Age Divides, Families, and Policy Preferences

By convention, we should expect young and old to demand different things from the government due to their distinct assets and vulnerabilities, with implications for how differently aged voters may evaluate rival parties’ policy offers (Vlandas Reference Vlandas2022).

In terms of policy preferences, older voters may show higher support for welfare programmes of which they are net beneficiaries (Bremer and Bürgisser Reference Bremer and Bürgisser2023; Meltzer and Richard Reference Meltzer and Richard1981) and that combat their life-cycle welfare risks (Rehm Reference Rehm2016). With some exceptions (Garritzmann et al. 2018, 856), older adults tend to be more enthusiastic about investment in retirement-oriented social consumption programmes such as pensions, rather than social investment policies such as state education and childcare (Banducci et al. Reference Banducci, Elder, Greene and Stevens2016; Bremer and Bürgisser Reference Bremer and Bürgisser2023; Busemeyer and Garritzmann Reference Busemeyer and Garritzmann2017; Busemeyer and Lober Reference Busemeyer and Lober2020; Cattaneo and Wolter Reference Cattaneo and Wolter2009; Chrisp and Pearce Reference Chrisp and Pearce2019; Elder and Greene Reference Elder and Greene2016; Häusermann et al. Reference Häusermann, Pinggera, Ares and Enggist2022; de Mello et al. Reference de Mello, Schotte, Tiongson and Winkler2017; Munger Reference Munger2022, 123–43; Vlandas Reference Vlandas2022, Reference Vlandas2023). Because of intergenerational inequalities in property access in many societies (Flynn Reference Flynn2020) and because of perceptions of one’s house value as retirement insurance, older adults also drive much of the negative association between homeownership and support for housing construction (Hankinson Reference Hankinson2018).

It follows that we should also expect age-related heterogeneity in economic voting, with old and young sanctioning parties based on a personal cost-benefit analysis according to their age-based economic exposure (Duch and Stevenson Reference Duch and Stevenson2008). Recent work depicts ‘ego-tropic’ economic voting by individuals beset by shocks linked to specific government policies or mismanagement of the national economy (Hobolt and Tilley Reference Hobolt and Tilley2016; Larsen Reference Larsen2021; Tilley, Neundorf, and Hobolt Reference Tilley, Neundorf and Hobolt2018). The growth of a retired electorate dependent on fixed incomes from assets and pensions rather than expanding employment opportunities and wage growth has consequences for holding the government accountable. Vlandas (Reference Vlandas2022) demonstrates that older electorates are more likely to reward incumbent governments for overseeing house price increases and low inflation and less likely to punish them for weaker economic growth and unemployment. Other studies link older electorates to higher budget deficits and debt burdens on future generations (Buchmeier and Vogt Reference Buchmeier and Vogt2024) and, cross-nationally, there is a negative association between a country’s share of retirees and its overall economic performance, with adverse consequences for younger workers (Vlandas Reference Vlandas2023).

A narrow reading of self-interest would, therefore, predict that the era of ‘grey power’ politics produced by ageing electorates could further exacerbate intergenerational inequality (Buchmeier and Vogt Reference Buchmeier and Vogt2024; Cantó et al. 2020; Chrisp and Pearce Reference Chrisp and Pearce2019; Dorothée et al. 2019; Flynn Reference Flynn2020; Joyce and Xu Reference Joyce and Xu2019; Munger Reference Munger2022; Rahman and Tomlinson Reference Rahman and Tomlinson2018; Vlandas Reference Vlandas2022, Reference Vlandas2023). However, we also question the idea that one can understand preferences or predict the consequences of ageing electorates through traditionally narrow readings of self-interest alone. One modification could be ‘enlightened self-interest’, given that pension pay-outs are directly tied to the productivity of younger taxpayers (Cattaneo and Wolter Reference Cattaneo and Wolter2009, 226). However, more fundamentally, it is important to realize that the quest for personal gratification is tempered by the considerations of one’s social ties. Group loyalty induces individual sacrifice for the welfare of the community (Rueda and Stegmueller Reference Rueda and Stegmueller2019, 137). Awareness of the needs of others may influence political attitudes by undermining negative stereotypes about the disadvantaged and generating a greater sense of empathy (Newman Reference Newman2014). Repeated interactions with those in a different economic position might present convincing arguments for how government might manage society differently, and the social pressure of this environment over an extended time period might produce conscious or unconscious imitation of their politics (Lindh et al. Reference Lindh, Andersson and Volker2021). Empirically, conditions in one’s local community (Bisgaard et al. 2016; Rogers Reference Rogers2014), the wellbeing of co-ethnics (Green et al. Reference Green, Hellwig and Fieldhouse2022) and friends (Lindh et al. Reference Lindh, Andersson and Volker2021; Newman Reference Newman2014) are predictive of vote choice and redistribution attitudes.

Here, we focus on the potential influence of close family members in different age groups. While not ruling out other contact effects, for instance, from friends or neighbours (Bisgaard et al. 2016; Lindh et al. Reference Lindh, Andersson and Volker2021; Newman Reference Newman2014; Rogers Reference Rogers2014), we theorize that, due to their ubiquity, duration, emotional intimacy, and interdependencies, familial relationships may dominate when it comes to capacity for generating intergenerational solidarity (Bengston and Oyama Reference Bengston, Oyama, Cruz-Saco and Zelenev2010; Soliz and Rittenour Reference Soliz, Rittenour and Giles2012). We argue that familial ties between older voters and their adult children, grandchildren, nieces, and nephews (et cetera) place a natural limit on the willingness of the former to over-tax and under-invest in the next generation (cf. Buchmeier and Vogt Reference Buchmeier and Vogt2024; Chrisp and Pearce Reference Chrisp and Pearce2019; Vlandas Reference Vlandas2022, Reference Vlandas2023) for two reasons: emotional bonds and shared risks.

