Study design

The study was based on a regularly conducted cross-sectional survey of the Swiss population called MOSAiCH (Measurement and Observation of Social Attitudes in Switzerland [CH]; [114]). MOSAiCH is connected to the current module of the International Social Survey Programme (ISSP), which has an international and a country-specific part. Participants were drawn from a probabilistic sample representing the Swiss population aged 18 and older.

Procedures

The results were based on a nationally representative sample of 3,349 permanent residents aged 18 and over who were living in private households in Switzerland. MOSAiCH was conducted by FORS, the Swiss Centre of Expertise in the Social Sciences based in Lausanne, Switzerland (www.forscenter.ch). Data from the survey is publicly available at www.swissubase.ch [115].

The survey consisted of two parts. The sampling of part 1, the ISSP part, took place between 26.02.2021 and 30.06.2021, while the follow-up survey sampling (part 2) started on 07.05.2021 and ended on 31.07.2021. The MOSAiCH sample was randomly drawn from the national population register (SRPH; German: Stichprobenrahmen für Personen- und Haushaltserhebungen) by the Swiss Federal Statistical Office. As of 31.12.2020, Switzerland had a total population of 7,114,731 permanent residents. MOSAiCH was designed to achieve a net sample of 1,000 participants for the follow-up interviews in part 2 of the survey.

Participants were contacted in writing or via email by FORS. All questions were implemented both online (Qualtrics-based) and on paper, and the survey could be conducted in German, French, or Italian. For a detailed description of the recruitment and sampling process, see Ernst Stähli and colleagues [114].

A total of 6,008 individuals were contacted for part 1, resulting in 5,991 eligible cases. A response rate of 55.7% (N = 3,349) was achieved. All the respondents who completed at least 50% of the web questionnaire or submitted a valid paper response (not blank) in part 1 were invited to part 2. Of the 3,259 participants who were contacted for the follow-up, 2,328 provided valid interviews (response rate of 38.7%; 69.5% of the respondents from part 1 also participated in part 2).

Measures

Data analyses

All analyses were based on the dataset retrieved from FORS [115]. The data of the online questionnaires was coded directly by the Qualtrics survey software, while the data of the paper questionnaires was manually entered and coded in office by FORS. Data checking and cleaning was also conducted by FORS; the detailed procedure can be found in the publication by Ernst Stähli and colleagues [114].

Theoretical models with outcome variables and predictors were prespecified according to hypotheses and findings from previous literature. For each variable, one questionnaire item was chosen to be included in the model. Since each construct, such as health behaviour, was measured by different items in the MOSAiCH questionnaire, we built a sum score out of all the items for each variable construct and chose the item for each construct that had the highest correlation with that score. Then we used lasso-regularised regression using the glmnet package [123] for each prespecified theoretical model to select the variables for the final models using an available case approach (n = 1,137). To select the optimal regularization parameter (λ), we employed 10-fold cross-validation via the cv.glmnet function. The best λ value was determined using lambda.min, which corresponds to the λ that minimizes the cross-validation error. Each model was run 10 times, and if an item was included in five or more of the runs, we included it in the final model. For these final models, we imputed missing data with the mice package [124]. In a next step, we used linear models to investigate the effect of the remaining factors within each model.

The following models were tested with data from all participants (N = 2,328) to investigate the differences in health status, health behaviour, healthcare use, psychological well-being, life satisfaction, loneliness, and human social support by pet ownership and regular animal contact status (question 1). The original models for health had health status (model 1.1), health behaviour (model 1.2), and healthcare use (model 1.3) as the dependent variables, while pet ownership, regular animal contact, loneliness, psychological well-being, and human social support were the predictors. These predictors were chosen based on literature showing that human health highly depends on perceived loneliness, human social support and psychological well-being [92,125]. All the models included gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type as the control variables before the lasso procedure. The socio-demographic variables were chosen based on previous literature identifying them as important factors for human health [1,2,4,7,79,92,93]. The original model with loneliness as the dependent variable (model 2) had pet ownership, regular animal contact, and human social support as the predictors, while gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type served as the control variables. The predictor was chosen based on literature showing that loneliness is often closely linked to human social support [126].

