefits. While the nature of the employment opportunities for older workers has been subject to investigation the character, and influences on, the monetary rewards for Third Age workers have not received the same degree of attention.

Source New Earnings Survey 1997, Part F, Table F7.

The Office of National Statistics publishes an annual earnings survey which includes cross sectional, and highly aggregated, data for differing age cohorts from which age-earnings profiles of the character shown in Diagram 1 may be generated, see also Disney (1982, p 136). Such age-earnings profiles indicate peak earnings for differing segments of the labour market are achieved prior to individuals entering the Third Age. A consequence of focusing on aggregated data of this character when considering the earnings experiences of older workers is that potential ethnic, sexual and educational differences across regions is ignored.

The significance of education, for example, on wage levels in the United States has been demonstrated by Murphy and Welch (1992) who noted that for individuals with five years work experience there was, in 1989, a 74% wage differential between those with a college degree compared with those whose highest qualification was an high school diploma. However, the same authors noted the `educational' differential did decline in significance over time, suggesting it fell to under 50% for those with between 26 and 35 years employment experience.

The objective of this paper is to use the micro-economic data for one quarter, June - August 1995, of the Quarterly Labour Force Survey (ONS (1995)s) to consider the relative significance of a number of identifiable socio-economic factors which are thought to influences the wage earnings of full-time employees in the Third Age. A review of the relevant literature will be undertaken in Section 2 and an explanation of the model constructed is given in Section 3. This model is initially used to explain net hourly wage behaviour for the full-time working population as a whole to establish the extent of any discrimination based on age and/or gender after adjusting for ethnicity, educational qualifications, regional location, work experience and other factors and the results are reported in Section 4. In the penultimate Section the model is used with the same variables to focus specifically upon the determination of the actual wages of those in the Third Age category and results are reported in Section 5. The concluding section will examine the results obtained for both all workers and Third Age workers category.

LITERATURE SURVEY

While the underlying assumptions explaining the level of wages that individuals receive are derived from the marginal productivity theory of distribution, subsequent work has often sought to offer more detailed explanations of why wages vary between individuals. As an example there has been a tendency, identified for example by Creedy and Hart (1979) & Elias and Gregory (1993), for the average wage levels of individuals to decline in the latter years of their employment. It is within this context that the contributions of Becker (1975) and Mincer (1974) in developing our understanding of human capital, both general and specific, require viewing. Models derived from the human capital concept (3) appear to offer a partial explanation to the life cycle wage patterns of individuals, who normally experience a rising trend in their earnings followed by a steady decline in the later years of employment.

Although the human capital explanation of life cycle earnings patterns may reflect individual experiences at any one point in time, further observation does indicate that for each generation there is a further tendency for individuals to receive higher real wage levels than was the case for previous generation. This is a feature that Carliner (1982) attributed to the ongoing, and continuous, rise experienced over the long term in productivity levels. Drawing these threads together Phelps Brown (1977) sought to explain the variation over time in individuals' life cycle earnings patterns in terms of both improvements in educational opportunities and parallel changes in occupational structure and industrial activity which resulted in productivity growth. Although not challenging this hypothesis there is extensive evidence, see Organisation for Economic Co-operation and Development (1996), that in parallel with the decline in their earnings there are significant changes in the economic activity of older workers.

Whether the decline in labour force participation influences, or is influenced by, the lower level of potential wages available to older workers has been the subject of a number of studies. Blackaby, Bladen-Hovell and Symonds (1991) and Manning (1994) in their respective studies on the impact of unemployment on inflation both found that the long term unemployed, whom Blackaby et al suggested effectively withdrew from the labour market, exerted little downward pressure on wages. Thus it would appear reasonable to suggest that the decline in labour force participation by older workers in the last twenty years does not, by itself, provide an adequate alternative explanation for the observed decline in the wages of those older workers remaining in employment. However, using US data, a study by Johnson and Neumark (1996) did appear to indicate that the existence of social security, and it is assumed pension income, might encourage workers previously receiving a high wage to either withdraw from employment or seek less demanding employment with corresponding lower wages.

A model may now be constructed of net hourly wages for full-time workers with their industrial sector, their highest qualification level, their age, their sex, their length of employment with their current employer, their ethnic group and their home region as independent variables.

The estimation technique that will be used here is not Ordinary Least Squares (OLS) but is the Tobit approach which is preferable in terms of providing estimates that are not biased by the fact that individuals are unable to receive negative wages (an issue which OLS estimation ignores) and is computationally straightforward. The Tobit technique constructs an independent variable (y^*) which in this case is bounded by 0 and infinity and then estimating

The resulting model is essentially a half-way house between a Probit and an Ordinary Least Squares regression. Typically, the coefficients that are obtained from such a Tobit estimation are larger in absolute value than those from an OLS regression.

