Helping the Poorest Help Themselves? Encouraging Employment Past 65 in England and the USA

Abstract

In the context of population ageing and low retirement incomes, the UK government is encouraging delayed retirement. However, the OECD has argued that UK means-tested benefits disincentivise employment for the poorest, and Vickerstaff (2006b) has suggested managers have typically controlled opportunities to work beyond 65. In the US, contrastingly, benefits are meagre and difficult to access, and age discrimination legislation protects individuals from forced retirement. Would a US ‘self-reliance’ policy approach increase employment amongst the poorest over 65s in the UK and enhance or diminish their financial position? The evidence suggests that extending UK age discrimination legislation and restricting benefits would increase overall employment past 65, although not necessarily to US levels. Analysis of the English Longitudinal Study of Ageing and the US Health and Retirement Study finds the poorest over 65s were more likely to work in the USA than in England in 2002. However, within the USA, employment amongst the poorest was still low, especially compared with wealthier groups; logistic regression analysis primarily attributes this to lower levels of health and education. A US policy approach would therefore most likely damage the financial position of the poorest in the UK, as increased employment would not sufficiently compensate for lost benefits.

Introduction

Across industrialised countries concern is growing about the future costs of supporting growing numbers of older people (OECD, 2006c). Many countries are attempting to increase the percentage of pre-retirement age people in employment to offset some of the costs (Ebbinghaus, 2006). Furthermore, people will have to wait longer to receive a state pension in countries such as the UK, where the state pension age at the current time is set to rise from 65 to 68 between 2024 and 2046 at the latest. In the lead up to this, the UK government has increasingly sought to encourage people to work longer (Vickerstaff, 2006b) and delay taking their state pension(s).

But has it been realistic to expect those in most financial need of earnings to work past their mid sixties? This question is particularly pertinent to the UK because employment rights for men and women are, at the time of writing, negligible after age 65 (Lain, 2009; EC, 2002). Qualitative research by Sarah Vickerstaff and her colleagues (Vickerstaff et al., 2003; Vickerstaff, 2006a, 2006b) has suggested that UK employees have not had the degree of choice over retirement timing assumed by policy; instead, line managers have often decided whether individuals will be allowed to work beyond 65, based on the perceived needs of the department (see also Flynn, 2010). Furthermore, the OECD (2006a: 74) argues that means-tested benefits, that are lost as a result of working, mean that ‘delaying retirement is a significantly less attractive option [for poorer people]’.

American policy might be viewed as offering solutions to these potential barriers to employment. In the US, access to means-tested benefits has been severely restricted, limiting employment disincentives for the poorest (Lain, 2009). At the same time, US age discrimination legislation without an upper age limit formally protects people from forced retirement, in principle enabling individuals to delay retirement until they have sufficient resources to stop working. The UK's former Labour government had strongly hinted that it would follow the US path of outlawing mandatory retirement ages (DWP, 2009: 9); with a new coalition government, it seems increasingly likely at the time of writing that this will happen, the rationale being that it will promote economic independence and choice (BIS, 2010). In this context, it is not inconceivable that questions will also be raised in some quarters about the desirability of providing means-tested benefits that potentially discourage employment.

This paper therefore examines whether adopting a US policy approach in the UK would increase employment of the poorest over 65s and enhance their overall financial position. Using the English Longitudinal Study of Ageing (ELSA) and US Health and Retirement Study (HRS), we examine how wealth influences employment, and how this influence is mediated by differences in health and education. We focus on the year 2002 because men and women in both countries had access to a ‘normal’ state pension at 65; after this date Americans had to wait longer.¹ We conclude by arguing that extending age discrimination legislation and reducing access to state benefits in the UK would increase the number of over 65s in employment, but damage the financial position of the poorest as a group.