The original model for life satisfaction (model 3) had pet ownership, regular animal contact, loneliness, health status, health behaviour, healthcare use, and human social support as the predictors. Predictors were chosen because they can all be closely linked to each other, as recent studies show [127,128]. Gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type served as the control variables. The original model for psychological well-being (model 4) included pet ownership, regular animal contact, loneliness, and human social support as the predictors, while gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type served as the control variables. The choice of the predictors was based on literature [92,129,130]. The original model for human social support (model 5) included pet ownership and regular animal contact as the predictors, while gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type served as the control variables. All the final models with the respective items and the reduction process can be found in S1 Supporting information (Tables S1–S7 in S1 File).

To investigate if the attachment to animals, the socio-emotional support from animals, the amount of animal contact and animal species were related to health, loneliness, life satisfaction and psychological well-being (question 2), we ran models with all the participants who indicated that they owned pets or had regular animal contact (n = 1,105). The choice of predictors had the same basis as mentioned above. The original models had health status (model 6.1), health behaviour (model 6.2), and healthcare use (model 6.3) as the dependent variables, while loneliness, psychological well-being, human social support, pet ownership, regular animal contact, the number of animal species, the animal species, the amount of animal contact, attachment to the animal, and socio-emotional support from the animal were used as the predictors. The models included gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type as the control variables. The original model for loneliness (model 7) included human social support, pet ownership, regular animal contact, the number of animal species, the animal species, the amount of animal contact, attachment to the animal, and socio-emotional support from the animal as the predictors and gender, age, marital status, education, employment status, household income, partnership status, the number of people in one household, and settlement type as the control variables. The original model for life satisfaction (model 8) included loneliness, health status, health behaviour, healthcare use, human social support, pet ownership, regular animal contact, the number of animal species, the animal species, the amount of contact with animal, attachment to the animal and socio-emotional support from the animal as the predictors with gender, age, marital status, education, employment status, partnership status, the number of people in one household, household income and settlement type as the control variables. The original model for psychological well-being (model 9) included loneliness, human social support, pet ownership, regular animal contact, the number of animal species, the animal species, the amount of animal contact, attachment to the animal, and socio-emotional support from the animal as the predictors, while gender, age, marital status, education, employment status, partnership status, the number of people in one household, household income, and settlement type were the control variables.

To investigate if the attachment to the animal and the socio-emotional support from the animal were related to the animal species and the amount of animal contact (question 3), we ran two models with attachment to the animal (model 10.1) and socio-emotional support from the animal (model 10.2) as the dependent variables. Both models included the amount of animal contact, the number of animal species, the animal species, human social support, and whether the chosen animal lived in the household or outside as the predictors, while gender, age, marital status, education, employment status, partnership status, the number of people in one household, household income, and settlement type were the control variables. All the models with the respective items and the reduction process can be found in S1 Supporting information (Tables S8–S15 in S1 File).

For the predictor marital status, we combined the variables separated, divorced, and widowed into the single category “separated”. For the predictor settlement type, we combined countryside and village to the single category “rural” and combined city and town into the single category “urban” to ensure adequate sample sizes within each category.

In the pet module, we combined the items “currently living with pets” and “lived with pets in the last 6 months but not now” to calculate the total population with pets at the moment or within the last 6 months. We also combined small mammals and birds into one category, animals living in a terrarium or aquarium into one category, and equines and farm animals into one category to ensure adequate sample sizes within each category. We calculated a sum score for socio-emotional support from the animal (two items aggregated, higher values indicating more socio-emotional support) and a sum score for attachment to the animal (three items aggregated, higher values indicating stronger attachment).