Emotional bonds: Connections between older citizens and their younger family members remain strong, pervasive, and emotionally meaningful (Bengston and Oyama Reference Bengston, Oyama, Cruz-Saco and Zelenev2010; Soliz and Rittenour Reference Soliz, Rittenour and Giles2012). Forty to sixty per cent of adult children with a living parent in Germany, Britain and the US report weekly face-to-face contact with them, rising to 80–90 per cent once letters, telephone calls and electronic contact are included (Kalmijn and De Vries Reference Kalmijn and De Vries2009, 267). Family networks provide an important arena for intergenerational contact and a medium by which signals about the difficulties facing younger generations in accessing affordable childcare, housing, and educational opportunities (et cetera) might be transmitted to older generations. Familial bonds also give older adults a reason to act on these signals. Intergenerational relationships remain a highly important source of companionship, reassurance, support, and encouragement (Bengston and Oyama Reference Bengston, Oyama, Cruz-Saco and Zelenev2010, 44–45; Dykstra and Fokkema Reference Dykstra and Fokkema2011; Fingerman et al. Reference Fingerman, Kim, Tennant, Birditt and Zarit2016; Swartz Reference Swartz2009, 194–196; Thomas et al. Reference Thomas, Liu and Umberson2017). Conflicted or ambivalent familial relations are less common (Boersch-Supan et al. Reference Boersch-Supan, Heller and Reil-Held2011; Katz Reference Katz2009). Research in gerontology indicates that quality of life for older adults is positively associated with active relations with one’s non-co-resident children and grandchildren, which provide an important sense of esteem and identity (Danielsbacka et al. 2022; Katz Reference Katz2009; Thomas et al. Reference Thomas, Liu and Umberson2017, 5). Conversely, worries about young adult relatives and the stress of having to support them are associated with poorer sleep quality for mothers and fathers in their fifties and sixties (Seidel et al. Reference Seidel, Yorgason, Polenick, Zarit and Fingerman2018). Exposure to children with poor economic and social circumstances has also been linked to rates of parental depression (Fingerman et al. Reference Fingerman, Cheng, Birditt and Zarit2012; Tosi Reference Tosi2020).

From this first perspective, it stands to reason that older adults would wish to protect the wellbeing of their loved ones, both by supporting policies targeted at younger adults and by punishing parties that fail to advocate for them at the ballot box.

Shared risks: Relatives of a struggling younger adult are subject to negative externalities. Declining homeownership rates among younger European adults have, for instance, increased the numbers filling the spare rooms of their parental homes or requiring financial contributions from parents to access the housing ladder (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022, 71; Flynn Reference Flynn2020, 327; Jabbour Reference Jabbour2024). ‘Dependents’ are remaining dependent for longer. The scale of financial assistance given by older adults is often very large. A 2019 British study found that the average parental contribution for homebuyers totalled over £24,000 (a £6,000 increase on 2018), enough to make ‘the bank of mum and dad’ the 10th largest mortgage lender (BBC 2019). In the USA, polling found that 45 per cent of parents with adult offspring gave their children money during the coronavirus pandemic for help with housing, debts, groceries, bills and other expenditures (Fox Reference Fox2021). One-fifth gave over $1,000, one in twelve gave over $5,000, and 79 per cent of givers said the funds would have otherwise gone towards their own personal finances. In one sense, we can interpret these trends as a sort of ‘new social risk’ (Taylor-Gooby Reference Taylor-Gooby and Taylor-Gooby2004) for older adults stemming from rising intergenerational inequality in combination with norms of familial obligation and caregiving. An increasing number of older adults are facing the real prospect of continuing to provide for their adult children for an extended period (Flynn Reference Flynn2020, 338). Knowing this, many older adults might sense an (indirect) self-interest in social provision for their close younger family members so that they may not have to provide substantial private support themselves.

Why would older voters want political interventions to assist their younger relatives rather than just passing on money directly (which higher taxes might interfere with)? Personal provision is not always the best or most practical option. As the aforementioned sums suggest, the amounts that get passed down to support young adults with living expenses (particularly housing) tend to be large. This makes it plausible that shared tax rises associated with a government expanding provision for the young (and by association one’s relatives) may prove cheaper to older voters. Furthermore, many older adults may not feel able to offer as much personal support as they would like. While working-class parents may have the fewest ‘spare’ resources, even wealthier parents may feel unable to give ‘enough’ due to their correspondingly higher aspirations for their children (Albertini and Radl Reference Albertini and Radl2012). Large sums, including one’s own property, can be passed through inheritance. However, longer lifespans mean that inheritance beneficiaries are most often in their 50s–60s rather than in their crucial young adult years (Sefton and Falkingham Reference Sefton and Falkingham2023). Older adults might also place a high value on self-determination and independence among their relatives (Albertini and Radl Reference Albertini and Radl2012, 111).

‘Emotional bonds’ and ‘shared risks’ are not inherently contradictory. They both stem from exposure to a struggling relative, and one must care about one’s family members to even consider assisting them financially where laws do not compel this. However, they are theoretically distinct. The former consideration is rooted in heightened concern for the younger family member, the latter more in financial self-interest. Ascertaining which of these motivations dominates would, in one small way, allow us to speak to the wider debate in political science about whether voters are fundamentally egotropic or other-regarding in their preferences (Alesina and Giuliano Reference Alesina, Giuliano, Benhabib, Bisin and Jackson2011).

In theorizing that family matters for the policy and party preferences of older adults, we build on emerging literature that highlights the role of childrearing in the formation of political attitudes. Parents of young children demonstrate greater support for government spending on childcare and education (Banducci et al. Reference Banducci, Elder, Greene and Stevens2016; Bremer and Bürgisser Reference Bremer and Bürgisser2023; Elder and Greene Reference Elder and Greene2016; Greenlee Reference Greenlee2014; Urbatsch Reference Urbatsch2014; Burlacu and Lühiste Reference Burlacu and Lühiste2021). Other research links the raising of boys to lower support for military intervention (Urbatsch Reference Urbatsch2014) and the raising of girls to more feminist attitudes (Glynn and Sen Reference Glynn and Sen2015; Greenlee et al. Reference Greenlee, Nteta, Rhodes and Sharrow2020). This implies that the impact of the family is not just about personal material self-interest but is connected to a deeper shift in one’s own identity and convictions, as well as the development of a (potentially life-long) sense of stewardship.