The institutional context of employment past age 65

We might expect employment past 65 to be similarly high in the UK and the USA: both are commonly classified as having weakly decommodifying ‘liberal’ welfare states, in which people are reliant to a significant degree on the labour market for meeting their welfare needs (following Esping-Andersen, 1990). At the same time, both are said to have comparatively unregulated liberal market economies (cf. Hall and Soskice, 2001) in which employers should have a high degree of latitude over employment decisions. However, a number of authors have questioned the liberal label. Castles and Mitchell (1993), for example, argue that whilst the USA might be categorised as ‘liberal’, the UK, Australia and New Zealand are too egalitarian for this category. Further, O'Connor et al. (1999) argue that Esping-Andersen's (1990) typology ignores the different gendered policy assumptions, or ‘policy logics’, found across English-speaking ‘liberal’ countries.

The present author has also argued that different US and UK ‘policy logics’ underpinned social and employment policy for older people in the early 2000s (Lain, 2009). A US policy logic of self-reliance was based on the expectation that older people should time their retirement according to their financial needs. The state's role was not to provide an adequate safety net, but to formally protect the employment of older people until they were in the financial position to leave work.

In the UK, we can identify a policy logic of ‘paternalism’ in the early 2000s (Lain, 2009). This policy logic assumed older people were a more vulnerable group, who were not expected to secure their welfare through employment. The role of the state was not, therefore, to formally protect the employment of those aged 65 and over, but to provide a financial safety net aimed at preventing poverty.

A political rhetoric of individual responsibility has developed in the UK that is somewhat at odds with this paternalistic policy logic. This will be reflected in the proposed introduction of the National Employment Savings Trust, the new name for Personal Accounts that are anticipated to encourage retirement saving amongst modest and low earners (DWP, 2006). Future generations will bear increasing financial risk with regard to pensions, given the shift occurring from defined benefit to defined contribution pensions (Bridgen and Meyer, 2005). Nevertheless, for older people reaching 65 in 2010 the basic parameters of the UK policy are fundamentally the same as in the early 2000s, particularly with regard to areas of potential importance to poorer people: means-tested benefits and age discrimination legislation.

Means-tested benefits

In neither the UK nor the US is state pension provision generous by international standards, particularly for those with formerly low earnings (Whitehouse, 2003: 35). According to Whitehouse (2003: 35), someone with a full contributions record earning half average earnings throughout their life would receive state pensions equivalent to 25 per cent of average earnings in the UK, and 29 per cent of average earnings in the USA. (This includes the Social Security pension in the US, and the Basic State Pension and State Earnings Related Pension in the UK.²)

When one considers that breaks in employment can dramatically reduce pension levels, and in the UK very low earnings do not contribute towards pension income (Ginn, 2003), it is obvious that means-tested benefits have a potentially important role in preventing individuals falling below subsistence levels. Given the greater emphasis on providing a financial safety net, means-tested benefits in the UK were easier to receive, more generous, and more widespread in the early 2000s (Lain, 2009). The need for means-tested provision will also continue into the future: the government has estimated that a third of pensioners will be entitled to means-tested benefits in 2050 (Bridgen and Meyer, 2007: 69; see also Price, 2007).

Both countries provided means-tested ‘income assistance’ benefits in 2002, guaranteeing minimum incomes for older people (the UK Minimum Income Guarantee and US Supplemental Security Income). However, the percentage of over 65s receiving income assistance was lower in the USA than the UK in the early 2000s, at around 3 per cent for those living alone or as a couple, compared with 12 per cent in the UK (Lain, 2009: 146).³ Partly this was because US politicians had marginalised the benefit to only the very poorest, by failing to uprate the amount of assets recipients were allowed to have. In 2002, recipients were allowed assets of around $2,000 excluding housing (SSA, 2003b: 2), the same amount as in 1989, and in real terms only 37 per cent of the amount allowed in 1974 (Elder and Powers, 2006: 343). This compares with £12,000 of non-housing assets allowable in the UK (McConaghy et al., 2003: 2).

Income assistance in the UK was also slightly more generous, guaranteeing a single pensioner an income equivalent to 26 per cent of average earnings, compared with 20 per cent in the USA scheme.⁴ Likewise, couples in the UK could receive income assistance equivalent to 40 per cent of average earnings, compared with 31 per cent in the USA.⁵ In 23 out of the 52 states in the USA, supplements increased income assistance levels, but typically by a trivial amount: the median supplement was only $31 a month for an unmarried person (US Committee on Ways and Means, 2004: 3–28–3–29, Table 3–5).

