The demographics of loneliness among older people in Perth, Western Australia


Correspondence to: Dr Lyndall Steed, Curtin University of Technology. Email:


Objectives: To determine the prevalence and demographic correlates of loneliness in a sample of older people in Perth, Western Australia.

Methods: People aged 65 years and over living in private dwellings were recruited randomly, stratified by socioeconomic area, sex and 5-year age groups to 85 years. A total of 353 people with mean age of 77.5 years responded to a mailed questionnaire comprising demographic questions and three measures of loneliness.

Results: Severe loneliness was reported by 7.0% of the sample and feeling lonely sometimes by 31.5%. Higher levels of loneliness were reported by single participants, those who lived alone and those with worse self-rated health. The protective value against loneliness of social networks appears to be, in order of importance: friends, relatives, neighbours and children.

Conclusions: Although loneliness is not universally reported by older Perth residents, its prevalence is still considerable and worthy of attention from mental health practitioners and policy-makers.


It is widely believed that loneliness is universal and inevitable in older adults [1] and, given the significant relationship between loneliness and depression [2], poses a major problem for mental health. Studies frequently report that up to a third of older people experience some degree of loneliness in later life. For example, Wenger and Burholt reported a study of elderly Welsh people in which being very lonely and moderately lonely were indicated by 9% and 29% of the sample, respectively [3]. Similar statistics are reported for elderly people in Finland [4] and Great Britain [5]. These prevalence rates suggest that the majority of older adults are not, in fact, lonely, and thus it cannot be argued that loneliness is universal. However, after reviewing European and Canadian studies, Perlman concluded that ‘Loneliness occurs in all, or virtually all, cultures. Nonetheless, I suspect that loneliness is significantly influenced by cultural factors . . . that shape loneliness's prevalence, intensity, and antecedents; perhaps culture even shapes the very nature of the phenomenon itself. In this sense, loneliness is not universal; it is culture bound’[6, p. 186]. This being the case, and given the dearth of Australian statistics, it is considered timely to identify the prevalence and correlates of loneliness in the Australian context.

Definitions and measures of loneliness

Loneliness, while being variously conceptualised and measured, can be simply defined as the experience of negative feelings about missing relationships [1]. Proponents of its unidimensionality argue that loneliness is the same across situations and causes and can be captured by a single global measure. Based on this assumption, Russell [7] developed the UCLA Loneliness Scale that taps both the frequency and the intensity of salient aspects of being lonely. A unitary concept of loneliness also lends itself to measurement by a single direct question about people's frequency of feeling lonely [5,8]. We used both of these approaches to measurement.

Based on a multidimensional approach, de Jong Gierveld and Kamphuis [9] designed a scale tapping the intensity, time perspective and emotional characteristics of loneliness. After further refinement, the de Jong-Gierveld Loneliness Scale now consists of 11 items whose scores can be totalled to provide a single global measure of loneliness [10]. This measure was also included in the present study.

Demographic correlates of loneliness

Studies designed to determine the demographic correlates of loneliness are consensual on some findings and equivocal on others. For example, elderly people with a partner consistently report lower levels of loneliness than those without a partner – either through bereavement, divorce or remaining single [4,8,11]. In addition, while it is acknowledged that being alone does not inevitably lead to loneliness, living alone is a predictor of loneliness [4,8].

The relationship between age and loneliness appears to be dependent on whether cross-sectional or longitudinal data are analysed. Using both, Jylha [8] reported that loneliness increases with age. This relationship is frequently reported [3,4,12], but the protective properties of advanced age have also been reported [5]. Jylha suggested that age is confounded with widowhood and poor health, such that ‘loneliness increases with age, not because of age per se, but because of increasing disability and weakening social integration’[8, p. 166].

Possible confounding could also explain the discrepant results pertaining to gender and loneliness. While women are generally lonelier than men [e.g. 4], once marital status, age and living arrangement are controlled, the relationship between gender and loneliness disappears [13]. Gender difference also depends on the measurement used. For example, using the scale developed by de Jong Gierveld and Kamphuis [9], men appear to be less emotionally lonely than women, but more socially lonely [14]. Using the direct measure of loneliness, Jylha [8] found that women were more likely to report being lonely than men. However, when measures that avoid the word loneliness are used, gender differences are much less pronounced. For example, using the UCLA Loneliness Scale, in which the word ‘loneliness’ does not appear, no gender differences are found [2]. Victor et al. [15] provide a comparison of loneliness estimates for key demographic characteristics, based on self-report (direct measure) and a scale (aggregate measure) [9], and discuss possible reasons for differences.

