Economcs Presentatons, Posters and Proceedngs Economcs 2008 The Rse of Obesty n Transton Economes: Theory and Evdence from the Russan Longtudnal Montorng Survey Sonya K. Huffman Iowa State Unversty, skostova@astate.edu Maran Rzov Mddlesex Unversty Busness School Follow ths and addtonal works at: http://lb.dr.astate.edu/econ_las_conf Part of the Agrcultural and Resource Economcs Commons, Eastern European Studes Commons, and the Health Economcs Commons Recommended Ctaton Huffman, Sonya K. and Rzov, Maran, "The Rse of Obesty n Transton Economes: Theory and Evdence from the Russan Longtudnal Montorng Survey" (2008). Economcs Presentatons, Posters and Proceedngs. 14. http://lb.dr.astate.edu/econ_las_conf/14 Ths Presentaton s brought to you for free and open access by the Economcs at Iowa State Unversty Dgtal Repostory. It has been accepted for ncluson n Economcs Presentatons, Posters and Proceedngs by an authorzed admnstrator of Iowa State Unversty Dgtal Repostory. For more nformaton, please contact dgrep@astate.edu.
The Rse of Obesty n Transton Economes: Theory and Evdence from the Russan Longtudnal Montorng Survey 1 Sonya K. Huffman and Maran Rzov Iowa State Unversty and Mddlesex Unversty Busness School (London) Selected Paper prepared for presentaton at the Amercan Agrcultural Economcs Assocaton Annual Meetng, Orlando, FL, July 27 29, 2008. Copyrght 2008 by Sonya K. Huffman and Maran Rzov. All rghts reserved. Readers may make verbatm copes of ths document for non commercal purposes by any means, provded that ths copyrght notce appears on all such copes. 1 We acknowledge helpful comments and suggestons from Wally Huffman, Alson Burrell, partcpants at the Economcs Semnar at Wagenngen Unversty, The Netherlands. We would lke to thanks the NC Populaton Center for provdng the data.
The Rse of Obesty n Transton Economes: Theory and Evdence from the Russan Longtudnal Montorng Survey Abstract Ths study ntegrates theoretcal and emprcal models to facltate understandng of human obesty and the factors contrbutng to rsng obesty n Russa durng the transton from a planned to a market economy. Recent ndvdual level data from the Russan Longtudnal Montorng Survey for 1994 and 2004 show that det/calorc ntake, smokng, gender and educaton are mportant determnants of obesty n Russa. Emprcal results strongly support our model for producton of BMI and demand for nputs n the BMI producton functon. The analyss provdes nformaton on the lnk between detary patterns and other factors of obesty n Russa whch s mportant for formulaton, mplementaton and montorng of effectve polces desgned to mprove overall nutrtonal wellbeng and reduce obesty and mortalty of the Russan populaton. Interventons, whch enhance educaton toward healthy lfestyles and healthy det, could play a vtal role n preventng obesty n Russa. Introducton Snce the early 1990 s seres of reforms have been mplemented n transton economes. Economc reforms amng at ncrease n effcency comprsed prce lberalzaton, prvatzaton and enterprse restructurng. However, the reforms also brought dramatc changes n all areas of the populaton s lfe. Important sde effects were ncrease n unemployment and poverty, addtonal stress and uncertanty, rsng crme, and fall n lvng standards, for certan groups of the populaton. As a result, the populaton n transton economes experenced dramatc changes n lfestyle and a sgnfcant declne n lfe expectancy. The adverse effects of transton were most severe n the Former Sovet Unon. Several studes examne the reasons for the mortalty crss n Russa and other former Sovet republcs (Branerd and Cutler, 2005; Cockerham, 2000; Shkolnkov et al., 2004). Brenerd and Cutler show that durng the 1990s greater alcohol consumpton was an mportant 1
determnant of hgher mortalty rates n Russa. The ncreased stress from the transton to a market economy had dramatcally affected the lfestyle and the det of the populaton as well. Furthermore, the authors fnd that across households rsng human obesty has mportant health consequences and s a sgnfcant predctor of mortalty, however, the magntude of the effect s small. Stllman (2006) revews the lterature examnng health outcomes n Eastern Europe and the Former Sovet Unon durng the transton perod and also ponts out to the lnk between obesty and health outcomes. Huffman and Rzov (2007) dscuss the determnants of obesty n Russa, whle Oglobln and Brock (2003) nvestgate the rsk factors and the economcs of the decson to smoke n Russa and fnd that obesty s negatvely assocated wth smokng. Obesty has reached epdemc proportons globally, wth more than 1 bllon adults overweght, and at least 300 mllon of them clncally obese (WHO, 2006). Obesty has become a major contrbutor to the global burden of chronc dseases and dsablty. The health consequences range from ncreased rsk of premature death to serous chronc condtons that reduce the overall qualty of lfe. The emergng and transton economes, ncludng Russa, had the hghest number of dabetcs a condton closely assocated wth obesty n 1995 (WHO, 2006). Therefore, a greater understandng of the rse n obesty and ts causes n transton economes could lead to mportant polcy recommendatons for reducng the problem and mprovng the health of the populaton. The rsk of obesty s strongly nfluenced by det and lfestyle whch have been changng dramatcally as a result of economc and nutrtonal transtons. However, very few studes have examned the causes of obesty n transton economes n contrast to the large lterature on hgh ncome countres such as the USA (Chou et al., 2004; Huffman et al., 2006; Komlos 2