We can see a similar, if more extreme, difference between the countries in terms of the generosity and availability of housing benefits. In the UK, Housing Benefit had a simple rationale: to ensure that no one in rented accommodation had an income of less than income assistance, once reasonable rental costs had been deducted (Priemus et al., 2005: 587). For those aged over 60, this therefore meant that a person receiving income assistance in rented accommodation received a benefit equivalent to their full rental costs (assuming rental levels were deemed reasonable; see Zebedee and Ward, 2002). The benefit was received as of right, as long as the income and asset conditions were met, and could be used for private or public sector rented accommodation.

In the USA, housing assistance was a much more fragmented system, split between vouchers for private sector accommodation, and low-rent housing programmes (OECD, 2004b: 5). Nevertheless, the common principle was that recipients had to contribute 30 per cent of their income, minus some deductions, to the rent; this strongly weakened the generosity of the benefit (Lain, 2009). In addition, housing assistance in the USA was a budgeted scheme, and the number of ‘places’ available severely restricted by budget constraints and the limited stock of suitable rental accommodation (OECD, 2004b; Priemus et al., 2005). In order to apply, individuals had to join queues for housing assistance that could last for many years (Priemus et al., 2005: 592). Once individuals reached the front of the queue, they were given 60 to 120 days to find a home within the allowed price range meeting the required quality standards. Failure to find such accommodation ranged from 30 to 50 per cent across different metropolitan areas (ibid.: 582). As a result, whereas just under fifth of people over 65 living alone or in couples received housing assistance in the UK, in the USA it was a twentieth (Lain, 2009: 146).⁶ Three quarters of housing vouchers went to the very poor (30 per cent of median income; OECD, 2004b: 5). However, in 1999 only ‘29 per cent of extremely low and very low income renter households were receiving a housing subsidy’ (Priemus et al., 2005: 582).

As noted in the introduction, the OECD (2006a) argues that means-tested benefits in the UK disincentivise continued employment; the evidence provides some support for this assessment. For single people over 65 in 2002, median means-tested benefits received totalled £54 per week (DWP, 2004: 33). With the exception of £5 earnings disregards (OECD, 2004a), these benefits would be lost entirely as a result of earning a wage of £54 or above. Given the difficulty of obtaining means-tested benefits in the US, on the other hand, it is interesting that the OECD (2006b) does not deem it necessary to discuss means-tested benefits in its analysis of financial incentives to retire.

Age discrimination legislation

With very modest state pensions for low earners, and access to means-tested benefits severely restricted, policy in the USA in the last 25 years has arguably been one of encouraging older people to time their retirements according to their financial needs. As a formal means to do this, the upper age limit on US age discrimination legislation was lifted in 1986. This built on the 1967 Age Discrimination in Employment Act (ADEA), which ‘prohibit(ed) arbitrary age discrimination in employment’ between the ages of 40 and 65 (Neumark, 2003: 298), and an ADEA amendment extending coverage up to age 70 from 1978.

By 1978, 15 US states had already introduced their own age discrimination legislation without upper age limits (Neumark and Stock, 1999: 1090–1101). However, the 1986 ADEA amendment was significant in forbidding employers across the US from using mandatory retirement ages for all but a few occupations where safety was an issue.⁷ Enforcement is overseen by the Equal Employment Opportunities Commission, and employers found to have discriminated on the basis of age potentially face having to reinstate or hire the individual(s) involved, and pay legal costs and damages (which may be punitively high) (Neumark, 2003: 300).

According to Macnicol (2006: 255), the under secretary of Labor, Malcolm R. Lovell, made three justifications for the 1986 amendment:

older workers who remained in employment would contribute to Social Security, that income from work was better than an inadequate pension, and that technological advances made the labour process less physically demanding.