Research consistently reports a negative relationship between loneliness and education [e.g. 4,5] and between income and loneliness [e.g. 4,16]. Similarly consistent are reports of the negative relationship between perceived health and loneliness [4,5,7,8,11,16].

Although size [14] and quality [17] of social network are known to be associated with loneliness, there is no consensus as to which aspects of the network are most salient. For example, Dykstra and de Jong Gierveld [14] reported that having weekly contact with children is protective of loneliness in women but not in men. Van Tilburg et al. [11] found that frequent contact with kin and friends is protective of social but not of emotional loneliness.

Data regarding the living arrangements of elderly people and the amount of time they spend alone are available in the Australian context [18], and social isolation has been identified as a priority concern [19]. However, there are little data available on the prevalence, risk and protective factors of loneliness in Australian samples. The aim of this study was to determine the prevalence of loneliness in an urban Western Australian sample and to identify the demographic factors that may place individuals at risk of, or provide protection against, loneliness.


Study sample

Based on socioeconomic status, representing high, middle and low, three local government areas in Perth were randomly selected. Within these strata, using data supplied from the Australian Electoral Commission, equal numbers of men and women who lived in private dwellings were sampled, in five 5-year age groups from 65 to 85 years. A total of 547 surveys were mailed, of which 353 were returned. Telephone follow up of non-responders revealed that a further 96 were unable to participate because of illness, absence or difficulty with English, yielding a response rate of 78.3%. The sample comprised 175 men (49.6%) with mean age of 77.5 years (SD = 7.2 years) and 167 women (47.3%) with mean age of 76.5 years (SD = 7.2). The remaining 11 respondents failed to state their gender. The demographic characteristics of the sample are shown in Table 1.

Table 1. Sample demographic characteristics (n = 353)
CharacteristicRelevant categoriesPercentage
Age (years)65–6919.7
Marital statusMarried/de facto64.8
Widowed, divorced, separated, never married35.2
EducationPrimary school16.5
Some secondary school28.5
Completed secondary school22.6
Trade, technical, apprenticeship17.9
University degree, diploma14.4
Socioeconomic statusLow29.5
Employment statusRetired81.0
Living arrangementLives alone31.4
Type of housingHouse75.8
Retirement village/other 5.4
Tenure in local areaLess than 4 years11.1
5–9 years 9.7
10–14 years 9.1
20 years or more58.6
TransportCar (they drive)65.6
Car (someone else drives)17.5
Taxi 6.0
Able to use public transportYes85.0
Manage on incomeRarely difficult49.4
Country of originAustralia58.2
English at homeYes95.3
English fluencyHigh48.8
Time aloneAlways 3.7

Ethics clearance for the study was obtained from Curtin University of Technology and all participants were treated in accordance with the national statement on ethical conduct in research involving humans [20].


The data reported here were collected as part of a larger study investigating social connections and health. Only the measures relevant to this paper are described.

Loneliness was measured in three ways, the first being a direct question in which respondents were asked how often they felt lonely on a four-point scale, namely always, often, sometimes or never. The experience of loneliness was explored with questions asking the days and times of greatest loneliness. In addition, respondents were asked if their experience of loneliness had increased over the past year and past 10 years.

The UCLA loneliness scale [7] is the most commonly used self-report measure of loneliness. The 20 items assess subjective feelings of loneliness and invite response on a four-point scale ranging from 1 (never) to 4 (always). Considered to be unidimensional, internal consistency has been reported as 0.89 in a sample of older people [7].

The De Jong Gierveld loneliness scale is an 11-item measure tapping both the social and the emotional aspects of loneliness. Respondents are asked to state the degree to which each statement applies to them on a five-point scale ranging from 0 (Yes!) to 4 (No!). Responses are dichotomised as described by the authors [10]. Internal consistencies of 0.8–0.9 have been reported [21].

