The effect of forest-environment construction on the physical activity among Chinese adults

The Chinese government attaches great importance to the construction of the National Forest City (NFC), aiming at creating a pleasant living environment beneficial for people’s health and well-being. This study aimed to estimate the effect of the NFC policy on the frequency of physical activity (PA) for people in mainland China. Using the nationally representative data from the China Family Panel Studies in 2010 and its follow-up in 2018, 17 704 participants aged 18 and older were analyzed in our study. We employed the two-way fixed estimator in the difference in difference setting to analyze the PA of the adults in the cohort with different development levels of the forest-environment construction measured by the number of NFCs, and explore the relative differences before and after the construction of NFC. Subgroup analysis of age group, gender and living areas and a series of robustness tests were further conducted. The mean age of 17 704 participants in 2010 was (45.03 ± 13.67) years old, 51.74% of which were female and 48.26% were male. DID results showed that after the construction of NFC, the frequency of PA increased by 1.35 (β = 1.35, 95% CI:0.73–1.98) per week compared to those in provinces without NFC after adjusting covariates. And the improving effect of the forest-environment construction on PA frequency was intensified in middle-aged and older adults aged 45 and above (β = 2.03, 95% CI:1.09–2.96), males (β = 1.26, 95% CI:0.39–2.12) and females (β = 1.49, 95% CI:0.58–2.40) and those living in urban areas (β = 1.29, 95% CI:0.57–2.00). Our findings showed that the forest-environment construction under the NFC strategy has a positive role in the PA of older adults, indicating the health-promoting effects of the national green low-carbon environmental strategy. Considering the heterogeneity, it is essential to promote the benefits of green exercise for young and rural adults.


Introduction
Forests around the world are increasingly impacted by factors related to climate change [1,2], along with industrialization and lifestyle changes, quantitatively and qualitatively diminishing possibilities for human contact with nature and thus causing worse health outcomes [3,4].Recently, more efforts on forest management including the increase of urban green space have been made to adapt and mitigate climate change [5].Aimed at reducing vulnerability to increasing threats from climate change as well as improving the health and well-being of people, China has been adhering to multiple measures including promoting the construction of NFC since 2004 [6].By the end of 2022, the number of NFC in China had increased to 219 from 25 provinces [7], forming a cross-regional and urban-rural construction system.
A city can be awarded the title of NFC after passing the inspection of the national forestry authorities, for both its urban areas and rural areas indicating that (i) the city's urban forest coverage is more than 35% in southern areas and 25% in northern areas; (ii) along rivers and other water bodies as well as railways and other roads of the city have a greening rate of more than 80%; (iii) nature reserves and important forest and wetland ecosystems of the city have been properly protected; (iv) the city has constructed many residential recreational greenspaces with various parks for the public where most citizens can get to within 500 meters on average; (v) the rural green area of the city has increased year by year in the past five years.Only a city that met all the conditions according to the criteria of NFC evaluation [8] can be identified as an NFC.As a result, the construction of NFC has integrated the forest into cities, resulting in a pleasant living environment, multifunctional greenspaces and facilities accessible to people.
The extent to which population exposure to the forest environment may be causally related to beneficial health outcomes has become the focus of an emerging field during the past few decades [9,10].Especially in recent years, there has been a fall in the amount of PA for people because of the development of densely populated cities, rising car ownership, and the reductions in the availability of environments dedicated to PA [11][12][13].And inactivity has been identified as the fourth leading factor in mortality worldwide [14] which shows that an increase in PA levels among the population is a priority for public health [15].Many scholarly efforts have been made to identify individual and household factors that shape PA behaviors, including socio-economic status [16,17], demographic features [18,19] and genetic predisposition [20].Previous studies have also found a strongly positive linkage between environmental features such as the availability of greenspaces [21,22], the presence of recreational facilities [23][24][25][26] and PA behaviors.In general, the higher the daily forest-environment exposure, the higher the PA level and the better the overall health including general health [27,28], mental health [29,30], obesity [31,32], and mortality [33][34][35].Based on these studies, we hypothesized that the construction of NFCs had a promoting effect on the PA of adults through forestenvironment construction including increasing forest coverage and offering suitable greenspaces for different types of activities.
Despite the boom in research on the forest environment and PA over the last few decades, major gaps in knowledge remained.Most of the studies were cross-sectional or only focused on the environment of a particular city or region [25,[36][37][38][39].And research on the effects of forest-environment construction on people's health in China was still limited.NFC, as a national strategy aimed at adapting to the impacts of climate change on residents' health for more than ten years, its effect has not been well elucidated with reliable empirical evidence.In the present study, we aimed to explore the effects of the forestenvironment construction through the NFC policy on PA among adults with different sociodemographic features in mainland China with nationally representative data and robust research methods.The findings would provide specific information and shed light on future studies regarding the impact of national environmental policy on the health-related behaviors and outcomes of adults.