Formal rights to continued employment were therefore legislated as a remedy to pensions that the government acknowledged could be inadequate. To ensure that continued work results in increased retirement incomes, an amendment to the ADEA further requires that employers provide ‘continued contributions, credits and accruals under pension plans for service beyond normal retirement age [i.e. beyond age 65]’ (Quadagno and Hardy, 1991: 473).

The situation in the UK in the early 2000s was very different to the US, as there was no age discrimination legislation (Macnicol, 2006). Employees had no right to continue working past age 65, and very limited rights to claim unfair dismissal or redundancy payments after this age (EC, 2002). Age discrimination legislation was passed in October 2006, but only gives individuals the opportunity to request continued employment past age 65 (Sergeant, 2006: 226); recent research by Flynn (2010) suggests this has done little to change the dominance of line managers in decisions about continued employment. Clearly, the UK has operated under a very different set of assumptions to the US about the potential role of older people as workers and earners.

Comparing employment by wealth and human capital: what we know

With limited access to state benefits but greater potential opportunity to work, we would expect the poorest Americans to be much more likely to work than their British counterparts. However, it is not possible to draw a firm conclusion on this from the literature, which makes it difficult to assess the possible impact of adopting US policies in the UK. The body of research on workers past typical retirement age is comparatively small for both countries (Smeaton and McKay, 2003: 5; McNamara and Williamson, 2004: 259) and has led to some mixed impressions about the financial position of these workers. For example, Smeaton and McKay (2003) perform logistic regression analysis on the factors associated with working past state pension age in the UK, and conclude:

Working [past state pension age] is associated with financial hardship such as: the absence of an occupational pension, income below £100 for men, still paying a mortgage and, for women, being separated or divorced . . . [However] Those in more extreme need with very low savings have often experienced a lifetime of disadvantage with attenuated employment opportunities leading to reliance upon state benefits, which in turn can function as an employment disincentive.

With the exception of those in most extreme need, working past state pension age is therefore said to be most commonly associated with financial disadvantage. However, it may be the case that the employment amongst wealthier individuals is not entirely captured, because of the primary focus on income received, not wealth held. For example, ‘absence of an occupational pension’ refers to receipt of occupational pension income, and many workers over 65 will have postponed pensions until retirement; indeed, it was not legally possible to draw a pension from a current employer unless you had pension rights dating back before 1989 (Meadows, 2003: 31). Likewise, ‘income’ refers to non-earnings income, which is likely to be low for workers who have deferred their occupational, personal or state pension.

Parries and Sommers (1994) also find that working in older age is associated with low levels of non-wage income in the US. Haider and Loughran (2001), on the other hand, find that: ‘labour supply [in the USA] is concentrated among the most educated, wealthiest and healthiest elderly’ (ibid. from the abstract). This conclusion is drawn on the basis of wealth rather than non-wage income, which avoids the problem of pension deferral found in other studies. Closer inspection of Haider and Loughran's (2001) results suggests that employment was not concentrated amongst the wealthiest, however, in the sense that it was confined predominantly to the wealthiest strata. Smeaton and McKay's (2003) UK analysis further shows that those with low savings were less likely to work. Without comparable cross-country studies, however, it is difficult to compare the role of wealth on employment between the countries.

Research has emphasised the fact that people are likely to work past state pension age for a range of reasons beyond the financial (Barnes et al., 2004; Tanner, 1997; Parries and Sommers, 1994; Parry and Taylor, 2007), so it is perhaps unwise to assume that older workers are concentrated at either the top or bottom of the economic hierarchy. However, poorer Americans should be more likely to work than their English counterparts given the nature of American policy; whether or not this is the case may depend on the degree to which levels of education and health limit employment. Studies from the UK and USA have consistently found that, compared with the economically inactive, workers past typical retirement age are likely to be healthier and better educated on average (Smeaton and McKay, 2003; Parries and Sommers, 1994; McNamara and Williamson, 2004; Haider and Loughran, 2001). Consequently, poorer older people with low levels of education and health may find it particularly difficult to compete for or retain employment. This is partly because of their lower range of work capabilities, but also because they are likely to be confined to precarious jobs at earlier ages.

