Content area
Current public obesity intervention focuses on promoting programs that encourage exercise and healthy eating. Our study emphasizes that rapid technological changes may also have the potential to lead to obesity epidemics. This research investigates whether household technology launched in China during the last two decades has the potential to cause increases in body mass index (BMI). We hypothesize that adopting household technology is a contributory factor in BMI increase, independent of daily calorie consumption and energy expenditure in exercise. To test this hypothesis, we use longitudinal data from individuals aged 18-55 who participated in the 1997-2009 China Health and Nutrition Survey. Linear fixed-effects regression captures the effects of the dynamic processes of adopting household technology on BMI. All analyses are stratified by gender. The results show that adopting computers or air conditioners is associated with BMI increases in men, while adopting washing machines promotes BMI increases in women. Having a computer is associated with a decrease in BMI for women. Food-preparation technologies, such as refrigerators, microwaves, rice makers, and pressure cookers, are associated with BMI increases for both men and women. This study suggests that household technology ownership and BMI increases are linked, whereas changes in overall energy intake and exercise may not function as mediators for this relationship. Future public health policy may evaluate interventions focused on increasing low-intensity activities impacted by household technologies.
http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = Popul Res Policy Rev (2015) 34:877899
DOI 10.1007/s11113-015-9371-z
Chih-Chien Huang1 Scott T. Yabiku2
Jennie J. Kronenfeld2
Received: 3 October 2013 / Accepted: 10 June 2015 / Published online: 19 June 2015 Springer Science+Business Media Dordrecht 2015
Abstract Current public obesity intervention focuses on promoting programs that encourage exercise and healthy eating. Our study emphasizes that rapid technological changes may also have the potential to lead to obesity epidemics. This research investigates whether household technology launched in China during the last two decades has the potential to cause increases in body mass index (BMI). We hypothesize that adopting household technology is a contributory factor in BMI increase, independent of daily calorie consumption and energy expenditure in exercise. To test this hypothesis, we use longitudinal data from individuals aged 1855 who participated in the 19972009 China Health and Nutrition Survey. Linear xed-effects regression captures the effects of the dynamic processes of adopting household technology on BMI. All analyses are stratied by gender. The results show that adopting computers or air conditioners is associated with BMI increases in men, while adopting washing machines promotes BMI increases in women. Having a computer is associated with a decrease in BMI for women. Food-preparation technologies, such as refrigerators, microwaves, rice makers, and pressure cookers, are associated with BMI increases for both men and women. This study suggests that household technology ownership and BMI increases are linked, whereas changes in overall energy intake and exercise may not function as mediators for this relationship. Future public health policy may evaluate interventions focused on increasing low-intensity activities impacted by household technologies.
& Chih-Chien Huang [email protected]
Scott T. Yabiku [email protected]
Jennie J. Kronenfeld
1 Department of Sociology, Saint Anselm College, Manchester, USA
2 Department of Sociology, Arizona State University, Tempe, USA
http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11113-015-9371-z&domain=pdf
Web End = The Effects of Household Technology on Body Mass Index among Chinese Adults
123
878 C.-C. Huang et al.
Keywords Body mass index Household technology Fixed-effects approach
Obesity Longitudinal study China Health and Nutrition Survey (CHNS)
China has experienced an extraordinary economic and social transitionalong with an ever-increasing obesity rate due to the radical transformation of peoples lifestyles, dietary habits, and health-related behaviors that occurred in the last decade of the twentieth century (Du et al. 2002; Ma et al. 2005; Popkin 2001; Popkin and Gordon-Larsen 2004; Wang et al. 2007). The rapid increase of the obesity rate in China is a matter of great concern. For example, Wang et al. (2007) show a substantial increase in overweight and obesity rates for Chinese adults in both genders and in rural and urban areas, from 14.6 to 21.8 %, over the 10-year period during 19922002; this increasing rate is similar to what the United States experienced in the 40 years during 19602000 (Flegal et al. 2002). More recent data show a sharper rise from previous studies: 45 % of male and 32 % of female adults were categorized as overweight or obese in 2010 (Patterson 2011); this implies that approximately one-fth of the worlds one billion overweight or obese people are Chinese (Wu 2006).
Obesity has been found to be an independent risk factor for many health conditions and is associated with increased hazard ratios for mortality (Adams et al. 2006; Berrington de Gonzalez et al. 2010). The relationship between obesity and the risks of hypertension, coronary heart disease, and type 2 diabetes are well recognized (Li et al. 2002). In addition, obese people have a greater risk of depression, low self-esteem, and poor body image as well as social prejudice, bias, discrimination, stereotype, and stigma. Researchers believe that the increasing prevalence of being overweight and obese is largely attributed to an increase in the population adopting riskier health lifestyles, including a Westernized diet, sedentary routines, and technology use. There are decades of social science research documenting obesity (Cawley 2014); however, most studies have been conducted in Western cultural contexts (Wang and Beydoun 2007) with limited understanding of the nature of social determinants on obesity in the developing world (Monteiro et al. 2004). Many important questions still remain unexplained in non-Western contexts and deserve more thorough study.
Household Technological Development and Obesity
Finkelstein et al. (2005) argue that technological improvement is one of the primary causes of the increased obesity following the 1980s. Developing technology allows people to rearrange their schedules, which affects dietary habits, activities, and nutritional status. Past Western studies have drawn upon the relationship between a variety of modern technologies and obesity prevalence. For example, scholars argue that agricultural innovation has resulted in increased food availability and reduced food prices; these factors encourage higher calorie consumption, which has been found to be related to obesity (Cutler et al. 2003; Lakdawalla and Philipson 2009).
123
The Effects of Household Technology on Body Mass Index 879
Mechanization promotes sedentary forms of lifestyles; studies show that each additional hour spent in a car per day is associated with a 6 % increase in the likelihood of obesity (Frank et al. 2004). People who perform daily tasks with modern devices, including using washing machines, dishwashers, cars, and elevators, spent 111 kcal less than they would when completing those tasks manually, which potentially adds 10 lbs. (4.5 kg) of weight per person yearly (Lanningham-Foster et al. 2003). On the other hand, studies examining the association between household technology adoption and the potential risk of obesity have been limited to Western populations, and the direct evidence for this association is still inconclusive in recent developing countries, where the initial adoption and spread of household technology has occurred much more recently. For example, television (TV) was initially adopted by Chinese households in the 1980s, increased signicantly in the 1990s, and universalized after 2008 (Italian Trade Commission 2011). This occurred about 30 years after it did in the United States where TV implementation in homes started in the late 1940s and has been widespread since 1975 (Bowden and Offer 1994; World Bank 2014).
The Growth of Household Technology in China
Recent empirical data collections in China provide an unprecedented opportunity to rigorously evaluate the association between household technology and obesity and the extent to which it will be possible to extract meaningful information from the Chinese experience for recently developing countries. In addition, the relationship between the adoption of household technology and obesity is likely to be different in rapidly developing countries than in settings that already have completed the household technology transition, such as in Western societies. Chinese household technology launched during the last two decades seems to have the potential to transform lifestyles in major ways. For example, recent patterns in China show a large rise in average viewing hours consistent with the initial spread of television. In contrast to the spread of television viewing in America in the 1950s, however, Chinas widespread adoption of TV is accompanied by large TVs connected to DVD players and videogames. Thus, the TV that has recently spread through China is potentially more immersive, engaging, and likely to draw viewers into sedentary behavior. For example, between 1997 and 2004, the average time spent watching TV among teens and children increased from 2.78 to 3.12 h in the United States and from .95 to 1.61 h in China, or a 11.85 and 68.95 % increase, respectively (Television Bureau of Advertising 2010; Zhang et al. 2012). Although absolute changes in TV-viewing hours would seem most relevant to absolute changes in BMI, the large percentage growth in TV-viewing time in China still suggests a potential for rising risks of obesity.
