On Housing-Related Financial Fears of Baby Boomer Women Living Alone in Switzerland

Abstract

The ageing population and rising housing costs in Switzerland are increasing the number of older adults facing financial housing concerns. Older women have particularly limited housing choices because they, on average, earn less, live longer, and are more likely to live alone. This study explores potential levers to alleviate housing-related financial fears among baby boomer women (aged 55–75) living alone in Switzerland, a subject with limited academic coverage. Using regression and random forest models on unique 2023 survey data ($N=371$), we examine the influence of socio-demographic, financial, well-being, and housing factors on fears related to affordability, price increases, and lack of housing supply. Key findings show that ownership status, perceived financial situation, and concerns about maintaining one’s lifestyle significantly drive these fears. The fear of unsuitable housing strongly influences perceived lack of supply. These results highlight the importance of retirement planning and support the consideration of measures such as reverse mortgages, co-housing, subsidies, and rent-controlled units.

1. Introduction

As the demographic transition unfolds worldwide, the population in Europe over 64 is expected to increase by 25% by 2050 (World Health Organization 2023). The post-World War II baby boom increases the number of individuals retiring today. Women in Europe have the highest life expectancy, compared to other continents, and live about five years longer than men (Eurostat 2024a). According to the Swiss Federal Statistical Office, in 2022, 2.1 million people aged 55 to 75 lived in Switzerland and 286,279 women in this age group lived alone, representing 26.5% of the baby boomer women.

Despite legal protections for tenants and a robust three-pillar pension system, Switzerland faces increasing rental costs and disparities between pension benefits and living expenses. The Swiss housing market is characterized by a low home ownership rate (35.9% vs. 69.1% in the European Union) (Swiss Federal Statistical Office 2023a; Eurostat 2024b), trends towards living alone, and a scarcity of housing options, which limits the moving choices for retired people, who rely on lower incomes (Federal Social Insurance Office 2024b; Zimmerli 2016).

The housing situation of older adults is an extensively studied topic (see, e.g., Roy et al. 2018). However, there is a lack of specific research on Swiss baby boomer women living alone (BBWA) and the factors contributing to their housing-related financial fears. According to Höpflinger et al. (2019), retired individuals living alone in Switzerland spend more than a third of their income on housing and are less likely to be homeowners than elderly couples. Baby boomers are reaching retirement age, with nearly a quarter living alone, primarily women (Swiss Federal Statistical Office 2023b). These women face financial constraints due to lower incomes, which limit their housing choices. Many occupy large housing with favorable rent, reducing housing options for younger families (Zimmerli 2016). Thus, analyzing housing-related financial fears is essential for Swiss policymakers and real estate developers to improve the living conditions of BBWA.

In this study, we aim to examine fears related to housing by investigating the factors that impact three underlying concerns: financial, price (i.e., housing costs) increase, and lack of housing offer fears. We contribute to the literature by using unique data from the 371 records of a 2023 survey specifically designed to study Swiss BBWA. This questionnaire, based on previous research by Huggenberger et al. (2024) and Kopanidis et al. (2017), includes information on socio-demographic factors, financial and well-being aspects, as well as housing characteristics and expectations. To identify the key factors influencing the main housing-related concerns, we apply logistic regression models and confirm the resulting predictors using a random forest technique. Our results demonstrate that fear of a change in lifestyle, ownership status, and perceived financial situation are crucial in explaining all three housing concerns. We show that the fear of living in unsuitable accommodation significantly contributes to the lack of housing offer fears. These new empirical insights highlight the need to tailor financial literacy programs, provide financial and practical solutions for BBWA to age in place, and increase suitable housing options.

The remainder of this article is organized as follows: Section 3 introduces the available data, presents the variables, and provides descriptive statistics. Section 4 describes the methodology. We report and discuss our results in Section 5 and conclude in Section 6. The appendix contains the survey questions.

2. Literature Review and Background Information

Article 41 of the Federal Constitution states that both the Confederation and the cantons must ensure that anyone seeking accommodation can find suitable options at reasonable terms. Tenant rights are protected under the Swiss Civil Code (Art. 253–274g), which regulates lease agreements, protects against unfair rent increases, and provides safeguards against arbitrary evictions. Federal housing support is determined by the Housing Act (LOG, Federal Office for Housing 2024), and social housing is promoted at both municipal and cantonal levels through subsidies, housing cooperatives, and the construction of affordable units.

The transition to retirement involves a shift from earned income to reliance on a three-pillar old-age provision system, consisting of old-age and survivors’ insurance (OASI), occupational pensions, and private pension savings, with nearly all citizens receiving at least OASI benefits (Federal Social Insurance Office 2024b). If their income is insufficient, retired citizens may also qualify for financial relief through means-tested supplementary benefits (Federal Social Insurance Office 2024a).

Over the past few decades, the average monthly net rent in Switzerland has significantly increased, rising from CHF 1059 in 2000 to CHF 1451 in 2023. Rents in urban cantons such as Zurich and Geneva are substantially higher than those in rural areas like Jura or Valais (Swiss Federal Statistical Office 2025). At the same time, the OASI average monthly old-age pension for all beneficiaries increased from CHF 1566 in 2000 to CHF 1919 in 2023 (Federal Social Insurance Office 2024b). As a result, growth in pensions has not kept pace with rising living expenses and the decline in the availability of accommodations (Zimmerli 2016).

In this country, older women face more financial and housing vulnerabilities than men (Huggenberger et al. 2024; Kuhn 2020). There are significant gender disparities in access to and benefits from occupational and private pensions, resulting in lower pension amounts for women compared to men (Swiss Federal Statistical Office 2020). These differences are attributed to factors such as gender pay gaps, differing career paths, and lower savings (Comolli et al. 2022; Kuhn 2020; Speelman et al. 2013). Individuals over 64 years old living alone, especially women, have a higher poverty rate (15.2 vs. 11.9% for men, Swiss Federal Statistical Office 2020). Furthermore, women are more likely to live alone (41% vs. 24% for men, Swiss Federal Statistical Office 2020). Since the 1980s, the increase in housing costs has disproportionately affected older women, especially those living alone, making them less likely to own homes (Darab et al. 2018; Ehrler et al. 2016) and thereby increasing their housing expenses, as observed by Karlen et al. (2021).

