2.10 Appendix to Chapter 2
2.10.7 Sensitivity analysis on one-step and two-step estimations
Table 2.11 Measurement model estimates using the one-step and two-step approaches (2012/14)
One-step estimation Two-step estimation
Loadings P-value Loadings P-value
Measurement models (not standardised) Financial Resilience BY Run-out 1.000 999.000 1.000 999.000 Sustain 0.858 0.000 0.852 999.000 Bills 0.970 0.000 0.903 999.000 Money-left 0.756 0.000 0.708 999.000 Credit 0.211 0.000 0.248 999.000 Future Orientation BY Tomorrow 1.000 999.000 1.000 999.000 Long-term 1.022 0.000 1.021 999.000 Retirement 0.970 0.000 0.956 999.000 Credit 0.330 0.000 0.431 999.000 Money-left 0.463 0.000 0.460 999.000 Financial Knowledge BY Understand 1.000 999.000 1.000 999.000 WP Pension 0.450 0.000 0.481 999.000
Direct effects (of household income) ON
Run-out -0.526 0.000 -0.505 999.000
Credit -0.323 0.000 -0.227 999.000
Money-left -0.425 0.000 -0.259 999.000
WP Pension* 0.107 0.010 0.220 999.000
Goodness of fit statistics
Chi-Square statistic (degrees of freedom) 624.767(157) 0.000 784.454 (170) 0.000 RMSEA (prob RMSEA<=0.05) 0.023 (1.000) 0.025 (1.000)
CFI 0.971 0.962
TLI 0.963 0.955
SRMR 0.059 0.061
Note: N=5,755. For the survey questions corresponding to the names, see Table 2.3. P-values for loadings that are fixed are set as 999.000 in Mplus. * This model is not the final model; the household income effect on WP Pension is found significant at the 5% level but omitted in the final model as the coefficient was estimated to be very close to zero (0.04), which is difficult to meaningfully interpret. There was no change in the goodness of fit statistics; the Chi square test statistics were greater by 5 (with 1 more degree of freedom). Given the sample size, this change is considered marginal.
2.10 Appendix to Chapter 2 70
Table 2.12 Structure model estimates using the one-step and two-step approaches (2012/14)
One-step estimation (a) Two-step estimation (b) Coefficients P-value Coefficients P-value Structural part (Standardised
coefficients) Financial Resilience ON Future orientation 0.197 0.000 0.188 0.000 Household income 0.470 0.000 0.472 0.000 University degree 0.040 0.019 0.041 0.016 Homeownership 0.182 0.000 0.184 0.000 Female -0.066 0.000 -0.067 0.000 Financial Knowledge ON Financial Resilience 0.214 0.000 0.200 0.000
Thought about funding retirement 0.349 0.000 0.334 0.000
Age group 40-49 0.086 0.000 0.088 0.000
Female -0.215 0.000 -0.214 0.000
Future Orientation ON
University Degree 0.219 0.000 0.209 0.000
Age group 40-49 0.057 0.004 0.057 0.003
Consideration for funding retirement ON
Future Orientation 0.188 0.000 0.183 0.000
Household income 0.280 0.000 0.277 0.000
University degree 0.115 0.000 0.118 0.000
Homeownership 0.102 0.000 0.101 0.000
Age group 40-49 0.081 0.000 0.081 0.000
Inheritance (past 2 years) 0.052 0.011 0.053 0.010
Retirement saving confidence ON
Financial Knowledge 0.341 0.000 0.338 0.000 Financial Resilience 0.269 0.000 0.276 0.000 Having a DB pension 0.355 0.000 0.349 0.000 Retirement saver 0.237 0.000 0.239 0.000 Retirement saver ON Financial Resilience 0.601 0.000 0.603 0.000 Future Orientation 0.174 0.000 0.163 0.000
Consideration for funding retirement 0.224 0.000 0.229 0.000
Retirement saving confidence -0.139 0.015 -0.138 0.017
Control variables Household income ON
University degree 0.309 0.000 0.310 0.000
Female -0.065 0.000 -0.062 0.000
Homeownership 0.217 0.000 0.217 0.000
Having a DB pension scheme ON
Household income 0.103 0.000 0.100 0.000
Having a DC pension scheme ON
Household income 0.197 0.000 0.199 0.000
Goodness of fit statistics
Chi-Square statistic (degrees of freedom) 624.767(157) 0.000 784.454 (170) 0.000 RMSEA (prob RMSEA<=0.05) 0.023 (1.000) 0.025 (1.000)
CFI 0.971 0.962
TLI 0.963 0.955
SRMR 0.059 0.061
Note: N=5,755. For the survey questions corresponding to the names, see Table 2.3. P-values for loadings that are fixed are set as 999.000 in Mplus. * This model is not the final model; the household income effect on WP Pension is found significant at the 5% level but omitted in the final model as the coefficient was estimated to be very close to zero (0.04), which is difficult to meaningfully interpret. There was no change in the goodness of fit statistics; the Chi square test statistics were greater by 5 (with 1 more degree of freedom). Given the sample size, this change is considered marginal.