We report our outcomes according to the Consolidated Standards of Reporting Trials (CONSORT), which suggest using the estimated coefficient with the confidence interval. The mean difference (coefficient) was used as the effect size, the confidence interval was defined at 95%, and the significance level was set at  .05. All analyses were conducted with R Statistical Software (version, 4.2.1; R Core Team, 2021) using the glmnet package to run all the models [123].

Descriptive statistics

From the total sample of 3,349 participants, 2,328 indicated whether they had pets or regular animal contact or not. Out of these, 1,105 participants indicated that they had pets or regular animal contact, while 1,223 participants did not have pets or regular animal contact. Table 1 shows the descriptive statistics of participants with and without animal contact. Table 2 shows the attachment to and the socio-emotional support from the closest animal, as well as the time spent with the different animal species, either for pets living in the household or for animals with whom the participants had regular contact outside the household.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics for participants with pets or animal contact and those with neither.

https://doi.org/10.1371/journal.pone.0352778.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Attachment to the animal, socio-emotional support from the animal, and amount of animal contact by animal species for pet ownership and regular animal contact.

https://doi.org/10.1371/journal.pone.0352778.t002

The descriptive statistics show that pet owners or people with regular animal contact were, in general, four years younger and were more likely to be female compared to non-pet owners and people without regular animal contact. Their marital status was comparable, while pet owners or people with regular animal contact had completed vocational training slightly more often but possessed a degree from a university or a university of applied sciences less often. Pet owners or people with regular animal contact were also more likely to be employed or to be in education, and were less likely to be retired compared to non-pet owners and people without regular animal contact. The presence of a partner and household income were comparable, while pet owners or people with regular animal contact were less likely to live in a household with one or two people and more likely to live in a household with three or more people. Moreover, pet owners or people with regular animal contact were less likely to live in a city, suburb, or town and more likely to live in a village or the countryside than non-pet owners and people without regular animal contact.

Differences in health status, health behaviour, healthcare use, psychological well-being, life satisfaction, loneliness, and human social support by pet ownership and regular animal contact status

Health status.

People with pets had a significantly lower self-reported health status compared to people without pets when we controlled for gender, age, marital status, employment status, household income, partnership status, the number of people in the household, and settlement type (see Table 3). The effect was, however, small. People with regular animal contact and people without regular animal contact did not differ in their health status.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Health status, health behaviour, and healthcare use by pet ownership and regular animal contact status.

https://doi.org/10.1371/journal.pone.0352778.t003

Loneliness.

Pet owners did not differ in their self-reported loneliness compared to non-pet owners when we controlled for age, education, employment status, the number of people living in the household, and settlement type (see Table 4). Moreover, people with and without regular animal contact did not differ in self-reported loneliness.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Loneliness, life satisfaction, psychological well-being, and human social support by pet ownership and regular animal contact status.

https://doi.org/10.1371/journal.pone.0352778.t004

Attachment to animals, socio-emotional support from animals, amount of animal contact and animal species related to health, loneliness, life satisfaction and psychological well-being

Health status.

Attachment to the animal and the amount of animal contact did not have a relationship with health status when we controlled for gender, age, marital status, employment status, household income, partnership status, the number of people in the household, and settlement type (see Table 5). No relationship was found for having small mammals or birds or farm animals or equines as pets, nor was one found for having regular contact with cats or with animals kept in a terrarium or aquarium. The number of animal species and socio-emotional support from the closest animal were not included in the model after the lasso selection.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Attachment to animals, socio-emotional support from animals, amount of animal contact and animal species related to health.

https://doi.org/10.1371/journal.pone.0352778.t005

Loneliness.

