Loneliness in later life, its prevalence and risk factors, has long been a focus of research. According to cognitive discrepancy theory, loneliness is defined as an unwanted discrepancy between desired and achieved levels of social contact [1].

In North America, Australasia and Western Europe research has consistently reported the prevalence of severe loneliness of approximately 10% for those aged 65 years and older with a further 30% classified as moderately lonely whilst countries in Central and Eastern Europe report prevalence rates of severe loneliness of between 15% and 20% [2]. Loneliness has been shown to be associated with a range of negative health outcomes and health behaviours (which vary between different age groups) [3, 4].

Previous research has identified a range of risk factors for the onset of loneliness which also vary somewhat across age groups. Predictors of loneliness in younger ages have been summarized by Mahon et al [5]. A much wider range of risk factors have been identified for older adults including gender [6], being widowed or divorced [7], reporting poorer self-rated health than expected [8, 9], sensory impairments [10], disability/impaired mobility [11], poverty and low material resources [6], time spent alone and household composition [8]. Living arrangements, social resources and social participation have been identified as potential mediators between health status and loneliness [12]. In terms of health outcomes loneliness has been linked with cardiovascular disease [13, 14], depression [15] and Alzheimer disease [16], and has been proposed as a mortality accelerator [17–19]. According to review published by Holt-Lundstad et al. [20] the mortality excess associated with weak social relationships is similar to other established risk factors such as low physical activity, smoking or drinking. Loneliness is, therefore, an important public health issue, and it is thus important to identify factors which can protect against or reduce vulnerability to loneliness as a means of developing appropriate interventions. Some of the established risk factors have been used in interventions to reduce loneliness. Since 2000 at least five reviews and meta-analyses to evaluate the evidence for the effectiveness of interventions to reduce loneliness have been published. Four focussed on older adults (the reviews by Cattan, et al. [21], Findlay [22], Choi et al. [23] and Hagan et al. [24]) whilst Masi et al. [25] included adults of all ages. These reviews showed that only a limited number of interventions demonstrated any significant impact upon levels of loneliness [21].

A number of interventions attempting to prevent or reduce loneliness have used ‘pet therapy’ based upon the attachment theory of Bowlby which emphasized the human need to be attached to somebody, to be close, to form and maintain relationships and the need for a sense of belonging [26]. Such interventions also build on work by Lazarrus and Folkman [27] indicating that human-pet attachment could provide a unique and affordable source of social support [28]. In the UK it is estimated that there are approximately 27 million pets and 45% of British households own a pet (Pet Food Manufacturers Association; http://www.pfma.org.uk/pet-population/; accessed December 2013). The UK ranks second in Europe for dog ownership, and third for cat ownership [29]. It is claimed that people who own pets do so to improve their subjective well-being, for company and to feel loved, depended upon and wanted [30].

The evidence base to support the use of such pet-based interventions is weak. There are a number of studies focusing on pet ownership or on animal assisted therapy (AAT) for older adults, for adults with serious mental health problems [31], cardiovascular events [28, 32, 33], or living in care-homes [34] which demonstrate positive outcomes [35–37]. Several studies have explored how, and to what extent, feeling of loneliness and social exclusion can be remediated (or prevented) by pet ownership [38–40] based on the premise that this reduces the impact of stressors in everyday live and consequently symptoms of depression or anxiety [41–44]. It is also hypothesised that pets may substitute for missing attachment figure(s). However, those who are highly attached to their pet report higher levels of loneliness compared to those who do not have such close relationship with their pet [45].

The evidence for the beneficial impact of pet ownership on loneliness is inconsistent as some studies show no impact of pet-ownership on health status or on mortality [46]. Furthermore most studies looking at this relationship are cross sectional in design. There is lack of evidence from longitudinal and intervention studies of using animals to reduce loneliness and social isolation among older people [47] although some effective interventions have been identified [48]. Longitudinal studies examining the relationship between pet-ownership or frequent contact with animals, and health, well-being or loneliness are rare. Raina et al. [49], focusing on the relationship between pet ownership and the physical and mental health of older people, reported that those who owned a pet were more active at the end of the study period than non-pet-owners, and that pet ownership significantly modified the relationship between social support and change in mental health. [49] Guest et al. reported that hearing dogs had a big impact on reducing loneliness among hearing-impaired owners but they did not use any control group [50].

Given that the evidence is mixed and extremely limited, the effectiveness of the presence of home pets on the prevention of loneliness and social isolation and improvement of subjective well-being has been questioned [51–53]. Furthermore rather than pet ownership mediating against loneliness, it has been proposed that the true nature of the relationship is, in fact, reversed i.e. pet ownership is a response to loneliness. However there are few longitudinal studies assessing the relationship between loneliness and pet ownership in both directions (pet ownership as a response to loneliness or pet ownership as a protection against loneliness). [54] The aim of this paper is to contribute to the evidence examining the relationship between pet ownership and loneliness. We examine if pet ownership is a response to, or protection against, loneliness using the English Longitudinal Study of Ageing (ELSA) by considering four questions:

Question 1: Can pet ownership protect against future loneliness?