A weakness of many existing studies on household ownership of domestic technology and Chinese obesity is that they do not adequately capture the dynamic processes of adopting household technology on weight change (Bell et al. 2002), nor do they control all the time-invariant confounders, such as unmeasured individual predispositions, which makes it difcult to infer any rm causal judgments (Bell
123
880 C.-C. Huang et al.
et al. 2002; Lear et al. 2014; Monda et al. 2008; Qin et al. 2012). In addition, the China Health and Nutrition Survey (CHNS), as panel data, allows us to measure the effects of household technology adoption on an individuals current BMI, but not to measure the effects of an individuals BMI on his or her inclination to use household technology. Specically, our study focuses on a particular direction of causation that of the effect of ownership of household technology on BMIand not the reverse. Finally, past research studies in China did not emphasize gender differences with respect to the association of household technology adoption and the risk of obesity (Bell et al. 2002; Lear et al. 2014). Our study emphasizes gender differences in the use of household technology.
Gender Differences in the Use of Household Technology
Labor-saving devices and food-preparation technologies are used to reduce time spent on domestic chores and arduous household tasks primarily carried out by women. For example, rapid economic and social development has altered the value of time for women; food-preparation technologysuch as microwaves, refrigerators, rice makers, or pressure cookershas made cooking at home possible by speeding up the shopping, preparing, and cooking process. However, scholars argue that introducing labor-saving devices or food-preparation technologies does not reduce womens time in housework; instead, it only changes the way that women perform household tasks (Shehan and Moras 2006). As an example, Shehan and Moras (2006) review the history of laundry in the United States. They argue that after washing machines were widely used in homes, housewives took over the responsibility for a task that had previously been performed in the household by hiring paid laundresses, and spent more time themselves on laundry. Robinson (1985) also observes that to rejuvenate Chinas economy, Deng Xiaoping promoted the availability of modern household technology with the intention of liberating women from arduous housework to free them for participation in labor production; however, modern household technology continues to tie women to traditional gender roles. Modern Chinese women still perform most private duties and domestic-based work in the home (Robinson 1985). In short, whether labor-saving devices and food-preparation technologies save womens energy expenditures in household works is still unclear.
Other household technologies are also utilized differently by gender. For example, Sugiyama et al. (2008) nd that women spent less time than men on video watching, computer use, listening to music, talking on the phone, and driving in a car. A report from the Television Bureau of Advertising shows that American women spent more time viewing television than men, teens, and children during the years 19882009 (Television Bureau of Advertising 2010). In short, gender differences are likely to persist in settings outside the United States as well, including China. To better understand how obesity is affected by household technology adoption, it is important to take gender differences into account.
123
The Effects of Household Technology on Body Mass Index 881
The Mediator between Household Technology Adoption and BMI Increases
There are two main hypotheses to explain the adoption of household technology and its link to BMI increases: the rst hypothesis is that exercise and total dietary calorie intake may function as mediators to the extent that they account for the relation between owning household technology and BMI increases. That is, household technology may cut leisure time spent exercising and increase daily calorie consumption, thereby triggering BMI increases. For instance, Jakes et al. (2003) state that for both men and women, increased TV viewing greatly decreased participation rates in vigorous activities. Furthermore, TV viewing is positively associated with total caloric intake and calories from fat; people snack more when they are in front of the TV (French et al. 2001; Jeffery and French 1998). Kobayashi and Kobayashi (2006) mention that the invention of air conditioners brought lifestyle changes to Japan; they suggest that during the hot summer season, children now spend signicantly more time indoors playing video games, watching TV, or studying rather than taking part in outdoor activities. However, this hypothesis still lacks direct supporting evidence.
An alternative hypothesis is that adopting household technology is a factor independent of exercising and daily calorie consumption for BMI increases. Evidence has shown that applying labor-saving devices has resulted in less energy expenditure for women (Lanningham-Foster et al. 2003). Substituting TV viewing with walking around at home is sufcient to expend energy equal to a 6.61 lb (3 kg) weight loss during a year among adults (Buchowski and Sun 1996). Numerous scholars have found the relationship between TV viewing or computer use and obesity remains salient even after controlling for exercise and overall food intake (Coakley et al. 1998; Hu et al. 2003; Koh-Banerjee et al. 2003; Vandelanotte et al. 2009). In other words, a decline in low-intensity activities may be the reason that adopting household technology increases the risk of obesity. This decline in low-intensity activities, such as daily chores around the home and hand washing clothes, is relatively independent of discretionary activities that might account for the BMI increases. For example, food-preparation technology, such as refrigerators, may reduce daily energy expended while shopping in grocery stories. Communication devices including telephones and cell phones may decrease traveling and physical social interactions, such as walking door to door to visit relatives and neighbors. Entertainment devices including TV, DVD/VCD players, and computers can encourage watching sports, movies, and concerts at home rather commuting to stadiums, theaters, and concert halls.
The empirical evidence shows that daily calorie consumption has declined, while energy expenditure on exercise has increased, whereas the prevalence of obesity has risen in the last two decades in China (Du et al. 2002; Ng and Popkin 2012; Ng et al. 2009). To explain these diverging trends, we have three hypotheses:
Hypothesis 1 Adopting household technology will be associated with BMI increases.
123
882 C.-C. Huang et al.
Hypothesis 2 Adopting household technology is a factor independent of exercising and daily calorie consumption for BMI increases; in other words, daily calorie consumption and energy expenditure in exercise will not function as a mediator to the association of household technology adoption and BMI increases.
Hypothesis 3 Because household products are utilized differently by gender, there will be gender differences in the relationship between BMI increases and a variety of household technologies.
Methodology
Study Population
The CHNS is an ongoing longitudinal project that gathers data on health, nutrition, and socioeconomic indicators at the individual, household, and community levels. The survey began in China in 1989 with follow-ups in 1991, 1993, 1997, 2000, 2004, 2006, and 2009. The CHNS undertook the rst round of data in 1989 using a multistage, random cluster process to draw a sample from eight provinces (Guangxi, Guizhou, Henan, Hubei, Hunan, Jiangsu, Liaoning, and Shandong), which vary substantially in geography, stage of economic development, public resources, and health indicators. Counties in each province were stratied by income levels, and multistage random sampling was used to select four counties in each province based on per capita income reported by the National Bureau of Statistics. Within each county or urban area, neighborhoods were randomly selected from urban and suburban areas, townships, and villages. Twenty randomly selected households were chosen within each neighborhood. The provincial capital and a lower-income city were selected when feasible. Villages and townships within the counties and urban and suburban neighborhoods within the cities were selected randomly with a total of 190 primary sampling units from 1989 to 1993. A new province (Heilongjiang) and its sampling units were added in 1997, whereas Liaoning Province was unable to participate in the survey because of a natural disaster and for political and administrative reasons (Popkin et al. 2010). In 2000, Liaoning returned to the survey, and nine provinces, including Heilongjiang, continued to participate in all subsequent surveys (Popkin et al. 2010). Due to these events, the sample size varies over time.
There were 4020 households surveyed in the 1989 wave with a total of 15,927 individuals from eight provinces, and all individuals within a household were interviewed. Follow-up levels were high, but families that migrated from one community to a new one were not followed. Response rates were *88 % at the individual level and *90 % at the household level for participants of the previous year (Popkin et al. 2010). CHNS data are not nationally representative, because provinces vary substantially in geography, stage of economic development, public resources, and health status. However, previous research ndings on key physical composition and dietary data trends based on the CHNS are similar to those revealed by nationally representative data (Ge et al. 1994; Wang et al. 2006).
