Retirement? What retirement? A tale of choice and necessity

Abstract

Not every older person retires at the same age; some don’t want to, others simply can’t afford it. Takao Maruyama and Vincent Charles present the dual faces of retirement in the UK

Ash T Productions/shutterstock.com

Retirement is no longer what it used to be in the UK. The landscape is shifting, and the traditional notion of kicking back and relaxing is evolving. Recent data from the UK’s tax collector HM Revenue & Customs shows that substantial pension withdrawals have occurred over multiple quarters: £4.05 billion was withdrawn between April and June 2023, with a similar figure of £4.03 billion reported in the first quarter of 2024 (tinyurl.com/44eppb9d). While many of these withdrawals were driven by the need to manage living costs, mortgages, and debt repayments before reaching pension age, they could also lead to a longer-term consequence: some seniors may find themselves needing to work beyond retirement age in order to rebuild their savings for retirement. This raises a key question: do people in such situations choose to work, or are they forced to continue working because of their financial circumstances? In other words, are they driven by choice or by necessity?

In this article, we delve into the complex motivations and circumstances surrounding post-retirement work in the UK. We aim to explain the dual faces of retirement: one marked by autonomy and freedom, and the other by financial strain and obligation.

The rise of post-retirement employment

Retirement used to be associated with leisure and relaxation, but it now includes a variety of activities, including continued employment. As mentioned above, statistics reveal a steady surge in individuals extending their careers beyond the conventional retirement age, challenging preconceived notions about what life is like after leaving the workforce.

Two contrasting narratives surface in the midst of this trend, concerning those who are “forced to work” and those who “choose to work”. A significant number of retirees gladly welcome the chance to continue working, but others struggle with financial insecurity and must look for and continue to work in order to sustain their livelihoods.

For many retirees, the decision to prolong their careers stems from a desire for fulfilment, purpose, and social connection. By doing meaningful work, people can make the most of their abilities and knowledge and benefit society while maintaining a sense of identity and self-worth. In addition, the changing nature of work in the digital age gives retirees greater freedom and flexibility to pursue side projects, part-time jobs, or consulting positions on their terms. Research¹ suggests that staying employed helps with social integration, mental health, and cognitive function – all of which help to lessen the negative consequences of retirement and the isolation and inactivity that come with it.

Conversely, a significant segment of pensioners find themselves compelled to continue working past their retirement age as a result of limited financial resources and insufficient pension plans. Rising living costs, declining savings, and escalating health-care expenses converge to create an uncertain financial climate, leaving many retirees vulnerable to economic instability. For these individuals, work is a lifeline that enables them to maintain their standard of living amidst mounting expenses while still meeting their basic necessities. However, re-entering the workforce later in life presents its own set of difficulties, such as limited job possibilities, age discrimination, and skill obsolescence.

Certain groups of older workers are more likely to be “forced to work” beyond the state pension age compared to benchmarking groups

In this article, we examine the underlying motivations of people who continue working beyond their state pension age, which is 66 years for both men and women at the time of writing. We also explore potential factors which may classify these individuals into two groups – those who “choose to work” and those who may be “forced to work”.

Data

Our data is derived from the Annual Population Survey (APS) conducted in the United Kingdom (tinyurl.com/35d688m6; tinyurl.com/wnemub7p). The APS covers a 12-month period and includes data from approximately 320,000 individuals across 122,000 households. The survey captures a wide range of variables, including respondents’ demographics (e.g., age, sex, marital status, ethnicity, education, region), household characteristics (e.g., household structure, role within the household), economic status, employment activities (e.g., employment status, type of occupations, work locations, working hours), health, and income.

