Sanna Pasanen & Arho Toikka
There is a growing amount of empirical evidence stating that urban green areas can provide health and wellbeing for people. Several processes might explain the health effect of urban green space, one possibility being the encouragement of physical activity. With binary logistic regression analyses, this study analyzes the relationship between urban green space and physical activity, measuring total physical activity and non-recreational physical activity (meaning walking and cycling for transportation purposes). The findings of the analyses performed provide evidence for the argument that urban green space can particularly support non-recreational physical activity, when certain conditions exist, most importantly sufficient population density in the area.
Introduction
Prior studies (e.g. Schipperijn et al. 2010; Maas et al. 2006; Ulrich 2006; Aspinall et al. 2013; Korpela et al. 2008) find that urban green space has a positive influence on health, but what remains unclear is the mechanism through which people living near urban green space get healthier. Because sufficient physical activity has an evident effect on better health (e.g. Lee et al. 2012; e.g. Sallis et al. 2011), urban green areas might support good health through promoting physical activity. The results of the Finnish Sport Survey (Kansallinen liikuntatutkimus 2009-2010) show that public urban space and particularly urban green areas are important environments of physical activity: some outdoor environments such as walking and cycling lanes, nature trails or jogging or skiing tracks were the most common place for physical exercise for 46% of respondents.
However, a contradictory relationship between urban green areas and physical activity has been reported in earlier studies, with some studies finding a positive relationship (Nielsen & Hansen 2007, de Vries et al. 2007), while others observe no relationship (Maas et al. 2008, Hillsdon et al. 2006). The majority of studies are reporting at least some significant relationships between types of recreation areas and exercise or proximity to recreation areas and exercise (Kaczynski & Henderson 2007).
Much of the earlier research has focused on accessibility, either measured as distance to green space or amount of green space in a living area. The results are mixed: some studies observe a positive relationship between access and visits (Nielsen & Hansen 2007) or access and exercise (Coutts et al. 2013), declining as the distance grows, while others (Hillsdon et al. 2006) find no significant relationship between access and activity. It is also unclear how the positive effect of green space on health works, as studies find that it is not due to actual visits (Nielsen & Hansen 2007) or physical exercise (Maas et al. 2008). Based on this, Nielsen & Hansen conclude that instead of formal visits to green areas, “the general character of the neighborhood could be affected by green infrastructure and thus be more conducive to outdoor activities and ‘healthy’ modes of travel in everyday life such as walking and bicycling” (Nielsen & Hansen 2007, 849).
To look into this general character, we operationalize it as a combination of urban green space and population density. Green space within three kilometers has been found to be associated with better health (de Vries et al. 2003), so we look at the “general character” of areas of roughly that size. We use planned or zoned green space as a proxy for usable green space and analyze the relationship between urban green space and sufficient physical activity while controlling for one potential general confounding character, population density. According to the results of Sealens et al. (2003), the residential density of a neighborhood seems to increase time spent in physical activity and lower obesity prevalence. However, measure misses many factors of area character, such as quality of green infrastructure.
This study explores whether urban green space might affect the total physical activity or the non-recreational physical activity of people (meaning walking or cycling for transportation purposes). As suggested by Nielsen & Hansen (2007) above, the effect of urban green space on physical activity of people might be related particularly to walking or cycling for transportation. So the analyses of this study will give further information concerning this difference between recreational and non-recreational physical activity related to urban green space.
The primary research questions of this study are: a) “Does urban green space have an effect on total amount of physical activity of people?” b) “Does urban green space have an effect on non-recreational physical activity of people?” c) “How do the findings of the analyses of this study concerning the effect of urban green space on physical activity of people change when controlling for population density, the quality of pedestrian and cycling facilities and respondents´ sex, age and the level of education?”
Methods and data
The main data set used in this study is part of the Tampere Health and Social Survey 2008 (Luoto et al. 2008), available for research purposes from the Finnish Social Science Data Archive. The data of the Tampere Health and Social Survey 2008 is a representative dataset with 3500 respondents and 180 variables. The sample of 3500 persons was randomly selected from the population register including people aged 15 and over and excluding those living in institutions. The response rate was 61%. The data is collected in the city of Tampere in Finland during January and February of 2008, using structured postal and internet questionnaires. Tampere is the third largest city in Finland with a population of approximately 215 000 people. Tampere is located in southern Finland, where there are still relatively mild weather conditions.