Older women have specific housing needs (Beringhs et al. 2024; Höpflinger et al. 2019; Zimmerli and Vogel 2012) and a higher intention to move (Huggenberger et al. 2024), often due to constraints rather than choice (Caradec 2010; Ravazzini and Chesters 2018). The ageing population and the increasing number of people living alone lead to discussions on policies and research, such as that commissioned by the Federal Office for Housing, aimed at enhancing market transparency, improving housing supply, and supporting housing policy for retired households with reduced incomes (Federal Office for Housing 2020).

3. Survey, Available Data, and Descriptive Statistics

In this section, we first introduce the survey, outline the topics covered, and provide details on the data collection methodology. We then describe the variables involved in the present study. Finally, we present descriptive statistics on the sample and report the shares of respondents who face financial, price increase, and lack of housing offer fears.

3.1. Survey Setup

Our study uses unique and novel survey data detailed in Beringhs et al. (2024). The survey carried out in 2023 in Switzerland focuses on the housing concerns of BBWA and was distributed by municipalities via newsletters and by the Swiss Federal Office for Housing through LinkedIn. The survey gathered 388 responses from women aged between 55 and 75 years and living in Switzerland’s German- and French-speaking regions. We excluded 14 observations from women living in elderly care facilities due to missing data in questions not applicable to their situation and three responses from women with marital status “married”. We remain with $N=371$ observations for our analysis. Although the survey has been executed in German and French, we provide an English translation of the questions relevant to this research in the Appendix A.

The customized questionnaire covers aspects of social life, housing, health, finance, and demographics. In the present study, we use 18 selected questions from the survey to examine the factors affecting the financial housing fears of BBWA. From these survey questions, we derive 26 variables that we group along six topics, namely key variables, socio-demographic factors, financial aspects, well-being aspects, housing characteristics, and housing expectations. A summary is given in

Table 1. Summary of the variables.

Variable	Description	Categories	Question
Key variables
$FIN$	Financial housing fears	Yes, no	C7.7, E3.3
$PRI$	Price increase fears	Yes, no	C7.2, C7.3
$LAC$	Lack of housing offer fears	Yes, no	C7.6
Socio-demographic factors
$LNR$	Language region	French, German	A4
$AGE$	Age	55–60, 61–65, 66–70, 71–75	A1
$RET$	Retired	Yes, no	B1
$MAR$	Marital status	Sep./Div., single, widowed	F1
$CHI$	Children	Yes, no	F2
Financial aspects
$PER$	Perceived financial situation	Modest, below avg., above avg., comfortable	E1
$SEC$	Buyback years of pension savings	Yes, no	E2.4.3
$THI$	Private pension savings	Yes, no	E2.5.1, E2.6
$OWN$	Ownership status	Oneself, private, institution, other	C2.1, C2.2
$STO$	Declared career break	Yes, no	E2.1
Well-being aspects
$ILL$	Diagnosed illnesses	Yes, no	D2
$HEA$	Perceived health status	Good, bad	D1
$HEF$	Fear of losing independence	Yes, no	E4.4
$FLI$	Fear of feeling alone	Yes, no	E4.1
Housing characteristics
$HTY$	Housing type	Flat, house, other	C1
$ROO$	Number of rooms	1–2.5, 3–3.5, 4+ rooms	C4
$ATS$	Attributes satisfaction	Yes, no	C6.1–5
$ACC$	Accessibility satisfaction	Yes, no	C6.10–14
$ENV$	Environment satisfaction	Yes, no	C6.6–9
Housing expectations
$STA$	Wish to stay home	Yes, no	C3.1
$ATT$	Housing attachment	Yes, no	C5.4
$SUI$	Current housing unsuitable fears	Yes, no	C7.5
$LIC$	Lifestyle change fear	Yes, no	E3.2

Note: The abbreviation “avg.” stands for average.

.

3.2. Description of the Variables

For the first key variable, financial housing fears ($FIN$), participants indicated their level of agreement with the two following statements (see questions C7.7 and E3.3) on a five-level Likert scale ranging from “strongly disagree” to “strongly agree”:

• “I am afraid that my finances will not allow me to have a home that meets my needs.”
• “I fear that my finances do not allow me to find other accommodation that would be better suited to my needs.”

We report the distribution of the responses in Figure 1a,b. Based on the two statements, we construct the binary latent variable $FIN$ reflecting financial housing fears. Therefore, we rate the original responses on a numerical scale from one (strongly disagree) to five (strongly agree), and we determine the mean level of agreement. A score strictly above three is interpreted as a “yes” for financial concerns. The construct is characterized by a Cronbach’s alpha coefficient of 0.78 and a Spearman correlation of 0.66, indicating high degrees of consistency and reliability. In Figure 2, we report the individual observations and frame those combinations resulting in financial housing fears ($FIN$ equal “yes”). Overall, we observe that 51.2% of the sample expressed such fears.

We construct the second key variable, $PRI$, to measure price increase fears by combining responses from two statements, one directed at homeowners and the other at tenants (see questions C7.2 and C7.3). Participants expressed their level of agreement on a five-level Likert scale ranging from “strongly disagree” to “strongly agree”. We report a “yes” if respondents answered “agree” or “strongly agree” to either statement, depending on their homeowner or tenant status:

• For homeowners: “I am afraid I will not be able to keep my property.”
• For tenants: “I am afraid that the rent will increase.”

We report the distribution of the responses separately for homeowners and tenants in Figure 3a and Figure 3b, respectively. Among tenants, more BBWA fear price increases (60.0% of tenants), while the opposite is true among homeowners (18.9% of homeowners). Overall, 47.7% of the respondents express affordability fears linked to prices.

The third key variable focuses on the lack of housing offer fears ($LAC$). Participants indicated their level of agreement with the following statement in question C7.6: “I am afraid that I won’t find accommodation that suits my needs/wishes.” We report a “yes” if respondents answered “agree” or “strongly agree”. The distribution of the responses is illustrated in Figure 4. Overall, we observe that 54.7% of the sample have housing offer concerns.

