Attitudes towards free-roaming dogs and dog ownership practices in Bulgaria, Italy, and Ukraine

Lauren Margaret Smith; Rupert Quinnell; Alexandru Munteanu; Sabine Hartmann; Paolo Dalla Villa; Lisa Collins

doi:10.1371/journal.pone.0252368

. 2022 Mar 2;17(3):e0252368. doi: 10.1371/journal.pone.0252368

Attitudes towards free-roaming dogs and dog ownership practices in Bulgaria, Italy, and Ukraine

Lauren Margaret Smith ^1,^*, Rupert Quinnell ¹, Alexandru Munteanu ², Sabine Hartmann ², Paolo Dalla Villa ^3,⁴, Lisa Collins ^1,^*

Editor: I Anna S Olsson⁵

PMCID: PMC8890656 PMID: 35235582

Abstract

Free-roaming dog population management is conducted to mitigate risks to public health, livestock losses, wildlife conservation, and dog health and welfare. This study aimed to determine attitudes towards free-roaming dogs and their management and describe dog ownership practices in three European countries. We distributed an online questionnaire comprising questions relating to dog ownership practices and attitudes towards free-roaming dogs using social media. We used logistic regression and ordinal probit models to determine associations between demographic and other factors with ownership practices and attitudes towards free-roaming dogs. This study found that most surveyed respondents wanted to see a reduction in free-roaming dog numbers, and felt that this should be achieved through sheltering, catch-neuter-release, and by controlling owned dog breeding. We identified significant associations between both attitudes and ownership practices with gender, religious beliefs, age, education level, reason for dog ownership, previous experience with free-roaming dogs, and country of residence. Respondents who identified as: (i) male, (ii) holding religious beliefs, (iii) owning dogs for practical reasons, (iv) being young, or (v) having no schooling or primary education had a lower probability of neutering and a higher probability of allowing dogs to roam. Respondents who identified as: (i) female, (ii) feeling threatened by free-roaming dogs, (iii) older, or (iv) having more education had a higher probability of answering that increases in free-roaming dog numbers should be prevented. These findings can help to inform future dog population management interventions in these countries. We emphasise the importance of considering local attitudes and dog ownership practices in the development of effective dog population management approaches.

Introduction

Domestic dogs (Canis familiaris) are one of the most abundant species of canids in the world, with total population estimates between 700 million to 1 billion [1, 2]. Around 75% of this population are classified as “free-roaming”, indicating their ability to roam and reproduce freely [1]. Where free-roaming dogs exist in high densities, there are important implications for public health [3–5], livestock losses [6–10] and wildlife conservation [11–13], in addition to issues for the welfare of the dogs themselves [14–17]. The management methods applied to control population sizes involve culling, reproductive control (e.g. through catch-neuter-release; CNR), and the use of shelters to house unowned or unwanted dogs [18].

Human behaviour can shape the success of a population management programme. This includes actions of local communities, the teams involved in dog population management and the governments imposing management interventions. Indeed, the World Organisation for Animal Health (OIE) has identified that understanding public attitudes is important for developing effective dog population control [19]. In order for interventions to be successful, there must be public support for both the management method and its aims (e.g. reducing or stabilising the number of free-roaming dogs). Different communities may have different attitudes towards free-roaming dogs and management methods due to culture, religion, and the specific risks to humans, wildlife, livestock, and other companion animals in the area. Organisations involved in dog population management should consider these cultural, religious, and risk factors to ensure interventions are effective. For example, free-roaming dog populations can be an important part of a community, providing protection to people and livestock [17]. Where management methods aim to reduce free-roaming dog numbers, there may still be demand for dogs in a community. Unless this demand reduces, new dogs may be acquired to replace those removed by population management (either bought, adopted, or by uncontrolled immigration of free-roaming dogs). Prior to implementing management interventions, the level of acceptance of free-roaming dogs in the area should be gauged (i.e. determine whether the public prefer to have fewer free-roaming dogs in the community) so that those involved in dog population management can work towards a goal that benefits the community.

Dog ownership practices can also influence the success of population management. Encouraging responsible ownership practices is included as an objective in Chapter 7.7 of the OIE Terrestrial Animal Health Code on free-roaming dog population control [19]. The OIE describes the requirements of responsible ownership as: “When a person takes on the ownership of a dog, there should be an immediate acceptance of responsibility for the dog, and for any offspring it may produce, for the duration of its life or until a subsequent owner is found” [19]. Unrestricted owned dogs (free-roaming owned) and abandoned dogs are sources of free-roaming dog population increase [19]. Dog ownership practices that allow owned dogs to roam and do not prevent reproduction can hinder efforts to control free-roaming dog populations. Those involved in dog population management must determine the extent to which owned dogs contribute to the free-roaming dog population so that management interventions can be tailored appropriately (e.g. by enforcing responsible ownership through legislation and education programmes).