Unlike these previous studies, we shift focus from the impact of young children (under 18) to young adult relatives: close family members that voters are (in most countries) no longer legally responsible for but, nonetheless, have strong (emotional or financial) incentives to protect. The parents of young children’s support for government investment in childcare or education may be explainable by classic egocentric household-income maximising behaviour.Footnote ¹ Such services supplement the voter’s own income (directly or by preventing them from needing to temporarily withdraw from the labour market) rather than transferring resources directly to their child’s generation per se. Wanting government investment in services that one does not use oneself and has no legal obligation to provide another, however, is more challenging to conventional models of individualistic political preferences.

In this respect, we are following Goerres and Tepe (Reference Goerres and Tepe2010, Reference Goerres and Tepe2012), who find that older adults who endorse notions of familial duty and regularly visit their adult children show greater support for state childcare, and Jabbour (Reference Jabbour2024), who shows that older adults cohabiting with young adult children evaluate government and the national economy more negatively. While our own hypotheses, if accurate, will corroborate these papers regarding the power of familial influences to constrain age divisions in politics, we build on them in the following ways. We focus specifically on older adults with struggling young relatives rather than all adults with young relatives (as with Goerres and Tepe) or, more narrowly, on those cohabiting with their children (as with Jabbour). We unify methodologies by measuring the impact of familial considerations on both policy and party support for a tougher test of our central theory. If correct, it should be the older relatives of financially struggling young adults in particular that should become more supportive of both a wide range of ‘pro-youth’ policies (for example, childcare, education, and housing) and less tolerant of parties that are seen to poorly represent the interests of young adults. We also attempt to empirically distinguish between pure altruism and disguised financial self-interest in the motivations that older adults have for wanting the government to help young people. Finally, empirically, we also enhance the credibility of the familial voting hypothesis through strong controls for the subjective and objective wellbeing of the respondent/older adults themselves, as well as by presenting discriminant validity tests that rule out the possibility of respondents simply desiring a larger welfare state in general.

From our theoretical discussion, we derive the following core hypotheses:

To ascertain whether ‘emotional bonds’ or ‘shared risks’ drive these results, we also stipulate two further rival hypotheses that assert the importance of either mechanism:

We assume that, given these are the only two hypothesized explanations for why exposure to a struggling younger relative motivates support for pro-young policies and parties, if a variable tapping shared risks cannot explain the association between these two variables, emotional bonds must be crucial.

We focus on the attitudes of older adults toward the wellbeing of their younger relatives due to the existing patterns of intergenerational inequality, which favour the old, and the frequency of accusations of selfishness against older generations in our case country (see below). However, we make no assumption that the young never vote altruistically on behalf of older family members, particularly in other contexts where patterns of intergenerational inequality are reversed. Our hypotheses may be equally applicable to the study of when younger adults support spending on the old in times and places where this question is politically salient.

In what follows, we introduce the survey that we designed to test our hypotheses before carrying out the analysis itself. While we are reliant on a unique and rich observational dataset, we employ many robustness and discriminant validity tests to rule out the possibility of competing explanations for our findings.

Case Selection and Data

We designed a large, original survey [Intergenpol-GB] about attitudes to policies commonly cited in discussions of intergenerational equality and fielded it to a representative sample of 6,021 people in Britain in August 2022, including over 4,458 adults aged 40+, whom we focus on here.Footnote ² Respondents were selected from participants in the May 2022 wave of the long-running British Election Study Internet Panel [BESIP], making it possible to use certain variables from this study. Appendix Table A1 notes the precise source for each variable alongside descriptive statistics.

Britain is an informative case. Like many other Western countries, it has experienced high levels of intergenerational inequality in wealth and economic insecurity (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022; Flynn Reference Flynn2020; Green and de Geus Reference Green and de Geus2022; Joyce and Xu Reference Joyce and Xu2019; Rahman and Tomlinson Reference Rahman and Tomlinson2018). A near majority of British adults believe that ‘today’s youth will have a worse life than their parents’ (Rahman and Tomlinson Reference Rahman and Tomlinson2018, 10). There is also a heated debate about whether the political decisions of the incumbent Conservative government – and, by extension, their disproportionately older supporters – are responsible (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022; Pickard Reference Pickard2019; Toynbee Reference Toynbee2023). However, while the opportunity for older adults to demonstrate intergenerational altruism is just as evident as it would be in many other countries with similar economic circumstances, Britain could also be considered a ‘least likely’ case for uncovering ‘family-centric’ political behaviour (Eckstein Reference Eckstein, Greenstein and Polsby1975).

Along with the USA and other countries in northern Europe, Britain has an individualistic culture, with a weaker role for extended family ties and norms of familial obligation (Lowenstein Reference Lowenstein, Cruz-Saco and Zelenev2010; Reher Reference Reher1998). Unlike certain peer countries, parents in Britain have no legal obligation to support their children post-18 (Chevalier Reference Chevalier2016). Historically, the state has also provided less targeted support – for instance, vocational training schemes – to foster the independence and productivity of young labour market entrants than countries such as Germany or Sweden (Chevalier Reference Chevalier2016), so there may be fewer expectations that the government should provide these services. Furthermore, Britain has recently experienced acrimonious age-related polarization around a range of sociocultural issues concerning immigration, national identity, and the European Union (Sobolewska and Ford Reference Sobolewska and Ford2020), which may reduce the willingness of older adults to support younger cohorts that largely reject their social values. Such issues also cross-pressure nominally altruistic older adults, who must trade-off their desire to support an opposition party, Labour, that is generally seen to more closely represent young people’s interests (Grant et al. Reference Grant, Green and Evans2022). The timing of our survey, late 2022, came during a particularly difficult time, with a mounting inflation crisis prompting widespread anxiety (ONS 2022). While this may have exacerbated the financial difficulties of younger adults, the effects were also widespread in the population as a whole; the tendency towards self-interest may have been heightened in such circumstances. If family-centric voting occurs here, it likely does elsewhere.