The amount of animal contact and having animals in a terrarium or aquarium were not associated with loneliness when we controlled for age, education, employment status, the number of people in the household, and settlement type (see Table 6). Higher socio-emotional support from the animal was tendentially associated with slightly higher loneliness. The number of animal species, other species, and attachment to the animal were not included after the lasso selection.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Attachment to animals, socio-emotional support from animals, amount of animal contact, and animal species related to loneliness, life satisfaction, and psychological well-being.

https://doi.org/10.1371/journal.pone.0352778.t006

Animal species and the amount of animal contact related to the attachment to the animal and the perceived socio-emotional support from the animal

Attachment to animal.

The attachment strength to the closest animal was significantly associated with the amount of animal contact: more time spent with the animal was associated with stronger attachment (see Table 7). Whether the animal lived in the household or not was significantly associated with the attachment to the animal: higher attachment was reported for animals living in the household. Women differed significantly from men, with women indicating higher attachment (women [M = 12.93, SD = 5.43], men [M = 10.73, SD = 5.77]). And the number of people in the household was significantly associated with attachment to the animal: more people in a household were associated with lower attachment to the animal. Moreover, individuals who had never been married reported a higher attachment to the animal, whereas people who were retired reported a lower attachment to the animal. Further, participants indicated significantly stronger attachment to the animal if they lived with a cat or a dog in the same household, as well as when they had regular contact with a dog. Among participants with regular animal contact, attachment to the animal was significantly lower for farm animals and for animals kept in a terrarium or aquarium (see Table 7).

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Animal species and the amount of animal contact related to the attachment to the animal and the perceived socio-emotional support from the animal.

https://doi.org/10.1371/journal.pone.0352778.t007

The absence of the “pet effect”

The absence of associations between pet ownership and healthcare use, loneliness, life satisfaction, psychological well-being, or the amount of human social support is in line with several studies that have also documented no association between pet ownership and human health and well-being [7,36,45,55–74,76,125]. At the same time, we could not confirm previous studies documenting better health among pet owners compared to non-pet owners [29–32]. Also, our results contrast with previous research reporting better well-being among pet owners on comparable self-reported measures, such as loneliness, psychological well-being, and life satisfaction [7,18,29,39,41,43–50,53–55,102,103]. This discrepancy between our findings and previous research reporting positive associations between pet ownership and health and well-being may partly reflect socio-demographic selection processes rather than universal effects of pet ownership itself. In contrast to many previous studies, our analyses were based on a representative sample, and we controlled for the socio-demographic factors of gender, age, marital status, education, employment status, household income, partnership status, the number of people in the household, and settlement type. Some of these variables have been found to explain differences between pet owners and non-pet owners [7,89–91]. In our sample, pet owners and individuals with regular animal contact were younger, more likely to be female, more likely to be employed or in education, and more likely to live in larger households and in rural areas compared to non-pet owners and individuals without regular animal contact. Pet owners and individuals with regular animal contact also reported lower levels of higher education and were less likely to be retired.

These sample differences align with previous population-based studies reporting that American pet owners were more likely to be younger, to be single females, to live in homes in more rural areas, and to belong to households where everyone is employed full-time [7,79]. Fraser and colleagues [79], for example, observed that pet owners were more likely to have a lower mean level of education. However, our results contrast with the finding that pet owners were more likely to be married [7,47] and partnered [79]. We assume that while pet and non-pet owners seem to differ in demographics generally, the specific characteristics might highly depend on local and socio-cultural factors, including culturally shaped roles and functions of animals, such as companionship, protection, working purposes, or urban lifestyle and aesthetic practices.

Interestingly, even when controlling for relevant socio-demographic variables, we found an association between pet ownership and lower health status, as well as slightly unhealthier behaviour. Although both effects were small, they are consistent with previous studies reporting negative or mixed associations between pet ownership and health and well-being outcomes [71,77,79–81,85]. At the same time, the observed difference in smoking behaviour was minimal, amounting to only 0.19 additional cigarettes per day on average. The effect on unhealthy behaviour must be interpreted cautiously: If this is calculated over the period of a year, this makes a difference of 70 cigarettes, but such a small difference may not reflect a meaningful lifestyle disparity. We hypothesize that this negative association could reflect a selection effect. Individuals with pre-existing health issues or higher stress levels, which are correlated with smoking [131], might be more likely to adopt a pet for comfort or companionship as a compensatory mechanism [89].