Questionnaire surveys are frequently used to gain insight into attitudes, opinions, behaviours, and the demographic and sociological factors associated with these. In terms of dog population management, different attitudes, opinions and behaviours about and towards dogs have been associated with responder gender [20–24], age [20, 22, 25], education [21, 22, 26], and previous life experiences (e.g. experience of keeping dogs in childhood) [27]. Questionnaires aiming to describe dog ownership, attitudes, and knowledge have been conducted in many countries around the world, but few published studies have been carried out in European countries.

This study determines attitudes towards the presence of free-roaming dogs and of dog ownership practices in three European countries–Bulgaria, Italy, and Ukraine. We selected these focal countries due to the networks established with collaborating organisations (VIER PFOTEN International and Istituto Zooprofilattico Sperimentale dell’Abruzzo e del Molise “Giuseppe Caporale”; IZSAM) that provided local knowledge to facilitate data collection. The focal countries are culturally and environmentally distinct, allowing comparison of the collected data between different countries within Europe.

In this study, we define attitudes as the thoughts, feelings, and opinions of respondents, as reported in the questionnaire. We define ownership practices as the actions taken to acquire, provide care, and relinquish ownership of dogs, as reported in the questionnaire. The objectives of this study are to: (i) determine attitudes towards the presence of free-roaming dogs; (ii) determine local ownership practices and attitudes, including whether owned dogs were free-roaming or neutered, the level of dog abandonment, and the reasons for dog abandonment; and (iii) investigate whether demographic and other factors (including age, gender, education level, religious beliefs, and previous experience with dogs) influence ownership practices and attitudes towards free-roaming dogs. This information can inform interventions so that education campaigns can target groups who are at-risk of irresponsible dog ownership behaviours [20, 26], as well as provide a baseline for evaluating the impact of interventions on human behaviour and attitudes [26].

Materials and methods

Study design

This was a cross-sectional study, with target populations of Bulgaria, Italy, and Ukraine. The study populations were residents who used social media. Participants were recruited through social media using an online questionnaire (Online Surveys [28]) that was open between the 8^th of March 2019 and the 21^st of December 2019, available in four languages: Bulgarian, Italian, Ukrainian, and Russian. The social media outlets used to distribute the questionnaire included Facebook [29] and Twitter [30]. Facebook advertising was used to increase the visibility of the questionnaire to the study population and increase the number of respondents. Facebook advertising targeted Facebook users who: (i) were recorded in their online profile as living in Bulgaria, Italy, or Ukraine; and (ii) were over the age of 18. The Facebook adverts invited respondents to provide their opinion on free-roaming dogs and dog ownership practices (see S1 File for English translation of adverts). All adverts were used in all countries to an equal extent, though it is not possible to know which advert respondents had seen. Minimum sample sizes were calculated for the three study areas, using Eq 1. A sample size of 385 respondents per study country was necessary to provide estimates with a 95% confidence interval that was within a 5% margin of error.

Eq 1. Sample size calculation

S a m p l e s i z e = \frac{\frac{z^{2} \times p (1 - p)}{e^{2}}}{1 + (\frac{z^{2} \times p (1 - p)}{e^{2} N})}

Where N = population size, e = margin of error, z = z-score, p = population proportion.

Ethical approval

Prior to completing the questionnaire, all participants gave consent within the online survey through a digital signature by selecting yes in a tick box to confirm that they (i) had read and understood the information sheet explaining the project (S2 File), (ii) understood that they had the opportunity to ask questions about the project, (iii) knew they could withdraw from completing the questionnaire, prior to submitting, (iv) agreed for their responses to be collected, stored, and analysed in an anonymised form for the purpose of reports and publication (see S2 File for information). Those who did not consent were not able to complete the questionnaire and were therefore not included in the study. No directly identifiable information was collected; all data obtained remains anonymous. Participants were able to withdraw from the questionnaire prior to completion, but as the data was collected anonymously, participants could not withdraw after the questionnaire was submitted. No minors were included in this study, as respondents who reported being under the age of 18 were not able to complete the online questionnaire. The study was approved by the University of Leeds Ethical Committee (reference BIOSCI 17–003).