Regarding the socio-demographic factors, we allocate Swiss cantons to their linguistic regions ($LNR$), namely, German-speaking and French-speaking parts, to investigate the associated cultural differences. We categorize the respondents’ ages ($AGE$) into four groups: 55–60, 61–65, 66–70, and 71–75. Whether or not the respondent is retired is captured in the variable $RET$. After excluding married women and those in registered partnerships, we remain with three categories for marital status ($MAR$): separated/divorced, single, and widowed. We transform the original numerical variable referring to the number of children into the binary variable $CHI$ to indicate whether the respondent has children or not.

For the financial aspects, the perceived financial situation ($PER$) is assessed with the four levels: modest, below average, above average, and comfortable. Whether the participant declared having repurchased years of pension savings or not is captured in $SEC$. For the $THI$ variable, we report a “yes” for women who contributed to a third pillar and/or have savings in life insurance, i.e., private pension savings. The $OWN$ variable distinguishes between homeowners and tenants, with further specification for the latter regarding the owner of their dwelling. The categories of $OWN$ are oneself for homeowners, and private owner, institution (including municipality, institutional investors, and cooperative or non-profit organization), and other (including “I do not know”) for tenants. Due to the limited number of observations, the data did not allow us to distinguish between the different types of institutions. The $STO$ variable captures those who have interrupted their professional career for more than a year when the answer is “yes”. For $SEC$, $THI$, and $STO$, we categorize women responding with “I do not know” as “no”.

Concerning the well-being aspects, we report a “yes” for women who declare having been diagnosed with a physical or a mental illness ($ILL$). The original question about perceived health ($HEA$) also differentiated physical and mental health. After converting the five-level Likert scale from bad to good to numerical levels from one to five, we report a “good” for women who rate both statements above three and “bad” for the rest. $HEF$ measures the fear of losing independence due to unexpected life events and $FLI$ the fear of loneliness. “Yes” corresponds to women who are rather worried or worried, while “no” indicates neutral, rather not worried, or not worried.

As for the housing characteristics, we consider houses (single-family and semi-detached homes), flats, and other housing types for the accommodation type ($TYP$). The size of the home is measured by the number of rooms ($ROO$), categorized as 1–2.5, 3–3.5, and 4+. From question C.6, we derive three variables: satisfaction with attributes ($ATS$) regarding size, insulation, interior fittings, and costs; accessibility satisfaction ($ACC$) covering proximity to shops, public transport, healthcare facilities, activities, and relatives; and satisfaction with the environment ($ENV$) involving property management, immediate environment, neighborhood safety, and noise pollution. A “yes” is reported for these factors when the mean value is above six on a ten-level Likert scale.

Finally, we study housing expectations, where those who consider their current home as a place that appeals to them for their old age are represented by $STA$, while those who express attachment to their home are represented by $ATT$. Concerns about living in an unsuitable dwelling in the future are coded in $SUI$, and worries about insufficient finances to maintain the standard of living they envision in $LIC$.

3.3. Descriptive Statistics

In

Table 2. Sample distribution and share of BBWA with financial ($FIN$), price increase ($PRI$), and lack of housing offer ($LAC$) fears.

Variable	Category	N	(%)	$\mathit{FIN}$	$\mathit{PRI}$	$\mathit{LAC}$
All records		371	100.0	51.2	47.7	54.7
Socio-demographic factors
$LNR$	French	260	70.1	55.8	51.1	60.8
	German	111	29.9	40.5	39.6	40.5
$AGE$	55–60	70	18.9	52.9	54.3	57.1
	61–65	108	29.1	59.3	50.9	62.0
	66–70	107	28.8	42.1	39.2	47.7
	71–75	86	23.2	51.2	48.8	52.3
$RET$	Yes	243	65.5	47.3	44.0	54.7
	No	128	34.5	58.6	54.7	54.7
$MAR$	Sep./Div.	203	54.7	52.7	48.8	58.1
	Single	118	31.8	51.7	50.9	53.4
	Widowed	50	13.5	44.0	36.0	44.0
$CHI$	Yes	231	62.3	54.1	47.2	57.6
	No	140	37.7	46.4	48.6	50.0
Financial aspects
$PER$	Modest	93	25.1	76.3	60.2	60.2
	Below avg.	89	24.0	78.7	67.4	71.9
	Above avg.	142	38.2	28.9	37.3	47.9
	Comfortable	47	12.7	17.0	17.0	31.9
$SEC$	Yes	99	26.7	42.4	45.5	53.5
	No	272	73.3	54.4	48.5	55.1
$THI$	Yes	254	68.5	42.9	46.1	50.4
	No	117	31.5	69.2	51.3	64.1
$OWN$	Oneself	111	29.9	22.5	18.9	36.0
	Private	125	33.7	67.2	60.8	67.2
	Institution	60	16.2	65.0	61.7	66.7
	Other	75	20.2	56.0	57.3	52.0
$STO$	Yes	208	56.1	57.2	49.5	55.3
	No	163	43.9	43.6	45.4	54.0
Well-being aspects
$ILL$	Yes	206	55.5	56.3	51.9	58.2
	No	165	44.5	44.9	42.4	50.3
$HEA$	Good	274	73.9	47.5	45.3	53.6
	Bad	97	26.1	61.9	54.6	57.7
$HEF$	Yes	235	63.3	56.6	51.9	62.1
	No	136	36.7	41.9	40.4	41.9
$FLI$	Yes	144	38.8	63.9	52.1	65.3
	No	227	61.2	43.2	44.9	48.0
Housing characteristics
$HTY$	Flat	317	85.4	53.9	52.4	56.5
	House	46	12.4	28.3	21.7	43.5
	Other	8	2.2	75.0	12.5	50.0
$ROO$	1–2.5	122	32.9	64.8	49.2	56.6
	3–3.5	147	39.6	49.0	54.4	57.1
	4+	102	27.5	38.2	36.3	49.0
$ATS$	Yes	288	77.6	45.5	43.8	48.6
	No	83	22.4	71.1	61.5	75.9
$ACC$	Yes	311	83.8	48.5	48.5	53.0
	No	60	16.2	65.0	43.3	63.3
$ENV$	Yes	178	48.0	37.6	38.2	45.5
	No	193	52.0	63.7	56.5	63.2
Housing expectations
$STA$	Yes	187	50.4	43.3	49.7	44.9
	No	184	49.6	59.2	45.6	64.7
$ATT$	Yes	230	62.0	43.9	47.4	48.3
	No	141	38.0	63.1	48.2	65.2
$SUI$	Yes	146	39.4	63.7	51.4	75.3
	No	225	60.6	43.1	45.3	41.3
$LIC$	Yes	147	39.6	81.6	70.1	68.0
	No	224	60.4	31.2	33.0	46.0