In addition to this survey data, a supplementary dataset was created to measure the exact amounts of urban green space and population densities based on archival data. The statistics concerning these environmental characteristics were provided by city officials of Tampere. The variables used for the modelling are described in Table 1, and the following sections describe each variable in detail and in relation to our theoretical framework.
To answer the research questions of this study, a part of the original dataset with 2079 respondents is used. From the original data with 3500 respondents, the respondents who lived outside central Tampere were excluded from the sample. The sampled persons who did not respond but were included in the original dataset were excluded. There are no weight variables in the data. The original variable with postal code information identifying the exact city district of the respondent was removed from the data when archived. However, this postal code information was re-categorized into service unit areas and this study uses this re-categorized variable. There are 14 service unit areas, based on older units used by the city of Tampere for service provision. In this study, in the case of the category City Centre, two service units were combined to make it comparable with statistics concerning the amount of urban green space. These service unit areas do not correspond exactly to experienced neighborhoods and do not cover finer details of the immediate area around residence or place of employment, but they do offer a reasonable amount of variety in the measures we are interested in.
In our analysis, we choose to focus on a binary outcome: whether the respondent engages in sufficient physical activity, as defined by health standards, or not, and hence use binary logistic regression. The attraction of this approach is that it allows the estimation of the likeliness of how the respondents distribute into the two non-overlapping categories of the response variable, depending on the explanatory variables. Analyses are controlled for the variables measuring environmental characteristics of the population density of the service unit area, the quality of walking and cycling lanes, and with variables measuring respondents’ age, gender and level of education. All analyses are conducted using PASW statistics 18.0 for Windows.
Total physical activity and Non-recreational physical activity
We use two different measures for physical activity, total physical activity and non-recreational physical activity. Leisure time vigorous physical activity done for the purpose of exercise has been the focus of much research, as it is easier to measure than habitual activities and work-related exercise (Gidlow et al. 2006). This is also the case in the spatial literature, where e.g. Toftager et al. (2011) use moderate and vigorous activity as the outcome of interest, while de Vries et al. (2007) use moderate exercise. There is much more variety in how low-intensity everyday physical activity is measured; including walking and cycling to get to a destination (Hentilä et al. 2015), sometimes limited to commuting to work (Borodulin et al. 2007); sometimes measured more broadly as hobbies (fishing, foraging) or chores (cleaning, yard work) that require physical activity (Saarela et al. 2015); or by letting the respondent define what they consider exercise during everyday activities (Hentilä et al. 2015).
The questionnaire of our data asked: In your leisure time, how often do you take physical exercise that lasts at least 30 minutes and makes you breathe more heavily and sweat at least a little? and On average, how many minutes a day do you walk or cycle to get from one place to another (e.g. to go to work or shopping)?. Thus, we have access to non-recreational physical mobility data, not limited to commuting, but not including nature- or home-based hobbies. This is obviously an imperfect measure of all everyday physical activity, but should be justifiable in the context of this study: activities inside the home should not be affected by green space, and the central Tampere area is not used for things such as foraging berries. Also, data collection took place in the winter, when people likely cycle and walk less, so although the respondents were asked to report averages, the measure potentially underestimates non-recreational exercise somewhat.
The categories of sufficient and insufficient total and non-recreational physical activity, which we use, are from the Finnish recommendations for healthy amount of physical activity of Fogelholm et al. (2007, 3). The effect of exercise on health can be non-linear: at the most active end, competitive sports do not necessarily further health and at the less active, the effects of non-strenuous exercise can easily be missed, if using a continuous measure of activity (Fogelholm et al. 2007). Although a more recent global recommendation for the sufficient amount of physical activity given by World Health Organization (WHO 2010) is available, the recommendations of Fogelholm et al. are used because in the questionnaire of our data, the amount of respondents´ physical activity was asked following these old recommendations.
