1 Policy background

The state pension age is the earliest age at which a state pension can be claimed in the UK. In 1948, this was set at 60 for women and 65 for men, and this remained unchanged until April 2010. Receipt can be deferred – in return for an increased entitlement – but few individuals choose to do this. In 2015–16, a full basic state pension, for which most new claimants qualify, was worth £115.95 (around $151) per week. The level of a basic state pension that an individual receives is based on the number of years (but not the level) of contributions made. Periods in receipt of certain unemployment and disability benefits and periods spent caring for children or adults can also boost entitlement. For those reaching the state pension age between 2010–11 and 2015–16 (inclusive), there was no minimum number of years of contributions needed for an individual to be eligible for at least some state pension, while for those reaching the state pension age prior to 2010–11 individuals needed to have contributions for at least a quarter of their working life to receive any state pension.

Some also qualify for an additional pension, which is related to earnings (since 1978) and some formal caring arrangements (since 2002). This can be worth up to around £160 ($208) per week in addition to the basic state pension. However, in the past, the majority of employees opted out of this arrangement, choosing instead to build up their own private pension entitlement in return for lower payroll taxes. State pension income is subject to income tax and can reduce entitlement to means-tested benefits, but there is no earnings test.

The 1995 Pensions Act legislated to increase the female state pension age from 60 to 65 over the decade from April 2010. This was to be implemented by raising the state pension age by 1 month every 2 months, thereby achieving equalisation with the male state pension age at 65 in April 2020. This reform therefore directly affected all women born after April 1950. The reform is shown in Figure 1, alongside a more recent reform that sped up the increase in the female state pension age to 65, so that it would subsequently be increased (along with the male state pension age) to 66 by October 2020 (the increase from 65 to 66 had previously been legislated to take place in the middle of the 2020s). The data used in our study cover the period up to and including March 2016, by when the female state pension age had reached 63; therefore, the impact of the later legislation shown in Figure 1 is outside our period of interest. To our knowledge, no occupational pension schemes adjusted their normal pension ages in line with the change in the state pension age for women.

Figure 1. UK state pension age for women under different legislation. Note: The reason the state pension age increases in a ‘sawtooth’ pattern, rather than a smooth line or a ‘step’ pattern, is that women born in a given month are allocated a single ‘state pension date’ at which they are eligible for a state pension. Therefore, women born later in the month have a slightly lower state pension age than those born earlier in the month. Source: Pensions Act 1995, schedule 4 (http://www.legislation.gov.uk/ukpga/1995/26/schedule/4/enacted); Pensions Act 2007, schedule 3 (http://www.legislation.gov.uk/ukpga/2007/22/schedule/3); Pensions Act 2011, schedule 1 (http://www.legislation.gov.uk/ukpga/2011/19/schedule/1/enacted).

Household incomes are also affected by the personal tax and benefit system, some features of which become more generous when an individual reaches the state pension age. On the tax side, employee payroll taxes (National Insurance contributions) are payable on the earnings of those below the state pension age, but not on the earnings of those above the state pension age. On the benefits side, those under the state pension age who are not in paid work may be eligible for certain benefits (such as income support, jobseeker's allowance (JSA) and employment and support allowance (ESA)). When an individual (man or woman) reaches the female state pension age, they are no longer eligible for working-age benefits, but instead, if they are on a sufficiently low income, they may receive pension credit. This means-tested benefit is more generous than working-age benefits: not only is the maximum award much higher (in 2015–16, it was worth up to £151.20 ($196) per week for a single person, which is more than twice the maximum £73.10 ($95) per week paid to a single JSA recipient aged 25 or over) but it also comes with less ‘conditionality’ – no requirements for recipients to seek work or to attend work-focused interviews.

Finally, if one individual in a household is over the female state pension age then they become eligible for a winter fuel payment worth £200 ($260) per year per household (or £300 ($390) per year if at least one person is aged 80 or over).

