Investigation of dog ownership and physical activity on weekdays and weekends using longitudinal data from the SOEP Cohort

Study population

The German Socio-Economic Panel (SOEP) survey is the representative longitudinal dynamic cohort study of individuals living in private households in Germany³⁰. The samples were drawn either using a random route method or based on drawings from residents’ registration offices. The survey is organized by the German Institute for Economic Research (DIW Berlin) and is funded by the German Federal Government and the State of Berlin. SOEP started in 1984 and is conducted every year. Currently it covers approximately 30,000 people in 15,000 households. An advantage of the SOEP is the rich and diverse information it contains at individual and household levels on income, living conditions, employment, health status, household composition, education, social capital, and satisfaction^31,32. The dataset used in the present study is SOEP-CORE.v38. The SOEP study was approved by the Institutional Review Board of the SOEP. This study was a secondary analysis of anonymized data, and therefore required no ethics approval. Participants gave their informed consent prior to data collection. Detailed information on ethical clearance and informed consent given by the participants related to the SOEP can be found on the website of the German Institute for Economic Research (DIW), Berlin ( The study was performed in accordance with all relevant guidelines and regulations in relation to the use of SOEP data.

Definition of pet ownership

Participants were asked if they had any pet in 2011 and 2016. Those with current pet ownership were asked to indicate the pet species, i.e., dog, cat, rabbit, guinea pig/hamster/mouse, bird, fish, horse/pony, or other pets. These responses were used to classify dog ownership as “always”, “sometimes”, and “no ownership” over the 5-year period. Always ownership was defined as those who had dogs in both 2011 and 2016. Sometimes ownership was defined as those who had dogs either in 2011 or 2016 but not both, and no ownership was defined as those who did not have dogs in both 2011 and 2016.

Definition of physical activity

Participants were asked, every year from 2013 to 2018, on how many hours they spend on sports, fitness, and exercise on an average weekday. Moreover, they were also asked how many hours they spend on an average Saturday and Sunday, respectively, in 2013, 2015, and 2017. For all questions, participants responded by stating the number of hours spent as numerical values. In this study, physical activity during weekdays was assessed based on participants’ responses over the 6-year period (2013 to 2018) and physical activity on Saturday and Sunday was combined and evaluated as physical activity levels on weekends over the 5-year period (2013 to 2017).

Socio-demographic, physical, and psychological variables

Socio-demographic variables included age, sex, birth order, household members, house type (owner, main tenant, or sub-tenant), education (in school, primary education, lower secondary education, upper secondary education, post-secondary non-tertiary education, short-cycle tertiary education, bachelors or equivalent level, masters or equivalent level, or doctoral or equivalent level), annual post-government equivalised income, employment status (yes or no), and smoking (yes or no) and drinking habit (everyday, 4–6 times a week, 2–3 times a week, 2–4 times a month, less than once a month, or never). Physical variables were history of chronic disease (ever diagnosed a stroke, high blood pressure, diabetes, cancer, psychiatric problems or arthritis), disability, and Physical Component Score (PCS). Psychological variables were satisfaction with health (0 to 10 points), self-rated health (excellent, very good, good, fair, or poor), and Mental Component Score (MCS). PCS and MCS were calculated by the Short Form 12 (SOEP-SF-12) that includes 12 items or questions that assess functional health and well-being of an individual³³. Since not all covariates were collected in all survey years, information from 2012 was used for PCS and MCS, 2014 for smoking habits and 2016 for drinking habits, while all other variables refer to data collected in 2011. Further, not all covariates were included in all statistical models. Covariates were added in groups in a step-wise manner to build successively more comprehensive models. Covariates were selected based on a priori theoretical reasoning on how likely they would correlate with the ability (e.g., physical health, employment and income variables) or tendency (e.g., phycological health variables) to engage in physical activity.

Eligibility criteria

To be eligible for the study, individuals must have completed the questionnaire on dog ownership in both 2011 and 2016, and must have completed the physical activity measures at least once from 2013 to 2018. A total of 15,240 participants were included in this study. Observations with missing values were excluded from the analysis.

The data file of the SOEP is made available for this research by the German Institute for Economic Research (DIW) at doi: The use of anonymized SOEP data is subject to strict standards and only for research purposes. SOEP data are available free of charge for scientific use upon requesting a data distribution contract with DIW. This study did not require ethics approval as the analysis only used de-identified data in the form of unit record data from the SOEP Survey.

Statistical analysis

First, relationships between socio-demographic, physical, and psychological factors and the length of dog ownership were tested using chi-square tests or analysis of variance. Next, group-specific mean values of physical activity on weekdays and weekends were calculated by dog ownership status during the 5-year study period. The relationships between dog ownership status and physical activity during weekdays and weekends were modeled by generalized estimating equations (GEE). GEE models can account for the correlation of within-subject data. Follow-up years were included as covariates in the form of year dummy values. Separate GEE models were estimated for respondents in no ownership and sometimes ownership groups, and no ownership and always ownership groups, respectively. Covariates were selected by all authors based on a priori theoretical reasonings. Model-1 was adjusted for follow-up years, age, sex, household members, and equivalised income. In addition to the covariates in Model-1, Model-2 added house type, education, employment status, smoking habit, drinking habit, disability, satisfaction with health, self-rated health, stroke, high blood pressure, diabetes, cancer, psychiatric problems, arthritis, PCS, and MCS scores. Lastly, since dog ownership was not randomly assigned in the data, to identify causal relationship between dog ownership and physical activity, an instrumental variable approach with birth order as the instrument was estimated. Birth order is a potential instrument since it was likely to be correlated with dog ownership due to potential birth-order effects³⁴ on caring and responsibility. In addition, birth order should not affect physical activity directly. Covariates used in the instrumental variable model were identical to those in Model-2 (i.e., follow-up year, age, sex, household members, equivalised income, house type, education, employ status, smoking habit, drinking habit, disability, satisfaction with health, self-rated health, stroke, high blood pressure, diabetes, cancer, psychiatric problems, arthritis, PCS, and MCS scores). Statistical analyses were conducted using Stata SE (version 18; Stata Corp, College Station, TX, USA).