Chapter 3
Gender difference in British young
adults’ retirement saving: A multi-group
analysis using Structural Equation
Modelling (SEM)
3.1
Abstract
While an increasing gender disparity in pension wealth is widely recognised in Britain, few studies have investigated gender differences in younger adults’ retirement saving. This study examines whether men and women (aged 30-49) differ in their retirement saving decision-making process, and if so to what extent, based on the adapted version of the model of financial planningusing the Wealth and Assets Survey (2012/4). A multi-group analysis in the structural equation modelling framework was utilised to investigate the gender difference in-depth. Findings show that attitudinal and behavioural factors that are linked to identifying retirement savers are similar between men and women. However, how and to what extent these factors are associated with individuals’ current social and economic arrangements (represented by income, homeownership, marital status, offspring) vary by gender. The findings show that the male-breadwinner model is still applicable to the younger adults.
3.2 Introduction 72
3.2
Introduction
As discussed in Chapters 1 and 2, the recent changes in the state pension and workplace pension schemes have fundamentally changed the way the younger generation saves for retirement in Britain. Risks involved in accumulating and generating retirement income have been largely transferred to individuals from the state and the employers. As a result, individuals are increasingly expected to save via work and to make an additional provision during their working-age years.
Studies, however, have argued that women have systematic disadvantages in accumulating savings and pension rights, largely due to the current pension structure that favours continuous labour market participation (Grady, 2015; Price, 2007). Women’s employment patterns differ from men’s, which tend to be full-time and continuous employment. Women take time off their employment, work part-time or are unable to remain in the labour market because of family-care duties (van der Horst et al., 2017). Women’s interrupted work histories not only affect their state pension entitlement but also determine the level of workplace pension scheme savings, which leads to women having, on average, a smaller retirement income than men.
Not much is known, however, about whether men and women differ in terms of additional retirement saving activities outside the state or workplace pension schemes. Findings from the previous chapter hint at a potential gender difference in the decision-making process. It reported that retirement saving activity is an outcome of an interplay between internal (attitudinal and behavioural factors) and external characteristics (demographic and socio- economic factors) and that individuals’ perception and ability to save vary substantially depending on broader socio-economic arrangements (see Chapter 2). These socio-economic arrangements, however, may vary between men and women who tend to have different social and gender roles during the partnership-forming and family-growing stages of life. Such differences may influence men’s and women’s attitudes or behaviours in the retirement saving decision-making process.
This chapter examines the gender difference in two different ways. First, it aims to test the way in which the economic autonomy in retirement saving differs between men and women.
3.2 Introduction 73
Second, it focuses on gender difference in demographic and socio-economic circumstances and how these factors may influence men’s and women’s ability to organise everyday finance differently. In particular, it tests whether the traditional ‘male-breadwinner’ hypothesis applies to the younger generation. In other words, would men’s financial behaviour be influenced by the size of financial resources available more than women’s financial behaviour? And would women be more influenced by characteristics that indicate stability in family life? The decision-making process used in this chapter follows an adapted version of the model of financial planning, which is modified to include a behavioural measure and a set of socio-economic environmental factors (see Chapter 2). Using this modified version of the model of financial planning, this study tests whether and to what extent women’s and men’s retirement saving decision-making differs among British adults aged between 30 and 49. Additionally, it examines in what ways the gendered life course may influence individuals’ retirement saving behaviours. To do so, multi-group analysis is performed in the SEM framework using the fourth wave of WAS (ONS, 2018d). The multi-group analysis allows investigation of the gender difference in a more nuanced way compared to regression analysis with interaction effects (for more on interaction effects, see Van Der Weele and Knol, 2014). It also enables a direct comparison between the gender groups; this feature is an advantage compared to studies that provide models for men and women separately but only with an indirect assessment of meaningfulness of such differences.
The structure of this chapter is as follows. The next section provides an overview of how men’s and women’s entitlement and saving outcomes vary in the state pension, workplace pension scheme and additional saving channels. Then the discussion focuses on private saving and provides a literature review on gender differences in factors that are known to affect retirement saving behaviours, such as time perspective, financial resilience and financial knowledge. Data and analytical strategies are explained, followed by the interpretation of results. The study concludes with a short discussion of policy implications.