Demographic details included marital status, age, level of education, whether they lived alone, had pets, type of housing, tenure in their current location, current employment, country of origin, main language spoken and fluency, most common means of transport, ability to use public transport, and whether able to manage on their present income. Respondents also recorded the number of children they had, how many children, relatives, friends and neighbours they felt close to, and whether they had a confidant. The amount of time spent alone and self-rated overall health (poor to excellent) were also recorded.

Statistical analysis

Descriptive statistics are presented, and group differences identified using t-tests, anovas and non-parametric techniques as appropriate. Assumptions underpinning these tests were examined prior to analysis.


Prevalence of loneliness

Only 1.5% of the participants rated themselves as ‘always’ lonely, 5.5% as ‘often’ lonely, and 31.5% as ‘sometimes’ lonely. The majority (61.5%) reported never feeling lonely. Of those who reported any degree of loneliness, most (44.0%) did not identify a specific time of the year when they felt most lonely, while others reported greater loneliness during holiday periods (18.2%), at the weekends (15.4%), during the week (14.0%) and at other times of the year (8.4%). Half (50.4%) of participants reported that there was no specific time when they felt lonelier, while 23.0% reported that evenings were the worst times, followed by nights (15.9%), afternoons (7.1%) and mornings (3.5%).

When asked to compare their loneliness with that experienced 1 year ago, 86.5% of respondents reported that the levels were about the same, while 21.1% were lonelier and 8.5% less lonely. However, responding with a 10-year perspective, 70.4% reported static levels of loneliness while 21.1% reported being lonelier and 8.5% being less lonely.

The mean score on the UCLA Loneliness Scale was 35.61 (SD = 10.11, potential range 20–80) while the mean score for the de Jong scale was 3.15 (SD = 3.14, potential range 0–11). According to the reported cut-off scores for the de Jong scale [10], 52.0% of respondents were not lonely, 39.3% were moderately lonely, 6.7% were severely lonely and 2.0% were very severely lonely.

Demographic differences in loneliness

Since the distribution of the scores on the direct measure of loneliness was so skewed, participants were scored as lonely (38.5%, including always, often and sometimes) or not lonely (61.5%). Relationships were explored using chi-squared and t-tests depending on the nature of the demographic. The results of these analyses are reported in Table 2. Each of the significant differences shown is discussed in the next section.

Table 2. Demographic differences in loneliness (n = 353)
 UCLADe JongDirect measure
Gendert = 0.77P = 0.43t = 1.71P = 0.09χ2 = 4.52P = 0.04
Ager = 0.04P = 0.47r = 0.02P = 0.67t = 1.40P = 0.16
EducationF = 0.70P = 0.60F = 1.32P = 0.26χ2 = 0.30P = 0.99
Marital statust = 3.68P = 0.00t = 3.48P = 0.00χ2 = 50.15P = 0.00
Live alonet = 4.20P = 0.00t = 3.79P = 0.00χ2 = 52.89P = 0.00
Number of childrenr = −0.03P = 0.59r = −0.12P = 0.05t = 0.58P = 0.56
Children close tor = −0.12P = 0.04r = −0.16P = 0.01t = 0.66P = 0.51
Relatives close tor = −0.29P = 0.00r = −0.28P = 0.00t = 2.21P = 0.03
Close friendsr = −0.39P = 0.00r = −0.34P = 0.00t = 2.42P = 0.02
Neighbour contactr = −0.23P = 0.00r = −0.21P = 0.00t = 1.54P = 0.12
Confidantt = 2.78P = 0.01t = 3.87P = 0.00χ2 = 3.92P = 0.05
Petst = 0.99P = 0.32t = 0.99P = 0.32χ2 = 1.63P = 0.20
Current employmentF = 0.44P = 0.82F = 0.56P = 0.72χ2 = 0.47P = 0.45
Socioeconomic statusF = 0.56P = 0.58F = 0.84P = 0.43χ2 = 0.75P = 0.69
Tenurer = −0.10P = 0.07r = −0.04P = 0.47t = 1.08P = 0.28
HousingF = 1.98P = 0.14F = 1.64P = 0.19χ2 = 3.12P = 0.32
TransportF = 2.72P = 0.04F = 3.11P = 0.03χ2 = 13.07P = 0.02
Can use public transportt = 1.41P = 0.15t = 2.11P = 0.04χ2 = 5.81P = 0.06
Income manageF = 4.22P = 0.02F = 2.08P = 0.13χ2 = 0.88P = 0.64
Australian bornt = 2.60P = 0.01t = 0.96P = 0.34χ2 = 0.10P = 0.76
Mainly Englisht = 0.77P = 0.44t = 0.75P = 0.46χ2 = 1.12P = 0.29
English fluencyF = 0.08P = 0.92F = 0.39P = 0.68χ2 = 0.58P = 0.75
Self-rated healthr = −0.26P = 0.00r = −0.27P = 0.00t = 4.49P = 0.00
AlonenessF = 30.78P = 0.00F = 22.6P = 0.00χ2 = 81.78P = 0.00