In our analysis, we focus on those aged 40+, their perceptions of the wellbeing of their close younger relatives in their late teens, twenties, and thirties, as well as their support for policies aimed at that age group and their party preferences. While somewhat arbitrary, we choose 40 as the relevant cut-off point between ‘younger’ and ‘older’ respondents for two reasons. Firstly, the need for the ‘young adult-oriented policies’ that we canvas support for (childcare, education, and new housing opportunities) peaks among those in their late teens, twenties and thirties and is generally much lower thereafter (Figure A1 in the Appendix). The same also largely applies to actual support for these policies (Figure A2). This minimises the possibility that we are picking up on a direct self-interest in these policies among the 40+-year-old respondents themselves. Second, it is very uncommon to have young adult (that is, aged 18–39) children – the closest sort of young adult family member one can have – prior to 40, but most respondents in their late forties and over do have children or grandchildren in this age bracket (Figure A3). Supplementary analyses presented in our Appendix, described beneath each table below, demonstrate the robustness of our findings in narrowing our scope to those aged 60+ to better capture the behaviour of ‘grey voters’ specifically (Chrisp and Pearce Reference Chrisp and Pearce2019; Vlandas Reference Vlandas2022).

Our principal independent variable of interest, the hypothesized catalyst for family-centric voting, is the respondents’ evaluation of the financial wellbeing of their younger relatives. We asked: ‘thinking about your own close family, how well are family members in the following age groups doing financially, on average? … Close family members in their late teens, twenties, and thirties’. Respondents could place their evaluations on a scale between 0 (‘doing very badly’) to 10 (‘doing very well’) or, if they had no relatives in that age group, they could select ‘Not sure/not applicable, I do not know any family members in that age group’.

By asking about ‘close family members’ in general, rather than children specifically, we allow respondents to define their own familial ‘in-group’. Certain respondents (especially those without their own children) may feel highly attached to their nieces, nephews, or grandchildren. That said, it is conceivable that parents might feel more emotionally invested than (for instance) aunts or uncles who have not put in the years of intensive work required to raise a child and that this might have political consequences. Accordingly, Table A2 demonstrates that our findings are robust to limiting our sample just to parents of children aged 18–39. Table A3 tests whether having such a child moderates the impact of perceptions of younger relatives’ finances on our various political outcome measures, but we find only limited support for this.Footnote ³

While under one-third (30.9 per cent) of our sample stated that they had no close young adult relatives to evaluate, we avoided dropping this group from our analysis. Instead, we recoded our survey item into a categorical variable. This distinguished those who perceived their younger relatives to generally be doing badly financially (0–4/10 on the original scale) from those who perceived them to be doing well (6–10/10), neither badly nor well (5/10), or those that had no relatives to evaluate. It is only those in the first of these four categories that we expect to be particularly ‘pro-young’ politically. Our results are entirely robust to changing these thresholds to 0–3, 4–6, and 7–10.

An identical question about the perceived wellbeing of respondents’ close relatives ‘in their forties and fifties’ (that is, 40–59) and ‘in their sixties, seventies and above’ (that is, 60+) allows us to benchmark perceptions of the wellbeing of young adults. Figure 1 shows the percentage of our sample (by age) reporting relatives in each of three age groups (18–39, 40–59, 60+) that are doing ‘badly’ financially (that is, 0–4/10). The denominator includes those without close relatives in each age group. Figure 1 shows that respondents perceive patterns very consistent with aggregate statistics about increasing intergenerational economic inequality (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022; Cantó et al. 2020; Dorothée et al. 2019; Flynn Reference Flynn2020; Joyce and Xu Reference Joyce and Xu2019; Munger Reference Munger2022; Rahman and Tomlinson Reference Rahman and Tomlinson2018). Around 1-in-4 people aged 40+ (including 1-in-3 in their fifties to early sixties) have struggling younger family members, with this only being less common than having struggling relatives in other age groups among the over-70s, and, even then, the difference is statistically null.

Figure 1. The percentage of British adults with financially struggling relatives in different age cohorts.

Source: Intergenpol-GB (August 2022). N = 4,458 people aged 40+.

Note: The percentage of people with relatives aged 18–39 (solid black line), 40–59 (short dashed grey line), and 60+ (long dashed grey line) who are ‘doing badly’ financially by age, plus 95 per cent confidence intervals. Estimates were smoothed based on a local regression function.

While subjective evaluations of the wellbeing of others are fairly common in political science (for example, Green et al. Reference Green, Hellwig and Fieldhouse2022; Newman Reference Newman2014); Table 1 provides validation of our measure. Here, we present the results from OLS regressions of subjective assessments of younger relatives’ financial wellbeing (using the original 0–10 scale) on a categorical indicator noting if the respondent’s eldest child (if aged 18–39) holds objective assets predicting economic security in contemporary Britain – a university degree or property (Green and de Geus Reference Green and de Geus2022).

Table 1. The objective assets of one’s eldest child predict subjective financial evaluations of one’s younger relatives in general

* = p < 0.05; ** = p < 0.01; *** = p < 0.001

Source: Intergenpol-GB (August 2022).

Note: OLS Models. All respondents aged 40+. See Table A5 for full results. Interpretation: Positive coefficients indicate that this type of respondent has more positive evaluations of the financial wellbeing of their close relatives aged 18–39 (Models A-D), 40–59 (Model E) or 60+ (Model F) than those in the reference category (those whose eldest child aged 18-39 has neither property nor a degree). Models D-E are discriminant validity tests (DVTs) demonstrating that one’s child’s assets only influence evaluations of the wellbeing of their child’s age cohort rather than one’s relatives in general.

DVs: Models A-C: Evaluation (0–10, ‘doing badly’ to ‘doing well’) of ‘close family’ aged 18–39 [M = 5.2, SD = 2.1]. Model D: Evaluation of ‘close family’ aged 40–59 [M = 5.8; SD = 1.9]. Model E: Evaluation of ‘close family’ aged 60+ [M = 6.1; SD = 2.2].