Model

Multiple (binary) logistic regression is a statistical method used to estimate the probability of an event occurring based on known values of one or more predictor variables. While it shares some similarities with linear regression, logistic regression is specifically designed to model situations where the outcome (dependent variable) is binary. This means that the outcome can take only one of two possible values, such as 1 (representing the occurrence of an event) or 0 (representing its non-occurrence). Unlike linear regression, which predicts a continuous outcome, logistic regression models the relationship between the predictors and the probability of the event occurring. It does this by transforming the probability using a natural logarithm (logit function), allowing it to handle binary outcomes effectively.² The multiple logistic regression equation is given by

$$P(Y)=\frac{1}{1+e^{-(b_0+\sum^k_{i=1}b_iX_i)}}$$

where P(Y) is the probability of an event happening; e is the base of the natural logarithm (approx. 2.718); b₀ is the intercept, representing the predicted value when all the predictors are zero; and b_i is the regression coefficient for the predictor X_i, which shows how much that predictor influences the probability of the event happening.

The logistic regression model provides several useful statistical outputs, one of which is the odds ratio. Odds ratios can be calculated by taking the exponential of the logistic regression coefficients (often expressed as “Exp(B)”). To understand odds ratios, we first need to understand odds. Odds refer to the likelihood of an event happening compared to the likelihood of it not happening. Odds can be calculated using the following formula:

$$odds=\frac{P(Y)}{1-P(Y)}$$

where 1 – P(Y) is the probability of the event not happening.

In this article, we are interested in comparing the odds of being “forced to work” versus “choosing to work” among the various demographics. For example, if we are concerned with males (coded as 0) and females (coded as 1), the odds ratio can be calculated as

$$\text{Exp}(B)=\frac{\text{Odds of being forced to work among females}\,(1)}{\text{Odds of being forced to work among males}\,(0)}$$

Odds ratios measure how much the odds change with a one-unit change in a predictor. In this case, the predictor is biological sex (going from 0 for males to 1 for females). Accordingly, an odds ratio greater than 1 means that the odds of being forced to work are higher for females compared to males. For instance, if the odds ratio is 2, this means that the odds of being forced to work are twice as high for females as for males. Conversely, an odds ratio less than 1 indicates that the odds are lower for females than for males. An odds ratio of 1 indicates no difference, meaning females and males have the same odds of being forced to work. In our article, in Figures 2 to 8, the dotted vertical line at odds ratio=1 represents the reference point or the “no effect” line, indicating the threshold where there is no difference in the likelihood of being forced to work between the groups being compared.

The analyses presented in this article (excluding Figure 1) were conducted using 1,000 stratified bootstrap samples for each yearly period between April and March from 2019 to 2023. Stratified bootstrap sampling is a technique that helps ensure that all key demographic groups are fairly represented in the analysis, facilitating a more nuanced understanding of retirees’ motivations and choices. In this study, the demographic variables used for stratification included respondents’ sex, ethnic background, long-term illness, marital status, socioeconomic classification, UK regions of residence and accommodation type.

Figure 1:

Reasons for working beyond state pension age

Stratified bootstrap sampling combines two techniques: stratified sampling and bootstrap sampling. Stratified sampling is a method that ensures specific subgroups (e.g., ethnic groups like white, Asian, black and others) are fairly represented in a sample. This helps make the sample more accurate by reflecting the actual proportions of these subgroups in the population. Once the subgroups are formed, bootstrap sampling is applied. This involves randomly selecting a person from each subgroup, recording their information, and then placing that person back into the subgroup so they can be selected again. This process is repeated many times, creating a new sample each time, while maintaining the correct proportions of each subgroup. By using this technique, 1,000 stratified bootstrap samples were generated for the analysis, ensuring that each sample reflected the true make-up of the population.

The Annual Population Survey reveals six main reasons why individuals opt to continue working beyond their state pension age. These include (A) “To pay for desirable items (such as holidays)”, (B) “Not ready to stop work”, (C) “Employer needs your experience or you are needed in the family business”, (D) “Due to opportunities to work more flexible hours”, (E) “To pay for essential items (such as bills)”, and (F) “To boost pension pot”.