Amount of urban green space
Earlier studies have often used distance-based measures for measuring the accessibility of green space, like distance to closest green area (e.g. Schipperijn et al. 2010) or percentage of green space within a given radius (e.g. de Vries et al. 2003, Maas et al. 2006). Even though these measures are objectively defined, perceived or experienced access to facilities is actually a better predictor of use (Scott et al. 2007, Schipperijn et al. 2010). This makes sense: distance ignores physical (e.g. a hill) and social features (e.g. commonly used routes) of the city. Incorporating measures of the urban structure would improve pure distance-based measures.
To account for this, as well as for reasons of data availability, we use external statistics provided by city officials of Tampere concerning the actual amounts of urban green space of a service unit area (Tampere: Kantakaupungin ympäristö- ja maisemaselvitys 2008, 76). These statistics are available for service unit areas that are roughly similarly sized as the objective distance categories found significant for green space access, less than 3km from postal code center point (de Vries et al. 2003). While the official categorization is by no means a perfect measure of actual experienced access, it should be more like it than pure distance. Still, it is possible that some important leisure areas fall right outside of the service unit area, and the results have some uncertainty based on that. The statistics concerning the amounts of urban green space contain only the green areas included in the governmental process of urban planning. Thus some green areas, such as private gardens or green roadside settings, are excluded from this analysis. This exclusion serves the aim of the analyses performed in this study well. When including only the green areas in the governmental planning process, it can be assumed that those green areas are mostly accessible to all citizens.
Figure 1. Green space in the 13 service unit areas, with the location of Tampere at bottom right
Survey respondents were re-categorized to four categories according to the amount of green space in their service unit area. The service unit area with least green space had 24% of green areas from a total land area, and the service unit area with the largest share of green space had 62% of green space. We categorize urban green space in four categories, from less than 30% to more than 50%. Using the four categories allows us to test for non-linear effects, e.g. is there an amount of green space that is sufficient to produce physical activity.
Population density
One of the main aims of this study is to investigate the influence of population density on the relation between urban green space and physical activity. In the area covered by the study, population density is linked to urban green space, but not uniformly so: there are areas of low and high density with low and high amounts of green space, allowing us to map these effects separately. The variable measuring the population density (people per km2) was created using external statistics provided by city officials of Tampere (Statistical Yearbook of the City of Tampere 2008-2009, p. 11-13). Population density is categorized into four categories, from low to highest density. The re-categorization of the variable population density was made to secure sufficient amount of respondents from all four categories. Service unit areas categorized as low density neighborhoods had 699 or less people per km2; medium density service unit areas had 1377-1854 people per km2; high density service unit areas had 2067-2286 people per km2 and the category highest density refers only to the city center of Tampere, which has 4985 people per km2. It should be noted that even though this study uses the categories of high or highest density, the population densities everywhere in Tampere are quite low, especially when compared internationally. The categories of population densities, which this study uses, are comparable to each other but not to the categories used in many international studies with cities that have much more inhabitants and truly high population densities.
Figure 2. Population densities in the service unit areas
Pedestrian and cycle lanes
A relationship between the area and physical activity could be due to green space offering better facilities for physical exercise, so we control for infrastructure. To elaborate the influence of the quality of walking and cycling lanes on the relation between urban green space and physical activity, this study uses a self-reported measure of quality of pedestrian and cycle lanes. For this variable, the respondents who report that their living area has very good or fairly good pedestrian and cycle lanes were re-categorized in the category “good”. The respondent who consider the pedestrian and cycle lanes of their living areas as fairly poor or poor, were re-categorized in the category “poor”. It should be noted that this measure uses a purely perceived living area definition and the living area understood by the respondent does not necessarily match the service unit areas used for archival data. The data was also collected during the winter, when conditions are more likely to be experienced as poor due to snow and ice. However, the difference between summer and winter should be fairly similar across areas.
Socio-demographic factors
We add socio-demographic factors to our model as controls. Green space affects the health of the elderly and the young more than other groups, as well as those with secondary education (Maas et al. 2006). Consequentially, the findings of this study will be elaborated with age (in four categories) and education (in two categories, lower level education and higher level education). Respondents with polytechnic or lower academic degree (B.A. or equivalent) and higher academic degree (M.A. or equivalent or higher) were categorized in the group of higher level education. Respondents with primary or lower secondary education, upper secondary education (general), upper secondary education (vocational) or college level vocational education (post-secondary) were categorized in the group of lower level education. We also control for gender.