In order to investigate the influence of wealth on employment, and how this is mediated through levels of health and education, we present analyses of the English Longitudinal Study of Ageing (ELSA) and US Health and Retirement Study (HRS).

Data

ELSA was designed in conjunction with the HRS, and consequently the surveys have ‘comparable measures of health, income, and education’ (Banks et al., 2006: 2037), as well as wealth. The surveys interview people in their fifties upwards alongside their partners; this included large samples of people aged 65 and over in 2002 (5,538 in England and 10,428 in the USA). Because we focus on 2002, the first year of the ELSA survey, our analysis here is cross-sectional (in any case, longitudinal analysis using the three subsequent waves of data would be limited by the small number of movements into and out employment for the over 65s). We use cross-sectional weights provided with the surveys with the aim of ensuring the results are representative of the populations being examined.

For the USA, the analysis in this paper has been conducted using the Rand HRS Version F file, which has been constructed by the Rand organisation from the raw HRS data files (see St Clair et al., 2006). The Rand Organisation have cleaned and processed the data, as well as performing a considerable amount of wealth and income imputation (see ibid.: 2), as was also performed for ELSA (see ELSA, n.d.). The ELSA results should broadly reflect the UK situation, because the vast majority of the population live in England.⁸

The influence of wealth on employment: policy lessons from the USA?

Table 1 shows employment rates by wealth level based on ELSA and HRS. The surveys identified employment by asking whether the individual was doing any paid work or was self-employed currently (in the USA) or in the previous month (for the UK) (see Lain, 2009: 80–1).⁹ Wealth is presented at the individual level, and relates to assets – for example, property, businesses, savings and investments (but excludes pension assets).¹⁰ If a person lived with a partner, their combined wealth was equivalised to the individual level using the ‘OECD-modified scale’ (OECD, n.d.; see also Atkinson et al., 1995), which assumes a couple needs 1.5 times the income or wealth of a single person to have an equivalent standard of living. Quintiles were calculated from the equivalised wealth for those aged 65-plus.

Table 1. The percentages of those aged 65-plus in employment by equivilised wealth quintiles: England and USA, 2002

Note: Differences between countries were statistically significant at the p = >0.001 level, in the percentages employed at each level of wealth (Panels A and B).

Note: Wealth equivilised to the individual level.

Source: Author's analysis of the English Longitudinal Study of Ageing 2002, and the Health and Retirement Study 2002 (Rand Version F file).

The results for ‘wealth excluding pensions’ show that in both countries the richest were most likely to work (20.7 per cent in the USA, 11.5 per cent in the UK). In England, employment rates fell as wealth decreased, with only 2.7 per cent of the poorest quintile working. In contrast, in the USA employment rates across the middle three quintiles were fairly similar (17.4–18.9 per cent). Nevertheless, as in the UK, the poorest in the USA were by far the least likely to work (12.7 per cent). When we exclude housing wealth, which might exaggerate the financial prosperity of some individuals, we still see low levels of employment for the poorest (see Table 1, panel B).

The higher rates of employment at all wealth levels in the US suggest US policies would increase employment in the UK, although ironically to lower levels for the poorest. However, to what extent does the broader evidence corroborate that US policies would increase overall employment? Figure 1 shows that in the mid 1960s around a quarter of men¹¹ worked past 65 in both countries (23.7 per cent in the UK and 26.7 per cent in the USA). During the period, employment fell in both countries, as we would expect given that over 65s became older and wealthier on average and unemployment rose above the very low levels experienced in the earlier post-war years. However, the decline was smaller in the US, suggesting that age discrimination legislation did have an impact. Some of the divergence in employment between countries occurred before 1978 when the ADEA was extended to the age of 70. However, the 15 US states introducing age discrimination legislation without an upper age limit before 1978 contained 37 per cent of the US population aged 65 and over.¹² It is therefore reasonable to conclude that state-based legislation played a part in stemming the decline in employment past 65 in the US during this period.