2 Data

We used data from the UK's Family Resources Survey (FRS). This is an annual cross-sectional household survey which contains around 20,000 households per year across England, Scotland, Wales and Northern Ireland. It is available up to and including the financial year 2015–16. The survey contains a battery of questions about incomes from all sources, of all individuals in the household, alongside background information such as sex, age, education, housing tenure and household structure. Importantly, since 2008–09, it contains the exact date of birth for all individuals, which allows us to calculate their state pension age. We used the comprehensive and consistent measures of income that are derived from the FRS in order to produce the UK's official statistics on income poverty, and the distribution of income, known as the ‘Households Below Average Income’ (HBAI) statistics.Footnote ³

We also measured income poverty in the same way as the UK's official absolute poverty measure – defined as having a household net equivalised income that is below 60% of the median in 2010–11, after adjusting for household inflation.Footnote ⁴ Net household income is measured by summing all income from private sources (such as income from employment, private or occupational pensions, and investments), deducting direct taxes paid on this income, and finally adding all income from state transfers or benefits (including the state pension). We used measures of income poverty based on incomes both before housing costs are deducted (‘BHC’) and after housing costs are deducted (‘AHC’). When incomes are equivalised, they are equivalised using the OECD modified equivalence scale and expressed as equivalent levels of income for a couple with no dependent children (since this is the most common family type among women aged 60–62). For a childless couple in 2015–16, the (BHC) absolute poverty line is a net income of £278 ($361) per week, and for a single adult £186 ($242) per week.Footnote ⁵

We adjusted for inflation using a modified version of the Consumer Prices Index that also incorporates owner-occupied housing costs – the same measure of inflation as the UK's Department for Work and Pension uses when constructing their official statistics on incomes and poverty. All cash figures are expressed in 2015–16 prices. Given that there is known to be considerable measurement error in survey measures of income at the very bottom of the income distribution (Brewer et al., Reference Brewer, Etheridge and O'Dea2017) and the very top (Burkhauser et al., Reference Burkhauser, Hérault, Jenkins and Wilkins2017), in each year of the data we excluded individuals who are in households that have the lowest 1%, and the highest 1% of household incomes.

Important drivers of the effect of increasing the state pension age on women's incomes will be the extent to which state benefit incomes and income from employment respond to the reform. Figure 2a and b provide graphical evidence of the impact on these outcomes, showing how mean benefit and employment incomes have changed from 2004–05 to 2015–16, for women aged 58–63. Figure 2a shows that prior to April 2010, average incomes were much higher among those aged 60 and over than among those aged 58 or 59 (and hence below the state pension age). It shows that after April 2010/April 2012/April 2014, mean benefit incomes fell substantially (but not to zero) for women aged 60/61/62, respectively, coinciding with the times at which affected age groups could no longer receive the state pension. At the same time, Figure 2b shows that there were rises in mean employment income coinciding with women of each age no longer being eligible for the state pension, consistent with the evidence in Cribb et al. (Reference Cribb, Emmerson and Tetlow2016) that one effect of the reform is for employment income to increase to offset (at least partially) the fall in benefit income paid to households resulting from the reform.

Figure 2. (a) Mean state benefit income of women aged 58–63, 2004–05 to 2015–16. Source: Authors’ calculations using the Family Resources Survey. Number of observations: 24,493. (b) Mean gross income from employment for women aged 58–63, 2004–05 to 2015–16. Source: Authors’ calculations using the Family Resources Survey. Number of observations: 24,493.

In addition to the data on incomes, the FRS contains questions regarding families’ levels of ‘material deprivation’. These questions ask whether families are able to afford certain material items. Given that household incomes will not perfectly reflect the standard of living of a given household, we estimate whether increasing the state pension age affects the probability that families report not being able to afford certain items. This paper looks at the effect on six items that have been reported consistently since 2008–09. The FRS questions are designed explicitly to measure whether individuals are experiencing material deprivation, although it is possible that there might be effects on other indicators of deprivation/other items that are not measured in these data. The FRS survey asks whether families can afford to:

1. (1) keep their home in a good state of decoration;

2. (2) have home contents insurance;

3. (3) put away £10 savings per month for a rainy day;

4. (4) replace worn-out furniture;

5. (5) replace broken electrical items;

6. (6) have a small amount of money to spend on themselves.

These questions ask whether individuals can afford these things now. They are therefore likely to be a measure of current living conditions, rather than reflecting living standards over a longer (backward looking) period.Footnote ⁶ These ‘material deprivation’ questions are only asked to families if all adults are aged 64 or under; there is a separate set of ‘pensioner material deprivation’ questions for families with an adult aged 65 or over. However, from 2011 to 2012, families are only asked the material deprivation questions listed above if all adults are under the state pension age. We discuss how we estimate the impact on material deprivation in light of this in Section 4.