123
The Effects of Household Technology on Body Mass Index 883
The CHNS provides panel data that contain observations over multiple time periods; our models are able to incorporate all available measurements from each individual, which maximizes our analytic sample. For this research, we focused on data in which respondents aged 1855 years completed weight and height examinations by health care experts during 19972009. The 19891993 survey data are omitted because they lack information for energy expenditure in exercise.
In our analysis, we include only individuals observed over multiple time periods during 19972009 (85.27 %, 37,592 cases). Among the 37,592 cases, women who were pregnant or breastfeeding (1.50 %, 563 cases) are omitted. The measure of transportation activity contained one-fourth missing values in our subsample. We investigated the patterns of missing values of transportation activity and found that79.77 % of individuals were not currently working or going to school, and concluded that those data are missing because transportation activity is not relevant for individuals who do not commute to work/school. Therefore, we recoded missing values to 0 metabolic equivalent (METs) hours per week. Home activity contains10.83 % missing values, and 83.63 % of those missing were from males. We conclude these data are missing because home activity is not relevant to individuals who do not perform domestic work tasks; therefore, missing values are recoded as 0 METs. Finally, only 73.29 % (27,140 cases) of the records13,037 males and 14,103 females from 12,285 households, which have information about age, marital status, socioeconomic status, smoking status, and physical activities in exercise, occupation, transportation, and homeare included in our nal analysis. Note that we tried an alternative approach in which metabolic equivalent was measured categorically, and individuals with missing METs were coded into a separate category. The coefcients, as well as their standard errors and p-values are similar in this analysis to the one in which the missing values are coded as 0 METs, and the conclusions to our hypothesis tests are unchanged.
Measurements
Dependent Variable: BMI
Body mass index (BMI) is dened as Weightkg
Heightm
2 in its continuous form. In Caucasian
populations, BMI is divided into standard categories for underweight (BMI \ 18.5), normal weight (18.5 B BMI \ 25), overweight (25 B BMI \ 30), and obese (BMI C 30). James et al. (2001) examined the worldwide obesity epidemic and nd that Asian investigators have supported an alternative classication system, because the absolute levels of diabetes and hypertension on the age- and sex-specic basis are higher in people of Asian origin. This implies that in adult life, increases in BMI may have different impacts in different societies. Therefore, in this analysis, we use the WHO-Asian criteria when referring to underweight (18.5 \ BMI), normal weight (18.5 B BMI \ 23), overweight (23 B BMI \ 25), and obese (BMI C 25).
123
884 C.-C. Huang et al.
Independent Variables: Household Technologies
At each survey wave during 19972009, the CHNS collected information regarding household ownership of electrical appliances and equipment via the following questions: Does your household own this appliance (yes/no)? and Does your household own this type of transportation (yes/no)? In this study, we examine the relationship between BMI and seven types of household technologies. We believe that individuals who live in households most able to adopt a large variety of technologies are more likely to experience a greater increase in BMI.
Seven technological devices in our regression models were recorded as dummy variables: 1 represents owning the appliance, and 0 represents that the household did not own any such appliance. Because household income may be highly related to the number of adopted technological devices at home, one might expect there to be a strong collinearity between the ownership of each technological device. However, in our study, some household technologies such as TVs, washing machines, food-preparation technologies, and communication devices were extensively adopted among all the households during 19972009, regardless of household income. In addition, there are distinct relationships between each specic technology and obesity that depend on the nature of each device. A count variable would not allow us to explore the devices individually, and it would also assume that the acquisition of each device had an additive association with obesity. Thus, for these reasons, we use dummy variables representing the technologies, rather than an additive scale of devices.
Household technologies are categorized based on the type of function:(a) televisions, (b) computers, (c) washing machines, (d) communication devices (telephones and cell phones), (e) air conditioners, (f) motorized vehicles (cars and motorcycles), and (g) food-preparation technologies (refrigerators, microwave ovens, electric rice makers, and pressure cookers). Note that CHNS collected the household ownership of cell phones only after 2004.
Covariates/Control Variables
Studies have found that marital status and age are factors inuencing body weight (Gallagher et al. 1996; Ogden et al. 2006). For example, BMI increase is a common occurrence as people get older, and the most substantial BMI increases occur during middle age (Bennett et al. 2008). Previous studies also show BMI-for-age curves are nonlinear; studies suggest that an age-squared term should be included in an analysis (Flegal et al. 2010; Rzehak and Heinrich 2006). Smoking can impact body weight; the BMI of smokers is lower than that of nonsmokers (Compton et al. 2006; Johansson and Sundquist 1999), but gaining substantial weight after quitting smoking is quite common (Schwid et al. 1992). Past studies show socioeconomic status (SES) and obesity have a strong relationship, especially among women (McLaren 2007; Sobal and Stunkard 1989). For example, the Western literature documents that a lower educational attainment is often associated with a higher obesity rate, whereas developing countries present the opposite trend (Roskam et al. 2010).
123
The Effects of Household Technology on Body Mass Index 885
The CHNS asks individuals whether they are presently working, and if not, why; disability is one of the possible answers. Because this is an indirect question regarding individuals with disabilities, who represented less than 1 % of the cases in our sample, we do not consider disability as a factor in changes in BMI. In our models, we measured changes in individuals BMI before and after they implemented technological adoption. We have also controlled for individuals physical activity related to exercise, occupation, mode of transportation, and household activities; people with disabilities can also gain weight through adoption of household technologies (despite having a different cutoff value for BMI), so we think it is also proper to include people with disabilities.
Based on the discussion above, the time-varying covariates that we have controlled in this study include: (a) current age and age squared; (b) marital status, recoded as a binary variable: currently married and currently not married (included single, divorced, widowed, or separated); (c) smoking status, recoded as a binary variable: currently smokes versus currently does not smoke; and (d) SES. There are two SES variables used in this study. Education is recoded as no education, primary education, secondary education, and college or above. Household gross income is assessed by adding each households nine potential sources of income: business, farming, shing, gardening, livestock, nonretirement wages, retirement income, subsidies, and other income. When any component was incomplete, the CHNS project team attempted to impute the missing data. Details of the imputation have been described elsewhere (China Health and Nutrition Survey 2013a, b). For interpretability, household gross income was logged. In such models, the logged case refers to the proportional change in the household gross income for one BMI increase. Finally, we have also controlled for ve survey years: 1997, 2000, 2004, 2006, and 2009 in all regression models.
Mediating Variables: Caloric Intake and Energy Expenditure
Arithmetically, an individual gains weight through a positive energy balance; that is, when calories that an individual consumes exceed calories that he or she expends, weight increases. It is essential to consider changes in dietary consumption and energy expenditure in exercise as key components in weight change in China. In our study, daily calorie intake refers to the energy value of all food consumed within 24 h, averaged over 3 days and calculated using the 1991 Chinese food composition table (FCT) for the years 1997 and 2000 only; the 2002/2004 FCT is used in all years following 2000. Detailed descriptions of the dietary survey are presented elsewhere (China Health and Nutrition Survey 2013a, b). We also controlled the change of calories in daily fat, carbohydrate, and protein intake, but it was not found to signicantly alter the estimates and was excluded from the nal models.