Data source and study population
Information on NFC was from the official website of the State Forestry and Grassland Administration Government [40].Public announcement information including the time and the province of the city for each year from 2011 to 2017 was collected and combined with the PA frequency and other individual information obtained from the China Family Panel Studies (CFPS).The CFPS was managed by the Institute of Social Science Survey (ISSS), Peking University and initiated in 2010 and then the investigated cohort was followed up at regular intervals every two years in 2012, 2014, 2016, 2018 and 2020.CFPS applied a multistage, stratified cluster sampling representing 95% of the population in mainland China covering the 25 provinces for demographic, socioeconomic, health, and behavioral factors.The investigators, mostly college students, were recruited by the ISSS [41].In our main study, considering the construction of NFC began in 2004 and continued till now, 1st (2010) and 5th (2018) CFPS can cover the time frame of NFC construction to the maximum extent and measure its impact more accurately.The reason for not choosing the data of the 6th (2020) CFPS is that the measurement questions and criteria for PA in the 6th CFPS are different from those in the previous five periods.Thus the 1st (2010) and 5th (2018) waves of the CFPS were used and those fully followed up were combined into a panel dataset.And we added the 3rd (2014) CFPS and analyzed the 3rd and 5th periods of CFPS again in the robustness test 2 to enhance the credibility of the results.The panel data covered the cohort interviewed before and after the construction of NFC in seven years with other key variables available.We included participants living in both urban and rural areas and aged 18 years or above in 2010 based on the research purpose.Finally, 17 704 adults were included in the analysis.Figure 1 illustrates the derivation of our analytical sample.It was verified that after excluding participants with missing values, the samples in our study can effectively represent the distributions of most of the characteristics of the population.

Outcome variables
The study's outcome focused on the frequency of PA measured by the weekly count of PAs in adults and considered as a continuous variable.Within the CFPS questionnaire, respondents were asked to specify the number of their PA engagement during the last week.We assumed the respondents' PA level from the previous week as their average, and accordingly derived the variable for PA frequency from this question.

Exposures
In our study, the degree of forest-environment construction under the strategy of NFC at the provincial level was used as a quasi-experimental study.The focus in our study lies primarily on any potential exposure to NFC at the macro policy level, rather than individual participation in specific NFC services and it is assumed that the effect radiation range of NFC including publicity, education and science guidance is more likely to be around the city or within the province given the higher cost of inter-provincial mobility.As a result, being identified as a NFC can be seen as an intervention and the exposure to NFC (living in the NFC environment) was measured by both exposure degree relative to the number of NFC by areas at the provincial level and exposure time relative to the NFC title-awarding time.Geographically, the participants in our study were interviewed in areas with different degrees of forest-environment construction measured by the number of NFCs.Adults from NFC were considered as the exposure group while others were considered as the reference group.And the number of NFCs also reflected the different degrees of forest environment construction in the provinces.Temporally, we constructed the time variable (whether) using the time the participants were investigated relative to the time when a city was awarded the title of NFC, with the exposure group after the time of title-awarding and the reference group before awarding or no city has been awarded the title yet.