Results
The first analysis examines how urban green space affects respondents´ total physical activity. Second analysis explores how urban green space affects respondents´ non-recreational physical activity. Binary logistic regression analyses are used to explore respondents´ likelihood of being more or less physically active when living in service unit areas with more green space, compared to the respondents living in service unit areas with less green space. In the latter analyses, to evaluate the reliability of the findings, the influence of population density, quality of walking and cycling facilities, and socio-demographic factors will be controlled for.
First, we analyze the effects on total physical activity. The results of the Model 1.1 (Table 2) indicate that when comparing the respondents who live in an area with more urban green space to those with less urban green space, the odds are greater concerning insufficient total physical activity (OR 1,22, 1,13, 1,04). The trend is most visible when comparing the respondents who live in a service unit area with from 30% to 40% of urban green space to those with less than 30% of urban green space (OR 1,22, P=0,16).
It is important to note that many of the results are not statistically significant, so there are patterns of behavior to be found from the sample of this study, but these patterns lack the predictability in the formal confidence level of 95%. Despite this weakness in the analysis, the observed relation will be investigated further by elaborating it with population density, walking and cycling facilities, age, gender and level of education. Firstly, the variable measuring the effect of population density of the service unit area is added in Model 1.2. Secondly, the rest of the control variables will be added in Model 1.3. In all regression model tables, we follow the guidelines of American Statistical Association for reporting p-values (Wasserstein & Lazar 2016) and report all coefficients, whether they are significant or not.
When adding more variables in the Model 1.3, the odds change very little. Model 1.3 indicates that respondents aged 61 or over seem to have a particular risk in belonging in the group with insufficient total physical activity.
In the second analysis (Table 3), we look at non-recreational activity. As the results of the following Model 2.1 indicate, there is a similar trend concerning the likelihood of insufficient non-recreational physical activity related to more urban green space as when analyzing the relation between urban green space and total physical activity. However, when taking into consideration only the amount of non-recreational physical activity, the effect of urban green space on insufficient physical activity is more obvious. Strong evidence was found especially when comparing the category of 30% to 40% (OR 1.43, P=0.01) of urban green space to the category of less than 30% of urban green space. This result means that the respondents living in an area which has from 30% to 40% of urban green space, have a clearly increased likelihood of insufficient physical activity when compared to the respondents living in an area with less than 30% of urban green space.
However, when controlling for the influence of population density (Model 2.2), the effect of urban green space on insufficient non-recreational physical activity turns more random or is even reversed to negative odds. After controlling for the influence of population density in Model 2.2, when comparing the category with 30% to 40% of urban green space to the category with less than 30% of urban green space, there is still greater odds for insufficient physical activity to be seen but it is no longer significant. When comparing the rest of the categories to the category with less than 30% of urban green space the odds turn to negative. Found trend is similar with the previously presented, which considered the effect of urban green space on the amount of total physical activity.
This means that when we account for the positive influence of population density on non-recreational physical activity of people, urban green space seems to increase the likelihood of sufficient non-recreational physical activity. Though, as in the earlier analysis, it is important to notice that although there are trends to be seen, many of the results have no statistical significance in the formal confidence level of 95% and the degree of explanation is low.
In Model 2.3, the relation between urban green space and non-recreational physical activity is elaborated further by controlling for the influence of the quality of pedestrian and cycling lanes, age, gender and the level of education. In Model 2.3 the effect of urban green space on non-recreational physical activity is a bit stronger than in the Model 2.2 when more variables are controlled for.
The findings from the last part of the analysis draws attention to the factors through which urban green space seems to support sufficient non-recreational physical activity. Strong evidence still remains concerning particularly the necessity of sufficient population density. The main result emerging from the analyses of this study is that urban green space seems to encourage non-recreational physical activity when certain conditions exists, the most important of which is sufficient population density. The influence of population density on increased non-recreational physical activity can be seen already when comparing the respondents living in medium density service unit area to respondents living in low density service unit area, though the influence is most clearly seen when comparing the respondents living in highest density service unit area to respondents from low density service unit area. According to the findings from the analyses performed in this study, even medium population density seems enough to increase the odds for sufficient non-recreational physical activity associated with more urban green space.
