Home & Garden

Access to variety contributes to dietary diversity in China

Access to variety contributes to dietary diversity in China
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  Access to variety contributes to dietary diversity in China  Jing Liu a, ⇑ , Gerald E. Shively a,b,1 , James K. Binkley a,2 a Department of Agricultural Economics, Purdue University, 403 West State St., West Lafayette, IN 47907-2056, United States b School of Economics and Business, Norwegian University of Life Sciences, Ås, Norway a r t i c l e i n f o  Article history: Received 3 July 2013Received in revised form 15 August 2014Accepted 8 September 2014 Keywords: Dietary diversityRural–urban differenceVariety access cost a b s t r a c t In the canonical consumer demand problem, an agent makes a decision about quantities to consume,under the assumption that all varieties can be accessed at zero cost. In reality, the cost of accessing vari-ety may not be zero. In this paper we study the effect of variety access cost on the consumption of foodvariety and its role in explaining regional differences in dietary diversity in China. We find that a highercost of access negatively affects the individual’s ability to diversify her diet in terms of both the totalcounts and the balancing of varieties consumed. The primary policy implication of this research is thatattempts to create a healthy food environment in China must be differentiated along rural and urbanlines. In rural communities where consumers have been limited in their ability to diversify food basketsby high electricity and transportation costs, infrastructure development and modernization may effec-tively improve nutritional balance. For more urbanized communities where the cost of consuming addi-tional food variety is relatively low, food policies might instead focus on interventions that promotehealthy eating to mitigate the burden of over-nutrition.   2014 Elsevier Ltd. All rights reserved. Introduction Like many developing countries, China is experiencing a nutri-tion transition, in which income growth and urbanization begintoshiftdietsawayfromcoarsegrainsandlegumestowardsgreaterconsumption of sugar, edible oil, and animal protein (Popkin,2014). This shift raises the possibility that China is, or will soonbefacingadoublediseaseburdenresultingfrombothunder-nutri-tion among the poor and over-nutrition among the non-poor. Thelatter is of special concern given the rapid rise over the past twodecades in the rates of those in China who are overweight(Gordon-Larsen et al., 2014). Since increases in overweight preva-lence will continue to outpace reductions in underweight preva-lence (Dearth-Wesley et al., 2007), the focus of food policy inChina must evolve from food security to health-related consider-ations. As aresult, nutritionbalanceanddiet diversitywill becomemorerelevantdimensionsofhealthydietsinChina,comparedwithcalorie and nutrition adequacy. It is generally thought that anincrease in variety brings about nutritional improvements, a viewsupported by research (e.g. Kant et al., 1993; Lo et al., 2013), but improvements are not guaranteed. Outcomes depend on thechoices consumers make, which in turn depend on the choicesavailable to them.Studies of dietary diversity among Chinese consumers aresomewhat rare. Exceptions include Kim et al. (2003) who con-struct a Diet Quality Index-International (DQI-I) to assess dietquality (including variety) but do not account for regional distinc-tions, and Li et al. (2009) who compare rural–urban DQI-I butfocus on differences among families with and without youth. Liuet al. (2012) investigate urban–rural nutritional disparities, butdo not include all categories of food and study the status of children only. Moreover, most of these studies share a commoninterest in the role of particular household socio-economiccharacteristics such as income, household size, age, sex composi-tion, employment status, education level of the household head,but do not directly explore the idea that the cost of access to vari-ety and the technology-related cost of seeking and accessing addi-tional variety may be determined by the setting in whichconsumers make choices. Understanding this is important, sincevariety due to greater access may have different nutritional andpolicy implications than variety brought about by increases inincome or other demand-side factors. This is a key aspect of thedebate in high income countries, where some argue that taxesand restrictions on the food environment are required to improvenutrition (Miljkovic et al., 2008; Kuchler et al., 2005) while others emphasize the importance of consumer education (Freeland-Graves and Nitzke, 2002). http://dx.doi.org/10.1016/j.foodpol.2014.09.0070306-9192/   2014 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. Tel.: +1 765 494 4301. E-mail addresses:  liu207@purdue.edu (J. Liu), shivelyg@purdue.edu (G.E. Shively),  jbinkley@purdue.edu (J.K. Binkley). 