Prevalence of loneliness

Results for the direct loneliness question suggest that the prevalence of loneliness in elderly Perth residents is comparable with samples in other countries. Combining the ‘always’ (1.5%) and ‘often’ (5.5%) lonely categories, the resultant 7% is the same as the 7% experiencing ‘severe loneliness’ reported in Great Britain [5]. Similarly, the rates of ‘sometimes’ lonely were almost identical. Using the similar single question of ‘Do you suffer from loneliness’, 39% of elderly respondents from Finland reported that they suffered from loneliness in varying degrees [8], which is comparable to the 38.5% of respondents reporting any degree of loneliness in the present study. Using a measure comprising eight questions based on feelings or attitudes to levels of social contact, Wenger and Burholt reported a similar prevalence rate in elderly Welsh people [3].

As many as 41.8% of the present sample were born overseas. Further examination of the country of origin revealed that 18.5% were from the UK, 7.0% were from Italy and the remaining 16.3% were mainly from Asian countries. Given the large representation of British people in the sample, it could be argued that any comparisons with British studies are in some sense ‘compromised’. However, census data [22] indicate that in 2001, 47.6% of the Perth population in the age group of 65 years and over were born overseas, of which 21.1% were from the UK and a further 5.8% were from Italy. Therefore, our cultural mix of participants actually validates those comparisons, because our sample fairly accurately represents the multicultural nature of the population. It is also worth noting that, with only one exception, there is no significant difference in all three loneliness measures for each of the ‘ethnically related’ variables (Australian born, mainly English, English fluency) included in Table 2.

With regard to findings based on the UCLA Loneliness Scale, it could be argued that, overall, loneliness is not a problem for our older people. With a potential range of 20–80, a mean score of 50 could be assumed to reflect the average loneliness score. The mean score of 35.61 in the present sample is congruent with prevalence rates reported above in that less than half of the sample reported any degree of loneliness. However, a single sample t-test revealed that this mean is significantly greater (P = 0.002) than the 31.51 obtained from a comparable sample in the USA [7]. It is impossible to explain this difference from one study. If it were a true reflection of experienced loneliness then cultural differences in social networks would need to be considered. However, it is also possible that the higher scores may reflect a cultural difference in willingness to self-disclose levels of loneliness.

The mean score for the De Jong scale in the present study (3.15) compares variously with results from other studies. For example, in a large-scale study of 55- to 89-year-old adults in the Netherlands and Italy, mean scores were 2.09 and 3.24, respectively [23]. Another study of Dutch elderly men yielded scores between 2.2 and 2.8, depending on age [24]. Using recommended cut-off scores [10], 8.7% of the present sample were very severely or severely lonely, 39.3% moderately and 52.0% not lonely. In comparison to the norm data, the present sample had a greater degree of severe loneliness (by four percentage points) and a lower proportion of people not lonely (by 10%). Once again, without further exploration it is not possible to determine whether these are actual differences in loneliness or cultural differences in disclosure.

When reporting on specific times of the year or day when loneliness was maximal, it is clear that there is no consensus. While half did not report a specific time of day or year, for those who did, holiday periods were marginally more frequently cited as problematic. This is consistent with the literature that reports increased levels of depression in all age groups at holiday times [25]. Similarly, it is not surprising that evenings and nights are more problematic than day times, when social activities are more likely to be accessible.