Covariates
Due to our study design and the feature of panel data we used, two-way fixed effects would be included in the models we used.Thus, only time-varying variables will be the potential covariates in this study.
Based on the aforementioned evidence, we considered two types of time-varying explanatory covariates in our study.First, the demographic and socioeconomic variables consisted of age (treated as a continuous variable), living areas (urban or rural), marital status (in marriage or not in marriage), income including pension and employment income (treated as a continuous variable and logarithmic transformation was used due to the skewed distribution), education level (primary school and below, junior high school, senior high school and above) and employment (employed or not).Second, we considered the health conditions including having any doctor-diagnosed chronic disease (yes or no) and self-rated health status (very healthy, not very healthy).In addition, the province variable was also controlled.

Statistical analyses
From 2010 to 2018, 25 provinces added a different number of NFC cities each year depending on the NFC policy criteria.Such a data structure allows this study to treat the NFC as a 'quasi-natural experiment' and employ the two-way fixed estimator in the difference in difference (DID) setting to estimate the effect of the forest-environment construction on the PA frequency [42].The DID models focused on the relative difference in PA frequency before and after the construction of the forest environment as well as in areas with different degrees measured by the NFC number.And we also included two-way fixed effects to isolate the influence of time and individual heterogeneity.The models can be expressed as follows: Let Y ijk denote the PA frequency of individual i interviewed during period j in the area with forestenvironment construction degree k; β is the parameter of interest, which denotes the DID estimators of the effects of the construction degree on the PA frequency.Degree j is a binary variable that takes the value 1 if there was one or more NFC in the province and 0 otherwise (no NFC yet).And Time k is also a binary variable that takes the value 1 if periods of observations correspond to after title-awarding and 0 otherwise (before title-awarding or no city yet).X ijk denotes the covariates if any and ε ijk is the random error.Besides, time and individual fixed effects were also controlled.
Furthermore, we observed the potential heterogeneity of the effects across demographic characteristics in 2010 including age groups (adults aged 18-44 years or older adults aged 45 years and above), gender (male or female) and living areas (rural and urban) and applied the interaction variables to test whether the effects were evenly distributed.For all models, we first calculated the crude DID estimation coefficient (value) and its 95% confidence interval (CI) without any covariates, then involved confounding variables to obtain the adjusted estimators.A two-sided p-value of less than 0.05 was identified as statistically significant in this study.STATA V.17 (STATA, College Station, Texas, USA) was used for data analysis.

The characteristics of the participants
The final analytical sample included 17 704 with a mean age of 45.03 ± 13.67 years in 2010.51.74% of the participants were female and 48.26% were male.Other characteristics of the participants in the two survey waves were listed in table 1.
With the age increase during the 8 years, the proportion of adults better educated, employed, had a chronic disease, self-rated not very healthy and earned a lower income increased in 2018.From 2010 to 2018, the PA frequency of urban adults increased from 1.19 to 2.65 per week.The extent to which this increase was due to the forest-environment construction will be further explored in the following study.

The introduction of the NFCs
During the survey windows, 11 provinces established 1-3 NFCs, 13 provinces established more than 3 NFCs, and the remaining 6 provinces had not yet established NFCs.Table 2 shows that the differences of adults in demographic characteristics including marital status, educational level, employment and chronic disease in provinces with different numbers of NFCs in 2018 were statistically significant.Whether the difference among adults was caused by different degrees of forest-environment construction will be explored further.
Among the participants, 28.73% of adults were with a higher PA frequency of four times and above per week, 42.54% lower than that in 2018.According to the trend of the Alluvial diagram (figure 2), 23.07% of adults rose from lower PA to higher PA from 2010 to 2018, of which 77.99% come from provinces with NFC.Based on this trend of change, the effect of forest-environment construction in the process of PA change of adults deserves further exploration.