1 Tel.: +1 765 494 4218. 2 Tel.: +1 765 494 4261. Food Policy 49 (2014) 323–331 Contents lists available at ScienceDirect Food Policy journal homepage: www.elsevier.com/locate/foodpol  Understanding variety in food consumption is also of economicimportance. A diversified diet improves a consumer’s welfare bothbecause greater variety increases the likelihoodof matching a con-sumer’spreferenceswithproduct characteristics andbecausevari-ety counteracts diminishing returns to quantity (Li, 2013). Furthermore, understanding diversification in food demand canbeparticularlyimportanttounderstandingtheevolutionandstruc-tureoffoodmarketingsystems(Reardonetal.,2003;Pingali,2007). In this study we examine regional differences in dietary diver-sity in China and offer explanations for these differences. Herediversity is measured by both the total number of varieties con-sumedand the consumption balance in terms of quantity and foodcategories. 3 It is not our objective to measure the relationshipbetween dietary diversity and nutritional outcomes. Instead, weare interestedin the question‘‘What couldhave possiblycontributedto the evolution in dietary diversity among Chinese consumers?’’When approaching this question, we place a special emphasis onthe cost of accessing variety in urban/rural settings and on what thiscost implies for strategic attempts to promote healthy diets amongChinese consumers. Our contribution is primarily empirical. Ourresults suggest that a higher cost of access, driven by lower availabil-ity, higher transaction costs of seeking and obtaining variety, andlimited access to improved food storage, adds a marginal cost tothe market price of variety and creates friction in the individual’sability to diversify her diet. We also find that this friction is some-what greater in rural areas, resulting in less diversity. Consumingless variety may or may not imply lower nutrition, but as long asnutritional differences are associated with differences in diversity,an implication of the result is that the same food and nutrition pol-icies many not apply equally to rural and urban China. Conceptual framework  Traditional modeling of consumer choice, based on strict con-vexity, implies an inherent preference for variety, as reflected bytheindifferencecontoursthatareasymptotictotheaxesandbowedtowardtheorigin(Lancaster,1990).Itassumesthatconsumerscan accessallavailablevarietiesatnocostbeyondthepriceofthegood.Maximumutilityisachievedbyadjustingthequantityofeachgood,rather than the composition of varieties. Despite the popularity of thisapproach,however,theassumptionofcostlessandinfinitevari-ety is at odds with evidence (Bell et al., 1998). Because consuming more variety is constrained by limits on resources (information,time and monetary) and the marginal utility of variety diminishes,the consumer ends up with some but not all varieties.In an attempt to explain the consumption of variety, Jackson(1984) formally introduced the concept of hierarchical demand.He demonstrates how only a subset of all goods available is actu-ally consumed and how the variety of goods purchased increaseswith income. In the context of food consumption, a series of coun-try-specific studies confirm Jackson’s finding, and further identifywhich socio-economic characteristics determine preference fordietary diversity in addition to food expenditure (Thiele andWeiss, 2003; Torheim et al., 2004; Thorne-Lyman et al., 2010;Rashid et al., 2011; Bhagowalia et al., 2012; Drescher andGoddard, 2011; Shimokawa, 2013). In a recent paper, Li (2013)carefully derives the optimal choice of variety that equates mar-ginal benefit and marginal cost of consuming variety. He showsthat consumption of variety is positively correlated with expendi-ture but negatively correlated with variety accessing cost.A second branch of the literature focuses on why consumersmight exhibit a taste for variety. Consumers may seek varietybecause of intrinsic stimuli, such as a self-generated desire forchange, as in McAlister and Pessemier (1982), or because of inher-entuncertaintyregardingcurrentorfuturepreferences,asinWalsh(1995). On the other hand, extrinsic stimuli such as environmentalchange, promotion, word-of-mouth and external constraints con-tribute to patterns in which consumers may be willing to try newthings (Howard and Sheth, 1969; McAlister and Pessemier, 1982). Thedegreeofvariety-seekingcanalsobeassociatedwithcharacter-istics of goods (Adamowicz and Swait, 2013). Goods that are char- acterized by less concentrated market share distribution, morefrequentreplenishmentratesandlowerunitprices(implyingsmal-ler consequences of misjudgments) are more likely to expand theset of available and revealed choices (Adamowicz and Swait,2013). In addition, how variety is purchased – the ‘‘variety cycle’’– matters. Those who make fewer trips to satisfy their demandfor variety than to complete