We will now turn to a discussion of the expected coefficients of the variables in our model.

Our sample is taken from the June - August 1995 Quarterly Labour Force Survey which consists of over 150,000 individuals. We have removed those who either (a) did not work full-time (which resulted in the loss of over 67% of the sample) and / or (b) did not respond to one or more of the questions which generate the variables we wish to use, for example, almost 60,000 respondents did not give their highest qualification. Thus we are left with a usable sample of 3363 people of whom 543 are aged 50 years or over. Because of the 'snapshot' nature of the data, we feel that we are safe in not taking overall demand conditions into account. A time series or panel data analysis would, of course, have to take this factor into account.

RESULTS FOR ALL FULL-TIME WORKERS

All of the models are significant at the 99.9% level.

The full results are given in Tables I - III but the key findings of this model are (with statistical significance being at the 95% level) :-

• There is a statistically significant difference in the hourly wage rate between men and women of £1.10 per hour (net), ceteris paribus.
• The age dummy variables (for the twenties, thirties, forties and fifties) are significant for the overall model and all save the twenties dummy variable are significant for the male model. They indicate that earnings rise to a peak in the 40s and then decline for both the all workers and male models. However, none are significant for the female model but the indications are of an earnings peak during the 30s. These findings do tie in with Diagram 1 suggesting that the aggregate picture may not be that bad a guide for individuals as might be supposed but, as these results bear out, age and gender do not give the full picture on the determinants of wages received from employment.
• Ethnicity has no effect on the hourly wage rate in any model, ceteris paribus.
• The industry employed in (given the three broad categories) has no effect on the hourly wage rate in any model, ceteris paribus.
• The number of months employed is highly significant in each model with each month employed with the current employer adding ½p to the net hourly wage rate in the male model, almost 1p in the all workers model and almost 2p in the female model, ceteris paribus.
• Possession of a degree (both undergraduate and postgraduate) or of Further Education is significant in all models. Possession of an apprenticeship is significant in both the overall and male models whilst possession of GCSEs (or equivalent) is significant only in the overall model.

For the all workers model, those with a postgraduate qualification (as their highest qualification) earn £6.33 per hour (net) more than those with no qualifications, those with a degree (as their highest qualification) earn £3.55 more per hour (net), those with Further Education (as their highest qualification) earn £1.97 more per hour (net), those with an apprenticeship (as their highest qualification) earn £1.32 more per hour (net) and those with GCSEs (as their highest qualification) earn £1.01 more per hour (net).

For the male workers model, those with a postgraduate qualification (as their highest qualification) earn £6.45 more per hour (net) than those with no qualifications, those with a degree (as their highest qualification) earn £3.60 per hour (net) more, those with Further Education (as their highest qualification) earn £1.84 more per hour (net) and those with an apprenticeship (as their highest qualification) earn £1.36 more per hour (net).

For the female workers model, those with a postgraduate qualification (as their highest qualification) earn £5.24 more per hour (net) than those with no qualifications, those with a degree (as their highest qualification) earn £2.81 more per hour (net) and those with Further Education (as their highest qualification) earn £1.32 more per hour (net).

Thus, it would seem that the value of a postgraduate qualifications, ceteris paribus, is over £1.20 more per hour (net) for men than women and the value of an undergraduate qualification, ceteris paribus, is slightly less than 80p more per hour (net) for men than for women.

These results are after taking the ethnic, gender, industrial and educational effects into account.

• Those living in the Northern Region (excluding Tyne & Wear) or South Yorkshire or East Midlands or the South West or rest of the North West (excluding Merseyside and Greater Manchester) or Wales or Scotland (outside of Strathclyde) earn statistically significantly less than those living in Inner London in all models.
• Those living in Tyne & Wear or the West Midlands (excluding the Metropolitan County) or Greater Manchester earn statistically significantly less than those living in Inner London in the overall model.
• Those living in Yorkshire and Humberside (excluding South or West Yorkshire) or the West Midlands Metropolitan County earn statistically significantly less than those living in Inner London in the overall and Female models.
• Those living in Merseyside or East Anglia earn statistically significantly less than those living in Inner London in the overall and Male models.
• Those living in Outer London earn statistically significantly less than those living in Inner London in the Male model only.
• Those living in West Yorkshire, the rest of the South East (excluding Inner and Outer London) and Strathclyde do not earn statistically significantly less than those living in Inner London in any of the models

To summarise these findings, the regions where we can compare the performance of the sexes are (with differentials in net hourly wages from Inner London):-

So, females with a given education level and length of employment with their current employer in a given region can expect to earn up to £1.42 less per hour (net) than men in the same region with the same education level and length of employment with their current employer.

This model is estimated for both genders and is significant at the 99.9% level overall.