One MET is dened as the amount of oxygen consumed while at rest. As such, work at two METs requires twice the oxygen and, at three METs, three times the oxygen of a resting metabolism, and so on (Jette et al. 1990). In our study, exercise is recoded as a continuous variable summed up by MET hours per week of participation in activities including martial arts, jogging, swimming, dancing, aerobics, sports, and others. Other energy expenditure in non-exercise physical
123
886 C.-C. Huang et al.
activities controlled in our models includes occupational activity, transportation activity, and home activity. Occupational activity is categorized as very light/light (ofce worker, watch repairer, salesperson, laboratory, etc.), moderate (student, driver, electrician, metal worker, etc.), and heavy/very heavy (farmer, dancer, steel worker, athlete, loader, logger, miner, stonecutter, etc.). The total weekly energy expenditure is estimated by MET hours per week of participation in occupational activities. Transportation activity is based on whether or not the participants walked or biked to school/work and is summed up by MET hours per week for biking or walking. Home activity is based on whether or not the participants reported time spent preparing food, buying food, doing laundry, and on childcare. The total weekly energy expenditure is estimated by time spent on each home activity, multiplied by specic MET values based on the compendium of physical activities. Detailed descriptions of the measurements of metabolic equivalent for physical activities in exercise, occupation, transportation, and home are presented elsewhere (Ainsworth et al. 2000; Ng et al. 2009)
Statistical Analyses
We selected linear xed-effects (FEs) regression models because they have two attractive features: First, they are efcient in estimating the effect of variables that vary considerably within an individual. FE models in longitudinal data have been regarded as one of the best kinds of statistical methods for making causal inferences in nonexperimental social science research (Allison 2009). In other words, the FE method is similar to the experimental study setting: We compared the average BMIs from the experimental group (who adopted household technology in previous survey years) to the control group (who did not adopt household technology in previous survey years), assuming that household technologies are randomly adopted by households at a given point in time. Second, FE models are designed to study the causes of changes within an individual by controlling for potential unobserved heterogeneity bias. In other words, standard regression models are more likely to suffer from omitted variable bias and spuriousness because unmeasured factors are likely to be correlated with both the dependent variable and key independent predictors. In short, FE models offer stronger causal evidence because unmeasured factors that do not change within individuals over time cannot bias the relationship between the outcome and predictor variables (Allison 2009; Baltagi 2008; Wooldridge 2010). Specically, the linear regression model with time-invariant covariates in our study is written as
yBMIijt at X
K
k1
bkxkijt X
M
m1
cmzm ui vj eijt
where at is an intercept that may be different for each survey year; b and c are vectors of coefcients; and x represents time-varying covariates, including one of our independent variables and was recorded as a categorical variable: 1 represents a household that owned the appliance, and 0 represents the household that did not own the appliance. x also includes covariates that we have controlled in our study,
123
The Effects of Household Technology on Body Mass Index 887
such as survey year, age, age2, marital status, smoking, education, occupational activity, transportation activity, and home activity, as well as important mediators such as daily energy intake and exercise. Our FE models also incorporate SES (household gross income) at the household level.
Z stands for time-invariant covariates, including year-specic and person-specic effects that play a role in BMI increases. Covariates with person-specic effects are those that affect individuals in different ways but are constant across time, such as ethnicity, birthplace, childhood SES, and genetic factors. Covariates with year-specic effects are those that affect all individuals in the same way but change over time, such as food policy and obesity legislation. There are three error terms in this model: eijt represents the random variation of each individual i in household j at each survey year t, ui represents the variation of all unobserved variables on BMI that vary across individuals but are constant over time, and vj represents unobserved variations at the household level.
In estimating rst-difference equations, the factors that are constant over time, such as PMm1 rmzm;ui, and vj, are removed from the equation, so FE does not allow
us to assess time-invariant covariates (Allison 2009; Baltagi 2008; Wooldridge 2010). The nal equation is written as
DBMIijt Da X
K
k1
bkx1ijt xkijt 1 eijt
where D the rst-difference operator. This model is meant to test whether or not there is a net effect of household technology on BMI increases. Specically, we are assuming a particular direction of causation: that ownership of household technology affects BMI change and not the reverse. In FE models, the cluster refers to each individual when repeated measurements in each year are nested. All the individuals observed over time within a household are included in our analysis however, FE models rule out between variations, meaning that unobserved differences between households are removed, and they no longer bias our estimates.
Finally, we believe that household technology may have different impacts on men and women based on the previous discussion, and we ran the models separately by gender. To test if the effects of focal variables of interest varied signicantly by gender, we formally tested gender differences with fully interactive models that reproduced our gender-specic models. All FE models were run on Stata, Version 13 (Stata Corporation, College Station, TX, USA).
Results
Percent of Individuals with Each Form of Household Technology: 19972009
Table 1 shows the trend for individuals within households in the CHNS possessing each form of technology since 1997. Figure 1, which shows the change in the ownership of household technology over 13 years, reveals substantial variations in
123
888 C.-C. Huang et al.
Table 1 Percentage of individuals owning each household technology by wave (CHNS 19972009)
Observation (individuals) Year
1997 2000 2004 2006 2009 5091 5246 5781 5454 5568
Household technologies
Television (%) 53.09 72.21 88.51 94.83 99.10
Computer (%) 2.02 5.13 11.52 16.70 32.15
Washing machine (%) 51.80 57.51 63.92 70.02 79.35
Food-preparation technology (%) 61.66 74.29 81.02 87.07 94.27
Air conditioner (%) 5.15 9.21 18.85 22.52 31.03
Communication device (%) 30.84 51.94 82.04 87.64 97.04
Modern vehicle (%) 16.19 25.94 36.83 41.55 49.53
Observations here refer to individuals in the sample having two or more observation sessions of the technology during 19972009 and were included in our nal xed-effects linear regression
100
90
80
70
60
50
40
30
20
10
Fig. 1 Percentage of individuals owning each household technology by wave (19972009)
the adoption of each household technology between 1997 and 2009. The sample shows that Chinese individuals experienced a rapid rate of household technology transition. Most individual technology possession rates increased considerably. For example, only half of individuals had a TV at home in 1997, but 99.10 % of individuals had at least one by 2009. Although only about one third to three-fths of individuals owned a washing machine, food-preparation technology, or communication device in 1997, a large proportion of individuals possessed at least one washing machine (79.35 %), food-preparation technology (94.27 %), or
0 1996 1998 2000 2002 2004 2006 2008 2010
TVComputerWashing machineAir conditionerFood preparation technology Communication devices Modern vehicles
123
The Effects of Household Technology on Body Mass Index 889
communication device (97.04 %) in 2009. However, in terms of other household technologies, the number of individuals who owned a computer (32.15 %), air conditioner (31.03 %), or modern vehicle (49.53 %) was still relatively low in 2009.
In an additional analysis not described here, we looked at adoption rates of household technology by rural and urban areas, indicating that ownership of household technologies was relatively higher in urban areas at the beginning of the survey year, but more substantial increases occurred in household technology ownership in rural areas between 1997 and 2009. In general, our sample shows a higher rate of household technology ownership in urban areas during 19972009. Whereas the demand for modern vehicles in rural areas increased signicantly after 2004, the gap in modern vehicle ownership between the two types of areas may reect differences between rural and urban areas in the development of public transit. People who live in rural areas appear to have more interest in owning cars in years when household economies are relatively well off, and people appear to have less interest in general in owning cars in dense urban environments where both population and public transport availability are developing rapidly.
Because the purpose of our research is to identify how household technologies affect weight gain across a broad segment of Chinese society, we believe it is reasonable to provide regression models regardless of rural/urban area. We expected that BMI would increase more dramatically in rural areas because of faster rates of household technology ownership between 1997 and 2009, and our initial analyses found that BMI indeed did increase more noticeably in rural areas. However, we are unable to explore this in detail because the study lacks information to construct detailed migration histories and rural/urban status over time. We believe technological differences between urban and rural areas may help explain the rapid increase in BMI in rural areas, but that more detailed research is necessary to conrm this.