and Baur, 2004; Lakdawalla et al., 2005; Lakdawalla and Phlpson, 2002; Rashad, 2006; Rashad et al., 2006). Mendez and Popkn (2004) fnd that the populaton of low ncome countres has also become susceptble to obesty n the process of economc development. An nterestng study by Lefert (2004) examnes food securty n Russa and ponts that a serous health problem s overweght and obesty whch have ncreased durng transton and currently affect over half of the adult populaton. Zohoor et al. (1998) fnd that the prevalence of obesty, as well as the alcohol consumpton, has rsen sgnfcantly n Russa durng 1992 1996. The goal of ths paper s to develop theoretcal and emprcal models to examne human obesty and ts determnants n Russa, the largest transton economy. The analyss contrbutes to better understandng of how ratonale economc agents respond to external factors that are expected to nfluence food consumpton and weght. Russa s also one of the transton economes facng the most severe obesty and general health problems. Obesty has ncreased durng transton, rsng from 20.3 percent of the populaton n 1994 to 28.0 percent n 2004 a 38 percent ncrease, based on our weghted 1 sample. Indvdual and household level data from the Russan Longtudnal Montorng Survey (RLMS) for 1994 and 2004 s employed to study the factors contrbutng to rsng obesty n the framework of the productve household models and to emprcally test our hypotheses. The models are estmated for the pooled sample and for subsamples by gender and resdence. The analyss s stratfed by urban/rural resdence because the rural households often undertake extensve home producton whch has a large mpact on food avalablty and presumably nutrtonal outcomes. Emprcal results strongly support our models for producton of BMI and demand for nputs n the BMI producton functon. The rest of the paper s organzed as follows. 3
Frst, a conceptual framework s developed, based on health productve household models. Next, the data and econometrc technques are descrbed, followed by a dscusson of the estmaton results. Fnally, conclusons and polcy recommendatons are offered. Theoretcal model The productve household models of health developed by Rosenzweg and Schultz (1982) and Grossman (2000), and the agrcultural household models developed by Huffman (1991) provde a useful framework for analyzng overweght and obesty. An mportant proposton n the framework s that the health status of each household member s determned by the degree of overweght and obesty of that member. The ndvdual has a utlty functon U = U ( D, C, BMI, L ; O ). (1) Utlty s determned by consumpton of food (det), D; consumpton of other goods (excludng food) and servces (other than health nputs), C; body mass ndex, BMI, and lesure, L. Furthermore, utlty s affected by a vector O of fxed (observable) characterstcs, such as educaton, age, gender, place of resdence, etc. The ndvdual has a BMI producton functon BMI = B ( D, L, O, μ ), (2) where μ s the unobservable ndvdual characterstcs that affect the ndvdual s health; such characterstcs may nclude genetc factors. Food consumpton/det affects utlty drectly and ndrectly through BMI producton, provdng energy, proten, vtamns and mnerals. We assume that the ndvdual has a budget constrant P D + P C = W ( T L ) N, (3) D C + 4
where P D and P C denote the prces of food (D), and other goods and servces (C), respectvely. Further, W s the wage rate per unt of tme, T s the fxed tme endowment (T L=work), and N s the nonlabor ncome. For an nteror soluton of the model, substtute equaton (2) nto (1) and use the budget constrant (3). Then, the ndvdual chooses D, L and C by maxmzng hs/her utlty subject to the budget constrant. The utlty maxmzaton problem can be wrtten as L = U [ D, C, B ( D, L ; O, μ ), L ; O ] + λ ( WT + N P D P C WL ), (4) where λ s the Lagrange multpler representng the margnal utlty of ndvdual s full ncome. The frst order condtons for an optmal soluton are: D C U U B B B B D L + U = λ P, (5) D D + U = λ W, (6) L U C = λ P, (7) C WT + N = P D + P C WL, (8) D C + where U = U / B, B = B / D, U = U / D, B = B / L, = U / L and B D D L U L U C = U / C. For an nteror soluton, equatons (5) (8) yeld the ndvdual s optmal demand functons for D, L and C: * Φ = f Φ ( P, P, W, N, O, μ ), Φ = D, L, C. (9) D C Therefore, the demand for nputs nto the BMI producton functon depends on the prces of the purchased nputs (P D, P C ), the wage rate (W), non labor ncome (N), fxed factors (O) and unobserved factors (μ), whch are assumed to have zero expected mean. After substtutng the optmal demand functons D * and L * from equaton (9) nto the BMI producton functon (2), we obtan the ndvdual s BMI supply functon: 2 5
