Willingness to Purchase Hypothetical Private Long-Term Care Insurance Plans in a Super-ageing Society: Evidence from Hong Kong

ABSTRACT

Despite the soaring demand for long-term care (LTC) services in aging societies, dedicated risk-pooling mechanisms are largely absent. Private insurance has been advocated but the market remains small. This study seeks to unravel this paradox through an empirical study in Hong Kong, a super-aging society. We analyzed middle-aged individuals’ willingness to purchase hypothetical private LTC insurance plans derived from a discrete choice experiment. A survey was conducted in 2020 and sampled 1,105 respondents. We noted a fairly encouraging level of acceptance but also found clear barriers toward potential purchase. The desire for self-sufficiency and preference for formal care powerfully increased individuals’ interest. Cognitive difficulty, habitual adherence to out-of-pocket payment, and unfamiliarity with the LTC insurance market reduced such interest. We explained the results with reference to the changing social dynamics and drew policy implications for LTC reforms in Hong Kong and beyond.

KEYWORDS:

• Financing
• Hong Kong
• Insurance
• Long-term care

Introduction

Rapid aging populations, coupled with the increasing prevalence of chronic disability among the elderly, have created tremendous demands for long-term care (LTC) services. Unfortunately, low awareness of LTC cost leads to poor financial preparedness among the elderly, resulting in financial hardship. Even in societies where risk pooling arrangements are in place, the lion’s share of the total LTC cost is still paid out-of-pocket, representing one of the catastrophic financial risks for older adults (Brown & Finkelstein, 2009; Costa-Font et al., 2015). In many high-income societies, taxation constitutes another key source of LTC financing, but cost explosion and fiscal uncertainties have cast serious doubt on the financial sustainability of public financing (Colombo & Mercier, 2012). Social LTC insurance appears to be an optimal solution to pursue but launching such a mandatory system is easier said than done. As such, some forms of voluntary private insurance becomes a promising candidate in LTC financing.

Conventional economics theories suggest that LTC dependency is an ideally insurable risk as a result of its high-cost but low probability nature (Costa-Font, 2010; Gleckman, 2007), but paradoxically, the reality is that the private LTC insurance (LTCI) market is remarkably underdeveloped even in fledged market economies. Among the various hurdles, high price of private LTCI is typically found to limit the willingness-to-subscribe of prospective purchasers (Gleckman, 2007). Hence, calibrating an optimal level of premium to balance commercial profitability and price acceptability is of critical significance for nurturing the market. However, in most societies where such market is virtually non-existent, estimating premiums would be to essentially solve the chicken-and-egg dilemma, because the lack of historical data makes actuarial analysis extremely difficult. On the other hand, LTCI premium depends on the level of benefits. Some European countries link the coverage to the level of dependency. For example, in Spain, three levels of LTC benefits are established: (1) basic level (funded by the national government), (2) supplementary level (provided by the provincial authorities), and (3) improvement level (private insurance is envisaged). Hence, private LTCI providers can focus on the upper-level market (De La Peña et al., 2018; Fernández-Ramos et al., 2018). However, the problem is how to implement such a system. Unlike Spain, some regions do not have the Dependency Act, and they also lack technical bases and actuarial data (e.g., differential on mortality) for the determination of dependency insurance.

The project represented by this current paper overcomes this difficulty with an innovative two-stage strategy. Set out in the context of Hong Kong, this study is valuable for the international readership for two reasons. Hong Kong, as one of the super-aging societies in the world, has been grappling with daunting challenges in LTC financing. The Hong Kong Special Administrative Region Government has been contemplating private LTCI as a solution to diversify its financing base. Hong Kong as an international finance hub owns a highly developed insurance industry. Private insurance has a very high penetration rate in the society, reflecting wide public acceptance. Thus, Hong Kong could become an ideal place to experiment private LTCI and offer useful experience for other aging societies.

In the first stage of the project, actuarial projection computed the indicative attributes including premiums of hypothetical LTCI plans. We then conducted a discrete choice experiment to elicit the hypothetical plans that were most appealing to middle-aged participants. The findings of this stage of the project have been published in a distinguished international journal (He et al., 2021). In the second stage, we investigated the willingness to purchase such hypothetical plans among the middle-aged adults through a telephone survey. We focused on the middle-aged adults because they are at the stage of life when financial planning for old-age is being, and should be, made. Built on the well-known Anderson’s behavioral model of health services use, this paper seeks to examine the receptivity of hypothetical LTCI plans among the middle-aged adults in Hong Kong and to reveal the demand-side factors associated with the willingness to purchase such plans. This paper analyses data collected from a telephone survey that was conducted in 2020 on a sample of 1,105 middle-aged Hong Kong adults. The study finds encouraging interest in two hypothetical private LTCI plans. The willingness to purchase was significantly associated with a series of factors. We seek explanations from the social dynamics in the Hong Kong context that profoundly shape people’s LTC planning behaviors, as well as the key characteristics of the hypothetical plans.

Literature review

Drawing from a wide range of multidisciplinary studies in the international literature, this study synthesizes both demand- and supply-side factors that may affect individuals’ decision of LTCI purchase. The individual decision to purchase insurance is determined not only by a self-assessment of risk and anticipated needs, but also by the built-in characteristics of the insurance plan itself. High premium costs are found prohibitive for many prospective purchasers (Gleckman, 2007). Prior studies have identified that income, financial status, and bequest motive are closely related to individuals’ decision of LTCI purchase (Allaire et al., 2016). Having high income and assets are strong predictors of LTCI ownership (Caro et al., 2011), but the effect of the bequest motive appears theoretically ambiguous (Brown et al., 2012; Sloan & Norton, 1997). In a most relevant study conducted in Hong Kong, He and Chou (2020) found that the middle-aged and older adults in their sample who showed a demand for LTCI were relatively young, better educated, and of higher social classes.

The availability of informal care might crowd out the demand for LTCI. Specifically, the presence of children who can provide informal care may serve as a major deterrent to purchase (Favreault et al., 2015). However, evidence for this hypothesis has been mixed (Allaire et al., 2016; Brown et al., 2012). Individuals living with children are usually less likely to require paid formal LTC, whereas those living alone are more likely to rely on paid care (van Houtven et al., 2015). Many individuals are reluctant to buy LTCI for psychological and cognitive reasons. Attitude and beliefs play a major role in decision making. How highly people value insurance depends on their beliefs about their future need for care. Unsurprisingly, such beliefs are found to be strongly associated with LTCI coverage (Finkelstein & McGarry, 2006).