Background Characteristics
Table 2 presents our sample on the prevalence of obesity, BMI, and individual background characteristics by gender. We found that average BMI also increased substantially from 22.12 to 23.44 kg/m2 among males and from 22.47 to 23.12 kg/m2 among females during 19972009. The prevalence of obesity increased from14.57 to 30.36 % among males and 18.74 to 26.49 % among females from 1997 to
2009.
Reduced exercise and increased dietary consumption have been blamed as drivers of the obesity epidemic; however, our sample indicated that these trends moved in the opposite direction during 19972009. The total energy intake (kcal) decreased, while energy expenditure in exercise (MET) increased. This is similar to the ndings from other studies using the same database (Du et al. 2002; Ng et al. 2009; Ng and Popkin 2012). Our sample also shows that occupational physical levels have decreased considerately, especially for women, from 321.85 (METs) in 1997 to 193.63 (METs) in 2009. Finally, physical activity in transportation has decreased for both men and women, whereas home activity has increased. In
123
890 C.-C. Huang et al.
Table 2 Descriptive statistics of individual-level variables by gender (CHNS 1997 and 2009)
Observation (individuals) Male Female
1997 2009 1997 2009
2492 2665 2599 2903
Dependent variable
Body mass index (kg/m2) 22.12 (2.75) 23.44 (3.38) 22.47 (2.98) 23.12 (3.38)
Obesity (%; BMI C 25) 14.57 30.36 18.74 26.49
Control variables
Age (years) 37.21 (10.24) 40.69 (9.87) 37.88 (9.79) 41.20 (9.62)
Log household income 9.36 (.75) 10.20 (1.45) 9.37 (.78) 10.17 (1.37)
Household gross income (per1000)
Education (%)
No education 9.63 5.37 27.47 15.02
Primary education 63.96 60.11 52.64 57.18
Secondary education 22.87 26.75 17.62 21.77
College and above 3.53 7.77 2.27 6.03
Married (%) 79.65 83.26 84.88 89.08
Current smoker (%) 64.77 58.87 3.08 2.00
Mediating variables
Daily energy intake (kcal) 2.64 (.71) 2.46 (1.36) 2.26 (.60) 2.05 (.79)
Exercise (MET-h/week) 4.33 (14.78) 5.35 (19.13) 1.62 (9.10) 3.51 (14.24)
Physical activity (MET-h/week)
Occupation 342.07(213.65)
general, our sample shows that there were substantial changes in lifestyle and behavior among Chinese adults during the last 12 years.
Causal Effects of Ownership of Household Technology on an Individuals BMI: 19972009
In our study, FEs models provide an estimate of this relationship by controlling unmeasured and time-invariant variables such as sex, region of birth, and genetic makeup, but they do not control for time-variant variables such as age, household income, education, marriage status, daily energy intake, and physical consumption. These time-varying variables are handled through inclusion in our regression models. Table 3 results indicate that in the ordinary least squares model, almost
15.09 (12.60) 46.60 (75.58) 15.14 (12.46) 43.54 (61.66)
266.32
(222.41)
321.85
193.63
(210.31)
(191.25)
Transportation 2.02 (3.32) .78 (1.71) 1.84 (3.12) .83 (2.29)
Home 7.95 (15.80) 11.18 (21.49) 34.81 (24.46) 47.81 (52.77)
The table only presents samples from the baseline, 1997, and the nal wave, 2009
Standard deviations in parentheses
Observations here refer to individuals in the sample having two or more observation sessions of the technology during 19972009 and were included in our nal xed-effects linear regression
123
The Effects of Household Technology on Body Mass Index 891
every household technology signicantly increased BMI for men and women, whereas FE regressions eliminated the inuence of time-invariant unobserved heterogeneity on this relationship. In additional models not shown, we formally tested if the effects of household technology on BMI increases signicantly varied by gender. The household technology effects described above were signicantly different for both men and women for household ownership of computers (p \ .05 for test of difference), washing machines (p \ .05), and air conditioners (p \ .10).
The effect of food technologies did not differ by gender.
Model 1 indicates that food-preparation technologies increases BMI by .101 (p \ .10), a computer increases BMI by.110 (p \ .10), and an air conditioner increases BMI by .194 in men (**p \ .01). In other words, for a man of average height (164.8 cm) in China, adopting food-preparation technologies (refrigerator, microwave, rice maker, or pressure cooker) increases his weight by .17 kg[(1.65 m)2 9 (.101 kg/m2)] = .17 kg]. Adopting a computer increases his weight by .18 kg while adopting an air conditioner increases his weight by .32 kg. Model 2 indicates that after adjusting for physical activities and dietary intake, there is very little change in the coefcients of household technologies compared to those of Model 1, which means that exercise and dietary intake do not function as mediators for the association between household technology and BMI in men.
For women, owning a washing machine increases BMI by .099 kg/m2 (*p \ .05), food-preparation technologies increase BMI by .186 kg/m2 (***p \ .001), and a communication device increases BMI by .167 kg/m2 (**p \ .01). For a woman of average height (154.5 cm) in China, this gures are equivalent to adding .15, .29, and .26 kg over the period of 1997 through 2009, respectively. Unexpectedly, our sample also indicates that owning a computer at home decreases female BMI by .146 kg/m2 or .22 kg (*p \ .05). Model 4 indicates that exercise and dietary intake do not function as mediators for the association between household technology and BMI in women, but the changes of METs in home and occupational activity had signicant effects on BMI during 19972009.
Discussion and Conclusion
The initial household technology adoption took place about ve decades ago in Western countries, a time period for which highly accurate representative data on technology ownership does not exist, and it is unknown how this adoption affected BMI increases. In China, however, the initial adoption and spread of household technology occurred much more recently. Recent empirical data collections in the CHNS provide an unprecedented opportunity to rigorously evaluate the relationship between initial and rapid household technology adoption and BMI increases. It gives us an excellent opportunity to learn from Chinas experience of how household technology growth can trigger obesity epidemics in a rapidly developing world. However, this study focuses on adults only. Obesity among children may also be related to the introduction of technology into the home, but children may respond to household technology differently compared to adult men and women. The impacts of household technology on childhood obesity require further research.
123
892 C.-C. Huang et al.
Our results show that food-preparation technologies promote BMI increases for both men and women; we suppose that food-preparation technologies may change eating habits for the entire family. For example, the accessibility and convenience of microwave ovens and refrigerators may promote food cravings throughout the day. On the other hand, our study results suggest that household technology may have different impacts on BMI increases by gender due to diverse gendered household technological usage. We note several signicant gender differences in the association between household technology and BMI increases. For example, our data show that washing machines promote BMI increases in women but not in men, whereas air conditioners cause BMI increases in men but not in women. This may suggest that washing machines decrease the low-intensity activities at home that are primarily carried out by women, and air conditioners may encourage men to spend a larger portion of their time engaged in sedentary activity at home during the hot summer season. The most puzzling nding is that having a computer is associated with a decrease in weight for women but an increase in weight for men. Computer adoption in Chinese households is relatively new, and longer follow-up studies will help us understand the true nature of its effects. On the other hand, with the increasing social values that cite thinness as the ideal identity for Chinese women (Chen et al. 2007; Leung et al. 2001; Luo et al. 2005; Xu et al. 2010), adopting a computer at home may allow women to control their weight efciently by gaining knowledge and social support from Internet forums, weblogs, social blogs, etc. In short, future research may be directed to investigate gender differences in household technology related to health risk behaviors. Understanding the mechanisms of household technology that trigger BMI increases has important public health implications. Current public interventions against obesity in China have focused on promoting education programs that encourage exercise and healthy eating. Future public health policy may evaluate interventions focused on increasing low-intensity activities impacted by household technologies.