More broadly, our findings highlight the difficulty of disentangling the causal effects of pet ownership from pre-existing differences between pet owners and non-pet owners. It is possible that pet ownership leads to a lower health status and slightly more unhealthy behaviour. Another possibility is that unhealthier people are more likely to acquire pets as a compensatory mechanism and that the results can be explained by a selection effect, even though the literature largely argues that healthier people are more likely to acquire pets. Previous research has similarly demonstrated how strongly analytical approaches can influence observed “pet effects.” For example, Gmeiner and Gschwandtner [47] found that when comparing the raw data for pet carers and non-pet carers, the life satisfaction of pet carers appeared to be lower on average. However, when they used an instrumental variable approach, caring for a pet increased life satisfaction by 50% [47].

The role of the characteristics of the human-animal relationship

In addition to pet ownership, we also investigated regular animal contact and found that regular animal contact alone was not associated with health, loneliness, life satisfaction, psychological well-being, or the amount of human social support. Similarly, we also found that attachment to the closest animal was not associated with health, loneliness, life satisfaction, and psychological well-being. However, higher socio-emotional support from the animal was significantly associated with slightly higher loneliness and tendentially slightly associated with lower psychological well-being. The amount of animal contact was significantly associated with less healthy behaviour (a slightly higher number of cigarettes smoked per day) and tendentially associated with higher psychological well-being, while it was not associated with health status, healthcare use, loneliness, or life satisfaction. Our findings suggest that the functional quality of the relationship may be more relevant than ownership status or attachment strength alone.

Our results further indicate that the quality of human–animal relationships is shaped by interaction patterns and social context. The fact that attachment to the animal was not associated is surprising, as we hypothesised that the lack of a general association between pet ownership and health and well-being could be explained by the individual relationship characteristics, such as the strength of the attachment to the animal [63,95]. Previous research has shown that people with lower social support and higher loneliness, stress, anxiety, or depression indicated a stronger attachment to their pets [57,97], suggesting that pet ownership might be particularly beneficial for vulnerable individuals. At the same time, it has been suggested that a very strong attachment to the animal may be detrimental to health, while no attachment may yield no health benefits and medium attachment optimal benefits [95].

We were nevertheless able to show that the quality of the human–animal relationship depends on several relational and socio-demographic factors. Stronger attachment and higher socio-emotional support were associated with spending more time with the animal and living in the same household as the animal, suggesting that relationship closeness may depend on the frequency and intimacy of interaction. Moreover, women and individuals who had never been married reported stronger attachment, whereas larger household size and retirement status were associated with lower attachment to the animal. The finding that women reported higher relationship quality with their pets is consistent with previous research, showing that women tend to report higher levels of positive behaviour and attitudes toward animals [132] and form a stronger attachment to pets [133]. However, within-sex variation was much higher than between-sex variation [132], and an interaction between gender and age has been observed in a Chinese population [134]. In a population of Swiss pet owners aged 55 years and older, women reported a stronger attachment to their animals than men and more frequently reported that they love their pets, that they consider them as friends, and that they enjoy having them around [135].

Regarding the animal species, the species of the pet in the household was not associated with health, loneliness, life satisfaction, or psychological well-being, which contrasts with previous research, suggesting species-specific effects on mental health and cognitive functioning [38,84]. However, relationship quality appeared to differ across animal species. Participants mostly indicated a high relationship quality with dogs. In fact, a dog or a cat living in the same household, or a dog with whom participants had regular contact, was associated with stronger attachment. Weaker attachment was associated with a farm animal or an animal kept in a terrarium or aquarium that participants had regular contact with. Similarly, both a dog living in the household and a dog with whom people had regular contact were associated with more socio-emotional support from the animal. Small mammals or birds living in the household were likewise associated with higher socio-emotional support.