Given the complexity of planning for such a long-term risk and navigating the various financing instruments, consumer personality and competency may play a crucial role in understanding insuring behaviors. However, this aspect has not received much attention in the literature. Both numerical abilities and cognitive capacity are found to be important predictors of LTCI ownership even after controlling for a variety of determinants of insurance demand (McGarry et al., 2017). Moreover, being a “planner” or a person who believes “it is important to plan now for the possibility of needing LTC services” is also positively associated with LTCI ownership in the US context (American Health Insurance Plan, 2012). Financial literacy and risk aversion are also associated with a high tendency to buy LTCI (Chen, 2003; He & Chou, 2020; Lusardi & Mitchell, 2007).

The context

Hong Kong is one of the fastest-aging societies, with a very high life expectancy at birth (82.2 for men and 88.1 for women in 2019) and a very low fertility rate (1,051 per 1,000 female population). The number of citizens aged 65 and over reached 1.32 million (or 17.6% of the total population) in 2019; this percentage is projected to increase to 35.0% by 2069. The number of old-old people (i.e., those aged 85 and above) will grow faster than other cohorts, from 153,000 in 2014 to 724,400 in 2064. This subgroup of the elderly population tends to have greater need for LTC services. At present, around 70–80% of community-based and residential services are utilized by the old-old. Moreover, the drastic changes in household size, family structure, and living arrangements will lead to increased demand for formal LTC (Chui, 2011; He & Chou, 2019). A 2008 survey in Hong Kong estimated that approximately 280,500 community-dwelling elderly needed assistance in their daily living, but only 47.6% of them had caregivers. Of the latter group, 65.6% received care from family members, whereas 25.6% were supported by domestic helpers or nurses (Census and Statistics Department, HKSAR, 2009).

Current LTC expenditure in Hong Kong is substantial and expected to escalate in the coming decades, from 1.4% of GDP in 2004 to 3.0% by 2036 (Chung et al., 2009). Good-quality private LTC services are available, but the costs are substantial even for middle-class families. The annual cost of a room in a high-quality private nursing home was over HK$120,000 in 2014 (Hong Kong Consumer Council, 2015), and these fees may double or even triple depending on the user’s level of disability. As a result, paying for private LTC represents one of the largest financial risks faced by Hong Kong’s aging population. The government has been exploring alternative arrangements, including vouchers, co-payment, and private insurance, to diversify the financing base of Hong Kong’s LTC system (Elderly Commission, 2017). The government also expects to encourage those better-off elderly to seek care in the private sector so that government-subsidized services could be left to the worse-off. Yet, given the high charges in private facilities, any attempt to move the elderly users away from the government-subsidized facilities must be accompanied by a proper financing instrument that is sufficient to shoulder a significant proportion of LTC cost to be incurred in the private sector. It is proposed in some other societies to transform the old-age pension into a formal benefit to help pay for LTC costs (De La Peña et al., 2021). However, such integration is not feasible in Hong Kong at this moment as it does not have publicly managed mandatory contributory pension plans (i.e., the first pillar in the 5-pillar World Bank Model). The second pillar, known as the Mandatory Provident Fund (MPF), serves as one main source of required retirement income for the vast majority of Hong Kong citizens. The MPF is a defined contribution system.

Through a discrete choice experiment, the first stage of the project elicited two hypothetical LTCI plans most preferred by participants. The primary plan was associated with the following characteristics: 1) a monthly benefit of HK$20,000; 2) inflation protection (3% annual growth of benefit); 3) 15 years of premium contribution (from age 50 to 65); and 4) monthly premium of HK$1,237 for men and HK$2,150 for women. The secondary plan was associated with the following characteristics: 1) a monthly benefit of HK$20,000; 2) no inflation protection or inflation protection of 3%; 3) a three-month waiting period; 4) 35 years of premium payment (30 to 65 years old); and 5) monthly premium of HK$180 for men and HK$298 for women. While both plans carry a generous benefit package (the second highest monthly benefit offered in the experiment), they do demonstrate salient differences. Most important, the secondary plan evens out high premium payment throughout a longer contribution period in order to get a significantly lower monthly payment. This option may be preferred by those younger middle-aged individuals who are reluctant to shoulder a heavy financial commitment so long before they officially enter old-age.

Methodology

Conceptual model

According to the Andersen Model, the utilization of health-related services is determined by three groups of factors: predisposing, enabling, and needs factors. Predisposing factors, such as age, gender, and belief, exist prior to care uptake but may influence behaviors related to purchase decisions. Enabling factors facilitate or inhibit the purchase of LTCI; these include income, asset, financial literacy, and cognitive capacity of individuals. Needs factors represent the perceived and real need for LTC that exert a direct influence on the purchase decision (Andersen, 1995). The Andersen Model has been widely applied in studies of LTCI (He & Chou, 2020; Igarashi et al., 2014; McGarry et al., 2014) and is adopted in this study to account for the various factors that shape individuals’ decision of LTCI purchase.

Data and sampling

A standard household survey with random sampling was originally planned for this study. Unfortunately, the COVID-19 outbreak in Hong Kong made it virtually impossible to carry out face-to-face data collection. To address this problem, a telephone survey was eventually employed. Respondents were recruited using a two-stage stratified sampling strategy. In the first stage, a random sample of residential addresses was drawn from a sample of quarters and segments obtained from the Census and Statistics Department. Both public housing estates and private properties were included. An invitation letter was subsequently sent to the sampled households inviting participation. Within households, a member aged between 40 and 59 years old was chosen through the “next birthday” method, in order to ensure unbiased and random selection. Data collection was performed through phone calls. A total of 1,987 households were invited; 1,105 of them accepted the invitation. The response rate was 55.6%. Consent to participate was sought prior to every interview. Respondents were assured of anonymity. Ethical clearance was obtained from the first author’s university.

When benchmarking our sample with the Hong Kong population, one will find out that key demographic characteristics are very similar. For instance, by the end of 2020, the percentage of people in Hong Kong aged between 50 and 59 accounted for 51% of the middle-aged group (40–59). Women accounted for 57% of the entire middle-aged group.

The dependent variables measured the likelihood that a respondent would purchase the two hypothetical LTCI plans. We presented the key features of both plans to respondents and invited them to indicate their willingness to subscribe using a 0–100% scale. These two continuous variables were named “Plan 1” and “Plan 2.” In the event that respondents did not indicate a firm intention (100%) to subscribe, we asked them to indicate their reasons, including; 1) dissatisfaction with the terms and conditions of the insurance plan; 2) the availability of alternative funding to pay for LTC costs (e.g. children, savings, other commercial insurance, investment returns, annuities, etc.); 3) distrust of commercial insurers; 4) belief that it is the government’s responsibility to finance LTC; 5) wishing to leave more of a bequest to children; 6) unfamiliarity with private LTCI; and 7) other reason. Respondents were invited to indicate the primary reason for non-purchase intention in this single-choice question battery.