We further tested whether exercise and daily energy consumption functioned as mediators for the association between household technology and BMI increases and did not nd signicant mediation. In fact, our sample shows that the energy expenditure in exercise increased, whereas daily energy consumption decreased during 19972009; our ndings are similar to other studies (Du et al. 2002; Ng and Popkin 2012; Ng et al. 2009). We posit that adopting household technology may facilitate the decline of low intensity activity, such as doing chores around the house or hand washing clothes, and this decline is the mechanism that triggers BMI increases, independent from exercise and daily calorie consumption. Finally, past studies demonstrate that owning a washing machine and food-preparation technology is related to a signicant increase in weight for men and women (Monda et al. 2008). Our FE models have shown that adopting washing machines has no effect on men. Other scholars note that adopting motorized vehicles promotes weight gain (Bell et al. 2002; Qin et al. 2012). Our results do not nd a signicant impact from motorized vehicles on BMI increases in women or men. It may be that past studies investigating the associations between Chinese household technologies and BMI increases drew slightly different conclusions from us because the associations may be confounded by time-invariant variables (Bell et al. 2002; Lear et al. 2014; Monda
123
The Effects of Household Technology on Body Mass Index 893
-.002
(.058)
-.141*
(.058) (.088)
Washingmachine.510***
(.065)
.096*
(.048)
.184***
(.056)
.155**
(.052)
.067
(.064)
.011
(.044)
.257**
(.085)
-.001***
(.000)
-.010
(.097)
-.146*
(.058)
.099*
(.048)
-.001***
(.000)
.006
(.058)
.186***
(.056)
Table3Fixed-effectslinearregressionfortheownershipofhouseholdtechnologypredictingmeanBMIbygender(CHNS19972009)
VariablesMaleFemale
OLSFEOLSFE
(1)(2)(3)(4)
.167**
(.052)
.057
(.064)
.017
(.044)
.260**
(.085)
-.007
(.097)
.366***
(.089)
-.251**
(.088)
.537***
(.065)
.016
(.077)
.551***
(.080)
.117*
(.058)
.256***
(.024)
-.002***
(.000)
.090
(.077)
.164
(.096)
.023
(.060)
.110
(.060)
-.017
(.050)
.101
(.057)
-.031
(.054)
.192**
(.062)
.499***
(.091)
-.003***
(.000)
.009
(.045)
.012
(.079)
.024
(.060)
-.018
(.050)
.101
(.057)
-.030
(.054)
.194**
(.067)
.498***
(.091)
-.003***
(.000)
Computer.344***.110
(.060)
.012
(.045)
.015
(.079)
Foodtechnologies-.128
(.077)
Communicationdevice.636***
(.081)
Airconditioner.586***
(.079)
Married.392***
(.089)
Householdtechnologies
TV.398***
(.089)
Motorizedvehicle.186**
(.058)
Age2 -.003***
(.000)
Covariates
Age.310***
(.023)
123
894 C.-C. Huang et al.
-.178
(.155)
-.084
(.076)
-.462*
(.198)
-.034***
(.010)
.014
(.014)
-.128
(.122)
-.011
(.023)
-.002
(.001)
-.174
(.155)
.013
(.014)
-.084
(.076)
-.132
(.122)
-.474*
(.198)
-1.322***
(.152)
.111
(.163)
-.046*
(.021)
-.027
(.074)
-.815***
(.093)
-.226***
(.049)
.036*
(.014)
.061
(.105)
.105
(.136)
.052
(.185)
Currentsmoker-.449***
(.054)
Exercise-.0001
(.001)
-.227***
(.049)
.036*
(.014)
.060
(.105)
.103
(.136)
.047
(.185)
Loghouseholdincome.009
(.021)
Collegeandabove.666***
(.158)
Table3continued
VariablesMaleFemale
OLSFEOLSFE
(1)(2)(3)(4)
Education(ref.=Noeducation)
Primary.286**
(.109)
Secondary.541***
(.119)
Mediatingvariables
Dietaryintake-.002
(.017)
Physicalactivity
Occupation-.006
(.010)
123
The Effects of Household Technology on Body Mass Index 895
.005
(.007)
.001
(.000)
14.657***
(2.795)
Observations13,03713,03713,03714,10314,10314,103
R-Squared.12.90.90.11.90.90
Adj.R-squared.12.82.82.11.83.83
Two-tailedtests;standarddeviationsinparentheses***p\.001,**p\.01,*p\.05,
14.416***
(2.793)
15.977***
(.461)
p\.10
8.632**
(2.956)
Home.001
(.001)
8.611**
(2.954)
Constant14.666***
(.453)
Allmodelswereadjustedforsurveyyears19972009
Table3continued
VariablesMaleFemale
OLSFEOLSFE
(1)(2)(3)(4)
Transportation-.007
(.007)
123
896 C.-C. Huang et al.
et al. 2008; Qin et al. 2012). Our FE approach may have yielded more robust tests of these associations.
There are several limitations to this study. First, we are aware that the FE model improves on OLS, but it is not ideal if unobserved factors exist that vary over time in a way that can inuence both household technologies and BMI. For example, uncontrolled time-varying variables, such as medical insurance, Westernization, and development of public infrastructure systems can confound the results. Further, individuals who adopt a household technology device at a given point in time may be systematically different from individuals who do not, and these systematic differences may correlate with an individuals underlying characteristics, including sedentary tendencies. Consequently, endogeneity could be biasing our resultseven where individuals were not able to adopt household technology, their BMIs would likely have increased over the same time interval. If this is the case for an individuals BMI increase, the FE estimator will overstate the impact of the actual household technology on BMI. Therefore, to declare that a causal relationship exists is a strong statement.
Second, the CHNS survey followed up only with individuals who continued to reside in selected households in any given yearthat is, it did not follow individuals who moved out of their households. In addition, there are signicant development differences between urban and rural areas in China, including housing and availability of water, electricity, and fuel. However, we are unable to estimate the effects of migration from rural to urban areas or vice versa because the CHNS did not track information from individuals moving out of their households; the residential areas are a time-invariant variable and xed. However, we believe it is reasonable to provide regression models regardless of residential area because the difference in rates of BMI increase between rural and urban areas are more likely correlated with the increasing household technology ownership. In our study, we predicted that BMI would increase more dramatically in rural areas because of faster rates of household technology ownership between 1997 and 2009, and we nd that BMI indeed increased more noticeably in rural areas.
Third, owning a household technology does not mean every individual in household uses it to its full potential. Unfortunately, there is no information from the CHNS about the amount of time or frequency of each technologys use by each household member. Fourth, some time-varying variables are not taken into account due to the limitations of the survey. For example, most household chores such as cleaning the house and home maintenance are not included in the survey. However, our ultimate purpose is to emphasize that rapid technology transition in developing societies may have important inuences on obesity epidemics, along with changes in physical activity and diet behaviors.
Finally, our study may underestimate obesity prevalence because the CHNS is not nationally representative. For example, in our analytic sample, 30.36 % of males and 26.49 % of females were categorized as obese (BMI over 25) in 2009; this is lower than the national assessments, which showed that an average of 38.5 % of the 2010 population age 15 or over had a BMI of 25 or greater (Patterson 2011). Our aims in this paper, however, have been to examine the relationship between household technology and BMI increases rather than to obtain the most accurate
123
The Effects of Household Technology on Body Mass Index 897
national point estimates of obesity. It is likely that the associations we have found would apply to other regions of China, as well as to other parts of the world that have rapidly adopted these technologies.