Demographic Controls: Age; Gender; Education; Logged Equivalised HH Income; Logged HH Savings; Logged Property Value; NRS Social Class; Employment Status; Ethnicity; Region. Personal/National Economic Outlook Controls: Subjective Personal Financial Outlook; Subjective National Economic Outlook.

We present the coefficients from bivariate models (Model A) and two (Models B and C) controlling for the assets, characteristics, and personal/national economic evaluations of the respondents themselves in order to isolate evaluations of one’s relatives from personal pocketbook and broader sociotropic economic evaluations (Duch and Stevenson Reference Duch and Stevenson2008). Model B adds demographic controls for respondents’ age, gender, education (university degree vs no degree), logged household income equivalised for household size, logged household savings (no savings = £0), logged property value (non-homeowners = £0), occupational class (using Britain’s National Readership Survey [NRS] classification, see Table A1 for details), employment status (full-time worker; part-time worker; retired; unemployed; other), ethnicity (white British; other whites; Asian; Afro-Caribbean; other), and region (northern England, midlands England, southern England, London, Scotland, and Wales). Missing values were imputed.Footnote ⁴ Model C adds measures of respondents’ subjective personal and national economic outlooks. Personal financial outlook sums up two questions asking a) ‘How does the financial situation of your household now compare with what it was 12 months ago?’ and b) ‘How do you think the financial situation of your household will change over the next 12 months?’, both measured from 1–5 (‘got/get a lot worse’ to ‘get/got a lot better’). National economic outlook was captured by summing answers to two questions with identical response scales: a) ‘How does the general economic situation in this country now compare with what it was 12 months ago?’ and b) ‘How do you think the general economic situation in this country will change over the next 12 months?’ Further information and descriptive statistics for all variables can be found in Table A1, and the results of supplementary robustness tests (for example, controlling for respondent partisanship and ideology or limiting the sample to the over-60s) in Table A4. Table 1 displays only our primary independent variables of interest, but Table A5 provides control variable coefficients.

Subjective evaluations of the wellbeing of one’s younger relatives tap into objective reality. In the bivariate model (Model A), parents whose eldest child holds a degree and has property rate their younger relatives’ financial wellbeing 1.4 points (0.65 standard deviations [SDs]) more positively than parents whose child has neither. Once parental characteristics and other economic perceptions are held constant (Models B and C), the gap between the parents of non-propertied children with and without degrees is no longer significant, but parents whose children have both assets still evaluate them substantially (1 point/0.5 SDs) more positively than do those whose children have neither. Models D and E provide discriminant validity tests [DVTs] our findings: the assets of one’s children have no relationship to the perceived wellbeing of one’s middle-aged and older relatives.

The robustness of the coefficients associated with a child’s assets to controls for parental wealth suggests constraints on the extent to which parents believe they can guarantee the wellbeing of their younger relatives privately (Albertini and Radl Reference Albertini and Radl2012). We expect this to motivate support for greater government intervention instead. In the following three analytic steps, we demonstrate the implications of young adults’ economic insecurity for the political preferences and behaviour of their older family members.

Testing H1: Struggling Young Relatives and Policy Support

Our first hypothesis was that negative evaluations of younger family members’ finances positively predict support for greater state spending on young adults. To test this, we require survey items concerning support for increased investment in services predominantly used by young adults. Accordingly, we asked respondents: ‘Would you support or oppose increased government spending on the following, even if it leads to increasing taxes or a reduction of spending elsewhere?’ followed by:

1. a) ‘Increasing free vocational or technical education for non-University students’

2. b) ‘Increasing free education for University students’

3. c) ‘Increasing the amount of affordable housing built in your local area’

4. d) ‘Increasing the amount of council housing built in your local area’

5. e) ‘Increasing free childcare for pre-schoolers’

Responses were scaled from 1 (‘strongly oppose’) to 5 (‘strongly support’). These specific policies were chosen because young British adults and their advocates have repeatedly cited the lack of access to affordable university education, vocational training, childcare, and housing as major grievances (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022; Cominetti et al. Reference Cominetti, Costa, Eyles, Henehan and McNally2022; Topping Reference Topping2021). Our data also shows that the need – and, in most cases, support – for greater investment in these services is higher among our younger respondents (Appendix Figures A1 and A2). That said, we acknowledge the specific policies that are deemed ‘pro-youth’ will be different in contexts where the relationship between (for example) age and homeownership or participation in higher education looks different.

Figure 2. A summary of our key findings and the possibility of their mediation by perceptions of self-interest

Source: Intergenpol-GB (August 2022).

Note on Interpretation: The figure presents coefficient plots (Jann Reference Jann2014) that summarize our main findings (Models A–D in Tables 2–4, in addition to Table A15 for the final ‘potential self-interest mediator’ model). They show how many standard deviations more / less likely, on average, a person with a financially struggling relative is to support spending on YAOPs (left column), to prioritize spending on YAOPs (centre column), and to intend to vote Conservative (right column). This is in comparison with someone whose young adult relatives are ‘doing well’ (top row) or someone with no close young adult relatives (bottom). All respondents are aged 40+. See the original tables for model specification details. All lines are 95 per cent confidence intervals. Demographic Controls: Age; Gender; Education; Logged Equivalised HH Income; Logged HH Savings; Logged Property Value; NRS Occupation Social Class; Employment Status; Ethnicity; Region. Personal/National Economic Outlook Controls: Subjective Personal Economic Outlook; Subjective National Economic Outlook. General Tax-Spend Preference Controls: General Tax-Spend Attitudes.