Conceptualizing underlying mechanisms

Our findings further suggest that potential associations between human–animal relationships and well-being may depend on species-specific interaction styles and the psychological mechanisms they engage. In particular, more regular contact with other species was significantly associated with more life satisfaction, and more regular contact with an animal kept in a terrarium or aquarium was associated with lower psychological well-being. The finding regarding animals kept in a terrarium or aquarium raises important questions about the mechanisms underlying species-specific human-animal interactions. From an attachment and social support perspective, the benefits of animal contact are often attributed to social interaction, emotional reciprocity, and perceived responsiveness [136], which are more characteristic of socially interactive mammals than of animals typically kept in terrariums or aquariums. Consequently, contact with less interactive species may not engage the same attachment-related or social buffering mechanisms associated with psychological well-being. This aligns with developmental findings showing that children’s empathy toward animals tends to decrease with increasing phylogenetic distance [137], suggesting that social closeness and perceived similarity may shape emotional responses to other species.

An alternative explanation for the observed association may reflect selection effects: individuals who prefer contact with non-interactive animals may differ systematically in personality traits, social needs, or lifestyle factors that themselves might relate to lower well-being. Whereas keeping dogs or cats may serve compensatory social or emotional functions, keeping terrarium or aquarium animals may primarily fulfil other motives (e.g., fascination with exotic species, aesthetic appreciation, or a preference for low-maintenance companionship) [138], which may not offer the same socio-emotional benefits. Nevertheless, a survey of reptile owners found that an emotional attachment toward reptiles was within the range of those observed in traditional pets [139]. Together, our findings suggest that the associations between human–animal relationships and health and well-being depend on complex interactions among characteristics of the owner, the animal and the human-animal relationship, such as time spent with the animal, animal species, perceived social responsiveness, relationship quality, and the owner’s emotional needs. This may help explain inconsistencies in the literature regarding the association between pet ownership and human health, and underscores the importance of conceptualizing pet ownership as a multidimensional and context-dependent phenomenon rather than a uniform construct.

Limitations, strengths, and future directions

Several limitations of this study should be considered. First, due to the cross-sectional design of this study, it was only possible to identify associations of pet ownership and regular animal contact with health and well-being measures. Causal conclusions could not and cannot be made. The observed associations may reflect causal effects, selection processes, or compensatory mechanisms, whereby individuals with specific psychological or social characteristics are more likely to acquire pets or seek animal contact. Longitudinal and causal modelling approaches are therefore needed to better disentangle these mechanisms and clarify whether pet ownership and animal contact influence health and well-being, or whether pre-existing health characteristics influence the likelihood of acquiring pets. Moreover, caution should be taken in generalizing these results, as our analyses were exploratory, with several tested hypotheses and no correction for alpha error. Our results should be replicated and extended with other samples.

Second, we used self-reported measures for health and well-being, and each outcome was only operationalised by one item. This could lead to different results compared to objective measures of health. We acknowledge the ongoing debate regarding the reliability and validity of single-item measures [140] for assessing complex psychological constructs, such as loneliness and well-being. While some studies suggest that single-item measures have comparable validity to multiple-item scales [141,142] and may be suitable for assessing well-being [143], the limitations of single-item measures must also be considered. It is therefore possible that our measurement approach was not sufficiently sensitive to detect important differences. The lasso procedure reduces overfitting by avoiding overly complex models. However, given the number of variables, there is no guarantee that overfitting will be eliminated and – as always with variable selection – coefficients should be interpreted with care. Different items operationalizing the same construct might explain the heterogeneity in the results of the existing literature on the effects of pet ownership on health and well-being. We used items from validated scales to measure attachment to the animal and socio-emotional support from the animal, but socio-emotional support was operationalised by only two items, and attachment to the animal by only three items. Also, the amount of animal contact was measured by only one item. As we have replicated the finding that demographics seem to be relevant in investigations of the effect of pet ownership and regular animal contact on health and well-being, it must be noted that the data came from a Swiss population and that these results are not generalizable to other communities.