Questionnaire design

The questionnaire was developed in English and translated into Bulgarian, Italian, Ukrainian, and Russian. The questionnaire comprised closed questions regarding the respondents’ attitudes towards free-roaming dogs and their management. Respondents that reported owning a dog were asked to also complete questions relating to dog ownership practices. Likert-type scales were used to estimate the level of agreement with specific questions. The questionnaire consisted of three sections: (i) socio-demographic information of the respondent (all respondents); (ii) ownership practices (only dog owners); and (iii) attitudes towards the presence of free-roaming dogs and the management of the free-roaming dog population (all respondents). A copy of the questionnaire in English can be found in S2 File.

Statistical analyses

All predictor and response variables are described in Table 1. Bernoulli logistic regression models were used to test associations between demographic parameters and respondent experience on the binary response variables: (i) Neutering status of owned dogs (binary response: neutered/not neutered); and (ii) respondents’ answers to the question “Do you think an increase in dogs on the street should be prevented?” (binary response: Yes/no). Ordinal probit models [31] were used to test associations between demographic parameters and respondent experience on: (i) Roaming status of owned dogs; (ii) I do not like the presence of free-roaming dogs around my home or work; and (iii) respondents’ answers to the question “Would you prefer to see: no free-roaming dogs, fewer free-roaming dogs, do not mind free-roaming dogs, more free-roaming dogs”. Ordinal variables are categorical variables with a natural order, for example Likert-type scales [32]. Ordinal variables are assumed to have an underlying continuous latent variable that cannot be measured directly (e.g. the attitude of a respondent). This underlying latent variable is therefore split into discrete options that can be measured (e.g. Strongly agree or Agree). The intervals between these discrete options may not be equal (i.e. not equidistant), an assumption required by metric models [33], and responses to ordinal questions may have non-normal distributions. Ordinal predictor variables can be problematic if analysed metrically, leading to Type I (false positive) and Type II (false negative) errors [31]. Ordinal models deal with issues in potential non-equidistant responses and non-normal distributions.

Table 1. Response and predictor variables (self-reported responses to questions) included in the statistical analyses and their levels.

Variables	Levels
Age	^* 18–24, 25–34, 35–44, 45–54, 55–64, 65–74 and 75 and above
Children in household	Children in household, no children in household
Country	Bulgaria, Italy, Ukraine
Dog ownership	Dog owner, non-dog owner
Education status	^* No education, primary, secondary, tertiary
Gender	Male, Female, NA (including option Other)
Neutering status of owned dogs	Neutered, not neutered
Owning a dog for practical reasons	Practical, not practical
Religious belief	Religious, non-religious
Feeling physically threatened by dogs on the street	^* Strongly disagree, disagree, neither agree nor disagree, agree, strongly agree
Been attacked by dogs on the street	Been attacked, not been attacked
Respondent or family members have been bitten by dogs on the street in last 12 months	Been bitten, not been bitten
Roaming status of owned dogs	Never, Sometimes, Always
I do not like the presence of free-roaming dogs around my home or work	Strongly disagree, disagree, neither agree nor disagree, agree, strongly agree
Should an increase in dogs on the street be prevented?	Yes, No
Would you prefer to see dogs on the street?	No free-roaming dogs, fewer free-roaming dogs, do not mind free-roaming dogs, more free-roaming dogs

Open in a new tab

* Ordinal predictor variables analysed as continuous variables in statistical models.

Dog ownership practices

Model 1 tested associations between demographic parameters and respondent experience with neutering of owned dogs using a Bayesian Bernoulli logistic regression model. The response variable was neutering status of owned dogs with fixed effects of gender, age, education status, religious belief, owning a dog for practical reasons and country (Table 1). Model 2 tested associations between demographic parameters and respondent experience with the roaming status of owned dogs using a Bayesian ordinal probit model. The response variable was roaming status of owned dogs and fixed effects were the same as for Model 1.

Attitudes towards free-roaming dogs

Model 3 tested associations between demographic parameters and respondent experience with agreement to the statement I do not like the presence of free-roaming dogs around my home or work using a Bayesian ordinal probit model. The response variable was I do not like the presence of free-roaming dogs around my home or work and fixed effects were dog ownership, gender, age, education status, children in household, feeling physically threatened by dogs on the street, been attacked by dogs on street, respondent or family members have been bitten by dogs on the street in last 12 months, and country.

Model 4 tested associations between demographic parameters and respondent experience with the question Do you think an increase in dogs on the street should be prevented? using a Bayesian Bernoulli logistic regression model. The response variable was should an increase in dogs on the street be prevented, with fixed effects the same as in Model 3.