Note that we also prompt respondents to consider the personal costs of supporting these measures by referencing potential tax rises or spending cuts elsewhere. This is designed to maximize the validity of answers by forcing consideration of budgetary trade-offs. For similar reasons, we reference proposals for more housing ‘in your local area’ to induce considerations about the implications for one’s own property’s value (Hankinson Reference Hankinson2018). We further distinguish ‘university’ and ‘non-university’ tertiary education and ‘affordable’ and ‘council’ housing.Footnote ⁵ Thus, we offer policies that would seem more familiar and useful to less affluent, working-class respondents (that is, apprenticeships and council housing) but also the aspirational middle classes (that is, university education and a route to homeownership). Likewise, education is more relevant to those in their early twenties, and childcare and housing are more appealing to those in their thirties. Each respondent should then find items that would potentially assist their own family. Given that their internal consistency (Cronbach’s alpha = 0.78) produces a reliable scale, our dependent variable is the average support for each of the five ‘young adult-oriented policies’ [YAOPs]. However, Table A6 demonstrates that having struggling younger relatives predicts each item separately to show that we are not tapping a particular preference for social investment policies such as education over more consumption-based welfare programmes like subsidized housing (Häusermann et al. Reference Häusermann, Pinggera, Ares and Enggist2022).

Models A–D of Table 2 present the results of OLS regressions of average support for YAOPs on whether respondents have young adult relatives doing badly financially (the reference category), doing well, doing neither well nor badly, or if respondents have no young relatives at all. The coefficients for control variables are redacted but can be found in Table A7. Model A presents the bivariate associations, and then, to further isolate the effects of one’s relatives’ wellbeing from one’s own, Models B and C add controls for respondents’ own demographics and assets and their own personal and national subjective economic outlooks. Model D then also adds a control for respondents’ general attitudes towards government spending on health and social services – measured from 0 (‘cut taxes a lot and spend much less’) to 11 (‘raise taxes a lot and spend much more’) – to isolate support for YAOPs from being generally pro-state investment. Regardless of these controls, there is a clear association between the wellbeing of one’s relatives and one’s attitudes. In the bivariate model, respondents with relatives doing ‘badly’ are 0.32 and 0.39 points more supportive of YAOPs than those whose family members are doing ‘well’ or those without young relatives, respectively. This only declines to 0.24 and 0.32 in the highly conservative Model D. Table A8 demonstrates that these relationships are robust to various alternative model specifications, such as limiting analysis to the over-60s, listwise deletion of missing values, or additionally controlling for the respondents’ partisanship, and EU referendum stance.

Table 2. Financial evaluations of one’s younger relatives predict the extent to which one supports spending on young adult-oriented policies

* = p < 0.05; ** = p < 0.01; *** = p < 0.001

Source: Intergenpol-GB (August 2022).

Note: OLS Models. All respondents aged 40+. See Table A7 for full results. Interpretation: Negative coefficients indicate that this type of respondent is less supportive of YAOPs (Models A-E) or pension spending (Models F and G) than those in the reference category, that is, those whose close relatives in that age group are doing badly financially. Models D-E are discriminant validity tests (DVTs) demonstrating that evaluations of one’s older relatives do not also predict support for YAOPs and that evaluations of one’s young relatives do not predict pension support.

DVs: Models A–E: Average support for gov. raising spending on 5 YAOPs (1–5, ‘strongly oppose’ to ‘strongly support’) [M = 3.5, SD = 0.9]. Models F-G: support for raising pension spending (1–5, ‘strongly oppose’ to ‘strongly support’) [M = 4.0; SD = 1.1].

Demographic Controls: Age; Gender; Education; Logged Equivalised HH Income; Logged HH Savings; Logged Property Value; NRS Occupation Social Class; Employment Status; Ethnicity; Region. Personal/National Economic Outlook Controls: Subjective Personal Economic Outlook; Subjective National Economic Outlook. General Tax-Spend Preference Controls: General Tax-Spend Attitudes Scale.

Models E–G present further discriminant validity tests (DVTs) of our theory. As expected, financial evaluations of one’s older relatives do not influence support for YAOPs (Model E), and evaluations of one’s young adult relatives do not influence support for more spending on pensions (Model F), whereas the wellbeing of one’s older relatives does (Model G).Footnote ⁶ Consistent with our theory, concern about family members in a particular age group is associated with more support for state services appropriate to that stage in the life cycle. Interestingly, these concerns do not seem to make one less supportive of spending aimed at other age cohorts, suggesting that those with struggling relatives do not adopt a narrative of intergenerational conflict and do not view state investment in zero-sum terms.

Testing H2: Struggling Young Relatives and Policy Prioritization

While a voter might display familial altruism in a relatively unconstrained setting, it is possible that narrower self-interest motivations dominate in more realistic scenarios when presented with a trade-off between rival recipients of government investment (Bremer and Bürgisser Reference Bremer and Bürgisser2023; Busemeyer and Garritzmann Reference Busemeyer and Garritzmann2017; Busemeyer and Lober Reference Busemeyer and Lober2020; Garritzmann et al. 2018; Häusermann et al. Reference Häusermann, Pinggera, Ares and Enggist2022; de Mello et al. Reference de Mello, Schotte, Tiongson and Winkler2017). Our survey therefore included several young vs old policy trade-offs. The first was a forced-choice exercise that asked respondents to select three areas for additional government investment from a list of ten policies, four of which were YAOPs (childcare, housing, university and vocational education) and four were ‘older adult-oriented policies’ (pensions, elder care, winter fuel allowances for pensioners, and free pensioner transport). A further two non-age-related policies were also included (environmental spending and border force investment). The full question wording can be viewed in Table A1. Given that respondents had only three choices and ten options, it would be perfectly possible for them not to prioritize a single YAOP. In Models A-D of Table 3, we regress a dummy variable indicating that the respondents chose to prioritize a YAOP on their financial evaluations of their young adult relatives. Model A gives the bivariate associations, Model B adds controls for demographic characteristics, Model C personal and national financial outlooks, and Model D general tax-spend preferences. Coefficients for these controls are redacted but can be viewed in Table A9.

Table 3. Financial evaluations of one’s younger relatives predict whether one prioritises young adult-oriented policies in a trade-off task

* = p < 0.05; ** = p < 0.01; *** = p < 0.001

Source: Intergenpol-GB (August 2022).