Demographic differences in loneliness

No gender difference using the UCLA Loneliness Scale is consistently reported [2,7] and is supported in this study. Similarly, there is considerable agreement regarding gender differences when loneliness is measured directly, with women consistently scoring higher [4,8]. These conflicting results depending on the measure used have been explained in terms of men being more reluctant to directly admit to feelings of loneliness [24], partly because there is greater stigma attached to loneliness for men than women [6].

Three instances of gender differences were identified when using the de Jong scale to measure loneliness. These were: a stronger relationship between loneliness and ‘feel close to children’ for women than men; a stronger relationship between loneliness and ‘number of close friends’ for women than men; and a stronger relationship between loneliness and self-rated health for men than women.

The present data showed no evidence of a relationship between age and loneliness. To support this, when comparing levels of loneliness 1 year ago and 10 years ago, the vast majority reported no change. It is acknowledged that self-reports of change in loneliness across 1 year and 10 years are likely to lack reliability, but the two findings are congruent.

The significant relationship between loneliness and marital status is well documented [4,5,8,11] as is the relationship between loneliness and living alone. These findings were repeated in the present study with large differences on all measures of loneliness. There was also a strong relationship between amount of time spent alone and loneliness. While the literature repeatedly reports that being alone is not necessarily related to loneliness, it is clear that people who spend a lot of time alone are vulnerable to loneliness. The choice to spend a lot of time alone is very different from having solitude forced upon an individual by widowhood or other circumstances, but our measures were not sophisticated enough to allow for this.

When considering social networks, it appears that the order of importance, when protecting against loneliness, is: friends, relatives, neighbours and children. Having a confidant was also very important. It is interesting to note that it was not the number of children that was important, but the number of close relationships with children.

Because gender differences in these relationships had been noted in the literature, they were explored further. Such differences were found only when loneliness was measured by the de Jong scale. This result once again implicates measurement issues and, in particular, the notion of ‘public’ and ‘private’ experiences of loneliness [15]. The public experience is captured in the single item measure, whereas the private experience is quantified by the de Jong scale. Perhaps the private measure is the only one sensitive enough to enable the detection of gender differences in relationships between variables. Interestingly, the relationship between loneliness and ‘children close to’ was significant only for women. This importance of parenthood to women has been noted by Dutch older adults where it was found that ‘women who interacted with one or more of their children on at least a weekly basis were generally less socially lonely than those who did not’[14, p. 152]. The relationship between loneliness and ‘number of close friends’ was much stronger for women than men. This again emphasises the particular importance of close relationships for women.

Our results concur with the consensual finding that loneliness is negatively related to perceived health [4,5,7,8,11,15,24]. That is, people with poor perceived health tend to have higher loneliness scores. This relationship was found for all three measures of loneliness and, in the literature, has been found for both emotional and social loneliness [11,23]. From these analyses it is not possible to determine the direction of causality, but it is clear that self-rated health is an important component of quality of life for older people. Neither higher education nor ability to manage on current income was related to loneliness, whereas relationships with both of these variables have been reported in other studies [4,5,15,23].

In conclusion, although loneliness is not universally reported by older Perth residents, its prevalence is still considerable and worthy of attention from mental health practitioners and policy-makers, especially in regard to those people identified in this study as being particularly at risk.


We would like to acknowledge Healthway, the Western Australian Health Promotion Foundation, for funding the study from which the results presented in this paper were derived. We would also like to acknowledge the following colleagues from Curtin University of Technology, who were involved in the study: Thérèse Shaw, Peter Howat and Ros Morrow.

Key Points

  • • The estimated prevalence of severe loneliness among older people in Perth is 7.0%, with 31.5% feeling lonely sometimes.
  • • Single older people, those living alone and those with worse self-rated health, are the most likely to feel lonely, while social networks appear to be important in the order: friends, relatives, neighbours and children.
  • • Although loneliness is not universally reported by older Perth residents, its prevalence is still considerable and worthy of attention from mental health practitioners and policy-makers.