* BMI = S ( P, P, W, N, O, μ ). (10) B D C The ndvdual s BMI producton functon (equaton 2) s a technology relatonshp, whle the BMI supply functon (equaton 10) s the behavor relatonshp based on the optmal ndvdual s decsons. Data and econometrc specfcaton Data from the Russan Longtudnal Montorng Survey (RLMS) for 1994 and 2004 s employed to nvestgate the factors contrbutng to the rsng obesty n Russa. The RLMS s a natonally representatve household survey that annually samples the populaton of dwellng unts. 3 The RLMS s desgned to montor the effects of Russan reforms on the health and welfare of the Russan ndvduals and households and s coordnated by the Carolna Populaton Center at the Unversty of North Carolna ( http://www.cpc.unc.edu/projects/rlms). The survey s based on mult stage probablty samples of the Russan populaton. The collected data nclude a wde range of nformaton concernng household characterstcs such as demographc composton, ncome, and expendtures. Data on ndvduals ncludes employment, anthropometrc measures, health status, nutrton, alcohol consumpton, and medcal problems. We use round 5 (1994) and round 13 (2004) of the RLMS. Table 1 presents the defntons, means and standard devatons for all varables used n the econometrc analyss for the weghted data. The varables are based on the estmated sample of 6,424 ndvduals (age 18 and over) and 3,710 households that reman after observatons wth mssng values are deleted. 4 The male and female samples consst of 2,532 and 3,892 ndvduals respectvely, whle 3,910 ndvduals resde n urban areas and 2,514 resde n rural areas. 6
Based on the theoretcal model, we add emprcal content by provdng emprcal defntons of the varables and mposng a specfc algebrac form of equatons (2) and (11). We focus our efforts on two dependent varables: ndvdual s weght and body mass ndex (BMI). A standard measure of obesty s based on the BMI ndvdual weght n klograms dvded by heght n meters squared (kg/m 2 ). Accordng to the WHO, an ndvdual wth a BMI over 25 kg/m 2 s defned as overweght, and wth a BMI of over 30 kg/m 2 as obese. However, the BMI may overestmate body fat n athletes who have a muscular buld, and may underestmate body fat n older people who have lost muscle mass (NIDDKD, 1996). Hence, we choose both an ndvdual s weght and BMI as measure of obesty. The BMI ndex s constructed for each respondent from data collected by traned personnel on weght and heght. The average ndvdual weght was 71.9kg n 1994 and t ncreased to 74.4kg n 2004, whle the average BMI was 26.2 n 1994 and ncreased to 27.4 n 2004. Tables 2a 2e present the descrptve characterstcs of the Russan populaton by categores of normal weght, overweght and obese based on the defntons dscussed above for the whole sample, and by gender and resdence for each round (1994 and 2004) and all rounds together. The overweght and obese are on average older (50 and 54 years respectvely) as compared to ndvduals wth normal weght (43 years), shorter (165 cm and 162 cm as compared to 167 cm), wth less educaton, smoke sgnfcantly less, drnk alcohol less than the ndvduals wth normal weght. The average total expendture for the overweght group s hgher (9972 rubles/month) whle for the obese group s lower (9238 rubles/month) than the normal weght ndvdual s expendture average (9686 rubles/month). Controllng for three levels of educaton the overweght and obese males are more lkely to have hgher educaton compared to ndvduals wth normal weght (15% and 18% as 7
compared to 12%). Whle the opposte s true for the overweght and obese females that are less lkely to have hgher educaton than normal weght females (17% and 11% as compared to 22%). Overweght men are more lkely to be employed (68%) compared to the obese (60%) and normal weght men (64%) but obese men are less lkely to have a job compared to the normal weght men. The obese males are earnng less than the overweght and normal weght males, but ther wage s much hgher than the female s wage. Females are earnng less than males. Nnety percent of the obese males are marred whle only 74 percent of the normal weght male are marred; 71 percent of the obese females are marred compared to 63 percent of the normal weght females. Normal weght females and males are more lkely to have kds. Overweght men eat more fat whle overweght females eat less fat. Furthermore, overweght and obese men lve n households wth slghtly hgher real expendture than the normal weght men. They are slghtly less lkely to drnk alcohol than the normal weght men (69 % versus 72%) and they are sgnfcantly less lkely to smoke (52% of