Predisposing factors first included some binary variables: sex (1=female), age (1 = 50–59 years old), marital status (1=married), having children (1=having at least one), educational attainment (1=tertiary education), expected living arrangement (1=expected living alone), and availability of family care (1=agree with the statement that “One of my family members will look after me when I need LTC”). The dummy variable “insurance ownership” gauged if one owned any private insurance, including life insurance, health insurance, or accident insurance. We also incorporated four attitudinal statements about LTCI (Brown et al., 2012; He & Chou, 2020), namely 1) bequest motive; 2) preference for formal care; 3) the principle of familial obligation; and 4) concerns about insurers. Enabling variables included income, paying for LTCI with savings, paying LTC with welfare benefits, financial literacy, cognitive capacity, and ability to afford the premium. Needs variables included the existence of chronic physical conditions, subjective health status, and anticipated dependence in the future. The measurement of variables is detailed in Table 1.

Table 1. Measurement of selected explanatory variables.

Variable	Measurement
Bequest motive	“Here are four statements describing a parent’s offering of bequests to children. Please choose the one that best represents your view.” 1=“I shall leave my children a considerable amount if they look after me well in my old age;” 2=“I shall leave my children a considerable amount irrespective of how well they look after me in my old age;” 3=“I don’t have a firm plan concerning a bequest to children;” 4=“I don’t plan to leave anything to my children;” 5=Don’t know/refuse to answer. The first two answers were grouped into a binary variable to represent one’s bequest motive.
Preference for formal care	To what extent do you agree with the statement of “When I need long-term care services, I would expect to be looked after by professional staff, rather than my children or relatives”?’ 1=agree, otherwise 0.
Familial obligation	‘To what extent do you agree with the statement “It is the children/family members’ responsibility to shoulder elders’ long-term care costs”?’ 1=agree, otherwise 0.
Concerns about insurers	Three statements described common concerns about commercial insurers, and respondents were invited to rate their level of endorsement of each statement: 1) “I don’t think that the insurance company will operate long enough to pay for my long-term care expenses.” 2) “I’m worried that the insurance company may increase my premium after I join the plan.” 3) “I’m worried that the insurance company may find excuses to decline my payment request.” 1=strongly agree, 2=agree, 3=disagree, 4=strongly disagree. In the analysis, we first consolidated the first two categories into “1” and the other two into “0” and then created an additive index to measure respondents’ concerns on insurers. A higher value represents greater distrust of commercial insurers.
Insurance ownership	“Whether you purchase any private insurance products including life insurance, health insurance, accident insurance, etc.”
Monthly personal income	We used a 12-item categorical variable in ascending order from <HK$2,000 to ≥HK$60,000 and then created a dummy variable of “low income” for monthly income below HK$10,000 and another of “high income” for HK$30,000 or more.
Paying with savings	“To what extent do you agree with the statement ‘I have my own means of paying for long-term care (such as savings or assets) even without long-term care insurance’?” 1=agree, otherwise 0. This binary variable represents the ability to pay LTC costs with one’s own savings.
Paying with welfare	“To what extent do you agree with the statement ‘When I’m in need of long-term care, the CSSA and other welfare benefits will be sufficient to meet my financial needs’?” 1=agree, otherwise 0. This binary variable represents the intention of paying one’s LTC costs with social welfare benefits.
Financial literacy	A battery of three items was used to assess respondents’ basic knowledge of finance: knowledge related to the compound nature of interest rates, understanding of the effect of inflation on consumption power, and knowledge of risk diversification (see Appendix 1 for details). We coded every correct answer as 1 and generated an additive index to measure financial literacy, with higher values representing better basic finance knowledge.
Cognitive capacity	A six-item module was used to test the respondents’ cognitive capacity in relation to simple mathematical skills and memory (Appendix 1). Each question was associated with a single correct answer. We created a dummy variable of “cognitive difficulty” and assigned it the value of 1 if the respondent answered any of the questions incorrectly.
Ability to afford the premium	“To what extent do you agree with the statement ‘I don’t think that I could afford the premium for long-term care insurance.’?” 1=agree, otherwise 0. This binary variable represents one’s ability to afford the premium for LTC insurance.
Anticipated dependence	Respondents were invited to rate their likelihood of disability and the resultant dependence using a three-item module as follows: 1) “I will need long-term care in the last few years of my life.” 2) “From a certain day on, my daily life will need to be assisted, e.g. bathing, wearing, eating, and toileting.” 3) “From a certain day on, I will be unable to live independently due to health problems.” These items were scored using the scale 1=very likely, 2=quite likely, 3=quite unlikely, 4=very unlikely. When a respondent indicated a likelihood to any of the three items, s/he was considered to have an anticipated potential dependence resulting from chronic disability.
Chronic condition	A comprehensive list of chronic conditions was presented (see Appendix 2 for details). Respondents were invited to indicate those from which they were suffering. This item was coded as 1 if any condition was present.
Self-reported health status	“What do you think of your current health status?” This item was scored using the scale 1=very good, 2=quite good, 3=average, 4=quite bad, 5=very bad. In the analysis, we combined the first two categories into a dummy variable to represent one’s good subjective health status.

Results

Descriptive results

First and foremost, the mean score of the interest in purchasing Plan 1 was high at 61.59 (SD = 34.27) out of 100, while that of Plan 2 was 46.72 (SD = 30.75). As expected, the primary insurance product generated by the discrete choice experiment appealed more to the middle-aged adults in our sample. We then investigated the stated reasons for the intention of non-purchase. The distribution of reasons on both plans was largely similar (see Table 2). Around one-fifth of the respondents attributed low interest to their unfamiliarity with the LTCI market. Almost the same percentage of respondents indicated that their preference for alternative ways of financing (other than insurance) was the key reason. The third most popular reason was dissatisfaction with key features of the hypothetical products. Distrust of commercial insurance companies explained more than 10% of the non-purchase decision.

Table 2. Reasons for non-purchase intention of the hypothetical LTCI plans.

Reasons	Plan 1	Plan 2
Unfamiliarity with the LTC insurance market	175 (21.6%)	199 (20.2%)
Preference for alternative ways of LTC financing	159 (19.6%)	194 (19.6%)
Dissatisfaction with key product features	142 (17.5%)	183 (18.6%)
Distrust of commercial insurers	103 (12.7%)	129 (13.1%)
Wishing to leave a bequest to children	84 (10.4%)	108 (11.0%)
Government should take the responsibility to finance LTC	78 (9.6%)	77 (7.8%)
Other	69 (8.5%)	95 (9.6%)
Total	810	985

The full summary of descriptive statistics is reported in Table 3. In our sample, 60.7% of the respondents were women. 53% of the sample aged between 50 and 59, while the rest was younger middle-aged adults (40–49 years old). Nearly 87% of our respondents were married. More than one-fifth held tertiary education credentials. About 17% of the respondents were considered high-income earners with a monthly income beyond HK$30,000 whereas 28% belonged to the low-income group, earning less than HK$10,000 a month. More than one-third of the respondents reported at least one chronic condition. Around half of the sampled individuals perceived their health status to be good. 80% had at least one child. About 73% of respondents owned at least one private insurance policy.