In this study, these reasons were categorised into two groups: “choose to work” and “forced to work”. If respondents indicated that their reasons for working were driven by the need to cover basic living costs and necessities, they were classified as “forced to work”. On the other hand, if their motivations extended beyond covering essential needs, they were classified as “choose to work”. Thus, reasons A–D were categorised as “choose to work”, while reasons E and F were classified as “forced to work”.

Findings

Based on the most recent data available at the time of writing (April 2022 to March 2023), the most frequently selected reason for continuing to work beyond the state pension age was B, “Not ready to stop work”, chosen by 2,328 respondents (see Figure 1). The second most common reason was E, “To pay for essential items”, selected by 640 respondents. These results suggest that the majority of individuals working past the state pension age are doing so because they have no intention of leaving the workforce in the near future. Overall, approx. 77% (3,064) of respondents were classified as choosing to work, while 23% (910) were classified as being forced to work.

The analysis of various demographic and socioeconomic factors sheds light on the nuanced reasons why individuals continue to work beyond their retirement age. The following findings are based on sample sizes for the four-year period: 2,875 respondents for April 2022 to March 2023, 3,258 for April 2021 to March 2022, 2,964 for April 2020 to March 2021, and 3,263 for April 2019 to March 2020.

The model’s performance, based on the concordance statistic (C-statistic), or area under the receiver operating characteristic curve (AUC), over the four years, shows moderate discrimination between outcome classes. The C-statistic values ranged from 0.6464 to 0.6666 across the periods, indicating that the model performs better than chance, though with room for improvement. Accuracy ranged from 0.5942 to 0.6491, reflecting reasonable, if not optimal, predictive capability. These results suggest that the model can differentiate between outcomes but may benefit from further refinement. Full models over the four-year period can be found in Table 1.

Table 1:

“Choose to work” or “forced to work” beyond state pension age, UK, April 2019–March 2020 and April 2022–March 2023 (dependent variable: 0, “choose to work”; 1, “forced to Work”)

	Apr19–Mar20				Apr20–Mar21
	B	Exp(B)	Lower	Upper	B	Exp(B)	Lower	Upper
Gender - Ref: “Male” vs. “Female”	0.277**	1.319	1.1	1.582	0.249*	1.283	1.041	1.582
Ethnicity - Ref: “White”	–				–
Asian	−0.366ns	0.693	0.377	1.276	0.281ns	1.325	0.667	2.631
Black	0.226ns	1.253	0.507	3.095	0.881ns	2.414	0.969	6.014
Other	−0.175ns	0.84	0.329	2.143	0.456ns	1.578	0.616	4.038
Long term illness - Ref: “With Illness” vs. “Without Illness”	−0.168ns	0.846	0.708	1.01	−0.060ns	0.942	0.77	1.153
Marital status - Ref: “Married/Co-habiting/Civil Partners” vs. “Non married”	0.440***	1.553	1.28	1.884	0.437***	1.548	1.235	1.94
SEC - Ref: “Higher managerial, administrative and professional Occp.”	–				–
Intermediate Occp.	0.504***	1.655	1.33	2.058	0.313*	1.368	1.074	1.742
Routine & manual Occp.	0.631***	1.879	1.49	2.369	0.440***	1.553	1.198	2.015
Region - Ref: “South”	–				–
North	−0.207*	0.813	0.663	0.998	−0.009ns	0.991	0.786	1.25
Wales	−0.221ns	0.802	0.576	1.116	0.455**	1.575	1.125	2.207
Scotland	−0.231ns	0.794	0.579	1.087	−0.228ns	0.796	0.523	1.211
Northern Ireland	−0.383ns	0.682	0.44	1.056	0.144ns	1.155	0.704	1.896
Accommodation type - Ref: “Owned outright/ rent free/ squatting” vs. “Mortgage/ Rented/ part rent & mortgage”	0.921***	2.512	2.086	3.025	0.888***	2.431	1.955	3.023
Constant	−2.062***	0.127			−2.389***	0.092