Model 2.3 also indicates that population groups who have greater odds of belonging in the group with sufficient non-recreational physical activity (when controlled with the variables used in this analysis) are people aged from 45 to 60, people with lower level education and women (not presented: OR 0,78, P<0,05). Men seem to have a particular risk of belonging in the group with insufficient non-recreational physical activity.
Discussion
The literature on connections between physical activity and urban green space has found discrepant results. This paper adds to that knowledge by providing more empirical evidence from observational data on the link between activity and urban structure. This link is not easy to research. Both activity and urban structure are multi-dimensional phenomena, and simplifying them for measurement necessarily misses some important factors. The combination is not easily manipulated for experimental research either. Spatial correlations and autocorrelations always structure real world data on spatial structure of urban areas, making ascribing causality complicated (Nielsen & Hansen 2007). Like any observational models, ours are simplifications. Thus, important limitations remain, but the analysis provides direction for future research as well as policy implications.
In this discussion, we compare our results to earlier literature and discuss how more refined data could improve the results further. We start with implications on measurement on modelling activity, move onto measuring and modelling urban structure, and finish with general observations.
Hillsdon et al. (2006) find no link between recreational activities and green space. Our results imply that urban structure might have the strongest effect on non-recreational, non-vigorous activity. Studies focusing on purposeful exercise (e.g. Toftager et al. 2011, de Vries et al. 2007) may miss these links between urban green space and non-vigorous activity. Measuring physical activity and particularly non-vigorous activity is complicated. Often in physical activity measurement, physical activity is characterized by frequency, intensity, duration and mode (Gidlow et al. 2006). There are difficulties to measure the activity, which is more lifestyle than sport. Our two measures are rough combinations of these items to manage them with single questions, using old guidelines for activity due to data limitations. We suggest that future research places an emphasis on measuring physical activity in a way relevant to urban structure, giving more impact to non-vigorous and non-recreational physical activity.
How to capture the important factors of urban structure is at least just as complicated as measuring non-recreational physical activity. Much of the literature we contribute to has simply used distance, ignoring other encouragements and restrictions of physical and urban structure. While labor-intensive data collection devices, like quality audits (used by e.g. Hillsdon et al. 2006), improve the usability measurement of the green space itself, it is possible that the “general character” of area that effects on physical activity and health (Nielsen & Hansen 2007) is not captured by these instruments. To reach towards better understanding of this character, future research should strive to combine objective distance to data on everyday practices. This implies a complex resource-intensive data collection, but one that should be feasible using modern mobile technologies.
Another difficulty with modelling urban green space and physical activity, is that the areas which have the largest share of urban green space are usually located near the borders of the city. Although the environment is greener, there usually are fewer other activity-supporting environmental factors identified by Sealens et al. (2003, 1552-1553), such as greater residential density, better land use mix-access (e.g. local shopping possibilities) or better street connectivity (e.g. short distances between neighborhood intersections). Particularly, the effect of population density on physical activity of people gets support from the analyses of this study.
We are not making causal claims from our models. It is not possible to conclude whether the urban structure makes people more physically active or whether the statistical relationship is spurious and derived from a self-selection process, which means that people who are already more active end up living in the neighborhoods giving better facilities to this existing lifestyle. However, most studies researching how the characteristics of living environments affect physical activity face this same methodological problem of self-selection, which is practically impossible to completely solve (Sallis et al. 2011, 39). To get over the problem of self-selection, we should study a group of respondents who have moved from a less green area to a greener one. If their physical activity would have been measured before they moved and after some period of time after they moved, the differences between the amounts of physical activity might give us a chance to evaluate causal relation. This kind of data seems quite hard to collect, at least in the extent needed for statistical analyses.
Furthermore, in the context of our study, we have a reasonably small amount of variability in the urban structure. For green space, the results are limited by access to fairly coarse measurements in one somewhat homogenous city. Tampere has a fair amount of green space and relatively low population densities everywhere. A more refined measurement of green space access in more heterogeneous areas would give more accurate and generalizable results.