The effects of forest-environment construction on PA frequency
The DID estimations of the effects of forestenvironment construction on the PA frequency were shown in figure 3 and more details can be found in appendix table 1.In the overall sample of 17 704 participants, the forest-environment construction significantly increased the PA frequency of adults.After the construction of NFC, the PA frequency increased by 1.35 (β = 1.35, 95% CI:0.73-1.98)per week compared to those in provinces without NFC after adjusting covariates.
Then, we further investigated whether the effect of forest-environment construction on PA frequency varied across age groups, gender and living areas and found some noteworthy variations.For the adults, the improving effect of the forest-environment construction on PA frequency was intensified in middleaged and older adults aged 45 and above (β = 2.03, 95% CI:1.09-2.96),males (β = 1.26, 95% CI:0.39-2.12)and females (β = 1.49, 95% CI:0.58-2.40)and those living in urban areas (β = 1.29, 95% CI:0.57-2.00).The PA frequency of young adults aged 44 years and below also increased after forestenvironment construction but was not statistically significant.

Robustness test
We conducted several sensitivity analyses to check the robustness of the estimations and details were shown a Note: 1. Analyzed by X 2 test for categorical variables and ANOVA for continuous variables. 2. Considering that the indicator of self-rated health is more subjective and comprehensive [43,44], with the increase of sample's age, the assessment of their own health can be worse than actual health status changes such as chronic disease, shown as data in 2018 that does not appear to match (79.58% no chronic but 76.24% self-rated not very healthy).on PA frequency (β = 1.83, 95% CI:1.12-2.54).And in order to enhance the credibility of the results, we used the third and fifth periods of CFPS to analyze again and the result was still positive (β = 0.35, 95% CI:0.12-0.57).Besides, considering that the relationship between continuous independent variables such as age and income and dependent variables in the model may be non-linear, we used the multivariable fractional polynomial model and got consistent results (β = 1.20, 95% CI: 0.58-1.82).Next, we used the PA frequency as a nominal variable instead of a continuous variable to conduct the analysis again.Given that we are focusing on the increase in PA, the amount of PA from highest to lowest was ranked and the top quartile of PA frequency was chosen as the basis for transforming a continuous variable into a categorical one.If PA frequency ⩾4 per week, the value is 1; Otherwise is 0. And the results were basically consistent (β = 3.60, 95% CI: 1.79-7.21)after replacing PA measuring method.
In addition, considering that the forestenvironment construction of a city may be affected by macroeconomic conditions, we conducted a hierarchical linear model (HLM) using the GDP in 2018 of each province [45].And meteorological influencing factors including temperature, rainfall and PM2.5 were also controlled by HLM using the annual mean temperature, the number of rainy days and the number of haze days in 2018 [46].And the results still showed consistency (appendix table 3).
Lastly, considering the long-term period from 2010 to 2018 and the first NFC-exposure time varies in different provinces, we conducted DID with multiple periods estimator using the first time identifying a city as NFC in each province, which could help us rule out the bias caused by different NFCexposure time of exposure group (appendix table 4) [47].The results showed that regardless of the first exposure time in 2011, 2012, 2013, 2014 and 2016, NFC had significant positive effects on the PA of the participants.