Table 3. Summary statistics for the household survey.

Variable	Mean	Min	Max	SD	N
Plan 1	61.59	0	100	34.27	1,105
Plan 2	46.72	0	100	30.75	1,105
Female	0.607	0	1	0.489	1,105
Age 50–59	0.530	0	1	0.499	1,105
Married	0.866	0	1	0.341	1,105
Having children	0.802	0	1	0.399	1,105
University education	0.212	0	1	0.409	1,105
Expected living-alone	0.091	0	1	0.288	1,105
Availability of family care	0.663	0	1	0.473	1,105
Bequest motive	0.603	0	1	0.490	1,105
Preference for formal care	0.722	0	1	0.448	1,105
Family obligation	0.520	0	1	0.500	1,105
Concern on insurers	0.327	0	3	0.723	1,105
Low income	0.282	0	1	0.450	873
High income	0.168	0	1	0.374	873
Paying by saving	0.701	0	1	0.458	1,105
Paying by welfare	0.452	0	1	0.498	1,105
Financial literacy	2.482	0	3	0.744	1,105
Cognitive difficulty	0.138	0	1	0.345	1,105
Ability to afford premium	0.567	0	1	0.496	1,105
Anticipated dependence	0.768	0	1	0.422	1,105
Chronic condition	0.364	0	1	0.481	1,105
Self-reported health status	0.492	0	1	0.500	1,105
Insurance ownership	0.731	0	1	0.443	1,105

More than three-quarters of the respondents in the sample anticipated some degree of LTC needs in their old age. An interesting finding is that while almost two-thirds of the middle-aged adults in our study expected family care to be available when LTC needs arose and only 9.1% anticipated living alone in their old age, the large majority of them actually stated an explicit preference for formal care. This result echoes the finding of He and Chou (2020) that the current cohorts of middle-aged and older adults in Hong Kong show a very strong desire of self-reliance. Around half of the respondents believed that supporting LTC is primarily a familial obligation. Reflecting a habitual adherence to out-of-pocket payment of LTC expenses, the majority indicated personal savings to be their principal source of LTC financing and close to half reported that they would pay for LTC using social welfare benefits. More than half of the respondents said that they could afford the premium of the potential LTCI. Around 60% expressed a bequest motive as they wished to leave a considerable amount of money to their children. Cognitive capacity, on average, was high, as the mean score of cognitive difficulty was 0.138 (max = 1). Financial literacy was also on the high side, with a mean score of 2.48 out of 3.

Multivariate results

As the dependent variables, the likelihood that a respondent would purchase the two hypothetical plans, are continuous but their range is constrained (i.e., 0–100%), we used Tobit models in multivariate analysis. We regressed the dependent variables in a stepwise fashion in order to analyze the effect of the three sets of independent variables both individually and in combination. The regression results are reported in Tables 4 and 5. As expected, anticipated dependence in old age powerfully predicted interest in both plans. Unsurprisingly, the presence of cognitive difficulties hampered some respondents from making proper judgments about financial planning, which in turn led to significantly less interest in the hypothetical purchase.

Table 4. Regression results for interest in purchasing Plan 1.

	Model 1	Model 2	Model 3
Predisposing variables
Female	−8.353**(3.370)	−12.51***(3.956)	−10.51***(3.914)
Age 50–59	−8.485**(3.363)	−3.255(3.669)	−2.619(3.674)
Married	9.128(6.075)	10.53(6.835)	9.895(6.729)
Having children	−7.191(4.735)	−8.761*(5.248)	−8.641*(5.172)
University education	−14.50***(4.040)	−12.72***(4.618)	−12.25***(4.570)
Expected living alone	10.39(6.762)	7.223(7.393)	7.627(7.294)
Availability of family care	−1.475(3.452)	−4.350(3.874)	−3.539(3.816)
Bequest motive	9.443***(3.451)	5.350(3.793)	4.477(3.736)
Preference for formal care	10.65***(3.640)	12.88***(3.969)	11.40***(3.917)
Familial obligation	−11.50***(3.273)	−12.89***(3.621)	−12.81***(3.567)
Concerns about insurers	0.541(2.248)	0.233(2.422)	0.839(2.389)
Insurance ownership	14.01***(3.708)	20.36***(4.141)	20.32***(4.083)
Enabling variables
Low income		0.138(4.288)	0.288(4.226)
High income		−13.46***(5.176)	−11.56**(5.112)
Paying with savings		−10.48**(4.103)	−9.178**(4.048)
Paying with welfare		1.783(3.705)	2.808(3.656)
Financial literacy		0.655(2.431)	0.247(2.395)
Cognitive difficulty		−22.19***(5.354)	−22.24***(5.320)
Ability to afford premium		10.63***(3.761)	8.487**(3.741)
Needs variables
Anticipated dependence			19.70***(4.094)
Chronic condition			−1.176(4.034)
Self-reported health status			−0.376(3.802)
constant	59.66***(7.927)	66.23***(11.48)	51.93***(12.04)
# of observations	1105	873	873
Pseudo R2	0.0087	0.0193	0.0229

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Table 5. Regression results for interest in purchasing Plan 2.

	Model 1	Model 2	Model 3
Predisposing variables
Female	−9.169***(2.507)	−9.283***(2.995)	−7.969***(2.979)
Age 50–59	−2.813(2.505)	0.123(2.785)	0.352(2.802)
Married	9.614**(4.585)	9.071*(5.240)	9.021*(5.188)
Having children	−1.362(3.544)	1.705(4.002)	2.044(3.961)
University education	0.213(3.011)	−1.780(3.520)	−1.440(3.499)
Expected living alone	13.90***(5.082)	11.78**(5.644)	12.29**(5.589)
Availability of family care	4.402*(2.565)	3.835(2.923)	4.199(2.890)
Bequest motive	2.703(2.578)	−0.191(2.888)	−0.796(2.858)
Preference for formal care	7.976***(2.730)	7.875***(3.033)	6.932**(3.006)
Familial obligation	−6.622***(2.435)	−7.623***(2.742)	−7.587***(2.712)
Concerns about insurers	−5.950***(1.687)	−5.145***(1.844)	−4.704**(1.827)
Insurance ownership	5.100*(2.783)	5.812*(3.153)	5.987*(3.122)
Enabling variables
Low income		−3.599(3.253)	−3.733(3.219)
High income		−1.294(3.958)	−0.264(3.925)
Paying with savings		−5.517*(3.099)	−4.720(3.073)
Paying with welfare		−3.470(2.810)	−2.817(2.784)
Financial literacy		−3.233*(1.851)	−3.628**(1.833)
Cognitive difficulty		−7.276*(4.140)	−7.663*(4.133)
Ability to afford premium		0.998(2.858)	−0.300(2.856)
Needs variables
Anticipated dependence			13.71***(3.163)
Chronic condition			2.077(3.087)
Self-reported health status			3.196(2.897)
constant	35.65***(5.954)	52.21***(8.756)	39.34***(9.253)
# of observations	1105	873	873
Pseudo R2	0.0065	0.0083	0.0111

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

As shown in both tables, women were less likely to purchase either plan; and this association was highly significant. This result seems puzzling because women typically have a longer life expectancy and thus greater LTC needs. We posit that their reluctance was primarily owing to the high annual premium payable at HK$25,440 compared to HK$14,884 for men (the primary plan). Furthermore, when the data collection was being undertaken, the world was wrestling with the COVID-19 pandemic that caused major economic downturn and rising unemployment in Hong Kong (as it was elsewhere). Women may be understandably more conservative when considering such a hypothetical purchase in these circumstances.