	Apr21–Mar22				Apr22–Mar23
	B	Exp(B)	Lower	Upper	B	Exp(B)	Lower	Upper
Gender - Ref: “Male” vs. “Female”	0.035ns	1.036	0.85	1.262	0.225*	1.253	1.036	1.515
Ethnicity - Ref: “White”	–				–
Asian	0.603*	1.828	1.109	3.012	0.790**	2.203	1.236	3.926
Black	1.033*	2.81	1.217	6.488	0.296ns	1.345	0.614	2.944
Other	0.203ns	1.225	0.567	2.647	0.158ns	1.171	0.449	3.056
Long term illness - Ref: “With Illness” vs. “Without Illness”	−0.174ns	0.84	0.696	1.014	−0.404***	0.667	0.555	0.803
Marital status - Ref: “Married/Co-habiting/Civil Partners” vs. “Non-married”	0.438**	1.55	1.261	1.905	0.448***	1.565	1.286	1.903
SEC - Ref: “Higher managerial, administrative and professional Occp.”	–				–
Intermediate Occp.	0.301*	1.351	1.072	1.702	0.315**	1.371	1.099	1.709
Routine & manual Occp.	0.475***	1.608	1.27	2.035	0.513***	1.671	1.327	2.105
Region - Ref: “South”	–				–
North	−0.036ns	0.964	0.78	1.192	−0.135ns	0.874	0.712	1.073
Wales	0.188ns	1.206	0.863	1.687	−0.354ns	0.702	0.482	1.023
Scotland	−0.384ns	0.681	0.458	1.013	−0.511**	0.6	0.42	0.858
Northern Ireland	−0.293ns	0.746	0.462	1.205	−0.664**	0.515	0.325	0.815
Accommodation type - Ref: “Owned outright/ rent free/ squatting” vs. “Mortgage/ Rented/ part rent & mortgage”	0.836***	2.308	1.889	2.82	0.775***	2.171	1.779	2.65
Constant	−2.138***	0.118			−1.654***	0.191

Note. “ns” (non-significant) indicates a p-value above the 5% level, * indicates significance at the 5% level, ** indicates significance at the 1% level, and *** indicates significance at the 0.1% level.

Gender disparities. The odds ratios presented in Figure 2 suggest that females are more likely to be “forced to work” beyond their retirement age compared to males (as the reference group). This indicates a gendered aspect of post-retirement employment dynamics.

Figure 2:

Gender comparison

Some potential reasons for this gender disparity include the fact that women are more likely to take career breaks or work part-time to manage care-giving responsibilities, which often results in lower pension contributions. This situation can “force” some women to continue working beyond the state pension age of 66. In fact, a survey by HR magazine (tinyurl.com/jvycsj2n) found that 53% of female respondents aged 45 and over were concerned that their pension would not be sufficient to ensure financial independence. Additionally, half of the women in this age group reported that they would need to continue working beyond retirement age to manage their finances.

Ethnicity trends. Among Asian respondents, the odds of being “forced to work” have progressively increased over recent years. Specifically, the odds ratio rose from 1.325 in April 2020–March 2021, to 1.828 in April 2021–March 2022, and further to 2.203 in April 2022–March 2023, compared to their white counterparts (as the reference group; see Figure 3). These values indicate a marked increase, though only the confidence intervals from the two most recent years (April 2021–March 2022 and April 2022–March 2023) do not overlap with 1.0, suggesting statistical significance. For black respondents, the odds ratio in April 2021–March 2022 was notably high at 2.810 (confidence interval (CI) 1.217–6.488), suggesting increased likelihood compared to white respondents. However, the confidence interval is wide, indicating greater uncertainty in the estimate for this year. For respondents categorised as “other”, no statistically significant differences were found when compared to white respondents in any of the observed years, as the confidence intervals consistently overlap with 1.0.