overweght, 39% of obese as compared to 72% of normal weght ndvduals). Smlar characterstcs for the females, the urban and rural populaton pont that the overweght and obese females and ndvduals resdng n urban or rural areas, smoke and drnk less, and are less lkely to work. Our (weghted) data reveal that the share of the populaton that s overweght and obese has ncreased n Russa between 1994 and 2004. Overweght people accounted for 34.25 percent of the total n 1994 and for 36.07 percent n 2004. In 1994 the overweght rate of males was hgher than that of females 35.81 percent versus 33.16 percent. The ncrease n obesty was much more dramatc durng the transton perod, from 20.28 percent n 1994 to 28.00 percent n 2004. In 1994, the obesty rate was much hgher for females, at 27.82 percent, compared to only 9.49 percent for males. Based on our data, women are more lkely 8
to be obese n Russa, whch s a trend smlar to western countres. For women, the overweght rate ncreased slghtly from 33.16 to 34.59 percent, for men from 35.81 to 38.07 percent, for urban resdents from 34.83 to 35.40 percent, and for rural resdents from 34.21 to 36.75 percent. However, the ncreases n obesty rates were more sgnfcant n magntude, from 27.82 to 36.62 percent for women, from 9.49 to 16.34 percent for men, from 19.59 to 28.75 percent for urban resdents, and from 22.75 to 27.13 percent for rural resdents. The ncrease n obesty was much hgher for the males and for urban resdents durng the ten year perod. The overweght and obesty rates have ncreased for both genders and urban/rural resdents over 1994 to 2004. Therefore, t s mportant to dentfy and understand the factors that could have contrbuted to ths dramatc ncrease n obesty n Russa durng the economc transton. Followng our theoretcal model, we frst establsh the techncal relatonshp between BMI (and weght, n an alternatve specfcaton) and ts determnants, ncludng det as measured by calorc ntake and composton, and control for selected economc and socodemographc factors. The ndvdual s BMI producton functon (2) or the techncal relatonshp s specfed as ln BMI = γ 1 + γ 2 ln Calores + γ 3 Fat + γ 4 Pr oten + γ 5 Smo ker + γ 6 Male, (11) 2 + γ Educaton + γ Work + γ Age + γ ( Age ) + γ Year + ε 7 8 90 10 11 1 where subscrpt refers to an ndvdual. 5 Calores s the total ndvdual calores consumed per day; Fat s the share of fat ntake n the total calore ntake; Proten s the share of proten n the total calore ntake; Smoker s a dummy varable equal to one f the ndvdual smokes and zero otherwse; Male s a dummy varable equal to one f the ndvdual s male and zero otherwse (.e. female); Educaton s a set of dummy varables for three levels of educaton (basc, hgh, and hgher); Work s the labor force partcpaton (employment) ndcator equal 9
to one f the ndvdual works and zero otherwse; and Year s a dummy varable equal to one f the year s 2004 and 0 for year 1994. γ 1 γ 11 are parameters of the ndvdual s BMI producton functon to be estmated. The year dummy varable controls for changes over tme related to publc health and the organzaton of the health care system as transton progresses. The random dsturbance term ε represents the mpact of all other factors and has a zero mean. Larger calorc ntake, other thngs equal, s expected to lead to weght gan and eventually to obesty (γ 2 >0). Lkewse, an ncrease n fat n the det, beyond a certan threshold, s expected to accelerate obesty (γ 3 >0), and ncrease n proten possbly to accelerate obesty (γ 4 ><0). Smokers consume fewer calores than non smokers. Specfcally, cgarette smokng s assocated wth lower weght because smokng tends to ncrease metabolsm and suppress appette, thus havng a negatve effect on BMI weght (γ 5 <0). For other varables the a pror hypotheses are more complex and outcomes less clear; therefore, we do not state pror expectatons about the sgns of coeffcents for these varables. Next, we estmate demand equatons (eq. 9) for calores, meat and fsh, fruts and vegetables, and dary products. The emprcal specfcaton for these demand equatons s / h 2 2 ln Q I = α 1 + α 2 ln Expend / h + α 3 ln( Expend / h ) + α 4 Age + α 5 ( Age ) + α 6 Educaton 18, (12) + α Male + α Smo ker + α ln Heght + α ln Eqnum + α Year + α Re gon + ε 7 8 9 10 h 11 r = 12 r r 2 where /h Q I s the ndvdual s () demand for (I=) a) calores, or the household s (h) demand for (I=) b) meat and fsh, c) fruts and vegetables, and d) dary products. Expend s the total household expendtures, whch s used as a proxy for ncome. Heght s the ndvdual s heght measured n cm. The varable adult equvalent number (Eqnum) of household members s excluded from the ndvdual s demand for calores equaton 12a. Regonal 10