In Table 4, the older cohort of middle-aged adults (50–59 years old) demonstrated lower interest in Model 1 but this variable ceased to be statistically significant after the enabling and needs variables were included in the regression. No discernible age effect can be seen from Table 5 either. Neither having children nor the availability of family care provided a statistically robust and consistent explanation for the variance in the dependent variables. For both plans, people holding other private insurance policies were more likely to purchase LTCI, arguably because of their tendency of risk aversion. Overall, the statistical significance of “familial obligation” (negative), preference for formal care, and private insurance ownership appear to reflect a desire for self-sufficiency when it comes to paying for LTC expenses.

We also find that holding tertiary educational credentials did not necessarily lead to greater interest in either of the hypothetical products. In fact, higher education was actually associated with a significantly weaker intention of purchasing Plan 1. At first glance, this result appears puzzling, as higher education attainment was commonly found to be a strong predictor of LTCI ownership or willingness to subscribe (He & Chou, 2020; McGarry et al., 2014). The literature on financial literacy and health economics suggests that higher education increases not only people’s level or risk aversion, but also their willingness to delay gratification for the future, a phenomenon denoted as “temporal discounting” (Cramer & Jensen, 2006; Meier & Sprenger, 2013). In simple words, individuals with higher levels of education tend to prefer premium payment for a longer period of time when making such decisions. Associated with a fairly short payment period and a considerably high premium rate, Plan 1 may therefore be less appealing to the middle-aged respondents.

A similar result is noted on income, as high-income earners and those reporting an ability to pay for LTC with savings demonstrated less interest in the hypothetical products (although the statistical significance was lost in some models). In particular, high-income status was strongly associated with a negative acceptance of Plan 1. We speculate that this non-purchase intention may be partly explained by their confidence in absorbing the LTC expenses through retirement savings. The ability to afford the premium predicted purchase intention in only one model. Other than in Model 1 of Table 4, the variable “bequest motive” held no statistical significance in explaining purchase interest.

Robustness check

We performed a set of robustness check to corroborate our statistical results. First, we used alternative models (fractional probit model) to run regression analysis again. Second, we excluded those observations that attributed the reason for non-purchase to dissatisfaction with product features, because this group of people may be in favor of insurance but tend to purchase alternative plans. By excluding this subset of the observations, we were able to gain more robust insights as to why people were not interested in the two hypothetical plans. Third, we excluded subgroups including those who reported unfamiliarity with the LTCI market and those who thought that the state should take the responsibility for LTC. Results of robustness checks are reported in Appendixes 3 to 10. Overall, the statistical patterns strongly resemble those reported in the main regression tables. Hence, our main empirical findings are corroborated.

Discussions

Taken together, the results reported in Tables 4 and 5 suggest that the current generation of middle-aged adults in our sample (born between 1961 and 1980) show a remarkably strong preference for formal LTC, be it in residential or community settings. More importantly perhaps, this preference has reinforced their interest in purchasing private LTCI. In contrast, possible support from family members appears no longer to be important as it has not crowded out people’s interest in LTCI purchase. These results as a whole mirror the notable changes in family structure and intergenerational dynamics in families in Hong Kong. The current cohort of middle-aged Hong Kong adults is more economically independent and appears to have much lower expectations that their children, or their family as a whole, will be the chief source of LTC support (He & Chou, 2019). They tend to recognize that their children are unlikely to support them in monetary or in-kind terms, should chronic disability unfortunately occur. Another telling piece of evidence is the diminished importance of bequest motive, as observed in our empirical results. While many previous studies, including those carried out in Chinese societies, have found bequest motive to be significant in senior citizens’ financial planning behaviors, our results have not endorsed that effect. We would argue that some major changes in familial values, toward individualism and independence, may be shaping the financial planning behaviors of middle-aged adults in Hong Kong.

All these results appear to suggest that middle-aged adults at the middle and higher strata of the Hong Kong society are not, in general, interested in private LTCI as would be expected. Echoing the univariate results for the stated reasons for non-purchasing (as reported in Table 2), this low intention is arguable due to unfamiliarity with the LTCI market and habitual adherence to old ways of financing LTC costs such as savings and assets. Majority of these better-off middle-aged people in Hong Kong appear to hold the view that their LTC needs can be met by hiring a domestic helper or self-insurance. No evidence indicates that welfare benefits crowd out respondents’ intention to purchase.

We did observe certain discrepancies in the statistical patterns between Tables 4 and 5, most of which can be reasonably explained by the salient differences in the key features of the two hypothetical products. Plan 1 is characterized by a generous package, relatively short payment period, and higher monthly premium, while Plan 2 trades off a much longer payment period in exchange for a significantly lower monthly premium. Specifically, concerns about insurers and financial literacy significantly reduced respondents’ interest in Plan 2 while having no major effect on Plan 1. This is arguable because Plan 2 evens out the substantive amount of premium across a very period (35 years), which may naturally arouse people’s concern about the vast uncertainties with regard to financial risks and possible insolvency on the part of insurance companies in the course of such a long-term contractual relationship.

Going beyond the sheer plain statistics and the Hong Kong context, this study also draws useful implications for the international literature as well as LTC financing reforms in other high-income aging societies. First and foremost, although the literature suggests a dim prospect for private LTCI even in free market economies, individuals, especially those middle-aged, are in fact fairly receptive, should a proper insurance plan be presented to them. Such plans must be built on sound actuarial analysis in order to meet basic commercial principles, but more importantly, the product designs should take into account the socioeconomic and cultural dynamics of the society. Many post-industrialized societies such as Hong Kong have been experiencing major changes in societal values, family structure, and intergenerational relations. The value of self-reliance and the preference for formal care, are evidently on rise. These changes have provided good opportunities for the development of LTCI.