Figure 1. Employment rates of men aged 65-plus: UK and USA, 1965–2008

Source: Data retrieved from http://stats.oecd.org/Index.aspx?DatasetCode=ALFS_SUMTAB#(08/09/09), and OECD (1986) prior to 1984.

Other research also suggests that US age discrimination legislation had a positive impact on employment, but in relation to the retention rather than recruitment of older workers (Adams, 2004; Neumark and Stock, 1999; Issacharoff and Harris, 1997). For example, Adams (2004: 240) examines how variations in age discrimination legislation across US states influenced employment in the 1960s; he concludes that ‘age discrimination legislation has succeeded at boosting the employment of older individuals through allowing them to remain in the workforce longer’. However, the likelihood of finding work past 65 was reduced if they were covered by age discrimination legislation (ibid., 236). We cannot automatically infer that US policies would reduce recruitment of over 65s in the UK, given that US employees are more likely than their British counterparts to be in jobs started after 65 (Lain, 2009: 202). Nevertheless, from a policy perspective, there is little evidence that US policies will not automatically open up employment opportunities for those seeking work past age 65.

The primary impact of the legislation is therefore likely to be the way in which it prevents employers from organising employment exit around fixed retirement ages, whilst at the same time enhancing people's expectations of, and claims to, continued employment. However, while we may have good reason to assume US policies would increase employment in the UK, we cannot assume this will be to the levels found in the US. Wider factors, such as labour market conditions and financial incentives, contributed to the higher rate of employment in the US in the early 2000s. Unemployment levels¹³ and the sectoral composition of employment were similar in both countries in the early 2000s.¹⁴ In addition, male employment rates diverged in the period up to 1981 when unemployment levels were similar in both countries.¹⁵ However, a tighter US labour market in the years leading up to 2002 probably contributed to the higher level of US employment at that time (Lain, 2009).

By the early 2000s, it is also evident that some of the gap in employment past 65 could be attributed to differences in financial incentives to work. Americans were better off on average (Disney and Whitehouse, 2002: 3) and actuarial adjustments for late receipt of state pensions were identical in both countries (OECD, 2001: 156). However, salary-related occupational pensions were more common in the UK than US (Ashcroft, 2009), and this may have discouraged employment in situations where employers ceased contributions at 65 but forced employees to wait until retirement for their pension (Meadows, 2003). It could be argued that US age discrimination legislation, if applied to the UK, would mandate employers to contribute to pensions until retirement, reducing this disincentive effect on employment (see above). It could also be noted that, looking into the future, final salary pensions are in decline in the UK too. Nevertheless, in 2002 some of the increased employment in the US was probably attributable to differences in pension incentives (Meadows, 2003).

In both countries, the poorest were unlikely to have salary-related occupational pensions, so these disincentive effects are unlikely to have had much effect for these individuals. However, despite the difficulty of obtaining benefit or occupational pension income, and despite protection from forced retirement, the number of poorer Americans working was low. Indeed, Table 1 shows that the poorest were only around half as likely to work as the richest in the USA. If this can be attributed to low levels of health and education, it would raise questions about the ability of the poorest in the UK to secure or retain employment in older age, even if protection from age discrimination was extended past 65 and financial disincentives to work were reduced.

Logistic regression analysis

In order to investigate the degree to which education, health and age placed restrictions on the ability of the poorest to work, we perform logistic regression analysis (Table 2). The dependent variable is whether the individual aged 65-plus is employed (a value of 1) or inactive (a value of 0). We use the following independent variables:

• • non-pension wealth quintile,

• • number of health limitations,

• • qualifications, and

• • age in years.

Table 2. Logistic regression analysis: wealth and human capital effects on the probability of working past 65 in England and the USA, 2002 (odds ratios)

Notes: *** = p < 0.001, ** = p < 0.01, * = p < 0.05.

Health limitations: sitting for 2 hrs; getting up from a chair; climb several flights of stairs; stooping/kneeling/crouching; difficulty lift/carry 10 lbs; picking up coin; reaching/extending arms up; pushing pulling objects.