their quantity demand are less vari-ety-seekingthanthosewhocontemplatetwodemandsatthesametime (Berne and Mugica, 2010). Based on these theories, we hypothesize that: (1) wealthierpeople consume more varieties than their poorer counterparts,other things being equal; (2) consumer characteristics affect pref-erence for variety; (3) high cost of access, caused either by a lowdegree of modernization or by the household’s low productivity,generates friction in a consumer’s ability to consume variety. Wenow turn to an empirical investigation of these conjectures inthe context of our sample. Empirical strategy  Measurement of dietary diversity There are several ways that dietary variety can be measured. Inthe nutritional literature, count measures are frequently applied(Kant, 1996), whereby the number of consumed food items and food groups is recorded. Some well-known indices measuring die-tarydiversityandoveralldietqualitybasedonthismethodaredie-tary diversity score (DDS) developed by Kant et al. (1993), dietaryvariety score (DVS) by Drewnowski et al. (1997), Healthy EatingIndex (HEI) by Kennedy et al. (1995), Diet Quality Index (DQI) byPatterson et al. (1994), Diet Quality Index (DQI)-Revised byHaines et al. (1999), DQI-China by Stookey et al. (2000), and DQI- International by Kim et al. (2003). These indices, although handy for interpretation, have an important disadvantage: a minor foodcountsasmuchasanimportantcomponentoftheindividual’sdiet.The economic literature, however, tends to measure variety notonly by the number of foods but also by their distribution – for agiven number of foods, diversity increases as their shares of thediet are more evenly distributed. The most often applied measuresare Entropy ( E ), the Simpson Index ( SI  ) and the Cumulative Share( CS  ) (Lee and Brown, 1989; Theil and Finke, 1983; Jekanowskiand Binkley, 2000). The basic idea behind each of these measure-mentsisthatmaximumdiversityoccurswhenconsumptionsharesare equally distributed among varieties. Entropy is defined as afunction of the consumption share  w i : E  ¼ X ni ¼ 1 w  i log 1 w  i   ;  ð 1 Þ where high diet diversity corresponds to a large index value of   E . Amaximumof log n  is reached when consumption is evenly distribu-tionacross all varieties.  SI   is computed as one minus the Herfindahlindex, a commonly used measure for market concentration: 3 We use standard food categories that are linked to nutrition. In the srcinal dietdiary from which our data are derived, each food item is assigned a unique six-digitfood code. The first two digits indicate food categories. The four subsequent digitsindicate subgroups within the category. Among all twenty-one categories, the firsttwelve are major foods (cereals, tubers, beans, vegetables, fungi and algae, fruits, nuts,meat, poultry, dairy, egg and seafood). The remainder are minor foods (includingdesserts, snacks, spices, beverages, sauces and candy). 324  J. Liu et al./Food Policy 49 (2014) 323–331  SI  ¼ 1  X ni ¼ 1 w  2 i  :  ð 2 Þ SI   variesfromzero(whenasingleitemisconsumed)toamaximumof 1  1 n  (when all shares equal  1 n ).  CS   provides an indirect indicatorof diversity derived from the shape of the cumulative distributionofconsumptionshares.Ifconsumptionissufficientlydiverse,sharesfollowa uniformdistribution. At the other extreme, if consumptionis highly concentrated, the cumulative distribution of consumptionshares, based on a ranking of shares from highest to lowest, ishighly skewed to the left. In this study, we use different measuresin the empirical analysis as a robustness check. Data Our data come from the China Health and Nutrition Survey(CHNS), an ongoing open cohort study designed to measure howthesocialandeconomictransformationofChinesesocietyisaffect-ing the health and nutritional status of its population. The studypopulation is drawn from nine provinces of China. 4 A multistage,random cluster process is used to draw the sample in each province. 5 Eight survey waves were conducted over the period 1989–2009. 6 Duringeachwave, nutritionintakedata werecollectedvia threecon-secutive 24-h dietary recalls. For this analysis, each food code is trea-ted as an individual food variety. The unit of analysis is an adulthousehold member. 