Model 5 tested associations between demographic parameters and respondent experience with response to the question Would you prefer to see: no free-roaming dogs, fewer free-roaming dogs, do not mind free-roaming dogs, more free-roaming dogs using a Bayesian ordinal probit model. The response variable was Would you prefer to see dogs on the street, with fixed effects the same as in Model 3.

To fit the statistical models using a Bayesian analysis framework, the package “brms” version 2.12.0 [34] was used in R version 3.6.1 [35]. All models were run with four chains, each with 2000 iterations (1000 used for warmup and 1000 for sampling). Thinning was set to one. The total number of post-warmup samples was 4000. Where a response was missing (i.e. a respondent did not answer a question), the response was omitted from the statistical analysis (see S1 Table for number of no responses per variable).

Collinearity in the predictor variables was checked using the “vif” function in R package “car” [36] and values lower than three were considered not collinear. Model parameters were summarised by the mean and 95% credible intervals of the posterior distribution (CI; 95% most probable values). A significant association was determined if the 95% credible intervals of the posterior distribution did not contain zero on the log odds or probit scale. Probabilities were converted from the logit scale to the probability scale by $\frac{e x p (x)}{1 + (\exp (x))}$ , and are converted to odds using exp (x), where x is the posterior value on the logit scale.

Results

Descriptive analyses

Demographics

The numbers of respondents were 5,434 in Bulgaria, 3,468 in Italy, and 19,323 in Ukraine. All demographic information is provided in S2 Table. Respondents were from multiple regions within Bulgaria, Italy, and Ukraine (see S3 to S5 Tables). A broad range of ages between 18 and 64 were represented in all three study countries. Most respondents were female in all three study countries (87.5%, n = 4,754 of 5,434, in Bulgaria, 83.1%, n = 2,882 of 3,468 in Italy, and 87.1%, n = 16,832 of 19,323 in Ukraine). In Bulgaria 68.9% (n = 3,743 of 5,434), in Italy 42.0% (n = 1,457 of 3,468), and in Ukraine 67.3% (n = 13,011 of 19,323) of the respondents considered themselves to be religious. In Bulgaria 36% (n = 1,970 of 5,434), Italy 43% (n = 1,474 of 3,468) and Ukraine 57% (n = 10,928 of 19,323) of respondents lived in households with children.

Ownership practices

Sixty-five percent of respondents in Bulgaria (n = 3,528 of 5,434), 75% in Italy (n = 2,581 of 3,468) and 56% in Ukraine (n = 10,797 of 19,323) reported owning a dog. The main reason for dog ownership in all three study countries was for pleasure and company (Bulgaria 85.5%, n = 3,017 of 3528, Italy 87.7%, n = 2,263 of 2,581, Ukraine 70.7%, n = 7,631 of 10,797; see S6 Table for detailed responses on ownership practices). In Italy, a higher percentage of respondents acquired their dog from a dog shelter (38.1%, n = 986 of 2,581), compared to in Bulgaria (9.7%, n = 341 of 3,528) and Ukraine (9.9%, n = 1,070 of 10,797) (Fig 1). In Bulgaria and Ukraine, more respondents found their dog on the street (Bulgaria 35.5%, n = 1,252 of 3,528, and Ukraine 34.6%, n = 3,734 of 10,797) or received their dog from friends/family (Bulgaria 32.6%, n = 1,149 of 3,528 and Ukraine 27.9%, n = 3,015 of 10,797). More respondents in Italy answered that they prevent their dog from breeding through neutering (65.4%, n = 1,689 of 2,581), compared to 40.4% (n = 1,424 of 3,528) in Bulgaria and 35.4% (n = 3,828 of 10,797) in Ukraine. When asked the reason why respondents did not prevent breeding, 37.6% (n = 82 of 218) of respondents in Bulgaria, 34.6% (n = 75 of 217) in Italy, and 13.7% (n = 242 of 1770) in Ukraine answered: “A dog should reproduce at least once” (Fig 1). When respondents were asked if they allowed their dog to roam outside unsupervised, 59.0% (n = 2,082 of 3,528) in Bulgaria, 92.1% (n = 2,377 of 2,581) in Italy and 79.4% (n = 8,571 of 10,797) in Ukraine responded Never, and 29.5% (n = 1,039 of 3,528) in Bulgaria, 6.3% (n = 162 of 2,581) in Italy and 16.3% (n = 1764 of 10,797) in Ukraine responded Sometimes.