Notes: The columns N and “(\%)” indicate the number of observations and their share in each category, respectively; FIN, PRI, and LAC report the shares (in %) of “yes” responses.

, we present the sample distribution of the 371 records along the variables and the distribution of the responses for the key variables $FIN$, $PRI$, and $LAC$. Around half of the respondents expressed fears related to $FIN$ (51.2%), $PRI$ (47.7%), and $LAC$ (54.7%). In the following, we highlight selected notable characteristics observed in our data.

Participants from the French-speaking part of Switzerland ($LNR$, 70.1% of the sample) show higher levels of fear across all key variables, with 55.8% for $FIN$, 51.1% for $PRI$, and 60.8% for $LAC$, compared to women from the German-speaking part (40.5, 39.6, and 40.5%, respectively). Age also appears to have an impact ($AGE$). Participants aged 61–65 show the highest concerns regarding $FIN$ (59.3%) and $LAC$ (62.0%), while those aged 66–70 show lower concerns across all dimensions $FIN$ (42.1%), $PRI$ (39.2%), and $LAC$ (47.7%). Non-retired women express higher levels of $FIN$ (58.6%) and $PRI$ (54.7%) than those who are retired ($RET$). Widowed women show fewer fears across all dimensions ($MAR$, 36.0% for $PRI$ and 44.0% for $FIN$ and $LAC$).

As expected, respondents who perceive their financial situation ($PER$) as modest or below average tend to report higher fears across all key variables than those with an above-average or comfortable financial perception. We note that a large share of women in our sample (38.3%) perceive their financial situation as above average. Those who have repurchased years of pension savings ($SEC$) or contributed to private pension savings ($THI$) indicate lower financial housing fears than those who did not. A quarter of the women in the sample have repurchased years of pension savings, and almost one-third benefit from private pension savings. Regarding ownership ($OWN$), tenants living in housing owned by privates or institutions declare higher levels of fear across all dimensions (above 60%). Conversely, homeowners, representing almost one-third of the sample, have the lowest percentage in $FIN$ (22.5%), $PRI$ (18.9%), and $LAC$ (36.0%). Almost two-thirds of the sample declare having stopped working for at least a year ($STO$), which seems to increase $FIN$ (57.2 vs. 43.6%).

Women with a diagnosed illness ($ILL$, 55.5% of the sample) tend to feel more fears across all dimensions than those without one. Despite the relatively high proportion of diagnosed illnesses, 73.8% perceive themselves as healthy ($HEA$). We observe that the latter have fewer fears in all dimensions than those who judge their health as poor. Women afraid of losing their independence ($HEF$, 63.3% of the sample) also seem to be more concerned about $FIN$ (56.6%), $PRI$ (51.9%), and $LAC$ (62.1%) than those who are not (41.9, 40.4, and 41.9%, respectively). A similar observation holds for participants with a fear of loneliness ($FLI$, 38.8% of the sample) compared to those without this fear.

Most respondents live in a flat ($HTY$, 85.4%), which tends to increase $FIN$ (53.9%), $PRI$ (52.4%), and $LAC$ (56.5%). Those living in houses conversely seem to manifest lower fears (28.3, 21.7, and 43.5%, respectively). No major disparity is found for the number of rooms ($ROO$). BBWA unsatisfied with their housing attributes ($ATS$, 22.4%) report higher fears across all dimensions. Similarly, dissatisfaction with the environment ($ENV$) increases the fears. Except for $PRI$, the same occurs when respondents express dissatisfaction regarding accessibility ($ACC$).

Half of the respondents wish to stay in their current housing as they age ($STA$). Those without this desire tend to express more $FIN$ (59.2%) and $LAC$ (64.7%) compared to those who would prefer to stay (43.3 and 44.9%, respectively). Women without housing attachment ($ATT$) seem to experience more $FIN$ (63.1%) and $LAC$ (65.2%). Those apprehensive about the unsuitability of their current housing ($SUI$) report higher levels of fear for all key variables, especially for $LAC$ (75.3%). Women who fear not being able to afford their envisioned lifestyle show more $FIN$ (81.6%), $PRI$ (70.1%), and $LAC$ (68.0%), in contrast to those without such fears ($LIC$, 31.2, 33.0, and 46.0%, respectively).

4. Model Framework

Our methodological framework combines regression and random forest techniques to determine the most influential factors impacting the three binary dependent variables $FIN$, $PRI$, and $LAC$ among BBWA in Switzerland. Regression analysis allows us to estimate the direction and strength of associations through statistical inference, while random forest models capture complex and non-linear relationships. The complementary strengths of these methods enable a more comprehensive understanding of the influences of predictors.

First, we fit a generalized linear model (GLM) on the set of $K=23$ variables listed in Table 1 in a “full model.” Using Akaike’s Information Criterion (AIC) to select the link function that best fits, we determine that a logit function is preferable for $FIN$ and $LAC$, while probit is best for $PRI$. GLM is widely used for modeling binary outcomes, providing suitable link functions, and ensuring ease of interpretation (Ruíz et al. 2023). We estimate the following regression models:

$$g\left(y_i\right)={\beta }_0+\sum _{k=1}^K{\beta }_k{X}_{i,k}+{\epsilon }_i,$$

(1)

where $g(·)$ is the link function and $y_i\in \{0,1\}$ stands for the ith individual, $i=1,\dots ,N$. The coefficient ${\beta }_0$ denotes the regression intercept, ${\beta }_k$ is the kth parameter we estimate, and $X_{i,k}$ is the value of the kth independent variable for individual i. Finally, ${\epsilon }_i$ is the error term for the ith observation.