Figure 3:

Ethnicity comparison

Health status. As indicated in Figure 4, individuals without long-term illnesses have lower odds of being “forced to work” compared to those with such illnesses (as the reference group). This also suggests that individuals without long-term illnesses have higher odds of “choosing to work” beyond their state pension age. For example, during the period from April 2022 to March 2023, individuals with long-term illnesses had 1.50 times higher odds of “choosing to work” (derived from the reciprocal of 0.667, the odds ratio for being “forced to work”). Although the odds ratios remained consistently below 1 over the four-year period, the wide confidence intervals crossed 1, indicating that the finding was statistically significant only in the most recent year. This result has important implications for policies on retirement age and support systems for individuals with long-term illnesses.

Figure 4:

Health status comparison

Marital status. Individuals who are not married or living on their own consistently have higher odds of being “forced to work” beyond their state pension age, compared to married/co-habiting/civil partners (as the reference group). According to a recent report by the Pensions and Lifetime Savings Associations (tinyurl.com/mr4xu8ec), lone pensioners need £31,300 per year to secure a “moderate” living standard, compared to £43,100 for pension-aged couples. This suggests that lone pensioners need to secure a significantly larger pension on their own compared to retired couples. This could explain the consistently higher odds of being “forced to work” observed among older workers living alone (see Figure 5).

Figure 5:

Marital status comparison

Socioeconomic classification. Individuals in “intermediate” and “routine and manual” occupations have higher odds of being “forced to work” compared to those in “higher managerial, administrative, and professional” roles (as the reference group). Moreover, individuals in “routine and manual” occupations appear to face even higher odds than those in “intermediate” occupations (see Figure 6), potentially highlighting a gradient of increasing financial pressure or necessity for continued employment in lower-skilled roles.

Figure 6:

Occupation comparison

Regional differences. Figure 7 shows that individuals in Wales had higher odds of being “forced to work” than those in the South of England, with an odds ratio of 1.575 (95% CI 1.125–2.207) during the period from April 2020 to March 2021. In contrast, during the most recent year (April 2022 to March 2023), individuals in Scotland and Northern Ireland had lower odds of being “forced to work” than those in the South of England, with odds ratios of 0.600 for Scotland and 0.515 for Northern Ireland. This means that individuals in Scotland and Northern Ireland were 1.67 (1/0.600) and 1.94 (1/0.515) times more likely to “choose to work” than their counterparts in the South of England. However, in other cases where the confidence intervals for odds ratios crossed 1, no statistically significant differences were identified. Comparisons are made with England as the reference group.

Figure 7:

Region comparison

Accommodation type. As shown in Figure 8, individuals who still have mortgage payments are more likely to continue working beyond their retirement age. This financial obligation is frequently reported in the media as a significant factor contributing to post-retirement employment, indicating that housing and renting costs can be a major driver for older individuals remaining in the workforce. Comparisons are made with homeowners (those without mortgage payments) as the reference group.

Figure 8:

Housing status comparison

Summary

Overall, the distinction between “choosing to work” and being “forced to work” reveals the complex dynamics shaped by respondents’ unique situations and circumstances. Certain groups of older workers, such as females, single and non-cohabiting individuals, those in intermediate or routine and manual occupations, and those with mortgages or rent payments, are more likely to be “forced to work” beyond the state pension age compared to benchmarking groups. In an increasingly ageing society facing rising living costs, it seems sensible for policy-makers to consider designing and delivering comprehensive support tailored to these specific groups of older workers who may feel compelled to continue working beyond the retirement age of 66.

Conclusion: making retirement fairer

In the complex landscape of retirement in the UK, the threads of choice and necessity intertwine, painting a nuanced portrait of post-career life. As policymakers navigate this landscape, it seems reasonable to explore targeted interventions aimed at supporting vulnerable groups of older workers facing financial pressures. By better understanding the multifaceted motivations and circumstances driving post-retirement employment, we may pave the way for a more inclusive and equitable retirement experience for everyone.