dummy varables are used to capture the dfferences n real prces. 6 In equatons 12 (b, c, and d) heght, gender, age, educaton, and smokng habts are proxed by the characterstcs of the household head who s assumed to be the man decson maker n the household. Equaton 12a s estmated by ordnary least squares (OLS), whle for equatons 12 (b, c, and d) we use the nterval regresson estmator n STATA to deal wth the censorng of dependent varables. The probablty of an ndvdual consumng alcohol (A ) s estmated as a functon of exogenous demographc and soco economc varables (O), ncludng Heght, Calores, Smoker, Male, Educaton, Expendture, Expendture 2, Age, Age 2, Year and Regon, usng a probt model: A * ' = δ O + ε 3 where A =1 f A * > 0 and 0 otherwse. (13) Alcohol consumpton can affect obesty through calorc ntake n our analyss, as well as t may ncrease mortalty caused by accdents, cardovascular dsease, etc. The ndvdual s BMI supply functon (10) s specfed as ln BMI = β 1 + β 2 Smo ker + β 3 Male + β 4 Educaton + β 5 ln Wage 15, (14) 2 + β Age + β ( Age ) + β Year + β Re gon + ε 6 7 8 r = 9 r r 4 where Wage s the real ndvdual hourly wage, whch s the mplct cost of tme, and Regon s a dummy varable for each of the eght regons of the country that represent, largely, the regonal dfferences n real food prces. The predcted wage s used as the opportunty wage for each ndvdual. Because observed wages are also endogenous to the choce to work, we control for ths self selecton bas by constructng the varable for the lnwage usng a Heckman sample selecton model. We use the Heckman sample selecton model to construct the predcted ndvdual wage as a functon of ndvdual characterstcs used n prevous regressons plus non labor ncome, martal status, number of chldren as well as the set of 11
regonal dummes and year. 7 Regonal fxed effects wll control for relatve prces of food and other omtted varables that dffer by regon. Studes on transton economes by Gardner and Brooks (1994) and Huffman and Johnson (2004) have found geographcal prce dfferences. Wage (earned ncome) s expected to have a postve effect on the supply of good health. We expect ndvdual BMI (or weght) to ncrease wth age, at least up to mddle age, but eventually the dgestve system starts to lose ts effcency and capacty and the effects of fnte lfe set n. Therefore, we expect β 6 >0 and β 7 <0. The econometrc specfcatons n equaton (11) and (14) are estmated by OLS, both for the whole sample, by gender (male/female) and resdence (urban/rural) subsamples. We also test for homogenety of the BMI producton and supply functons across genders (male/female) and across resdence (urban/rural) and expect to reject homogenety. Estmaton results We estmate the BMI producton functon, ndvdual demand functons for calores and alcohol, and the household demand functons for other food types, and then the BMI supply functon as specfed n equatons (11), (12), (13) and (14). The ndvdual s BMI producton functon (equaton 11), demand functon for calores (equaton 12a) and BMI supply functon (equaton 14) are ftted by OLS. 8 The household s demand functons for meat and fsh (equaton 12b), fruts and vegetables (equaton 12c) and dary products (equaton 12d) are estmated by nterval regresson to account for censorng of the dependent varable. The parameters of the alcohol consumpton equaton (13) are estmated usng the probt estmator. The equatons are ftted on a balanced panel for ndvduals and households from 1994 and 12
2004. We report the robust standard errors that have been corrected for ndvdual clusterng. 9 Table 3 presents the OLS estmates of the BMI producton functon where the dependent varable s the natural log of wegth. It s ftted on pooled data and separately by gender and resdence. We test for equalty across gender and across resdence and reject the homogenety. 10 We fnd strong econometrc evdence for the BMI producton functon for both females and males, and n urban and rural subsamples. Weght ncreases wth an ndvdual s heght. Age has a postve and sgnfcant effect on weght, but the age effect s dmnshng at hgher ages. Total calores consumed postvely and sgnfcantly affect male s weght and rural resdent s weght, whle the proten ntake leads to an ncrease n the weght of both men and women, but only of urban resdent s weght. A ten percent ncrease n calorc ntake ncreases ndvdual s weght by 0.6 percent for males and by 0.4 percent for rural resdents. An ncrease n food fat content, holdng proten content and calore consumpton constant, ncreases the weght for both females and males, and urban and rural resdents. Increasng proten content, holdng fat content and calores consumed constant, ncreases an ndvdual s weght by gender and only the urban resdents weght, as well. Smokng decreases sgnfcantly the male s weght by 7.6 percent and the female s weght by 2.7 percent. Havng a hgher level of educaton has a strong and sgnfcant negatve effect on woman's weght a decrease of 3.8 percent compared to the basc educaton category, and on urban resdent s weght a decrease of 2.5 percent, but there s no sgnfcant effect of male s or rural resdent s educaton on ther weght. An ndvdual beng employed has a statstcally sgnfcant postve effect on male s (a 1.7 percent ncrease) and urban resdent s (a 1.6 percent ncrease) weght. However, there s no a sgnfcant effect for women. Beng 13