We then consider a “reduced model” built with the stepAIC function in R: using forward selection and backward elimination techniques, we optimize the variables included in the regression model.1 The algorithm identifies and retains only the factors that contribute the most to improving the models’ performance measured in terms of AIC, thereby reducing the risk of overfitting present in the “full model.”

To identify the relevance of the factors in the three response variables, we rank the independent variables by importance using likelihood ratio tests ($\lambda$). This method involves, for each estimate, comparing the maximum likelihood from the unrestricted model, ${\widehat{\beta }}_U$, with those from a restricted model, ${\widehat{\beta }}_R$, using the following formula:

$$\lambda =-2·\{ln\mathcal{L}\left({\widehat{\beta }}_R\right)-ln\mathcal{L}\left({\widehat{\beta }}_U\right)\}\sim {\chi }_{(k_U-k_R)}^2,$$

(2)

with $k_U$ and $k_R$ denoting the number of parameters estimated in the unrestricted and restricted models, respectively. In our setting, the former model is the reduced model derived above, while the restricted model mirrors the reduced model but excludes the tested parameter.

Finally, to confirm the GLM regression results’ robustness, we propose to compare the variable importance ranking from a random forest model (RFM) applied on the whole set of covariates with the ranking given by $\lambda$ for each reduced regression model. This comparison provides a consistency and reliability validation of the selected variables across two modeling approaches, ensuring that our findings are not model-specific but reflect patterns in the data. We produce graphs using the varImpPlot function from the randomForest package in R.2 For ranking the variables in the RFM, we use the mean decrease accuracy (MDA), which indicates how much accuracy the model loses by excluding each variable. This measure involves observing the effect on the model’s accuracy when the values of each factor are randomly shuffled. Since a RFM comprises multiple decision trees, this process helps us understand the importance of each variable for the model’s prediction accuracy by measuring the change in accuracy when the information provided by the variable is excluded. That is, we compute $A_k$, the mean decrease in accuracy of variable k, with

$$\overline{A_k}={\displaystyle \frac{{\sum }_{t=1}^T(A_{\mathrm{baseline}}-A_{\mathrm{permuted}}^t)}{T}},$$

(3)

where T denotes the total number of trees, $A_{\mathrm{baseline}}$ is the initial model accuracy, and $A_{\mathrm{permuted}}^t$ is the accuracy after permuting a variable in tree t. In our application, we set $T=1000$ and $k=1,\dots ,K$, including all independent variables.

5. Results, Robustness, and Discussion

We present the estimations of the GLM model (1) in Section 5.1 and compare the importance of the variables with the one derived from the RFM as a robustness test in Section 5.2. We provide a short discussion of the findings in Section 5.3.

5.1. Regression Results

In the tables below, we estimate the full GLM model using all variables from Table 1 and present the estimated coefficients, standard errors, and significance levels. Positive coefficients indicate a higher impact on the examined fear. We also report the coefficients for the variables remaining in the reduced model. The last column reports the rankings based on the likelihood ratio test applied to the reduced model (see Equation (2)). The variables are ordered by importance. In the following, we discuss the most important variables.

Table 3. Regression models for financial housing fears ($FIN$).

		Full Model			Reduced Model
Variable	Category	Coef.	Std. Err.	Sig.	Coef.	Std. Err.	Sig.	λ
Intercept		–3.459	1.098	**	–3.661	0.557	***
$LIC$	Yes	2.437	0.383	***	2.203	0.339	***	1
$PER$ (Above avg.)	Modest	2.193	0.461	***	1.996	0.395	***	2
	Below avg.	2.394	0.454	***	2.168	0.403	***
	Comfortable	–0.467	0.596		–0.342	0.524
$OWN$ (Oneself)	Private	1.685	0.503	***	1.702	0.400	***	3
	Institution	1.604	0.580	**	1.509	0.483	**
	Other	0.350	0.552		0.516	0.434
$MAR$ (Sep./Div.)	Single	0.776	0.443	.	0.566	0.337	.	4
	Widowed	0.956	0.523	.	0.920	0.488	.
$ENV$	Yes	–0.601	0.348	.	–0.509	0.307	.	5
$ILL$	Yes	0.803	0.346	*	0.608	0.302	*	6
$FLI$	Yes	0.850	0.358	*	0.939	0.318	**	7
$SUI$	Yes	0.636	0.349	.	0.645	0.309	*	8
$ACC$	Yes	–0.609	0.491
$AGE$ (55–60)	61–65	0.203	0.474
	66–70	–0.250	0.517
	71–75	–0.593	0.503
$ATS$	Yes	0.843	0.460	.
$ATT$	Yes	–0.606	0.358	.
$CHI$	Yes	0.238	0.412
$HEA$	Good	0.594	0.408
$HEF$	Yes	0.097	0.348
$HTY$ (Flat)	House	0.250	0.670
	Other	1.042	1.233
$LNR$ (French)	German	0.147	0.443
$RET$	Yes	–0.369	0.367
$ROO$ (1–2.5)	3–3.5	–0.222	0.378
	4+	0.008	0.526
$SEC$	Yes	0.395	0.373
$STA$	Yes	–0.404	0.351
$STO$	Yes	–0.108	0.339
$THI$	Yes	–0.465	0.369

Notes: The significance levels are $p<0.1$, * $p<0.05$, ** $p<0.01$, *** $p<0.001$. $\lambda$ reports the rank obtained from log-likelihood ratio test. Baseline levels are indicated in brackets. For the “Yes” levels, the baseline is “No”.