Discussion
Using the national, large-scale, follow-up survey data covering most parts of China, this study provides novel insights into the impact of the forestenvironment construction on the PA frequency of adults in China for the first time.The longitudinal data DID models and several robustness checks allow us to control other confounding variables and get a clean robust effect.Overall, we found that the construction of a forest environment had a positive effect on the PA of adults, which is consistent with the previous studies that proved the importance of natural environments and greenspaces to PA [25,39,48,49].The results showed that although urbanization and land use change are irrevocably altering the environment in which many adults live, forest-environment construction greatly reduced the distance between people and the natural environment and played a positive role in national fitness.
There may be many possible reasons for which forest-environment construction improved the PA frequency for adults.Subjectively, there is evidence to suggest that the natural environment might engage people in PA by facilitating the enjoyment of exercise and experiencing nature [32,50].For those people engaging in green exercise, the nature element not only increases enjoyment and escapism from everyday life but also helps achieve greater intensity and maintain adherence to PA without the perception of effort changing [36,51].For instance, it was found that walking outdoors, compared with walking indoors on a treadmill, was associated with more positive emotions [52], as well as a greater and firmer intention to engage in PA in the future [53].Objectively, compared with gyms or other indoor fitness settings, NFC provides forest environments and greenspaces with PA facilities easily accessible and well maintained where people can engage free of charge and at times that better suit their daily schedules [51,54].The safe, accessible and multifunctional ecological settings may also attract people outdoors to entail different forms of PA, such as ball games, riding bikes and square dancing.Taken all together, NFC could increase both the intention and opportunity for people to be physically active, helping to improve the PA behaviors of adults.Furthermore, it is worth noting that considering that the living environment has an increasing impact on people's migration decision and inter-provincial migration is relatively small compared with intra-provincial mobility under the administrative division system in China, the impact of NFC construction can even radiate beyond a city to a province by ways such as attracting more residents in the province and overall regional environmental development.
In addition, PA was found to be particularly improved in middle-aged and older adults, male, female and urban adults in our study.In terms of age, results in our study showed that forest-environment construction was beneficial to increase the participation of middle-aged and older adults, which was consistent with previous studies indicating that green exercise was especially popular among older adults because these individuals pay more attention to maintaining or enhancing their health profile through PA, perceive experiencing nature as a more important motive for PA and make less use of alternative indoor exercise settings such as gyms than young adults [32,55].Besides, our study reached the conclusion females' PA frequency was increased more impacted by NFC.Women are more concerned about the environment than men has become an increasingly widespread conclusion in the field of environmental sociology [56,57].That is to say, compared with men, women show more concern for the environment to a certain extent, which was also reflected in women's friendly behaviors towards the environment.Indeed although some previous studies have attributed the fact that females were less likely to take outdoor PA to the perception of safety [48,58], it can be concluded from our results that with the improvement of the environment brought by the forestenvironment construction, females are more likely to perceive the changes in their living environment and thus adopt more environmentally-friendly behaviors [32,59].In terms of living areas, it is not the lack of forest environment or greenspaces that limits rural residents to PA.Therefore, the NFC strategy pays more attention to the improvement of recreational greenspaces and environmental greening for urban areas while for rural areas, whose environment is less affected by urbanization deterioration, economic development through ecological industry is one of the important goals [8].This could explain why rural residents' PA was less affected by forest-environment construction.
The importance of natural environments to PA and the health of adults has been long advocated [9,48,49], especially when sedentary and inactive lifestyles have become major contributors to the disease in recent years [60].There is even evidence suggesting that PA performed in greenspaces produces greater psychological and physiological benefits than PA done in other settings [61][62][63].For more than ten years, the strategy of NFC has improved the living environment in China and is trying to translate the achievements of environmental construction into the ecological well-being of people.Based on our results, this strategy has played a positive role in inducing behavioral changes and encouraging active lifestyles among adults.However, in order to realize the ecological benefits of forest-environment construction shared by all, it is essential to promote Forest City publicity and the benefits of green exercise for young and rural adults, attracting these individuals to make more use of greenspaces for PA.Based on the NFC Development Plan (2018-2025) [64], more adults can benefit from the NFC strategy for wellbeing and health level by improving the quality of the living environment of Forest City, giving full play to the Forest City ecological functions, and realizing the basic equalization of Forest City ecological services [65].