employed could possbly ncrease the opportunty costs of off the job physcal exercse needed to mantan lower (optmal) weght. The estmated coeffcents for the year dummy are statstcally sgnfcant and postve for all subsamples, ndcatng that people are heaver n 2004 than 1994. Durng the transton, ndvdual weghts have ncreased by 1.2 percent, other thngs equal. Ths fndng suggests that the standard mechansm drven by technologcal change s also at work, as t s observed for other countres. Table 4 presents the OLS estmates of the BMI producton functon where the dependent varable s the natural log of BMI. It s ftted to data pooled over men and women, and separately by gender and by resdence. The results from the BMI equaton look smlar to those from the weght equaton. Age follows a nonlnear relatonshp wth both weght and BMI. BMI and obesty appear to rse wth age and then peak at 61 for females and 60 for males, thereafter lowerng agan for those n ther 60 s and 70 s. We tested for equalty across male and female subsamples and across rural and urban subsamples, and rejected the hypothess at the 1 percent level for both gender and resdence. A male s age, total calores and the fat and proten content of hs food consumpton have statstcally sgnfcant and postve effects on hs BMI. A female s and urban resdent s BMI ncreases wth ther detary fat and proten consumpton, whle rural resdent s BMI ncreases wth fat and calores consumpton. Havng hgher educaton decreases women s and urban resdents BMI. Whle beng employed ncreases only the male s BMI, beng a smoker sgnfcantly decreases hs BMI. However, the BMI for both genders and resdence locatons s hgher n 2004 than n 1994. Det (food consumpton) s an mportant determnant of an ndvdual beng overweght or obese. In Russa, the tradtonal det s hgh n sugar and lvestock products (meat and 14
dary) that contan fat, proten and cholesterol, but s low n consumpton of healther foods, such as vegetables and fruts, and has extremely low ntake of ctrus frut (Gnter 1995). Ths s probably due to the dffculty of growng fruts and vegetables n the Russan clmate, as well as to the state authortes food recommendatons durng the Sovet era that heavly favored meat and dary products. The orgnal Recommended Daly Intake n the Sovet Unon specfed that hgh proten ntakes were necessary for mantanng good health. But hgh anmal proten dets are lkely to be hgh n saturated fat also. How dd consumpton of meat and fsh, fruts and vegetables, dary, and alcohol change durng the transton? Tables 2a 2e present the changes n consumpton of fruts and vegetables, meat and fsh, and dary products from 1994 to 2004 by populaton categores. Consumpton of food products s measured by the household real (wth reference to June 1992) monthly expendtures. There s a trend of declnng consumpton of food products for all groups, wth the most dramatc declne n consumpton of meat and fsh, and fruts and vegetables. The declne was much hgher for the urban resdents compared to the rural resdents. For the pooled sample the consumpton of fruts and vegetables declned the most for the overweght people (49 percent), followed by declne for the normal weght people (45 percent) and for the obese (33 percent). The consumpton of meat and fsh declned the most for the normal weght people, by 50 percent, followed by declne for the overweght people, 43 percent and for the obese, 36 percent. The dary products consumpton declned by 37 percent for normal weght people and by 32 percent for the overweght and obese groups. The average Russan adult s calores consumpton s about 1870 calores per day. Proten ntake contrbutes around 13 percent of total calores, wth a very small ncrease durng the transton perod. Calorc ntake from fat declned from 34 percent for adult men and 15
women, to 32 percent only for the normal weght and overweght groups, possbly due to an overall mprovement n nutrtonal status, whle for the obese people t stayed the same at 33 percent. It s notable that the average calore ntake by obese s slghtly lower than the calore ntake by other categores. It s to suggest the composton of the det as well as factors representng varaton n lfestyle play more mportant role than smple calore counts. Tables 2a 2e show also the changes n consumpton of alcohol and cgarette smokng n Russa durng the perod 1994 2004. The general pattern durng the transton perod s that the number of people who consume alcohol has decreased, whle there was a slght ncrease n the number of people who smoke. The ncrease n the share of people wth normal weght who smoke s hgher than those resdng n rural areas (24 versus 13 percent) and s the hghest ncrease among all ndvduals. The decrease n the share of ndvduals consumng alcohol s the hghest for rural resdents, rangng from 38 percent for overweght ndvduals, 27 percent for obese ndvduals, to 19 