Several limitations also relate to the operationalization of animal species. To achieve more balanced category distributions, we aggregated certain animal species (e.g., small mammals and birds, as well as farm animals and equines) into broader categories. However, interactions with horses, for example, can differ substantially from interactions with other farm animals, which may have contributed to the null findings for species effects, despite controlling for time spent with the animal and attachment to the animal. Moreover, some less common species were represented by relatively small subsamples, and we combined them to ensure a more even distribution among species categories, limiting the generalisability of species-specific findings. Future research should therefore investigate larger and more diverse samples that allow a more differentiated analysis of species effects and interaction styles. Clarifying whether regular contact with animals outside one’s household is comparable to living with a pet also warrants further investigation. At the same time, the present study has several important strengths. We used a representative sample and included animal-related questions within a survey that people did not expect to ask about pets. This reduced the potential bias of investigating an animal-friendly population. Comparable surveys have usually only asked about the ownership of cats or dogs and have neglected other species [30], despite research that has suggested that insects and birds might also affect perceived companionship, social support, and health [45,144]. Moreover, we extended the investigation to people who have regular contact with animals outside their household and accounted for the relationship quality with the animal. To our knowledge, this was also among the first studies to combine measures of relationship characteristics, including attachment and perceived socio-emotional support from the animal, with indicators of human health and well-being in a representative population sample. This approach provides deeper insights into how different factors of pet ownership and animal contact might be associated with human health.

More broadly, our findings suggest that research on human–animal relationships should move beyond treating pet ownership as a simple binary variable. The associations between animal contact and health and well-being are complex. Future research should therefore investigate how humans interact with animals, what functions these relationships fulfil, and under which conditions they may contribute to health and well-being. For example, it might be important to include the type of interaction with the animal, such as playing, caretaking, walking, or cuddling, and pet ownership history, the date of pet adoption, the age, health, and breed of the pet, and attitudes towards animals. Longitudinal studies are needed to clarify whether the associations found here represent causal effects or selection processes. Moreover, mixed-methods approaches might enhance methodological robustness. To reliably assess the effects of different predictors, such as the animal species, and to control for variables like demographics, large datasets are needed to better account for specific effects of less prominent species. Moreover, objective health measures validated measures for well-being, attachment to the animal, and socio-emotional support from the animal should be included. Investigating these mechanisms across different cultural and social contexts may help explain the inconsistent findings regarding the associations between pet ownership and human health and well-being, and may contribute to a more theoretically integrated understanding of human–animal relationships.

Conclusion

We found little evidence for a general association between pet ownership and human health and well-being. While pet owners reported a slightly lower health status and slightly more unhealthy behaviour, no differences emerged for healthcare use, loneliness, life satisfaction, psychological well-being, or human social support. Similarly, regular animal contact alone was not associated with health and well-being outcomes. However, relationship-related factors, including socio-emotional support from the animal and the amount of contact with the animal, were sometimes associated with human well-being. Moreover, the quality of the human–animal relationship depends on different factors. Stronger Attachment to the animal and socio-emotional support from the animal were higher in women, in people living with the animal, in people spending more time with the animal, and in people interacting with a dog.

Together, these findings suggest that associations between pet ownership and human health and well-being cannot be explained by pet ownership status alone, but depend on interactions between the characteristics of the owner, the animal and the human–animal relationship itself. Our study therefore highlights that pet ownership is a complex and heterogeneous construct that should be conceptualised multidimensionally to better understand its possible positive or negative associations with human health and well-being.