reports eight key factors affecting financial housing fears ($FIN$). According to $\lambda$, lifestyle change fear ($LIC$) is the most important factor. Retirement implies living with lower revenues, and housing represents about one-third of them (Swiss Federal Statistical Office 2020). Thus, worries about insufficient finances to maintain the envisioned lifestyle induce $FIN$. Unsurprisingly, women with modest or below-average perceived financial situations have stronger financial housing concerns ($PER$). Ownership status ($OWN$) reveals that tenants living in housing owned by private individuals or institutions experience more financial housing fears than homeowners. With lower housing expenses (Karlen et al. 2021), homeowners may live with more financial peace of mind. Marital status ($MAR$) indicates that being widowed or single increases $FIN$, compared to separated or divorced women. Separation or divorce often occurs earlier in life, which may help BBWA plan their retirement, whereas widowhood is unexpected, and being single increases poverty rates (Swiss Federal Statistical Office 2020). Satisfaction with the environment ($ENV$) reduces fears, while BBWA with illnesses ($ILL$) report higher levels of concern. Fears of losing independence ($FLI$) and concerns about unsuitable housing ($SUI$) also contribute to higher levels of $FIN$. These findings suggest that the living situation, including environment, health, and related fears regarding the independence level the accommodation offers, plays a crucial role in explaining $FIN$ for BBWA who have specific housing needs (Beringhs et al. 2024).

Table 4. Regression models for price increase fears ($PRI$).

		Full Model			Reduced Model
Variable	Category	Coef.	Std. Err.	Sig.	Coef.	Std. Err.	Sig.	λ
Intercept		–2.758	0.556	***	–2.219	0.347	***
$OWN$ (Oneself)	Private	1.556	0.266	***	1.320	0.218	***	1
	Institution	1.476	0.299	***	1.322	0.255	***
	Other	1.280	0.290	***	1.063	0.242	***
$LIC$	Yes	0.811	0.182	***	0.798	0.162	***	2
$ROO$ (1–2.5)	3–3.5	0.523	0.188	**	0.473	0.178	**	3
	4+	0.433	0.249	.	0.488	0.220	*
$PER$ (Above avg.)	Modest	0.601	0.229	**	0.435	0.206	*	4
	Below avg.	0.703	0.221	**	0.638	0.205	**
	Comfortable	–0.457	0.308		–0.524	0.275	.
$THI$	Yes	0.423	0.191	*	0.347	0.170	*	5
$STA$	Yes	0.327	0.182	.	0.368	0.155	*	6
$ILL$	Yes	0.397	0.172	*	0.266	0.150	.	7
$CHI$	Yes	–0.052	0.212		–0.250	0.159		8
$ACC$	Yes	0.397	0.238	.
$AGE$ (55–60)	61–65	–0.561	0.239	*
	66–70	–0.319	0.260
	71–75	–0.340	0.245
$ATS$	Yes	–0.325	0.235
$ATT$	Yes	0.335	0.185	.
$ENV$	Yes	–0.144	0.181
$FLI$	Yes	0.253	0.174
$HEA$	Good	0.229	0.199
$HEF$	Yes	0.107	0.176
$HTY$ (Flat)	House	0.269	0.344
	Other	–1.031	0.604	.
$LNR$ (French)	German	–0.045	0.226
$MAR$ (Sep./Div.)	Single	0.264	0.220
	Widowed	0.049	0.258
$RET$	Yes	–0.066	0.182
$SEC$	Yes	0.029	0.193
$STO$	Yes	–0.203	0.176
$SUI$	Yes	–0.205	0.178

Notes: See Table 3.

highlights eight critical factors related to concerns about price increases ($PRI$). The first is ownership status ($OWN$). Tenants are much more worried about price increases than homeowners. This is likely because homeowners have more control over housing costs (Morris 2009), while costs for homeowners depend essentially on the lending market and maintenance responsibilities, which may be perceived as burdensome for the elderly (Angelini and Laferrere 2012). Expectedly, fears related to insufficient finances to cover the envisioned lifestyle ($LIC$) are positively associated with $PRI$. Occupants of larger accommodations (3+ rooms, $ROO$) experience more $PRI$, possibly due to higher associated costs. The perceived financial situation ($PER$) is crucial, as those who view their financial situation as modest or below-average are more susceptible to express concerns about price increases, undoubtedly due to their lower financial margin. Surprisingly, having private pension savings ($THI$) is associated with higher worries about price increases. These savings are intended to help individuals accumulate supplementary retirement savings (Federal Social Insurance Office 2024c). Those who desire to grow old in their current dwelling ($STA$) and those with diagnosed illnesses ($ILL$) report higher $PRI$. Ageing in place brings financial challenges, especially for older adults who are unwell (Beer et al. 2012). Conversely, having children ($CHI$) reduces fears of price increases, possibly due to the support they can provide (Reher and Requena 2017).

Table 5. Regression models for lack of housing offer fears ($LAC$).

		Full Model			Reduced Model
Variable	Category	Coef.	Std. Err.	Sig.	Coef.	Std. Err.	Sig.	λ
Intercept		–1.687	0.857	*	–1.018	0.344	**
$SUI$	Yes	1.051	0.278	***	1.228	0.258	***	1
$HEF$	Yes	0.548	0.277	*	0.471	0.249	.	2
$PER$ (Above avg.)	Modest	0.252	0.369		0.181	0.318		3
	Below avg.	0.941	0.365	**	0.921	0.334	**
	Comfortable	–0.301	0.446		–0.525	0.396
$OWN$ (Oneself)	Private	0.838	0.380	*	0.691	0.309	*	4
	Institution	0.931	0.452	*	0.849	0.375	*
	Other	–0.014	0.444		–0.036	0.353
$STA$	Yes	–0.470	0.280	.	–0.595	0.246	*	5
$LIC$	Yes	0.682	0.296	*	0.592	0.266	*	6
$ACC$	Yes	0.209	0.391
$AGE$ (55–60)	61–65	0.008	0.385
	66–70	0.009	0.412
	71–75	–0.217	0.398
$ATS$	Yes	–0.212	0.390
$ATT$	Yes	–0.358	0.293
$CHI$	Yes	0.282	0.338
$ENV$	Yes	–0.278	0.284
$FLI$	Yes	0.327	0.278
$HEA$	Good	0.405	0.321
$HTY$ (Flat)	House	0.185	0.509
	Other	–0.569	0.831
$ILL$	Yes	0.371	0.276
$LNR$ (French)	German	–0.142	0.341
$MAR$ (Sep./Div.)	Single	–0.019	0.352
	Widowed	–0.262	0.399
$RET$	Yes	0.206	0.289
$ROO$ (1–2.5)	3–3.5	0.421	0.307
	4+	0.471	0.409
$SEC$	Yes	0.240	0.300
$STO$	Yes	–0.311	0.279
$THI$	Yes	–0.269	0.301