percent for normal weght ndvduals, whle the declne of the share of ndvduals consumng alcohol s 11 percent for obese and overweght people and 12 percent for normal weght urban resdents. Next, we estmated the ndvdual demand for a) calores by OLS, the household demands for b) meat and fsh, c) fruts and vegetables, and d) dary products (equaton 12) by nterval regressons, 11 and the ndvdual demand for alcohol (equaton 13) usng a probt model. Frst, we ftted the demand equaton for ndvdual s total calores for the pooled sample over urban and rural resdents and separately by resdence. An ncrease n age and expendtures sgnfcantly ncreases the demand for calores but the effect of expendture s dmnshng at hgher expendture levels, gven the negatve coeffcent on the squared term of expendture. The expendture elastcty estmated at the sample mean s 0.015 for the pooled sample, and 16
0.010 for the urban and 0.04 for the rural samples. Therefore, the estmated expendture effects are qute small. Beng a male ncreases the demand for calores by about 10 percent. The demand for calores s ncreased by 6.2 percent for a male resdng n urban areas, whle the demand s ncreased by 14.3 percent for a male lvng n rural area. The hgher demand for calores for rural resdent s consstent snce the lkely more physcal work for rural resdents requres more energy. Durng the transton, the demand for (consumpton of) calores has ncreased by 2.7 percent and by 1.6 percent and 4.6 percent for urban and rural resdents, respectvely. Beng a smoker decreases the demand for calores by 3.2 percent. Ths effect s statstcally sgnfcant only for the rural sample and the calore demand s decreased by 5.7 percent. The results of the estmaton of the demand for calores equaton show sgnfcant regonal effects, as well. Compared to the base Moscow St. Petersburg metropoltan regons, the consumpton of calores n all less urbanzed regons s hgher by between 10.5 and 23.6 percent. In the urban sample the base s Moscow St. Petersburg metropoltan regons, whle ths regon s dropped from the rural sample estmaton because t does not nclude rural populaton. The base regon for the rural sample estmaton s the North and Northwest regon. The calore demand s sgnfcantly hgher n North Caucuses, Ural and West Sbera regons compared to the base North and Northwest regon. The household demand equatons for meat and fsh, fruts and vegetables, and dary products were ftted to the household data. The results are presented n Tables 5a, b and c. The adult equvalent number of household members s ncluded to control for the sze of the household, and the ndvdual characterstcs of the household head are used. The heght, year dummy varable, household expendture and the head s educaton are among the mportant factors that sgnfcantly ncrease the household s demand for meat and fsh, fruts 17
and vegetables, and dary products. The quantty demanded of these products s 82 to 106 percent hgher n 2004 than n 1994, or about 10 percent hgher per year. The demands for fruts and vegetables, meat and fsh, and dary products are also convex n age of the household head the quantty demanded declnes untl mddle age and ncreases later. However, the effects are statstcally sgnfcant only for the fruts and vegetables and dary demand equatons. The postve coeffcent on expendture and the negatve coeffcent on expendture squared ndcate that the demands for calores, meat and fsh, fruts and vegetables, and dary products are concave wth respect to expendture. The expendture elastctes computed at the mean values of expendtures are postve and hgher than unty 1.99 for fruts and vegetables, 1.70 for meat and fsh and 1.28 for dary products, ndcatng that as real ncome fell there was a major reducton n the consumpton of these products. For the urban resdents the expendture elastcty for dary products s less than unty (0.93), ndcatng a necessty good, whle the other elastcetes are 1.67 for fruts and vegges, and 1.43 for meat and fsh. The estmated elastctes are relatvely hgh compared to those found n developed countres but comparable to those found n other transton economes (Hossan and Jensen, 2000). The regonal dfferences n demands are nterestng as the pattern of demand for all types of food s just the opposte of the pattern of demand for calores. In all regons, demands are lower compared to the base Moscow St. Petersburg regons. The results suggest that detary patterns dffer substantally between metropoltan areas and rural provnces n Russa. Table 6 presents the margnal effects of the ndvdual s probablty of consumng alcohol for the pooled sample and by urban/rural resdence. The margnal effects are evaluated at the sample means. Beng a male, a smoker, and havng a hgher educaton, all ncrease the 18