Note: LPM Models. All respondents aged 40+. See Table A9 for full results. Interpretation: Negative coefficients indicate that this type of respondent is less likely to prioritize YAOPs (Models A-E) or older adult-oriented policies (Models F and G) than those in the reference category, that is, those whose close relatives in that age group are doing badly financially. Positive coefficients indicate that they are more likely to prioritize that type of policy. Models D–E are discriminant validity tests (DVTs) demonstrating that evaluations of one’s older relatives do not also predict prioritization of YAOPs and that evaluations of young relatives do not predict prioritization of spending on older adults (only evaluations of relatives of that age does).

DVs: Models A–E: Dummy variable indicating prioritization of at least one YAOP in a forced trade-off exercise [M = 0.65, SD = 0.48]. Models F-G: Dummy variable indicating prioritization of at least one ‘Older Adult-Oriented Policy’ in the same exercise [M = 0.83; SD = 0.38].

Demographic Controls: Age; Gender; Education; Logged Equivalised HH Income; Logged HH Savings; Logged Property Value; NRS Occupation Social Class; Employment Status; Ethnicity; Region. Personal/National Economic Outlook Controls: Subjective Personal Economic Outlook; Subjective National Economic Outlook. General Tax-Spend Preference Controls: General Tax-Spend Attitudes Scale.

Consistently, respondents who felt that their young adult relatives were doing badly financially were between 11 and 13 percentage points more likely to prioritize a YAOP than those who had no relatives of that age or had relatives doing well. Once again, this finding is substantially unaltered by limiting the analysis to the over-60s, listwise deletion of missing values, or additionally controlling for respondents’ partisanship or EU referendum stance alongside other alternative model specifications (Table A10). Models E–G present discriminant validity tests demonstrating that concerns about the wellbeing of relatives in a particular age group only enhance support for policy generally relevant to people at that stage in the life cycle and not social policy in general. Those with struggling older relatives were less likely to prioritize YAOPs (Model E). Conversely, those who had struggling younger relatives were less likely to prioritise services aimed at older adults (Model F), but those with struggling older adult relatives were more so (Model G).

While they are omitted here due to text constraints, Appendix Tables A11 and A12 present the association between concerns for younger family members and responses to two even more explicit trade-off questions. The first asks whether the respondent would prefer the government to ‘prioritise spending on the needs of younger adults’ or ‘the needs of older adults’; the second asks whether they want further spending on education at the cost of cutting pensions (Busemeyer and Garritzmann Reference Busemeyer and Garritzmann2017). In both cases, akin to all the results in this stage of our analysis, we find that adults over 40 (and those over 60 more specifically) with struggling younger relatives are more likely to prioritize state spending on youth-oriented services even at a cost to themselves.

Testing H3: Struggling Young Relatives and Vote Choice

H3 was that negative evaluations of younger family members’ finances positively predict anti-incumbent voting (H3a) and that this association is explained by negative evaluations of the incumbent’s ability to represent young people (H3b). That is, there exists a family-centric economic vote, and adults with struggling younger relatives will punish governing parties for failing to alleviate problems facing younger people. Familial considerations influence behaviour as well as attitudes.

To test this, we regressed vote intention on financial evaluations of younger adult relatives. For parsimony, we use a linear probability model that distinguishes supporters of the incumbent Conservative Party (1) from those of any opposition party (0). We present our results in Table 4, once again building up from a bivariate model to one simultaneously controlling for respondent demographics, personal and national economic outlook, and general tax-spend preferences. As usual, only the coefficients for our main independent variables are presented; see Table A13 for the full model.

Table 4. Financial evaluations of one’s younger relatives predict whether one supports the incumbent Conservative Party, and this association is mediated by perceptions of how well the Conservatives represent young people

* = p < 0.05; ** = p < 0.01; *** = p < 0.001

Source: Intergenpol-GB (August 2022).

Note: LPM Models. All respondents aged 40+. See Table A13 for full results. Interpretation: Positive coefficients for the ‘financial evaluations’ variable indicate that this type of respondent is more likely to intend to vote for the incumbent Conservative Party than those in the reference category, that is, those whose close relatives aged 18–39 are doing badly financially. Negative coefficients indicate that they are less likely to intend to vote for the Conservatives. The comparison between Models E and F/G demonstrates that any association between vote intention and the perceived wellbeing of one’s relatives disappears after controlling for perceptions of how well the Conservatives represent young people, but NOT after controlling for how well the party represents older people.

DV: Intended vote intention: incumbent Conservative Party (1) or any opposition parties (0) [M = 0.38; SD = 0.49].

Demographic Controls: Age; Gender; Education; Logged Equivalised HH Income; Logged HH Savings; Logged Property Value; NRS Occupation Social Class; Employment Status; Ethnicity; Region. Personal/National Economic Outlook Controls: Subjective Personal Economic Outlook; Subjective National Economic Outlook. General Tax-Spend Preference Controls: General Tax-Spend Attitudes.

Models A–D in Table 4 deliver support for H3a. Depending on the precise set of controls, older adults are between 0.19 and 0.08 percentage points more likely to reject the Conservatives if they believe that their younger relatives are struggling financially relative to when they believe that their younger relatives are doing well. In the abstract, over-40s with struggling younger relatives are also more likely to support opposition parties than those without younger relatives at all; however, this relationship is no longer statistically significant after adjusting for personal and national economic evaluations. This suggests that the primary gap of relevance in terms of vote choice is between those with more and less financially successful young relatives. Table A14 demonstrates the substantive robustness of our results to a range of alternative model specifications, including controlling for EU referendum and immigration preferences, two strong predictors of party choice in contemporary Britain (Sobolewska and Ford Reference Sobolewska and Ford2020).