Fig 1 — The percentage of respondents who answered each of the answer options regarding (A) acquisition of dog, (B) reasons for not preventing breeding, (C) the outcome of the dog, and (D) reason for relinquishment. * Multi answer question: Percentage *of respondents who selected each answer option (i*.e. *100% would indicate that all respondents chose this option)*.

Most respondents in all study countries responded that they had never given up a dog (Bulgaria 98.5%, n = 3,474 of 3,528, Italy 92.4%, n = 2,386 of 2,581, and Ukraine 92.2%, n = 9950 of 10,797). Those respondents who had given up a dog mostly answered that this was because of an Animal behavioural problem (Bulgaria 27.3%, n = 9 of 33, Italy 36.5%, n = 66 of 179, and Ukraine 23.8%, n = 161 of 676), or Other reason (Bulgaria 39.4%, n = 13 of 33, Italy 57.5%, n = 104 of 179, and Ukraine 45.3%, n = 306 of 676) (Fig 1), such as family illness; a change in circumstances (e.g. birth of new child in home); moving home; owners going on a long trip away; the dog having puppies; or the dog not getting along with other dogs in the household.

Attitudes

In Bulgaria and Ukraine, high percentages of respondents had seen a free-roaming dog on the day they filled in the questionnaire (73.3%, n = 3,983 of 5,434, and 77.3%, n = 14,934 of 19,323 respectively), compared to only 15.4% (n = 534 of 3,468) of respondents in Italy (Fig 2; see S7 Table for detailed results). A higher percentage of respondents in Bulgaria (21.6%, n = 1,174 of 5,434) and Ukraine (26.5%, n = 5,129 of 19,323) had been attacked by dogs on the street ever in their lifetime, compared to few (4.2%, n = 147 of 3,468) in Italy. Higher percentages of respondents in Bulgaria answered that they provided care to free-roaming dogs by giving food (90.6%, n = 4,911 of 5,434), water (71.0%, n = 3,847 of 5,434), and shelter (34.8%, n = 1,886 of 5,434), compared to Italy (53.7%, n = 1831 food, 44.2%, n = 1508 water, 19.0%, n = 647 of 3468 shelter) and Ukraine (67.5%, n = 13,045 food, 29.6%, n = 5,721 water and 9.7%, n = 1,882 of 19,323 shelter) (S6 Table).

Fig 2 — The percentage of respondents who answered each of the answer options regarding (A) observation of free-roaming dogs and (B) agreement with statement “*I do not like free-roaming dogs being present around my home or work*”.

When respondents were asked their level of agreement with the statement “I do not like free-roaming dogs being present in the streets around my home or work”, responses were varied across the full range of options between strongly disagree and strongly agree in Bulgaria and Ukraine (varying between 14 and 25% for all answer options) (Fig 2). Most respondents in Italy strongly disagreed with this statement (35.8%, n = 1,242 of 3,468). In all three study countries, most respondents disagreed (Bulgaria 20.8%, n = 1,130 of 5,434, Italy 19.3%, n = 671 of 3,468, and Ukraine 25.3%, n = 4,890 of 19,323) and strongly disagreed (Bulgaria 42.2%, n = 2,295 of 5,434, Italy 56.4%, n = 1,956 of 3,468 and Ukraine 31.6%, n = 6,097 of 19,323) with the statement “I feel physically threatened by free-roaming dogs”.

Respondents answered most often that the municipality government and volunteer organisations should be responsible for managing the free-roaming dog population (Fig 3; S6 Table). Respondents most often answered that they would like to see no (Bulgaria 52.4%, n = 2,848 of 5,434, Italy 70.2%, n = 2,435 of 3,468, and Ukraine 45.2%, n = 8,740 of 19,323) and fewer (Bulgaria 32.8%, n = 1,780 of 5,434, Italy 24.3%, n = 841 of 3,468, and Ukraine 40.6%, n = 7,846 of 19,323) free-roaming dogs. Respondents who answered that they would like to see no or fewer free-roaming dogs answered that this should be achieved through sheltering, CNR, and controlling the breeding of owned dogs (Fig 3). Few answered that the free-roaming dog population should be reduced through culling (Bulgaria 1.7%, n = 92 of 4628, Italy 1.6%, n = 56 of 3276, and Ukraine 7.3%, n = 1,216 of 16586).

Fig 3 — The percentage of respondents who answered each of the answer options for: (A) who should be responsible for dog population management? and (B) how should free-roaming dogs be reduced? These were multi answer questions: Percentage of respondents who selected each answer option (i.e. 100% would indicate that all respondents chose this option).