Furthermore, although the LTCI paradox may partly result from the built-in characteristics of insurance plans themselves – especially high premium cost – demand-side issues are also responsible – perhaps to a larger extent – for the underdevelopment of the market. Low awareness of LTC cost, myopia, and habitual adherence to out-of-pocket payment still appear robust, even in Hong Kong, a society with a highly developed finance industry and wide coverage of commercial insurance. Low familiarity with LTCI and concerns on commercial insurers are likely to hinder consumers’ acceptance. Therefore, public education programs and information campaigns are necessary.

Limitations

This study is certainly not without limitations and we fully acknowledge them. The sweeping COVID-19 pandemic precluded a standard household survey, so we had to opt for a telephone survey. In addition, the very basis of this survey study was the discrete choice experiment that was built on actuarial projection. Several important assumptions were made when performing the actuarial analysis that may not fully hold in reality. The hypothetical plans we presented to respondents were illustrative in nature and may not necessarily reflect what a real product – if launched – may look like in the actual market. Therefore, the results of this study should be interpreted with caution.

Conclusions

The empirical results presented above lead us to three major conclusions. First, given the voluntary nature of the hypothetical plans, the figure of 60% of respondents in Hong Kong expressing an interest is high. Second, poor understanding of the LTCI market has clearly limited people’s interest. Although these respondents showed a basic level of financial literacy in general, they did not appear to be familiar with the market. Even middle- and upper-class individuals with ability to pay for care tended to habitually adhere to the option of making out-of-pocket payment of LTC costs from savings or assets instead of subscribing to a prepaid insurance plan that offers a fairly generous package. Third, as seen in other post-industrialized societies in East Asia, changes in intra-family relations and intergenerational values are also very evident in Hong Kong. In our study, those middle-aged adults who showed a higher interest in LTCI demonstrated a number of salient characteristics with regard to their LTC planning. They showed a high level of awareness of the potential LTC risks in old age and relatively low expectations that family would provide care. They did not seem to be bounded by traditional values such as leaving money to children, and strongly preferred formal LTC services.

Key Points

• The survey identified strong interest in purchasing hypothetical LTC insurance plans among middle-aged adults, particularly those who already owned other private insurance policies.

• The desire for self-sufficiency and the preference for formal care powerfully increased individuals’ interest.

• Cognitive difficulty, habitual adherence to out-of-pocket payment, and unfamiliarity with the LTC insurance market reduced such interest.

• The experience gained in Hong Kong is of great value for other societies that are contemplating the possibility of private insurance in their LTC reforms.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

The work was supported by the The Policy Innovation and Co-ordination Office, The Hong Kong Special Administrative Region Government [2017.A5.034.18B].

Appendix 1.

Question module used in measuring cognitive difficulties and financial literacy

Table

B35 Read out the questions below and record if the answer is correct, based on the reference answer	Reference
a. Suppose the infection rate of a disease is 10%, how many people would get infected out of 1000 people?	a. 100
b. Suppose five people win a lottery and the total award is HK$2 million, how much would each of them get?	b. HK$400,000
c. Suppose you have HK$200 on your bank account and the annual interest rate is 10%, how much would there be on your account two year later? Note, interest is offered on a yearly basis.	c. HK$242
d. I shall read out three words and ask you to repeat five minutes later: blue, sad, train.	Record
e. Please read these numbers in a reverse order:4 2 7 3 1	e. 1 3 7 2 4
f. Please deduct 7 from 100 until you are told to stop (until f5）	f1. 93
	f2. 86
	f3. 79
	f4. 72
	f5. 65

Appendix 2.

Full list of chronic conditions included in the survey instrument

Table

B18 Do you suffer from any of the following conditions that have been diagnosed by a qualified doctor?	0=No 1=Yes
a. Arthritis (and related conditions; omitted here)	□
b. Heart diseases (and related conditions; omitted here)	□
c. Hypertension	□
d. Asthma (and related conditions; omitted here)	□
e. Stomach disease (and related conditions; omitted here)	□
f. Diabetes	□
g. Fracture	□
h. Parkinson’s disease	□
i. Cognitive impairment	□
j. Mental health problems	□
k. Stoke (and related conditions; omitted here)	□
i. Cancer	□
m. Eye diseases (and related conditions; omitted here)	□
n. High cholesterol	□
o. Gout	□
p. Major/urinary incontinence	□
q. Epidermal ulcers	□
r. Limb paralysis	□
s. Renal diseases	□
t. Liver diseases	□
u. Lung diseases	□
v. Other chronic conditions requiring more than one year of treatment	□
w. Other conditions	□

Appendix 3.

Regression results for interest in purchasing Plan 1 (fractional response model, probit)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−0.132**(0.057)	−0.204***(0.070)	−0.170**(0.070)
Age 50–59	−0.158***(0.056)	−0.0772(0.064)	−0.0689(0.066)
Married	0.197*(0.105)	0.231**(0.116)	0.227*(0.118)
Having children	−0.124(0.082)	−0.164*(0.092)	−0.165*(0.091)
University education	−0.241***(0.071)	−0.213***(0.082)	−0.204**(0.083)
Expected living alone	0.218*(0.112)	0.160(0.118)	0.168(0.118)
Availability of family care	−0.0266(0.058)	−0.0753(0.066)	−0.0628(0.066)
Bequest motive	0.189***(0.059)	0.128*(0.068)	0.113*(0.067)
Preference for formal care	0.189***(0.059)	0.254***(0.068)	0.231***(0.068)
Familial obligation	−0.205***(0.054)	−0.230***(0.062)	−0.231***(0.063)
Concerns about insurers	−0.00403(0.038)	−0.00689(0.043)	0.00593(0.044)
Insurance ownership	0.211***(0.061)	0.319***(0.071)	0.321***(0.071)
Enabling variables
Low income		−0.00526(0.074)	−0.00430(0.074)
High income		−0.245***(0.092)	−0.216**(0.092)
Paying with savings		−0.191***(0.068)	−0.170**(0.068)
Paying with welfare		−0.00454(0.066)	0.0136(0.066)
Financial literacy		0.0132(0.041)	0.00314(0.040)
Cognitive difficulty		−0.383***(0.093)	−0.391***(0.095)
Ability to afford premium		0.173***(0.064)	0.134**(0.064)
Needs variables
Anticipated dependence			0.359***(0.071)
Chronic condition			−0.00242(0.069)
Self-reported health status			0.00913(0.065)
constant	0.148(0.135)	0.280(0.190)	0.0145(0.206)
# of observations	1105	873	873
Pseudo R2	0.0256	0.0523	0.0625

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 4.