Despite its weaknesses, the importance of this study is that its findings give suggestions for indicators that should be evaluated in similar future studies. Most importantly the effect of population density on the relation between urban green areas and physical activity. As we show, when controlling the influence of population density, the relation between urban green space and physical activity changes substantially. When the analyses are not controlled for population density, urban green space seems to decrease the amount of physical activity. However, when controlled for population density, urban green space appears to have an opposed effect, to increase the amount of physical activity. This result is particularly visible when taking into consideration non-recreational physical activity alone.
Based on earlier studies and analyses performed in this study, when there is sufficient population density, encouragement of non-recreational physical activity might be one process through which urban green space could support better health of people. We do not suggest that population density is the final causal factor, but that it is an enabling factor. There are other processes working behind this known relation between urban green areas and health, and more research is still needed to determine the processes through which green environment supports better health.
The effect of urban green space on health is an important issue for future research because in our time of increasing urbanization there is pressure to build more residences at the cost of green areas. We go along with the opinion of Nielsen and Hansen (2007, 849), that a green health perspective in urban planning could have an important role in promoting healthy activities and preventing obesity and diseases related to insufficient physical activity. If it could be scientifically clearly demonstrated that green areas support better health in many ways, there might be more political will to preserve or even build more green areas in our cities. As Gehl (2010) stated “First we shape the cities – then they shape us”.
Acknowledgements
This article is based on a master´s thesis “Urban Green Space and Physical Activity: Can urban green space advance health by encouraging physical activity?” done in co-operation with University of Helsinki. The thesis can be found from e-thesis archive of University of Helsinki and it provides more detailed categories, frequencies and distributions of all variables used.
References
Aspinall, Peter & Panagiotis, Mavros & Coyne, Richard & Roe, Jenny (2013). The urban brain: analysing outdoor physical activity with mobile EEG. British Journal of Sports Medicine, 49:4, 272-276.
Borodulin, Katja & Mäkinen, Tomi & Fogelholm, Mikael & Lahti-Koski, Marjaana, & Prättälä, Ritva. (2007). Trends and socioeconomic differences in overweight among physically active and inactive Finns in 1978–2002. Preventive medicine 45:2, 157–162.
Coutts, Christopher & Chapin, Timothy & Horner, Mark & Taylor, Crystal (2013). County-level effects of green space access on physical activity. Journal of Physical Activity and Health 10:2, 232-240.
Fogelholm Mikael & Paronen Olavi & Miettinen Mari (2007). Liikunta – hyvinvointipoliittinen mahdollisuus. Suomalaisen terveysliikunnan tila ja kehittyminen 2006. Sosiaali- ja terveysministeriön selvityksiä 2007:1. Helsinki: STM, OPM, UKK-instituutti 2007.
Gehl, Jan (2010). Cities for people. Island Press, Washington.
Gidlow, Christopher & Johnston, Lynne Halley & Crone, Diane & Ellis, Naomi, & James, David (2006). A systematic review of the relationship between socio-economic position and physical activity. Health Education Journal 65:4, 338–367.
Hentilä, Aki & Miettinen, Ismo & Kunttu, Kristina & Tammelin, Tuija & Venojärvi, Mika, & Korpelainen, Raija (2015). Muutokset korkeakouluopiskelijoiden kunto-ja hyötyliikunnassa vuosina 2000–2012. Liikunta & Tiede 52:1, 64–69.
Hillsdon, M. & Panter, J. & Foster, C. & Jones, A. (2006). The relationship between access and quality of urban green space with population physical activity. Public Health 120, 1127-1132.
Kansallinen liikuntatutkimus 2009-2010. Aikuisliikunta. SLU:n julkaisusarja 6/2010.
Kaczynski, Andrew T., & Henderson, Karla A. (2007). Environmental correlates of physical activity: a review of evidence about parks and recreation. Leisure Sciences 29:4, 315-354.
Korpela, Kalevi & Ylén, Matti & Tyrväinen, Liisa & Silvennoinen, Harri (2008). Determinants of restorative experiences in everyday favorite places. Health and Place 14, 636-652.
Lee, I-Min. & Shiroma, Eric J. & Lobelo, Felipe & Puska, Pekka & Blair, Steven N. & Katzmarzyk, Peter T. (2012). Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. British Dental Journal 213:7, 359-359.