3 Empirical methodology

We estimate the impact of increasing the state pension age on different measures of household income, exploiting the fact that we have data on the incomes of otherwise similar women who face different state pension ages. We utilise a difference-in-differences methodology in the same way as Cribb et al. (Reference Cribb, Emmerson and Tetlow2016), who estimate the impact of the same reform on labour market behaviour using similarly structured data. This is done by estimating the following model:

$$y_{ict} = \alpha ({\rm underSPA})_{ict} + \; \gamma _t + \lambda _c + \mathop \sum \limits_a \delta _a\left[ {{\rm ag}{\rm e}_{ict} = a} \right] + X_{ict}\beta + \varepsilon _{ict},$$

where the outcome of interest (y _ict) for individual i, from cohort c, and observed in time period t is allowed to vary by whether the woman is aged below or above the state pension age (underSPA) plus a set of controls. The dummy ‘underSPA’ measures whether, at the point they are observed, they have reached their state pension age. This is constructed by comparing the date of interview in the survey, to the individuals’ state pension date (which depends on their date of birth, as shown in Figure 1). Given the nature of the reform, this is an interaction term between an individual's cohort and age.

We use 8 years of FRS data, from 2008–09 to 2015–16 (so two financial years of data prior to the state pension age starting to rise, up to the latest available data by when the female state pension age had reached 63). The coefficient on underSPA (α) can therefore be interpreted as the average effect of being under, rather than over, the state pension age, for 60–62 years old women over the period from 2010–11 to 2015–16.

Our sample includes women born from 1948–49 to 1955–56 (inclusive). This means that the difference-in-differences estimator essentially compares the outcomes of two cohorts (1948–49 and 1949–50) who are unaffected by the reform, with the following six cohorts, who are affected by the gradually increasing state pension age. The exact sample used is shown in Appendix Figure A1a, and we have a sample size of 19,086 women.Footnote ⁷

In Equation (1), since both age and calendar time are important determinants of private income (such as earnings and private pension income), we control for these flexibly with dummy variables for age in years and quarters (48 dummies included in the model) and time period (γ _t) in years and quarters (31 dummies included in the model). We also control for the financial year of birth (which in the UK runs from April until March, λ _c) (six dummies included in the model). We cannot control for cohort using fixed effects for year and quarter of birth, as that would be perfectly co-linear with the age and time fixed effects. Using (financial) year of birth dummies allows us to control flexibly for fixed differences between cohorts, without perfect co-linearity.Footnote ⁸

Our identifying assumption here is that any age effects on the outcomes of interest are constant across time and cohort, that any cohort effects are constant across time and age, and that any time effects are constant across cohort and age (which is the standard ‘common trends’ assumption used in difference-in-differences estimation). This seems likely to hold, and indeed, pre-reform Figure 2a and b shows very similar changes in the two of the main components of income – state benefits and employment income – for different age groups. This supports our common trends assumption. No other reforms to the pension, tax or welfare system changed in a way that affected different cohorts at different ages in the way that the increase in the state pension age has affected women.Footnote ⁹

We also include controls for a set of individual characteristics (X _ict). These include education, relationship status, homeownership, region and – for those with a partner – partner's age and partner's education.Footnote ¹⁰

For continuous outcomes – specifically, different measures of income – we estimate Equation (1) using ordinary least squares. For dichotomous outcomes – such as whether a household is below or above the income poverty line, or different measures of deprivation – we estimate the equation using a probit model. Using the probit models, we calculate the average marginal effect of the treatmentFootnote ¹¹ and estimate standard errors by bootstrapping with 1,000 replications. For all models, we cluster standard errors at the year-and-month of birth level as those born in a given month all have the same state pension date. If there are serially correlated shocks at the cohort-time level, then not clustering at this level would bias standard errors (Moulton, Reference Moulton1990; Donald and Lang, Reference Donald and Lang2007).

As described in Section 2, we cannot estimate the impact of increasing the female state pension age on material deprivation using the same sample as for the income measures since the relevant questions are not asked of women with a partner aged 65 or over or, since 2011–12, of those women who are over the state pension age. However, because we still observe families where the woman is aged 60 or over but also under the state pension age, we can still use a difference-in-differences methodology to estimate the impact on material deprivation. Specifically, we take women aged 57–62 who answer the material deprivation questions and contrast how material deprivation rates change among 60–62 years old between periods when they are over the state pension age and under the state pension age, with how material deprivation changes among 57–59 years old, who are under the state pension age in all years. To do this, we can simply estimate Equation (1) using the sample of women aged 57–62 who do not have a partner aged 65 or over and who answer the material deprivation questions. The exact sample used is shown in Appendix Table A1 and has a sample size of 10,947 women.