Regression results for interest in purchasing Plan 2 (Fractional response model, probit)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−0.183***(0.048)	−0.179***(0.059)	−0.154***(0.059)
Age 50–59	−0.0583(0.048)	−0.00113(0.055)	0.00208(0.055)
Married	0.188**(0.092)	0.160(0.107)	0.162(0.108)
Having children	−0.0389(0.069)	0.0249(0.078)	0.0310(0.077)
University education	0.0308(0.060)	−0.00659(0.071)	0.00124(0.072)
Expected living alone	0.275***(0.101)	0.220**(0.112)	0.231**(0.111)
Availability of family care	0.0824*(0.048)	0.0720(0.056)	0.0797(0.056)
Bequest motive	0.0631(0.049)	0.00861(0.057)	−0.00386(0.056)
Preference for formal care	0.156***(0.051)	0.160***(0.058)	0.142**(0.058)
Familial obligation	−0.133***(0.046)	−0.156***(0.053)	−0.155***(0.054)
Concerns about insurers	−0.122***(0.036)	−0.107***(0.039)	−0.0982**(0.040)
Insurance ownership	0.104**(0.053)	0.110*(0.061)	0.114*(0.060)
Enabling variables
Low income		−0.076(0.062)	−0.079(0.062)
High income		−0.020(0.084)	0.001(0.083)
Paying with savings		−0.112*(0.058)	−0.098*(0.058)
Paying with welfare		−0.072(0.055)	−0.060(0.055)
Financial literacy		−0.073*(0.038)	−0.082**(0.037)
Cognitive difficulty		−0.142*(0.083)	−0.151*(0.085)
Ability to afford premium		0.006(0.057)	−0.020(0.057)
Needs variables
Anticipated dependence			0.269***(0.065)
Chronic condition			0.047(0.059)
Self-reported health status			0.065(0.055)
constant	−0.277**(0.117)	0.096(0.171)	−0.159(0.186)
# of observations	1105	873	873
Pseudo R2	0.0159	0.0192	0.0251

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 5.

Robustness check of intention to purchase Plan 1 (excluding those dissatisfied with product features)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−9.723**(4.014)	−14.73***(4.678)	−12.66***(4.608)
Age 50–59	−8.741**(4.003)	−2.802(4.379)	−2.076(4.361)
Married	9.967(6.989)	14.05*(7.897)	13.12*(7.740)
Having children	−9.164*(5.564)	−11.98*(6.245)	−11.84*(6.133)
University education	−17.43***(4.763)	−15.95***(5.454)	−15.18***(5.394)
Expected living alone	8.517(7.806)	4.384(8.479)	5.012(8.348)
Availability of family care	−0.645(4.096)	−3.090(4.675)	−2.122(4.593)
Bequest motive	11.53***(4.063)	6.571(4.479)	5.168(4.400)
Preference for formal care	13.15***(4.292)	14.97***(4.690)	12.80***(4.619)
Familial obligation	−13.17***(3.889)	−16.77***(4.324)	−17.12***(4.249)
Concerns about insurers	−0.0715(2.654)	−0.272(2.879)	0.748(2.833)
Insurance ownership	16.76***(4.404)	24.43***(4.942)	24.11***(4.860)
Enabling variables
Low income		1.716(5.103)	2.730(5.021)
High income		−13.05**(6.088)	−11.21*(5.985)
Paying with savings		−13.44***(4.946)	−11.90**(4.860)
Paying with welfare		2.576(4.397)	3.770(4.328)
Financial literacy		1.792(2.777)	1.265(2.729)
Cognitive difficulty		−25.26***(6.100)	−25.42***(6.054)
Ability to afford premium		14.30***(4.494)	11.93***(4.456)
Needs variables
Anticipated dependence			24.05***(4.695)
Chronic condition			−4.033(4.835)
Self-reported health status			0.328(4.534)
constant	60.37***(9.318)	65.22***(13.36)	49.79***(13.91)
# of observations	963	754	754
Pseudo R2	0.0102	0.0244	0.0295

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 6.

Robustness check of intention to purchase Plan 2 (excluding those dissatisfied with product features)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−9.691***(2.942)	−10.91***(3.554)	−9.836***(3.525)
Age 50–59	−2.567(2.946)	0.924(3.314)	0.684(3.311)
Married	8.066(5.273)	6.908(6.113)	6.771(6.038)
Having children	−1.127(4.182)	1.425(4.908)	1.446(4.853)
University education	−1.703(3.460)	−3.368(4.128)	−3.177(4.099)
Expected living alone	15.74***(5.845)	13.39**(6.574)	13.80**(6.506)
Availability of family care	5.001(3.038)	3.689(3.557)	3.807(3.515)
Bequest motive	3.580(2.995)	0.0358(3.391)	−0.707(3.355)
Preference for formal care	9.069***(3.155)	8.984**(3.557)	8.161**(3.519)
Familial obligation	−7.420***(2.864)	−8.675***(3.263)	−8.846***(3.225)
Concerns about insurers	−4.930**(2.011)	−4.142*(2.221)	−3.416(2.202)
Insurance ownership	4.943(3.251)	5.628(3.757)	6.097(3.721)
Enabling variables
Low income		−3.387(3.893)	−3.197(3.851)
High income		−1.819(4.584)	−1.200(4.538)
Paying with savings		−4.603(3.694)	−3.226(3.667)
Paying with welfare		−4.050(3.380)	−3.090(3.352)
Financial literacy		−2.908(2.144)	−3.220(2.122)
Cognitive difficulty		−7.525(4.703)	−8.560*(4.710)
Ability to afford premium		4.153(3.431)	2.845(3.431)
Needs variables
Anticipated dependence			14.59***(3.659)
Chronic condition			3.082(3.643)
Self-reported health status			4.275(3.442)
constant	36.71***(6.977)	52.28***(10.16)	38.17***(10.72)
# of observations	922	724	724
Pseudo R2	0.0064	0.0086	0.0115

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 7.

Regression results for interest in purchasing Plan 1 (excluding dissatisfied with the products and those believing the state should take the responsibility)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−11.18**(4.417)	−16.92***(5.037)	−14.32***(4.975)
Age 50–59	−9.549**(4.407)	−3.476(4.693)	−3.003(4.692)
Married	9.916(7.719)	15.68*(8.474)	14.45*(8.320)
Having children	−11.63*(6.202)	−15.55**(6.778)	−15.06**(6.668)
University education	−17.37***(5.225)	−13.13**(5.808)	−12.39**(5.748)
Expected living alone	4.886(8.411)	0.00947(8.927)	0.248(8.802)
Availability of family care	0.270(4.462)	−2.602(4.965)	−1.317(4.886)
Bequest motive	9.745**(4.466)	4.880(4.798)	4.327(4.714)
Preference for formal care	15.22***(4.690)	16.98***(5.000)	14.73***(4.941)
Familial obligation	−13.65***(4.264)	−17.31***(4.614)	−17.98***(4.544)
Concerns about insurers	−1.128(2.894)	−1.707(3.011)	−0.669(2.969)
Insurance ownership	17.71***(4.873)	26.03***(5.308)	25.58***(5.224)
Enabling variables
Low income		2.144(5.462)	2.598(5.387)
High income		−18.34***(6.440)	−15.91**(6.350)
Paying with savings		−14.40***(5.336)	−12.38**(5.255)
Paying with welfare		3.867(4.709)	4.948(4.638)
Financial literacy		2.250(2.958)	1.732(2.910)
Cognitive difficulty		−27.23***(6.560)	−27.72***(6.517)
Ability to afford premium		16.76***(4.838)	14.77***(4.787)
Needs variables
Anticipated dependence			23.92***(5.040)
Chronic condition			−3.192(5.159)
Self-reported health status			−1.391(4.849)
constant	66.30***(10.12)	69.63***(14.23)	53.68***(14.87)
# of observations	885	700	700
Pseudo R2	0.0104	0.0283	0.0331

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 8.