Notes: See Table 3.

shows that six factors significantly affect the lack of housing offer fears ($LAC$). The predominant contributing factor is the fear of living in unsuitable housing ($SUI$). The fear of losing independence ($HEF$) is also a significant predictor for $LAC$. Switzerland lacks suitable housing options for the increasing population of older adults (Zimmerli and Schmidiger 2016), which influences the availability of choices for remaining independent. With even fewer options available, BBWA with modest and below-average financial perceptions ($PER$) live with higher fears. Regarding ownership ($OWN$), tenants in housing owned by private individuals or institutions more frequently fear a lack of housing offers. Home ownership facilitates housing adaptations to ageing (Höpflinger et al. 2019), making a larger housing supply more crucial for tenants if they need to relocate. Conversely, respondents who prefer to stay in their current accommodation ($STA$) show reduced $LAC$, possibly because their plan to age at home alleviates concerns about housing options. Finally, lifestyle change fear ($LIC$) are significantly associated with these concerns. Housing choices are vital for BBWA who face a substantially higher risk of reduced post-retirement standard of living (Lee 2003).

When comparing the results from both full and reduced models in Table 3, Table 4 and Table 5, we observe that the significance levels remain similar in the reduced models. This consistency suggests that the excluded variables do not influence the included variables. Furthermore, we observe comparable coefficient values in both models, indicating the stability and robustness of our results.

5.2. Robustness of the Results

To enhance the robustness of our findings, we use the RFM (see Equation (3)) to validate the variables’ selection by evaluating the relative contribution of all variables from Table 1 to the model.

We report the variables’ importance using the mean decrease in accuracy in Figure 5. Regarding financial housing fears ($FIN$), five variables stand out in Figure 5a. Apart from the perceived financial situation ($PER$) and lifestyle change fear ($LIC$), which significantly impact $FIN$, ownership status ($OWN$), current housing unsuitable fears ($SUI$), and being retired ($RET$) also emerge. Figure 5b shows that price increase fears ($PRI$) mainly relate to four factors: ownership status ($OWN$), lifestyle change fear ($LIC$), perceived financial situation ($PER$), and housing type ($HTY$), with $OWN$ and $LIC$ being most important. Meanwhile, for the lack of housing offer fears ($LAC$), Figure 5c highlights the strong contribution of fears related to the current housing becoming unsuitable ($SUI$). Four additional factors are crucial: ownership status ($OWN$), lifestyle change fear ($LIC$), attributes satisfaction ($ATS$), and the fear of losing independence ($HEF$).

Table 6. Comparison of the variable importance rankings in the reduced GLM and RFM for financial housing fears $FIN$, price increase fears $PRI$, and lack of housing fears $LAC$.

		$\mathit{FIN}$		$\mathit{PRI}$		$\mathit{LAC}$
Variable	Description	λ	RF	λ	RF	λ	RF
Socio-demographic factors
$LNR$	Language region		10		13		12
$AGE$	Age		22		15		17
$RET$	Retired		5		22		20
$MAR$	Marital status	4	23		7		9
$CHI$	Children		19	8	6		6
Financial aspects
$PER$	Perceived financial situation	2	1	4	3	3	10
$SEC$	Buyback years of pension savings		14		23		15
$THI$	Private pension savings		8	5	18		18
$OWN$	Ownership	3	3	1	1	4	2
$STO$	Declared career break		13		17		16
Well-being aspects
$ILL$	Diagnosed illnesses	6	20	7	8		19
$HEA$	Perceived health status		18		21		23
$HEF$	Fear of losing independence		15		19	2	5
$FLI$	Fear of feeling alone	7	9		14		11
Housing characteristics
$HTY$	Housing type		16		4		14
$ROO$	Number of rooms		12	3	11		22
$ATS$	Attributes satisfaction		11		5		4
$ACC$	Accessibility satisfaction		17		20		21
$ENV$	Environment satisfaction	5	7		10		13
Housing expectations
$STA$	Wish to stay home		6	6	16	5	7
$ATT$	Housing attachment		21		12		8
$SUI$	Current housing unsuitable fears	8	4		9	1	1
$LIC$	Lifestyle change fear	1	2	2	2	6	3

Notes: The column “$\lambda$” reports the ranking obtained from the log-likelihood ratio test on the reduced model; the column “RF” reports the importance ranking from the mean decrease in accuracy measure in the RFM. The gradient shades of grey highlight the five most important factors in each model.

summarizes the variable rankings obtained from the likelihood ratio tests in the GLM and the mean decrease in accuracy in the RFM for the three key variables $FIN$, $PRI$, and $LAC$. For several variables, the results from the RFM confirm the reduced GLMs discussed in Section 5.1. In what follows, we compare each key variable’s five most relevant predictors and conclude by identifying the factors repeatedly significant across all housing-related financial variables.

Regarding $FIN$, there are similarities between the maximum likelihood ($\lambda$) and mean decrease in accuracy (RF) rankings, especially for the variables $LIC$, $OWN$, and $PER$, which consistently appear in the top tier. With a lower importance level (see Figure 5a), $OWN$ constantly ranks third. $SUI$ belongs to the $FIN$ reduced GLM in Table 3 and ranks fourth in the RFM. Nevertheless, the reduced model does not include $RET$, which ranks fifth in the RFM. $ENV$ remains important in both models (fifth position in $\lambda$ and seventh for RF), and $MAR$ ranks last in the RF model, although it is fourth for $\lambda$.