7 All variables are described in Table 1. Theseconsist of demographics, as have been used in previous studies, andvariables measuring access costs, which most previous studies havenot considered. Note that refrigerator ownership is much lower inrural areas, likely due to lower availability and higher cost of a reli-able power supply. Rural consumers also need to travel a longerdistance to reach the nearest supermarket. These suggest a highercost of accessing additional variety associated with rural areas. We first use diet data observed for individuals in 1989 (at thestartoftheCHNSsurvey)and2009(themostrecentroundavailableto us) to illustrate the overall pattern of changing consumption attheindividuallevel.Fig.1displaysthebivariatekerneldensitycon-tours of the number of items consumedper personper day for thispanel of individuals, separated into urban and rural samples. Inbothplots,themajorityofobservationslieabovethe45degreeline,indicatingthatmostpeoplewereconsumingmoreitemsperdayin2009 than in 1989. In addition, the tighter density of contour linesin the rural sample corresponds to the steeper slope of the ruraldensityfunction,suggestingthatdietarydiversityismorehomoge-neous and has been evolving more slowly in rural areas.Next we use cross-sectional CHNS data from 2006 to comparethe composition of rural and urban diets. Fig. 2 plots the distribu-tion of food consumption, organized by the USDA food pyramidcategories. Compared to urban diets, rural diets are more concen-trated on grains and vegetables, and exhibit less consumption of fruit, meat and dairy products. These latter commodities are pre-dominantly subject to spoilage and characterized by higher prices,higherstoragecosts, orboth. Toobtainaquantitativeevaluationof the degree of dietary diversity, we measure dietary diversity usingthe three methods outlined above in Section ‘Measurement of die-tary diversity’. The share  w i  is calculated in two different ways,basedonweight 8 and based on category. 9 We use the shares to com-pute the overall average level of three indices –  E  ,  SI   and  CS  10 – forthe rural and urban sub-samples, respectively. These index valuesare reported in Table 2, where a larger number indicates a higherlevel of diversity. Urban diets exhibit greater diversity, regardlessof the metric used.  Table 1 Summary statistics (2006). Urban RuralMean Std Min Max Mean Std Min MaxNumber of item consumed/day 10.413 2.699 1.667 19.667 9.019 2.352 3.333 19.333Age 49.383 15.835 18 95 48.679 14.719 18 96Female 0.528 0.499 0 1 0.524 0.499 0 1High school 0.341 0.474 0 1 0.157 0.364 0 1Married 0.806 0.396 0 1 0.858 0.349 0 1Income (1000 CNY) 33.944 33.869 0.005 251.093 24.993 30.495 0.103 460.617Income squared (1000 CNY) 2298.760 5973.742 0 63047.523 1554.268 8366.334 0.011 212167.594Household size 3.444 1.382 1 10 3.878 1.712 1 14Refrigerator 0.694 0.461 0 1 0.396 0.489 0 1Transportation 0.699 0.459 0 1 0.784 0.412 0 1Population density (1000/km 2 ) 4.836 6.208 0.001 24.213 1.388 2.074 0.016 14.317Restaurants (per 1000 ppl.) 3.932 5.834 0 30.075 5.028 7.902 0 48.157Bus stop 0.811 0.392 0 1 0.662 0.473 0 1Distance to market (km) 0.776 1.238 0 6 10.372 12.863 0 60.5Weekend 0.562 0.496 0 1 0.570 0.495 0 1Survey wave 1 0.082 0.275 0 1 0.065 0.246 0 1Survey wave 2 0.450 0.498 0 1 0.393 0.489 0 1Survey wave 3 0.387 0.487 0 1 0.377 0.485 0 1Home farming 0.228 0.420 0 1 0.691 0.462 0 1 4 They are Guangxi, Guizhou, Heilongjiang, Henan, Hubei, Hunan, Jiangsu, Liaoning,and Shandong. 5 Counties in the nine provinces were stratified by income (low, middle, and high).A weighted sampling scheme was used to randomly select four counties from eachprovince. Villages and townships within counties and urban/suburban neighborhoodswithin the cities were selected randomly. See http://www.cpc.unc.edu/projects/china/about/design/survey for details of the survey design and sampling methods. 6 Survey years were 1989, 1991, 1993, 1997, 2000, 2004, 2006 and 2009. 7 A household may be represented multiple times in the dataset depending on thenumber of adults residing in the household. Table 1 reports the mean, minimum andmaximum household size. Our regression results are invariant in sign, magnitude andsignificance to whether we conduct the analysis using all household members or arandom draw of one member from each household. We report results for the fullsample. 8 For example, if two items weighing exactly the same amount were consumed, theshare will be 1/2 each. 9 To make the grouping consistent with the USDA nutrition guidance, we furtheraggregate the srcinal twelve major food categories in footnote 3 into six broadcategories – grains, vegetables, fruits, meat/poultry/seafood, dairy, beans/eggs/nuts.The seventh category is created to encompass all the other minor food mentionedabove. This group accounts for only a very small portion of the total foodconsumption. To calculate the share by category, for example, if the consumerconsumes one item from each USDA category, the category share will be 1/7 each. 10 We first rank the two types of shares in descending order, and compute the sumof shares at a level representing approximately 75% of total consumption. We thencount the number of items that contribute to this cumulative share of 75%. A largernumber of counts indicates a higher degree of diversity.  