Question 2: Is current pet ownership a response to previous feelings of loneliness?

Question 3: How do different pathways of loneliness influence current pet ownership?

Question 4: What role do socio-demographic characteristics, known to be connected with loneliness, play in the association between pet ownership and loneliness?

The answers to these questions will extend the evidence about the potential long-term effects of the presence of domestic animals in the lives of older people on the feelings of loneliness and vice versa.

Levels of reported loneliness in ELSA are approximately 19% and these are roughly stable over time. Women reported loneliness more frequently than men (a differential of around 7%), and this difference is also approximately stable over time and supports previous studies using the revised UCLA scale [61] but no other studies using other scales to measure loneliness which reports higher rates of loneliness for men [62]. Loneliness in ELSA is reported more frequently in comparison with European data from the SHARE study conducted across a range of European countries (more information on http://www.share-project.org/). The reported rates of loneliness in SHARE in European countries (measured by the same instrument as in ELSA study and in comparable years) are somewhat lower oscillating between 5% (in Denmark or Switzerland), 13% in France and 18% in Hungary [2] compared with 20% for ELSA. These differences in the prevalence of loneliness across Europe support the hypothesis that loneliness is culturally defined and is associated with expectations. For example the prevalence of loneliness in Greece is traditionally reported as one of the highest in the Europe but only about 5% of older people live alone and about 60% reported that they had close daily contacts with family members or friends [1, 63].

The goal of our study was to assess the relationship between loneliness and having a pet. In particular we wanted to explore the direction of the association and answer the question whether pet ownership is a protection against or response to loneliness? Having a pet was reported by 41% of respondents at baseline (Wave 0) and by nearly 30% in Wave 5 and more than one pet by 24% and 18% of respondents respectively. We do not know why pet ownership decreased but it seems plausible that this may reflect reduced income resulting from retirement; deteriorating health resulting in pet care being too demanding and the death of pets.

We demonstrate that those who reported that they had a pet at the beginning of the study were 1.2-1.4 more likely to report loneliness compared to those who did not. This overall association masks a significant gender effect being confined to women but not men. Looking at the relationship the other way around and focusing upon loneliness as a predictor of pet ownership we see a similar association of a similar magnitude. In pooled analysis those who reported loneliness were 1.2-1.5 times more likely to have a pet at follow up. Again the gender interaction was significant in all analyses demonstrating that the association between loneliness and pet ownership is confined to women. Our pathway analysis demonstrated that women who always reported being lonely and those who moved out of loneliness were more likely to report pet ownership than their non-lonely contemporaries (odds ratios of 2.4 and 1.8 respectively). These results mean, perhaps, that pet ownership can be a response to loneliness for the always lonely and a protection for those who recovered from loneliness.

Gender is the key factor in our analysis. Our results suggest that the association between pet ownership and loneliness is particularly strong in women. Age, social, demographic and health behaviours variables including established risk factors for loneliness (such as age or marital status) do not substantially affect the magnitude or direction of the association between loneliness and pet ownership.

There are, of course, some limitations to our study. Loss to follow-up of individuals between the waves of ELSA data collection might have introduced selection bias. Recent articles using ELSA data suggest that sample attrition is, for example, greater among those who were in a disadvantaged socioeconomic position at the start of the study but any bias due to attrition might be only small [64]. There was a relatively small subsample of participants, who were asked about pet ownership in Wave 0 of ELSA. We do not have information about how long participants had owned a pet or whether they looked after somebody else’s pet. We also do not have sufficient consecutive information to determine the relation between the initial reporting of loneliness, possible acquisition of a pet and the subsequent loneliness status to see the whole sequence of events to explore reverse causality in full detail. We could not adjust for seasonality and although loneliness is higher in spring and winter [65, 66] we do not think that this would significantly alter our results. Finally, as we performed relatively large number of hypothesis tests, we focused more on the magnitude of the effects when interpreting the results rather than just purely focusing on significance of findings.

It is commonly assumed that pet ownership ‘protects’ older people against loneliness. Our analysis has demonstrated that, for women, this may be a plausible hypothesis as it is associated with recovery from loneliness. However we have also demonstrated that for women who are always lonely pet ownership may be a response to their loneliness. Our results contribute to research on loneliness by demonstrating the complexity of the link between pet ownership and loneliness-it can be both a response to loneliness and a potential pathway out of loneliness. We also demonstrate that these relationships are only demonstrated by women and are not moderated by established loneliness risk factors or confounders.

These results suggest a number of areas for future research and have implications for policy and practice. There is considerable scope for qualitative research examining the issue of pet ownership in later life in more detail and how older people see this as a response or pathway out of loneliness. There is a clear need for such research to explore the important gender dimension identified in our analysis. Quantitative studies can demonstrate a link between gender, loneliness and pet ownership but we need to conduct qualitative research to explore the factors that account for these relationships. In addition these results caution us as to the appropriateness of pet based therapies and interventions against loneliness. We may speculate that, based on our findings, that such interventions may be more appropriate and acceptable to women than men. Again we need further research to explore the nature of the relationships between gender, loneliness and pet ownership in order to develop interventions that are appropriate, acceptable and effective.