Wealth equivilised to individual level.

Source: Author's analysis of the English Longitudinal Study of Ageing 2002, and the Health and Retirement Study 2002 (Rand Version F file).

Before we examine the results, it is important to discuss how health and education are measured. Previous research examining health and employment in older age often uses the self-reported health ratings of survey respondents, typically ranging from ‘poor’ to ‘excellent’. Problematically, however, individuals are unlikely to use consistent criteria for what constitutes ‘poor’ or ‘good’ health (see McNamara and Williamson, 2004: 269–70).

There are two alternative health measures used in both surveys we can draw on. First, both surveys ask if respondents have ever been diagnosed with a range of medical conditions. Analysis by the author (Lain, 2009: 161–4) has shown that Americans over 65 were more likely than their English counterparts to be diagnosed with all of the conditions examined, in some cases by a large margin. This is consistent with previous research examining people from their fifties upwards using these surveys (Banks et al., 2006). However, diagnoses may have been made some time ago, and may not necessarily reflect current physical work capabilities. In addition, institutional and cultural differences between healthcare systems in each country may influence the way in which some illnesses are classified or diagnosed.¹⁶

We will use the second measure of health collected in both surveys, which asks respondents whether they have ‘any difficulties’ doing a range of everyday activities; the activities selected are listed in the notes to Table 2. These indicators relate to current problems, and avoid the possible influences of institutional/cultural differences in healthcare systems. Correspondingly, differences between countries in reports of difficulties with activities were smaller than was the case for diagnoses (see Lain, 2009: 162–3). The everyday difficulty health measure is self-reported, and cannot therefore be considered without some degree of subjectivity. Crucially, however, there is no reason for assuming Americans will be more likely to claim difficulties than Britons, all else being equal. The regression analysis in Table 2 presents health as the number of limitations a person has.

The educational classification is based on ISCED 97 qualification levels:

• • ‘high’, representing degree or above (ISCED 5–6);

• • ‘medium’, secondary level to below degree (ISCED 3–4); and

• • ‘low’, below secondary (ISCED 0–2).

Qualifications have been allocated as closely as possible to ISCED 97 using information in OECD (2004c). Slightly fewer Americans were categorised as being highly educated, compared with OECD (2004c), because the original US qualifications variable was not sufficiently disaggregated to recode to ISCED 97 exactly (see Lain, 2009: 165). Nevertheless, as in OECD (2004c) the classification used here found Americans to be more highly educated.¹⁷

We now turn to the regression analysis itself. In model 1, we only include wealth alongside age, as we want to see the influence of these factors before introducing health and education in model 2. The reference category for wealth is the richest quintile (quintile 1), so the odds ratio for this group is set to 1. The odds ratios below this indicate the probability of being in employment relative to the reference category. We present odds ratios as they are reasonably straightforward to interpret.

From model 1, we can see that, controlling for age, being in the poorest quintile reduced the probability of employment in both countries, but to a greater degree in England. Being in the poorest quintile reduced the probability of employment by 73 per cent in England compared with 46 per cent in the USA (relative to the richest quintile in each country).¹⁸ The odds ratios for the middle three quintiles were closer to 1 in the USA than in England, which reflects the fact that there was less variation in employment levels across wealth levels.

From model 2, we can see the effect of wealth on employment, once we take into account the effect of education and wealth. In both countries, the wealth odds ratio increases between models 1 and 2, indicating that some of the variation in employment between wealth levels is attributable to health and education. The odds ratio for the lowest quintile in the United States rises from 0.54 in model 1 to 0.82 in model 2. The fact that the odds ratio moves close to 1 in model 2 suggests that health and education were important in preventing the poorest from working in the same numbers as the wealthiest. Once we take the effects of health, education and age into account, being in the poorest quintile only reduced the probability of being in employment by 18 per cent (relative to the richest quintile).¹⁹ When the poorest had the capacity to work in the USA, they worked in similar numbers to the richest.