J. Liu et al./Food Policy 49 (2014) 323–331  325  Method Fig. 3 provides a longitudinal view of dietary diversity in ruraland urban China and its interaction with income based on theChinaHealthandNutritionSurvey(CHNS).Clearly,urbanresidentson average always reported consuming a more diverse diet thantheir rural cohorts. Over time, daily consumption exhibits increas-ing diversity in both urban and rural areas. This is consistent withthe evidence reported by Hovhannisyan and Gould (2014), whoargue that Chinese food preferences have undergone a structuralchange. These patterns coincide with the observation that house-holdincomegraduallyincreasedover thepast twodecades inbothregions.Interestingly,ifweregressanindicatorofeachindividual’sdietary variety on household income, while controlling for timeperiod, location (urban/rural) and their interaction, we find thatthe variety gap between urban and rural diets has grown largersince the 1990s. This suggests that outcomes are being driven byforces other than income.We parameterize the empirical model as follows to test theextent to which variety is correlated with income and cost of access, while controlling for potential confounders. V   ¼ c 0 þ Y  c 1 þ Y  2 c 2 þ  A c 3 þ  X  c 4 þ U c 5 þ u :  ð 3 Þ Inthiscross-sectionalstudy,thedependentvariable V   isameasure-ment of variety. To measure variety, we use both the count of vari-eties and indices ( E ,  SI  , and  CS  ) derived from individualconsumptions.  Y   represents observed annual household income inthousand CNY.  A  is the cost of accessing variety, which we assumeto be a linear function of the distance to the nearest market, com-munity population density, the number of restaurants per capitain a community, and a set of dichotomous variables indicating thepresence of a bus stop in the neighborhood, refrigerator ownership,and household transportation ownership.  X   is a set of demographicvariables including covariates that specify the timing of the survey.Although the data used in the study do not constitute a fully ran-dom sample, survey weights are not provided. We attempted topartially correct for this by including indicator variables for the‘‘waves’’. If the sampling is random, or at least of similar structureacross waves, the expected coefficients are zeroes. Otherwise, theycorrectfordifferences,atleastdifferencesinmean. U isasetofbin-ary indicators for province of household residence.  u  is the errorterm, assumed to be normally distributed.It is possible that some of these variables are jointly deter-mined, which could have implications for statistical inference inour regressions. For example, restaurants may locate where theybelieve demand for variety is high (although restaurants typicallyuse population as a measure of potential demand, and would havedifficultymeasuringdemandfor varietyper se). More troublesomeis that a consumer may purchase a refrigerator from a desire toincrease variety, which could lead to an overestimate of the effectof refrigerator ownership in our regressions. However, refrigeratorownershipislikelydeterminedbyincome,whichweincludeinthemodel as a control, and by electricity availability, which is anaspect of access. Nevertheless, we acknowledgethat some endoge-neity may be present in our regressions. Unfortunately, we areworking with a limited set of variables, making the selection orconstruction of adequate instrumental variables difficult, if notimpossible. Because the use of weak or invalid instruments canlead to results that are even more biased than standard OLS, andbecause weak instruments do not overcome the potential bias of OLS and can also mislead one regarding the size of standard errors(see Murray, 2006), we avoid an IV approach. Given that our focusis not on the causal effects of specific factors but rather overall dif-ferencesbetweenrural andurbanpopulations, webelieveitis bet-ter to adhere to a robust method like OLS than to attempt to Fig. 1.  Bivariate kernel density contours of daily items consumed per personoverlaid with scatter plot of data, for urban (top) and rural (bottom) samples. Eachdot represents one individual in the sample. Observations above the 45 degree lineindicate consuming more varieties in 2009 than in 1989.  Table 2 Alternative measures of dietary diversity (2006). Region Share by Entropy Simpson Cumulative shareRural Weight 2.00 0.81 6.87Category 1.22 0.66 1.77Urban Weight 2.33 0.86 10.45Category 1.38 0.71 2.28326  J. Liu et al./Food Policy 49 (2014) 323–331  Fig. 2.  Distribution of consumption by food category (based on 2006 sample). Percentage indicates the share of each food category in total consumption (in terms of itemcounts). Fig. 3.  Trends in dietary diversity and household income, 1989–2009. Left: fractional-polynomial prediction plot of average number of items consumed per person per day,with 95 percent confidence intervals; Middle: average annual household income in CNY; Right: rural–urban gap in number of items consumed per person per day, whilecontrolling for time period, location (urban/rural) and their interaction.  J. Liu et al./Food Policy 49 (2014) 323–331  327
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!