References
Adams, K. F., Schatzkin, A., Harris, T. B., Kipnis, V., Mouw, T., Ballard-Barbash, R., et al. (2006). Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. New England Journal of Medicine, 355(8), 763778.
Ainsworth, B. E., Haskell, W. L., Whitt, M. C., Irwin, M. L., Swartz, A. M., Strath, S. J., et al. (2000). Compendium of physical activities: an update of activity codes and MET intensities. Medicine and Science in Sports and Exercise, 32(9; SUPP/1), S498S504.
Allison, P. D. (2009). Fixed effects regression models (Vol. 160)., Quantitative applications in the social sciences Los Angeles: SAGE.
Baltagi, B. H. (2008). Econometric analysis of panel data (4th ed.). New York, NY: Wiley.
Bell, A. C., Ge, K., & Popkin, B. M. (2002). The road to obesity or the path to prevention: motorized transportation and obesity in China. Obesity Research, 10(4), 277283.
Bennett, G. G., Wolin, K. Y., & Duncan, D. T. (2008). Social determinants of Obesity. In F. Hu (Ed.),Obesity epidemiology (pp. 342376). New York, NY: Oxford University.
Berrington de Gonzalez, A., Hartge, P., Cerhan, J. R., Flint, A. J., Hannan, L., MacInnis, R. J., et al. (2010). Body-mass index and mortality among 1.46 million white adults. New England Journal of Medicine, 363(23), 22112219.
Bowden, S., & Offer, A. (1994). Household appliances and the use of time: the United States and Britain since the 1920s. Economic History Review, 47(4), 725748.
Buchowski, M. S., & Sun, M. (1996). Energy expenditure, television viewing and obesity. InternationalJournal of Obesity and Related Metabolic Disorders, 20(3), 236244.
Cawley, J. (2014). Introduction. In J. Cawley (Ed.), The oxford handbook of the social science of obesity(pp. 15). New York, NY: Oxford University.
Chen, H., Gao, X., & Jackson, T. (2007). Predictive models for understanding body dissatisfaction among young males and females in China. Behaviour Research and Therapy, 45(6), 13451356.
China Health and Nutrition Survey (2013a). Household income variable construction. Retrieved from http://www.cpc.unc.edu/projects/china/data/datasets/Household%2520Income%2520Variable%2520Construction.pdf
Web End =http:// http://www.cpc.unc.edu/projects/china/data/datasets/Household%2520Income%2520Variable%2520Construction.pdf
Web End =www.cpc.unc.edu/projects/china/data/datasets/Household%20Income%20Variable%20Construction. http://www.cpc.unc.edu/projects/china/data/datasets/Household%2520Income%2520Variable%2520Construction.pdf
Web End =pdf .
China Health and Nutrition Survey (2013b). Chinese food composition table. Retrieved from http://www.cpc.unc.edu/projects/china/data/faq.html
Web End =http://www. http://www.cpc.unc.edu/projects/china/data/faq.html
Web End =cpc.unc.edu/projects/china/data/faq.html .
Coakley, E. H., Rimm, E. B., Colditz, G., Kawachi, I., & Willett, W. (1998). Predictors of weight change in men: results from the Health Professionals Follow-up Study. International Journal of Obesity, 22(2), 8996.
Compton, M. T., Daumit, G. L., & Druss, B. G. (2006). Cigarette smoking and overweight/obesity among individuals with serious mental illnesses: a preventive perspective. Harvard Review of Psychiatry, 14(4), 212222.
Cutler, D., Glaeser, E., & Shapiro, J. (2003). Why have Americans become more obese? (Working Papers No. 9446). Retrieved from the National Bureau of Economic Research website: http://www.nber.org.ezproxy1.lib.asu.edu/papers/w9446
Web End =http://www.nber. http://www.nber.org.ezproxy1.lib.asu.edu/papers/w9446
Web End =org.ezproxy1.lib.asu.edu/papers/w9446 .
Du, S., Lu, B., Zhai, F., & Popkin, B. M. (2002). A new stage of the nutrition transition in China. PublicHealth Nutrition, 5(1A), 169174.
Finkelstein, E. A., Ruhm, C. J., & Kosa, K. M. (2005). Economic causes and consequences of obesity.Annual Review of Public Health, 26, 239257.
Flegal, K. M., Carroll, M. D., Ogden, C. L., & Curtin, L. R. (2010). Prevalence and trends in obesity among
US adults, 19992008. JAMA: The Journal of the American Medical Association, 303(3), 235241. Flegal, K. M., Carroll, M. D., Ogden, C. L., & Johnson, C. L. (2002). Prevalence and trends in obesity among US adults, 19992000. Journal of the American Medical Association, 288(14), 17231727. Frank, L. D., Andresen, M. A., & Schmid, T. L. (2004). Obesity relationships with community design, physical activity, and time spent in cars. American Journal of Preventive Medicine, 27(2), 8796. doi:http://dx.doi.org/10.1016/j.amepre.2004.04.011
Web End =10.1016/j.amepre.2004.04.011 .
123
898 C.-C. Huang et al.
French, S. A., Story, M., & Jeffery, R. W. (2001). Environmental inuences on eating and physical activity. Annual Review of Public Health, 22(1), 309335.
Gallagher, D., Visser, M., Sepulveda, D., Pierson, R. N., Harris, T., & Heymseld, S. B. (1996). How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? American Journal of Epidemiology, 143(3), 228239.
Ge, K., Weisell, R., Guo, X., Cheng, L., Ma, H., Zhai, F., & Popkin, B. M. (1994). The body mass index of Chinese adults in the 1980s. European Journal of Clinical Nutrition, 48, S148.
Hu, F. B., Li, T. Y., Colditz, G. A., Willett, W. C., & Manson, J. E. (2003). Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA: The Journal of the American Medical Association, 289(14), 17851791.
Italian Trade Commission (2011). China television industry market report. Shanghai ofce.
Jakes, R. W., Day, N. E., Khaw, K. T., Luben, R., Oakes, S., Welch, A., et al. (2003). Television viewing and low participation in vigorous recreation are independently associated with obesity and markers of cardiovascular disease risk: EPIC-Norfolk population-based study. European Journal of Clinical Nutrition, 57(9), 10891096.
James, P. T., Leach, R., Kalamara, E., & Shayeghi, M. (2001). The worldwide obesity epidemic. ObesityResearch, 9(S4), 228S233S.
Jeffery, R. W., & French, S. A. (1998). Epidemic obesity in the United States: are fast foods and television viewing contributing? American Journal of Public Health, 88(2), 277280.
Jette, M., Sidney, K., & Bluemchen, G. (1990). Metabolic equivalents (METs) in exercise testing, exercise prescription, and evaluation of functional capacity. Clinical Cardiology, 13(8), 555565. Johansson, S.-E., & Sundquist, J. (1999). Change in lifestyle factors and their inuence on health status and all-cause mortality. International Journal of Epidemiology, 28(6), 10731080.
Kobayashi, M., & Kobayashi, M. (2006). The relationship between obesity and seasonal variation in body weight among elementary school children in Tokyo. Economics & Human Biology, 4(2), 253261. Koh-Banerjee, P., Chu, N.-F., Spiegelman, D., Rosner, B., Colditz, G., Willett, W., & Rimm, E. (2003).
Prospective study of the association of changes in dietary intake, physical activity, alcohol consumption, and smoking with 9-y gain in waist circumference among 16 587 US men. The American Journal of Clinical Nutrition, 78(4), 719727.
Lakdawalla, D., & Philipson, T. (2009). The growth of obesity and technological change. Economics &Human Biology, 7(3), 283293.