In England, on the other hand, after we take health, education and age into account, being in the poorest quintile still reduced the probability of working by more than half (55 per cent).²⁰ This suggests that means-tested benefits did discourage employment among the poorest in the UK to a degree (otherwise, we would expect a greater similarity of employment across wealth levels after accounting for educational and health differences). However, the example of the USA suggests that any increases in employment amongst the poorest as a result of removing these benefits would be minimised by the barriers of low levels of education and health.

Moving on to the odds ratios for health limitations, it is evident that the probability of working decreased in both countries as the number of health limitations increased, controlling for other factors. The probability of working reduced faster in England as the number of health limitations increased, but was clearly important in reducing the likelihood of working in both countries.

It is interesting that qualifications appear to have a similar impact on employment in both countries. Taking into account the other factors included, having low qualifications roughly halves the chances of being in employment relative to the highly qualified. Likewise, having medium-level qualifications reduces the probability of employment by around 40 per cent in both countries.

Conclusion and discussion

American policy encourages individuals to time their retirement according to their financial needs, and therefore appears to offer attractive solutions to the problems caused by population ageing and low pension incomes in the UK. In the US, employers are forbidden from forcing individuals to retire because of their age, a move the UK looks increasingly likely to follow, and means-tested benefits have been severely restricted, reducing incentives to stop working amongst the poorest.

The much higher level of employment amongst over 65s in the US suggests that such a policy shift would increase employment amongst this age group in the UK. However, it is debateable whether these reforms would increase employment to US levels. Some of the higher US employment rate in the early 2000s can be attributed to a longer period of labour market demand, and to the lesser availability of salary-related pensions that encourage earlier retirement.

More fundamentally, there is little evidence that US policies would financially benefit the poorest. Given the difficulty of obtaining benefits in the US, the financial consequences of the poorest not working can be harsh. However, employment rates of the poorest fifth in the US were low, especially when compared with the wealthiest. Our regression analysis primarily attributes this to low levels of health and education. Reducing means-tested provision and strengthening employment protection in the UK would be likely to increase the employment rates of a relatively small number of poorer people with good health and reasonable levels of education. This might, for example, include educated and healthy divorced women with small pensions because of years spent out of the labour market for family reasons; this is something for future research to examine. However, the broader and much more significant impact of such a policy shift would be to damage the overall financial position of the poorest, as increased employment rates would not come close to compensating for the decrease in the availability of benefits. This conclusion is supported by the fact that poverty rates amongst older people were higher in the USA than the UK in the early 2000s. Poverty rates are high amongst older people in the United Kingdom too (at around a tenth of the population over 65), but we should not look to the USA for solutions (where a sixth live in poverty; Smeeding and Sandström,2005: 18).

Research suggests that older people want greater choice about retirement timing (Vickerstaff, 2006a, 2006b; Vickerstaff et al., 2003; Loretto and White, 2006), and a considerable minority of over 65s would like to work (e.g. Smith 2000: 21). Abolishing mandatory retirement ages could play a part in increasing opportunities for older people to continue working in the UK. It would enhance employee claims to continued employment and diminish the capacity of employers to organise their activities on the basis of fixed retirement ages. However, the consequences for delivering employee choice should not be over-stated. Those with low retirement incomes who have the capability and desire to work would continue to be financially penalised by loss of benefits. Addressing this would require a universal state pension above means-tested levels that could be taken in full whilst working.

More broadly, despite greater formal rights to employment, there is evidence of unmet demand for jobs amongst over 65s in the US. Smith's (2000: 21) analysis of the ISSP 1997 Work Orientation Study found that around two fifths of over 65s in the USA stated they would like a job even if they did not need the money, which is higher than the proportion actually working. Revealingly, Abraham and Houseman's (2005) longitudinal analysis of the HRS found that although a significant minority of older workers planned to reduce their hours or change jobs prior to, or instead of, retirement, few actually did so. This suggests that discrimination legislation cannot override the fact that for older people there are not enough jobs available matching their preferences, skills and experience. The state therefore has an important responsibility to ensure that pensions provide a decent standard of living without recourse to earnings.