Limitations
The study is subject to some limitations.Firstly, considering the long interval between the two investigations, it may be not adequate to take the NFCs' number of the seven years as the only measure of the construction degree as the NFCs of each province accumulated year by year between the two surveys.The nexus of NFCs and PA could be more accurate when city-level or microscopic databases such as green space proximity indicators are available in the future.Besides, it is not enough to tell whether the increase in PA is due to the accessibility of greenspaces and recreational facilities or environmental improvements more suitable for exercise such as air quality.However, the results of the robustness test showed that after taking into account the difference in air pollution of each province, forest city construction still has a significant positive effect.Additionally, although we have controlled the macro-factors of GDP at the provincial level, there may still be other potential macro-influencing factors.Furthermore, only the frequency of PA was analyzed in our study but not the levels or the duration because of the data limitation.Thus, further research is necessary when data is updated in the future.Despite the aforementioned limitations, this study contributed to the literature pool by examining the impact of the construction of forest-environment construction on adults' PA frequency on a national level as appropriately and accurately as possible.

Conclusions
Our findings provide new evidence of the change in PA of Chinese adults impacted by forest-environment construction under the NFC strategy.With the advancement of the NFC Strategy, the PA frequency of adults increased significantly, suggesting the health-promoting effects of the national environmental strategy.Considering the heterogeneity, more attention needs to be paid to young, male and rural adults and realizing the equalization of Forest City ecological facilities and services.These findings will help to bring cognizance to the effects of forestenvironment construction at the national level on PA and promote the simultaneous achievement of a green, low-carbon lifestyle and the high-quality development of the population.

Note:
a The DID estimation (β), 95%CI (lower to upper) were presented in the table.b Heterogeneity of the effects across age groups was based on adults' age in 2010.c In order to avoid the effect of rural-urban mobility, heterogeneity of the effects across living area was carried out without rural-urban mobility samples (n = 15 693).d The adjusted models were adjusted for a number of covariates, including age, living areas, province, marriage status, income, educational level, employment, chronic disease and self-rated health.

Figure 2 .
Figure 2. Alluvial diagram of physical exercise frequency change among Chinese adults by forest-environment construction in different provinces.

Figure 3 .
Figure 3.The effects of forest-environment construction on the physical exercise frequency of adults.Note: 1. Heterogeneity of the effects across age groups was based on older adults' age in 2010.2. The adjusted models were adjusted for a number of covariates, including age, living areas, province, marriage status, income, educational level, employment, chronic disease and self-rated health.

Table 1 .
Characteristics of the participants a .

Table 2 .
Characteristics of the adults in provinces with different National Forest City number in 2018 a .
a Note: Analyzed by Х2 test for categorical variables and ANOVA for continuous variables.in appendix table 2. Firstly, to control the effect of different degrees of forest-environment construction in different provinces, we excluded participants with provincial mobility at the time of two investigations and re-conducted the study.We still observed similar effects of the construction of forest environment

Table 2 .
The effects of forest-environment construction on the PA frequency: robustness checks a,b,c,d,e .

Table 3 .
The effects of forest-environment construction on the PA frequency: robustness checks a,b,c,d .The DID estimation (β), 95%CI (lower to upper)were presented in the table.b R5: conducted hierarchical linear model (HLM) using the GDP in 2018 of each province; R6: conducted hierarchical linear model (HLM) using the annual mean temperature in 2018 of each province; R7: conducted hierarchical linear model (HLM) using the number of rainy days in 2018 of each province; R8: conducted hierarchical linear model (HLM) using the number of haze days in 2018 of each province.The adjusted models were adjusted for a number of covariates, including age, living areas, province, marriage status, income, educational level, employment, chronic disease and self-rated health.

Table 4 .
The effects of forest-environment construction on the PA frequency: robustness checks using DID with multiple periods estimator a,b,c .The DID estimation (β), 95%CI (lower to upper)were presented in the table.b * * * p < 0.001, * * p < 0.01, * p < 0.05.c The adjusted models were adjusted for a number of covariates, including age, living areas, province, marriage status, income, educational level, employment, chronic disease and self-rated health.
Note: a