Luoto, Riitta & Paronen, Olavi & Vuori, Mika (2008). Tampere Health and Social Survey 2008 [computer file]. FSD2550, version 1 (2010-11-09). Tampere: Innolink Research & UKK Institute [data collection], 2008. Tampere: City of Tampere & UKK Institute [producers], 2008. Tampere: Finnish Social Science Data Archive [distributor], 2010.
Maas, Jolanda & Verheij, Robert A & Spreeuwenberg, Peter & Groenewegen, Peter P. (2008). Physical activity as a possible mechanism behind the relationship between green space and health: A multilevel analysis. BMC Public Health 8:206.
Maas, Jolanda & Verheij, Robert A. & Groenewegen, Peter P. & de Vries, Sjerp & Spreeuwenberg, Peter (2006). Green space, urbanity, and health: how strong is the relation? Journal of Epidemiology & Community Health 60:7, 587-592.
Nielsen, Thomas Sick & Hansen, Karlsten Bruun (2007). Do green areas affect health? Results from a Danish survey on the use of green areas and health indicators. Health & Place 13, 839-850.
Saarela, Heidi & Mäki-Opas, Tomi & Silventoinen, Karri, & Borodulin, Katja (2015). Sosiodemografiset erot suomalaisten hyötyliikunnassa. Liikunta & Tiede 52:1, 55–63.
Sallis, James F. & Millstein, Rachel A. & Carlson, Jordan A. (2011). Community Design for Physical Activity. In: Dannenberg, Andrew L. & Frumkin, Howard & Jackson, Richard J. (edited) Making healthy places: designing and building for health, well-being, and sustainability. Island Press, Washington, D.C., 33-49.
Schipperijn, Jasper & Stigsdotter, Ulrika K. & Randrup, Thomas B. & Troelsen Jens (2010). Influences on the use of urban green space – A case study in Odense, Denmark. Urban Forestry & Urban Greening 9, 25-32.
Scott, Molly M. & Evenson, Kelly R. & Cohen, Deborah A., & Cox, Christine E. (2007). Comparing perceived and objectively measured access to recreational facilities as predictors of physical activity in adolescent girls. Journal of Urban Health 84:3, 346-360.
Sealens, Brian E. &. Sallis, James F & Black, Jennifer B. & Chen, Diana (2003). Neighborhood-Based Differences in Physical Activity: An Environment Scale Evaluation. American Journal of Public Health 93:9, 1552-1558.
Statistical Yearbook of the City of Tampere 2008-2009. 45th volyme. City of Tampere. Mayor´s office, Tampere.
Tampere: Kantakaupungin ympäristö- ja maisemaselvitys 2008. Tampereen kaupunki: Suunnittelupalvelut / Selvityksen ja arvioinnit 2008. Kaupunkiympäristön kehittäminen. Maankäytön suunnittelu. Julkaisuja 1. Tampere.
Toftager, Mette & Ekholm, Ola & Schipperijn, Jasper & Stigsdotter, Ulrika & Bentsen, Peter & Kamper-Jørgensen, Finn (2011). Distance to green space and physical activity: a Danish national representative survey. Journal of Physical Activity and Health 8:6, 741–749.
Ulrich, Roger S. (2006). Evidence-based health-care architecture. Lancet 368, 38-39.
de Vries, Sanne I. & Bakker, Ingrid & Van Mechelen, Willem, & Hopman-Rock, Marijke (2007). Determinants of activity-friendly neighborhoods for children: results from the SPACE study. American journal of health promotion 21:4_suppl, 312–316.
de Vries, Sjerp & Verheij, Robert A. & Groenewegen, Peter P. & Spreeuwenberg, Peter (2003). Natural environments – healthy environments? An exploratory analysis of the relationship between greenspace and health. Environment and Planning A 35:10, 1717-1731.
Wasserstein, Ronald L., & Lazar, Nicole A. (2016). The ASA’s Statement on p-Values: Context, Process, and Purpose. The American Statistician 70:2, 129–133.
WHO (2010) Global recommendation on physical activity for health. World Health Organization.
[1] Odds ratio represents relative probabilities, so that 1 means the two groups compared have the same probability for insufficient activity; smaller numbers mean a lower probability in the group being compared, bigger higher.