H3b stated that the reason why adults with struggling younger relatives punish incumbent parties is because they deem them ineffective at representing the interests of younger people (and, by extension, their family members). Can we be sure that the relatives of struggling young adults are really shunning the Conservatives because of concerns about that party’s ability to help young adults rather than, say, concerns about the party’s general capabilities? We add credibility to our assumptions about the motivation behind anti-Conservative sentiment via a simple mediation analysis. Our respondents were asked, ‘Some people say that all political parties look after certain groups and are not so concerned about others. How closely do you think the Conservative Party looks after the interests of young people’ (1–4, ‘not at all closely’ to ‘very closely’)? If the impact of exposure to a struggling young family member on non-support for the incumbent party really is mediated by concern about their record on representing one’s relatives’ age group, the difference between having a younger family member doing ‘badly’ or ‘well’ should be nullified after the inclusion of this new variable. To ensure a clean comparison, we first repeat Model D using the subset of respondents who are non-missing on the representation question (Model E). Importantly, when we add this variable to our model (Model F), the previously robust association between having a struggling younger relative (v. a more successful one) and party choice disappears. This suggests that the relationship between anti-incumbent voting and having young adult relatives in financial difficulty really is reducible to concern about the government’s track records on representing young people rather than general dissatisfaction. Model G further enhances the credibility of this assessment: perceptions of how closely the Conservative Party ‘looks after the interests of retired people’ do not mediate the impact of having a struggling young relative, as we would expect.

A Summary of Our Tests of H1-H3

Figure 2 presents coefficient plots (Jann Reference Jann2014) that summarize our main findings from Tables 2–4. They show how many standard deviations (SDs) different on average, a person with a financially struggling relative (that is, the relative is ‘doing badly’/<4/10) is regarding their average support for spending on YAOPs (left column), their likelihood of prioritising spending on YAOPs (centre column), and of intending to vote Conservative (right column). This is in comparison with someone whose young adult relatives are ‘doing well’ (>5/10) (top row) and someone with no close young adult relatives (bottom). The top symbol in each grid represents the associations from bivariate models, with the following three symbols demonstrating the extent of change once the sequential addition of different groups of controls (as with models A–D in Tables 2–4).

The plot provides a visual representation of our evidence corroborating hypotheses H1–H3, as well as the substantive strength of the associations that have been uncovered. Relative to those with young adult relatives who are doing well, 40+ year-olds with financially struggling younger relatives are between 0.36 and 0.27 SDs more likely to support investment in YAOPs, 0.28 to 0.24 SDs more likely to prioritize this sort of investment, and 0.40 to 0.16 SDs less likely to support the Conservatives. Such adults are also 0.44 to 0.36 SDs and 0.28 to 0.26 SDs more likely to support and prioritize (respectively) investment in YAOPs than those without any young relatives at all; however, the vote choice gap between these groups is null after respondents’ broader economic perceptions are accounted for. Overall, then, we have robust evidence that having a financially precarious family member in their late teens, twenties, and thirties is associated with having uniquely pro-young political attitudes as well as, conditional upon having younger relatives, voting behaviour.

Testing H4: Are Family-Centric Preferences about Altruism or Pure Self-Interest?

While our selection of control variables and discriminant validity tests allow us to be relatively confident that we are not simply tapping respondents’ own economic anxiety, perceptions of the national economy, or generally pro-tax/spend inclinations, there may still be a role for self-interest. As we discussed, rather than a purely altruistic concern for their loved ones’ wellbeing due to the emotional bonds of family, our respondents may have been more wary of the ‘shared risks’ and financial obligations that they may incur themselves.

These two considerations are not necessarily mutually exclusive. However, to avoid fully ascribing to altruism what might more accurately be considered self-interest and to provide a formal test of H4a and H4b, we repeated the analysis in Model D of Tables 2–4 with the addition of a new variable capturing the perceived risk of having to personally support a struggling relative. This was measured by a survey item asking: ‘During the next 10 years or so, how likely or unlikely is it that you will need to give significant financial or practical assistance to a close family member in their twenties and thirties with childcare or housing needs’ (1–5, ‘very unlikely’ to ‘very likely’). This question does not cover all of the challenges that a young adult might face in modern Britain, but these are two of the most regularly discussed problems facing British Millennials (Broome et al. Reference Broome, Hale, Cominetti, Corlett, Handscomb, Murphy, Slaughter and Try2022; Cominetti et al. Reference Cominetti, Costa, Eyles, Henehan and McNally2022; Topping Reference Topping2021). Including this variable should nullify the coefficient associated with exposure to a struggling relative if H4b (shared risk/self-interest) is accurate.

The final symbol (marked ‘X’) in each grid of the plot of Figure 2 demonstrates the remaining effect of having a struggling young adult relative net of this new proxy for self-interest in greater state assistance for young people (see Table A15 for the full model).Footnote ⁷ Each of the five previously statistically significant relationships remains so after the addition of this new potential mediator. The reduction in the size of these associations (relative to the previous model) is also mild, ranging from a 37 per cent decline (the difference in support for YAOPs stemming from having a struggling relative versus not having a young relative at all) to just 1 per cent (difference in propensity to vote for the Conservatives between those with young relatives doing badly or well). In all cases, then, most of the net association between exposure to struggling relatives and ‘pro-youth’ political preferences seems more attributable to altruistic considerations than financial self-interest. We therefore consider this evidence for H4a and the crucial importance of emotional bonds rather than shared risk-style mechanisms. Of course, the wellbeing of children may be related to the psychological wellbeing of parents and grandparents (for example, Seidel et al. Reference Seidel, Yorgason, Polenick, Zarit and Fingerman2018; Tossi Reference Tosi2020), which might be a broader definition of ‘self-interest’, but this evidence at least suggests that the demand for government investment in younger generations is not purely reducible to removing the financial burdens on older adults.

Notes on Contributors

Zack Grant is a Postdoctoral Research Fellow at Nuffield College, University of Oxford. ORCID ID: 0000-0002-9839-5447.

Jane Green is a Professor of Political Science and British Politics at Nuffield College, University of Oxford. ORCID ID: 0000-0003-3975-824.

Geoffrey Evans is Professor of the Sociology of Politics at Nuffield College, University of Oxford. ORCID ID: 0000-0003-0237-1481.