For $PRI$, consistency is observed in $OWN$, $LIC$, and $PER$, as they all hold top positions in RF (1, 2, and 3) and $\lambda$ (1, 2, and 4). A disparity is found for $HTY$, which emerges as an important factor in the RFM (4) but is absent in the reduced GLM (see Table 4). Furthermore, the RFM does not confirm the importance of the variables $ROO$ (3 for $\lambda$ and 11 for RF) and $THI$ (5 for $\lambda$ and 18 for RF). Conversely, $HTY$ and $ATS$ are relatively important in the RFM (4 and 5) but absent in the reduced GLM.

When considering $LAC$, $SUI$ remains the most significant variable across both rankings. $OWN$ and $LIC$ also maintain high rankings in both models (GLM: 4 and 6; RFM: 2 and 3). While ranked second in $\lambda$, $HEF$ also ranks high in the RFM (5). $ATS$ is absent in the reduced GLM (see Table 5) but ranks fourth in the RFM. Although the reduced model includes $PER$ and $STA$ (3 and 5 in terms of log-likelihood), the RFM ranks them only in positions 10 and 7, respectively.

5.3. Discussion

Overall, the analysis highlights the crucial role of lifestyle change fear ($LIC$), ownership status ($OWN$), and perceived financial situation ($PER$) in affecting all three key variables among respondents. The consistency of these factors, along with those described below, across all housing-related financial fears and their confirmation through the log-likelihood ratio tests and the RFM importance rankings underlines their central role in understanding these concerns.

Our results find support in the literature insofar as Darab et al. (2018) and Kopanidis et al. (2017) found that fears over potential lifestyle adjustments are closely associated with housing financial insecurities, stressing the need of better informing baby boomer women about their options and developing tailored financial literacy programs for the following generations to improve their retirement planning (Giesecke and Yang 2018; Lusardi and Mitchell 2006).

Our study confirms that home ownership significantly impacts housing-related financial fears, aligning with findings from Garten et al. (2024), who indicate that ownership provides security and well-being in old age. While tenants may benefit from low rents due to rent control and the deterioration of their accommodation, they do not benefit from the protective aspects of owning a home and thus face potential price increases (Swiss Federal Statistical Office 2018). Being a homeowner reduces housing costs (Karlen et al. 2021), and women living alone are more frequently tenants (Darab et al. 2018). Retired individuals may struggle to keep their mortgage due to banks’ strict affordability criteria (UBS 2024). Thus, homeowners should benefit from housing equity initiatives, such as reverse mortgage programs, enabling them to leverage their housing wealth (Nakajima and Telyukova 2017) and supportive housing models to address their specific needs (PwC Switzerland 2021).

Studies by the Swiss Federal Statistical Office (2020) and Kopanidis et al. (2017) also demonstrate that the perceived financial situation plays a crucial role in shaping the post-retirement housing concerns of older adults, especially for those living alone. As a result, the pension system must allow for a sustainable retirement (Lewis and Ollivaud 2020). Subsidies, rent-controlled units, or co-housing should be offered, as practiced in cooperatives and the private rental market (Karlen et al. 2021). Implementing these measures could also help reduce fears among BBWA who desire to stay home.

In Table 6, fears related to the current housing becoming unsuitable ($SUI$) are especially relevant for lack of housing offer fears ($LAC$) and, to a lesser extent, for financial housing fears ($FIN$). Angelini and Laferrere (2012) observed that home ownership facilitates the adaptation of living spaces for ageing at home. The associated management and maintenance duties, seen as a burden for the elderly (Angelini and Laferrere 2012), do not seem to impact the financial fears in our study. It is recommended to facilitate flat exchanges, assist in finding suitable accommodations, extend caretaker services, enhance social and neighborhood support, strengthen community engagement, and implement technological solutions (Althaus and Birrer 2021). Removing barriers to enable free and easy access to home modifications would be beneficial (Aplin and Petersen 2023).

The wish to stay in the current home ($STA$) affects the reduced models of price increase fears ($PRI$) and lack of housing offer fears ($LAC$) and ranked sixth in the financial housing fears ($FIN$) RF model. Despite the fact that half of our respondents do not desire to stay in their current home, barriers to accessing information and tools for planning and changing housing persist (Kopanidis et al. 2017). The financial constraints due to retirement are significant motivators for relocations among older adults (Gobillon and Wolff 2011; Hasu 2018; Huggenberger et al. 2023; Ravazzini and Chesters 2018) and especially BBWA (Beringhs et al. 2024). The lack of suitable housing options discourages older individuals from moving, given relocation costs and the rising housing prices (Statistique Vaud 2018).

6. Conclusions

As populations in developed countries like Switzerland age, the number of older women living alone increases, with significant implications for social structures. In this paper, we examine the housing-related financial fears among Swiss BBWA, focusing on three primary concerns regarding housing: financial, price increase, and lack of housing offer fears. We base our study on a dataset containing 371 BBWA with information on socio-demographic factors, financial and well-being aspects, as well as housing characteristics and expectations. Through logistic regression and random forest models, we find factors that significantly affect these fears.

Our analysis underlines the role of lifestyle change fear, ownership status, and perceived financial situation in affecting all three examined housing-related financial fears. Concerns about the suitability of current housing are especially relevant for the lack of housing and financial fears. The desire to stay in their current accommodation also increases fears related to the price increase and the lack of housing offer. The consistency of these factors across all housing-related financial fears and their significance in different statistical models highlights their central role in understanding these concerns.

The current study has limitations due to the dataset’s limited scope and sample composition. Although the survey was specifically designed to investigate the housing situation of BBWA, information regarding revenue decrease following retirement is missing. The lack of data on housing expenses further restricts the analytical scope of this study. This scarcity of data can be attributed to the small sample size, which, in turn, reduces the statistical robustness and the broader applicability of our findings. Furthermore, the Swiss population is not fully represented in the sample composition, as French-speaking BBWA are over-represented. To mitigate these issues, future efforts would require a larger and more representative sample alongside a comprehensive set of financial information. Our research guides policymakers in reflecting on mitigating BBWA’s housing financial fears and the decline in post-retirement living standards. While our analysis helps to identify critical factors influencing these concerns, further research should be conducted to ascertain regional disparities in Switzerland and gather valuable information about housing options and support programs designed to address BBWA’s financial insecurities.