The full results are given in Table V but the key findings are (with statistical significance being at the 95% level) :-

• There is a statistically significant difference in the hourly wage rate between men and women of £1.70 per hour (net), ceteris paribus, with men earning more.
• Ethnicity has no effect on the hourly wage rate in any model, ceteris paribus.
• The industry employed in (given the three broad categories) has no effect on the hourly wage rate, ceteris paribus.
• The number of months employed is highly significant in each model with each additional month employed adding over ½p to the net hourly wage rate for all 'Third Age' workers, ceteris paribus.
• Possession of a degree adds considerably to wages; an undergraduate degree as the highest education level adds £3.45 per hour (net) whereas a postgraduate degree as the highest education level adds £9.23 per hour (net), ceteris paribus.

• Perhaps, somewhat surprisingly, there are no regions where individuals earn significantly less (in a statistical sense) than those living in Inner London. This might be explained by (a) the importance of education in explaining net hourly earnings as shown above and (b) the finding that almost 33% of those in the 50 - 69 years old sample with a graduate qualification or higher live in London or the South East. This ties in with the literature on male activity rates with Collis & Mallier (1996) showing that London and the Rest of the South East had high 'Third Age' male activity rates of over 67%; it appears that in the case of London this may be due to its higher than average graduate population - almost 39% compared to 30% nationally.

The importance of education for the labour market prospects of the 'Third Age' workers is corroborated by an examination of the relationship between the labour market status and the highest educational qualification achieved of the 'Third Age' cohort for the Summer 1995 Quarterly Labour Force Survey (ONS(1995)).

These figures suggest that the probability of a 'Third Age' person with a postgraduate qualification having a full-time job is almost twice as high as the probability of a 'Third Age' person with no qualifications having a full-time job, 55.12% as compared to 27.94%.

CONCLUSIONS

We have found that there is evidence of a statistically significant difference between the sexes that is independent of education, ethnicity, the industry employed in, the length of continuous employment and the region lived in. This difference amounts to £1.10 per hour (net), ceteris paribus, for all workers and £1.70 per hour (net) for those in the 'Third Age'. There is some evidence of a pattern of rising and falling earnings as an employee ages which is also in agreement with the literature.

Perhaps somewhat surprisingly, given traditional neo-classical discrimination analysis, ethnicity has no effect on the net hourly wage rate in any model. However, recent work by the Policy Studies Institute (Modood et al (1997)) bears out this finding pointing to a lack of educational opportunity as being the major determinant of ethnic underachievement in the labour market. This is borne out by the importance of education in this model, with the possession of a degree adding at least £2.80 to net hourly earnings. Again we notice a differential between males and females with the value of a postgraduate qualification, ceteris paribus, is over £1.20 more for men than women and the value of an undergraduate qualification, ceteris paribus, is slightly less than 80p more for men than for women.

One other finding of note, before considering the Regional effects, is that the model does provide evidence that remaining with an employer is rewarded with each additional month employed resulting in an increase in net hourly earnings of between ½p and 2p depending on the model.

The importance of both education and experience in our findings tie in with both the "firm specific" and "general" elements of the human capital approach to wage determination.

The regional findings suggest that females can expect to earn up to £1.42 less than men in the same region with the same education level and length of employment with their current employer. However, when looking at the 'Third Age' there are no such regional differences. This might be explained by the importance of education in explaining net hourly earnings for this group with almost 33% of the 50 - 69 years old sample with a graduate qualification or higher living in London or the South East.

This finding may be indicative of a link between possession of an higher education and labour force participation in the 'Third Age' which is a suggestion strongly reinforced by our noting of a link between educational qualification and the probability of employment for the 'Third Age' cohort. For example, the probability of full-time employment is 20% higher for those 'Third Age' cohort members with either a postgraduate qualification or a further education qualification or an apprenticeship as their highest qualification than it is for someone with no qualifications.

These findings on the 'Third Age' labour force need not be in conflict with that of Johnson and Neumark (1996), who suggested that social security provision may encourage early exit from the labour force, as almost 70% of the UK working population are not graduates and we have found no statistically significant evidence of non-graduate qualifications having an effect on the net hourly wage for the "Third Age".

Finally, we suggest that this link between possession of an higher education and labour force participation in the 'Third Age' may be due to either (a) an higher economic rent being received by this type of worker which encourages them to remain in the labour force and / or (b) such workers being less likely to be employed in jobs which require substantial physical labour thus enabling them not to have such sharply diminishing productivity with age as might be expected for manual workers.

BIBLIOGRAPHY

Becker, G. S. (1975), Human Capital (2nd Edition), University of Chicago Press, Chicago.

Blackaby, D. H., Bladen-Hovell, R. C., and Symonds, E. J. (1991), "Unemployment, Duration and Wage Determination in the UK : Evidence From the FES 1980 - 1986", Oxford Bulletin of Economics and Statistics, Vol. 53 No. 4, pp. 377 - 399.