Regression results for interest in purchasing Plan 2 (excluding those dissatisfied with the products and those believing the state should take the responsibility)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−10.31***(3.197)	−12.14***(3.859)	−11.05***(3.821)
Age 50–59	−2.193(3.192)	0.683(3.573)	0.278(3.562)
Married	7.707(5.682)	6.722(6.508)	6.385(6.416)
Having children	−2.580(4.586)	−1.546(5.372)	−0.963(5.304)
University education	−0.534(3.746)	−2.348(4.431)	−2.060(4.387)
Expected living alone	15.72**(6.297)	13.73*(7.033)	13.93**(6.947)
Availability of family care	5.602*(3.284)	4.563(3.845)	4.866(3.790)
Bequest motive	4.137(3.249)	0.928(3.664)	−0.110(3.618)
Preference for formal care	9.666***(3.385)	9.990***(3.809)	8.858**(3.766)
Familial obligation	−8.890***(3.098)	−10.01***(3.529)	−10.19***(3.479)
Concerns about insurers	−5.794***(2.155)	−4.879**(2.358)	−4.055*(2.334)
Insurance ownership	4.153(3.519)	5.895(4.032)	6.208(3.986)
Enabling variables
Low income		−2.816(4.185)	−2.641(4.130)
High income		−2.251(4.921)	−1.641(4.860)
Paying with savings		−4.851(4.055)	−2.896(4.025)
Paying with welfare		−4.055(3.657)	−3.403(3.615)
Financial literacy		−2.000(2.298)	−2.498(2.271)
Cognitive difficulty		−7.058(5.136)	−7.776(5.108)
Ability to afford premium		5.563(3.712)	4.376(3.693)
Needs variables
Anticipated dependence			16.68***(3.991)
Chronic condition			3.099(3.914)
Self-reported health status			4.146(3.709)
constant	39.43***(7.540)	52.18***(10.93)	36.61***(11.49)
# of observations	845	670	670
Pseudo R2	0.0071	0.0095	0.0130

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 9. Regression results for interest in purchasing Plan 1 (excluding those dissatisfied with the products and those unfamiliar with LTCI)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−13.21**(5.190)	−20.42***(6.208)	−17.81***(6.100)
Age 50–59	−6.882(5.121)	0.103(5.776)	0.820(5.752)
Married	13.93(9.217)	16.72(10.58)	18.12*(10.33)
Having children	−13.12*(7.332)	−17.72**(8.395)	−16.29**(8.229)
University education	−21.50***(6.161)	−24.23***(7.235)	−21.97***(7.138)
Expected living alone	14.56(10.70)	8.259(11.89)	9.625(11.69)
Availability of family care	−2.117(5.246)	−5.596(6.152)	−4.512(6.022)
Bequest motive	12.49**(5.345)	3.151(6.079)	0.453(5.976)
Preference for formal care	15.44***(5.599)	15.54**(6.321)	13.86**(6.201)
Familial obligation	−15.02***(5.009)	−18.30***(5.706)	−17.48***(5.588)
Concerns about insurers	0.0270(3.492)	−1.352(3.765)	0.333(3.706)
Insurance ownership	20.40***(5.741)	32.73***(6.688)	31.69***(6.544)
Enabling variables
Low income		3.240(6.755)	3.410(6.618)
High income		−11.39(8.151)	−10.06(7.984)
Paying with savings		−14.00**(6.538)	−11.96*(6.408)
Paying with welfare		1.460(5.854)	2.950(5.748)
Financial literacy		3.870(3.610)	3.205(3.538)
Cognitive difficulty		−29.93***(8.088)	−29.59***(7.989)
Ability to afford premium		20.14***(5.907)	16.41***(5.852)
Needs variables
Anticipated dependence			28.37***(6.304)
Chronic condition			−5.995(6.417)
Self-reported health status			0.492(6.016)
constant	65.46***(12.73)	69.80***(17.97)	48.51**(18.87)
# of observations	788	616	616
Pseudo R2	0.0111	0.0237	0.0269

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.

Appendix 10. Regression results for interest in purchasing Plan 2 (excluding those dissatisfied with the products and those unfamiliar with LTCI)

Table

	Model 1	Model 2	Model 3
Predisposing variables
Female	−10.91***(3.760)	−12.68***(4.642)	−11.18**(4.614)
Age 50–59	−0.990(3.750)	4.892(4.325)	4.171(4.338)
Married	8.391(7.015)	7.162(8.152)	7.476(8.047)
Having children	−1.521(5.493)	0.906(6.479)	0.928(6.406)
University education	−3.401(4.392)	−5.952(5.397)	−5.398(5.357)
Expected living alone	17.35**(7.938)	13.28(9.012)	13.87(8.913)
Availability of family care	5.828(3.901)	3.644(4.640)	4.497(4.590)
Bequest motive	1.066(3.901)	−5.098(4.524)	−6.162(4.484)
Preference for formal care	9.617**(4.097)	9.126*(4.754)	8.953*(4.702)
Familial obligation	−7.713**(3.680)	−9.760**(4.270)	−9.475**(4.223)
Concerns about insurers	−5.767**(2.540)	−5.670**(2.844)	−4.706*(2.826)
Insurance ownership	6.694(4.229)	7.553(5.013)	7.436(4.963)
Enabling variables
Low income		−2.743(5.144)	−3.162(5.095)
High income		−0.966(6.122)	−0.693(6.060)
Paying with savings		−7.094(4.852)	−6.014(4.802)
Paying with welfare		−6.782(4.421)	−5.617(4.392)
Financial literacy		−5.029*(2.765)	−5.558**(2.753)
Cognitive difficulty		−8.403(6.101)	−9.270(6.111)
Ability to afford premium		5.277(4.442)	3.181(4.453)
Needs variables
Anticipated dependence			15.96***(4.769)
Chronic condition			4.107(4.870)
Self-reported health status			4.907(4.566)
constant	37.92***(9.288)	64.05***(13.31)	48.36***(13.97)
# of observations	723	558	558
Pseudo R2	0.0064	0.0108	0.0137

Note: Standard errors in parentheses; *p < .1, **p < .05, ***p < .01.