Lanningham-Foster, L., Nysse, L. J., & Levine, J. A. (2003). Labor saved, calories lost: the energetic impact of domestic labor-saving devices. Obesity Research, 11(10), 11781181. doi:http://dx.doi.org/10.1038/oby.2003.162
Web End =10.1038/oby.2003.162 . Lear, S. A., Teo, K., Gasevic, D., Zhang, X., Poirier, P. P., Rangarajan, S., et al. (2014). The association between ownership of common household devices and obesity and diabetes in high, middle and low income countries. Canadian Medical Association Journal, 186(4), 258266.
Leung, F., Lam, S., & Sze, S. (2001). Cultural expectations of thinness in Chinese women. EatingDisorders, 9(4), 339350.
Li, G., Chen, X., Jang, Y., Wang, J., Xing, X., Yang, W., & Hu, Y. (2002). Obesity, coronary heart disease risk factors and diabetes in Chinese: an approach to the criteria of obesity in the Chinese population. Obesity Reviews, 3(3), 167172.
Luo, Y., Parish, W. L., & Laumann, E. O. (2005). A population-based study of body image concerns among urban Chinese adults. Body Image, 2(4), 333345.
Ma, G. S., Li, Y. P., Wu, Y. F., Zhai, F. Y., Cui, Z. H., Hu, X. Q., et al. (2005). The prevalence of body overweight and obesity and its changes among Chinese people during 1992 to 2002. Chinese Journal of Preventive Medicine, 39(5), 311.
McLaren, L. (2007). Socioeconomic status and obesity. Epidemiologic Reviews, 29(1), 2948.
Monda, K. L., Adair, L. S., Zhai, F., & Popkin, B. M. (2008). Longitudinal relationships between occupational and domestic physical activity patterns and body weight in China. European Journal of Clinical Nutrition, 62(11), 13181325.
Monteiro, C. A., Moura, E. C., Conde, W. L., & Popkin, B. M. (2004). Socioeconomic status and obesity in adult populations of developing countries: a review. Bulletin of the World Health Organization, 82(12), 940946.
Ng, S. W., Norton, E. C., & Popkin, B. M. (2009). Why have physical activity levels declined among Chinese adults? Findings from the 19912006 China health and nutrition surveys. Social Science and Medicine, 68(7), 13051314. doi:http://dx.doi.org/10.1016/j.socscimed.2009.01.035
Web End =10.1016/j.socscimed.2009.01.035 .
Ng, S. W., & Popkin, B. M. (2012). Time use and physical activity: a shift away from movement across the globe. Obesity Reviews, 13(8), 659680.
123
The Effects of Household Technology on Body Mass Index 899
Ogden, C. L., Carroll, M. D., Curtin, L. R., McDowell, M. A., Tabak, C. J., & Flegal, K. M. (2006). Prevalence of overweight and obesity in the United States, 19992004. Journal of American Medical Association, 295(13), 15491555.
Patterson, S. (2011). Obesity in China: Waistlines are expanding twice as fast as GDP 2011. Retrieved from University of South California, US-China Today website: http://www.uschina.usc.edu/w_usci/showarticle.aspx%3farticleID%3d16595%26AspxAutoDetectCookieSupport%3d1
Web End =http://www.uschina.usc.edu/w_usci/ http://www.uschina.usc.edu/w_usci/showarticle.aspx%3farticleID%3d16595%26AspxAutoDetectCookieSupport%3d1
Web End =showarticle.aspx?articleID=16595&AspxAutoDetectCookieSupport=1 .
Popkin, B. M. (2001). The nutrition transition and obesity in the developing world. The Journal ofNutrition, 131(3), 871S873S.
Popkin, B. M., Du, S., Zhai, F., & Zhang, B. (2010). Cohort Prole: The China Health and Nutrition Surveymonitoring and understanding socio-economic and health change in China, 19892011. International Journal of Epidemiology, 39(6), 14351440.
Popkin, B. M., & Gordon-Larsen, P. (2004). The nutrition transition: worldwide obesity dynamics and their determinants. International Journal of Obesity, 28, S2S9.
Qin, L., Stolk, R. P., & Corpeleijn, E. (2012). Motorized transportation, social status, and adiposity: The
China Health and Nutrition Survey. American Journal of Preventive Medicine, 43(1), 110. Robinson, J. C. (1985). Of women and washing machines: Employment, housework, and the reproduction of motherhood in socialist China. The China Quarterly, 101, 3257.
Roskam, A.-J. R., Kunst, A. E., Van Oyen, H., Demarest, S., Klumbiene, J., Regidor, E., et al. (2010). Comparative appraisal of educational inequalities in overweight and obesity among adults in 19 European countries. International Journal of Epidemiology, 39(2), 392404.
Rzehak, P., & Heinrich, J. (2006). Development of relative weight, overweight and obesity from childhood to young adulthood. A longitudinal analysis of individual change of height and weight. European Journal of Epidemiology, 21(9), 661672.
Schwid, S. R., Hirvonen, M. D., & Keesey, R. E. (1992). Nicotine effects on body weight: a regulatory perspective. The American Journal of Clinical Nutrition, 55(4), 878884.
Shehan, C. L., & Moras, A. B. (2006). Deconstructing laundry: Gendered technologies and the reluctant redesign of household labor. Michigan Family Review, 11(1), 3954.
Sobal, J., & Stunkard, A. J. (1989). Socioeconomic status and obesity: A review of the literature.Psychological Bulletin, 105(2), 260.
Sugiyama, T., Healy, G. N., Dunstan, D. W., Salmon, J., & Owen, N. (2008). Is television viewing time a marker of a broader pattern of sedentary behavior? Annals of Behavioral Medicine, 35(2), 245250. Television Bureau of Advertising (2010). Local media marketing solutions. Association of Americas broadcast television industry.
Vandelanotte, C., Sugiyama, T., Gardiner, P., & Owen, N. (2009). Associations of leisure-time internet and computer use with overweight and obesity, physical activity and sedentary behaviors: Cross-sectional study. Journal of Medical Internet Research, 11(3), e28. doi:http://dx.doi.org/10.2196/jmir.1084
Web End =10.2196/jmir.1084 .
Wang, Y., & Beydoun, M. A. (2007). The obesity epidemic in the United Statesgender, age, socioeconomic, racial/ethnic, and geographic characteristics: A systematic review and meta-regression analysis. Epidemiologic Reviews, 29(1), 628.
Wang, H., Du, S., Zhai, F., & Popkin, B. M. (2006). Trends in the distribution of body mass index among
Chinese adults, aged 2045 years (19892000). International Journal of Obesity, 31(2), 272278. Wang, Y., Mi, J., Shan, X. Y., Wang, Q. J., & Ge, K. Y. (2007). Is China facing an obesity epidemic and the consequences? The trends in obesity and chronic disease in China. International Journal of Obesity, 31(1), 177188.
Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data (2nd ed.). Cambridge,MA: MIT Press.
World Bank (2014). Infrastructure (newspapers and television). Retrieved March 2, 2014, from WorldDataBank: World Development Indicators database.
Wu, Y. (2006). Overweight and obesity in China: the once lean giant has a weight problem that is increasing rapidly. British Medical Journal, 333(7564), 362.
Xu, X., Mellor, D., Kiehne, M., Ricciardelli, L. A., McCabe, M. P., & Xu, Y. (2010). Body dissatisfaction, engagement in body change behaviors and sociocultural inuences on body image among Chinese adolescents. Body Image, 7(2), 156164.
Zhang, J., Seo, D.-C., Kolbe, L., Middlestadt, S., & Zhao, W. (2012). Associated trends in sedentary behavior and BMI among Chinese school children and adolescents in seven diverse Chinese provinces. International Journal of Behavioral Medicine, 19(3), 342350.
123
Springer Science+Business Media Dordrecht 2015