Blackaby, D. H., Clark, K., Leslie, D. G. and Murphy, P. D. (1997), "The Distribution of Male and Female Earnings 1973-91 : Evidence for Britain", Oxford Economic Papers, Vol. 49 No. 2, pp. 256 - 272.

Carliner, G. (1982), "The Wages of Older Men", The Journal of Human Resources, Vol. 17 No. 1, pp. 25 - 38.

Collis, C. and Mallier, T. (1996), "Third Age Male Activity Rates in Britain and its Regions", Regional Studies, Vol. 30 No. 8, pp. 803 - 809.

Creedy, J., and Hart, P. E. (1979), "Age and the distribution of earnings", Economic Journal, Vol. 89, pp. 280 - 293.

Department of Social Security (1994), Security, Equality, Choice : The Future for Pensions, HMSO, London.

Disney, R. (1982), "The Structure of Pay" in Creedy, J., and Thomas, B. (eds), The Economics Of Labour, Butterworths, London.

Elias, P. and Gregory, M. (1993), The Changing Structure of Occupation and Earnings in Great Britain, 1975 - 1990 : An Analysis Based on the New Earnings Survey Panel Data Set, Research Series No. 27, Department of Employment, London.

Harris, R. I. D. (1989), "Regional earnings differentials in Great Britain 1970 - 1982", in Thompson, A. and Gregory, M. (Ed.), A Portrait of Pay in the 1970s : Using the New Earnings Survey, Oxford University Press, Oxford.

Johnson, R. W. and Neumark, D. (1996) "Wage declines among older men", Review of Economics and Statistics, Vol. 78 No. 4, pp. 740 - 748.

Manning, N. (1994), "Are Higher Long-Term Unemployment Rates Associated With Lower Earnings?", Oxford Bulletin of Economics and Statistics, Vol. 56 No. 4, pp. 383 - 397.

Mincer, J. (1974), Schooling Experience and Earnings, National Bureau of Economic Research : Columbia University Press, New York.

Modood, T., Berthoud R., Lakey, J., Nazroo, J., Smith, P., Virdee, S., Beishon, S. (1997) Ethnic Minorities in Britain: Diversity and Disadvantage (The Fourth National Survey of Ethnic Minorities), Policy Studies Institute, London.

Murphy, K. M. and Welch, F. (1992), "The Structure of Wages", The Quarterly Journal of Economics, Vol. 107 No. 1, pp. 285 - 325.

Organisation for Economic Co-operation and Development (1996), Labour Force Statistics, OECD, Paris.

Office for National Statistics (1995), June - August 1995 Quarterly Labour Force Survey.

Phelps Brown, H., (1977), The Inequality Of Pay, Oxford University Press, Oxford.

Sloane, P. J. (1985), "Discrimination in the Labour Market" in D. Carline, C. A. Pissarides, W. S. Siebert and P. J. Sloane (Ed.), Labour Economics, Longman, London.

FOOTNOTES

1. Defined as those aged between 50 and 69 years inclusive as the payment of private pension can be taken in the UK from 50 years of age onwards (Department of Social Security (1994) and the taking of the state pension may be delayed until 70 years of age.
2. Material from the June - August 1995 Quarterly Labour Force Survey is Crown Copyright; has been made available from the Office for National Statistics (ONS) through the Data Archive and has been used by permission. Neither the ONS nor the Data Archive bear any responsibility for the analysis or interpretation of the data reported here.
3. The human capital concept acknowledges both "firm specific" human capital acquired through continuing employment over the life cycle and "general" human capital which is often acquired prior to employment and is thought to depreciate over time.

Table I - All Full-Time Working Adults (N = 3363)

* = statistically significant at the 95% level ( z ³ 1.96)

** = statistically significant at the 99% level ( z ³ 2.58)

*** = statistically significant at the 99.9% level (z ³ 3.29)

Table II - All Full-Time Working Adult Females (N = 1101)

* = statistically significant at the 95% level ( z ³ 1.96)

** = statistically significant at the 99% level ( z ³ 2.58)

*** = statistically significant at the 99.9% level (z ³ 3.29)

There is no Pakistani? variable in this model as there were no female Pakistani full-time workers in the usable Summer 1995 Quarterly Labour Force Survey Sample.

Table III - All Full-Time Working Adult Males (N = 2262)

* = statistically significant at the 95% level ( z ³ 1.96)

** = statistically significant at the 99% level ( z ³ 2.58)

*** = statistically significant at the 99.9% level (z ³ 3.29)

Table IV

All Full-Time Workers Aged 50 - 69 (N = 543)

* = statistically significant at the 95% level ( z ³ 1.96)

** = statistically significant at the 99% level ( z ³ 2.58)

*** = statistically significant at the 99.9% level (z ³ 3.29)

TABLE VI