probablty of consumng alcohol by 20.6, 12.5 and 11.3 percent, respectvely, for the whole sample. The magntude of the effects of male, smoker and hgher educaton are larger for the rural sample, beng 27.0, 12.2, and 12.4 percent versus 19.3, 10.4 and 7.6 percent for the urban sample. The margnal effect of age evaluated at the sample mean s negatve. The margnal effect of expendture at the mean s postve and statstcally sgnfcant for the pooled and rural sample, but small n magntude. The probablty of alcohol consumpton s sgnfcantly hgher n Ural (by 11.9 percent) regon and sgnfcantly lower (by 9.9 percent) n the North Caucuses regon compared to the Moscow St. Petersburg regons. Urban resdents from the North and Northwest, Central and Ural regons have hgher probablty of alcohol consumpton compared to the base Moscow St. Petersburg regons, whle for the rural sample the probablty of consumng alcohol s lower n the North Caucuses rural areas and hgher n the Ural regon compared to the North and Northwest regon. The probablty of consumng alcohol s lower n 2004 compared to 1994, by 5.3 percent for the whole sample and by 12.3 percent for the rural resdents sample, whle the effect for the urban sample was not statstcally dfferent from zero, at conventonal levels. The results from the OLS estmaton of the BMI supply functon usng lnweght as the dependent varable for the pooled sample, and by gender and resdence are presented n Table 7. The estmated coeffcent for the predcted wage s postve suggestng that the more sedentary nature of hgher wage jobs as well as hgher opportunty cost of off the job exercsng ncreases the ndvdual weghts. The estmated coeffcent s statstcally sgnfcant at the 5 percent level for the pooled and rural samples. Urban resdents weghts were hgher n 2004 than 1994, by 2.3 percent, and rural resdents weghts were 2.8 percent hgher. Only the weght of ndvduals lvng n the Central, North Caucasus, Sbera and 19
East Sbera regons are statstcally sgnfcantly hgher relatve to the weght of ndvduals resdng n the base Moscow St. Petersburg regons. Table 8 presents the OLS estmates of the ndvdual s BMI supply functon where the dependent varable s lnbmi. The results are very smlar to the results of the specfcaton where the dependent varable s lnweght. We ftted the models separately for the 1994 and 2004 subsamples and also for each year by gender. 12 The effects of the factors affectng obesty are smlar, wth the only dfference beng the mpact of wages (ncome), whch s larger n magntude and more sgnfcant n 2004. Ths s an mportant result pontng to the fact that economc forces have begun to play an ncreasngly mportant role n ndvdual choces wth the unfoldng transton to a market economy. Conclusons Ths paper develops both theoretcal and emprcal models to facltate understandng of the ncreased human obesty, measured as weght and BMI, and the factors contrbutng to t n Russa durng the transton from a planned to a market economy. Durng ten years of transton there was a sgnfcant rse n obesty n Russa a 38 percent ncrease by 2004. Emprcal results strongly support our models for BMI producton and demand for nputs n the BMI producton functon. Det/calorc ntake, smokng, gender and educaton are mportant determnants of obesty n Russa. The study fnds a strong postve effect of calorc ntake and a strong negatve effect of smokng on weght and BMI fndngs smlar to those n developed market economes (Chou et al., 2004; Rashad, 2006; and Rashad et al., 2006). 20
We employ a balanced panel from the RLMS for 1994 and 2004 for both households and ndvduals n order to evaluate the changes n overweght and obesty durng the transton and the effects of varous factors. The ndvdual BMI producton and supply functons are the man focus of ths analyss. Demographc and anthropometrc characterstcs such as heght, gender, and age postvely and sgnfcantly nfluence the degree of overweght and obesty n Russa, whle age has a nonlnear effect, and better educated ndvduals are less overweght and obese. Economc and detary factors such as calorc ntake and composton of fat and proten also affect postvely and sgnfcantly the ndvdual s weght, and therefore contrbute to obesty, but smokng deters overweght. These fndngs are smlar to fndngs for developed economes, ncludng the US (Chou et al., 2004; Lakdawalla and Phlpson, 2002). Beng employed ncreases ndvdual s weght only for males and urban resdents, also. Hgher ncome, measured as expendture n our study, s assocated wth ncrease n calore demand, suggestng that calores are normal good and a major contrbutng factor to obesty wth rsng ncomes, a trend currently observed n developed countres. For example, n 1970s the obesty rate of the populaton n the Unted States was heavly concentrated n low ncome households, but over the past three decades obesty has spread throughout the mddle and hgh ncome populaton groups as well (Maheshwar et al., 2005). Unbalanced det and unfavorable health behavor such as large ncreases n alcohol and tobacco consumpton amongst certan groups of the populaton are mportant determnants of health. Snce det s an essental factor affectng obesty, we ftted demand equatons for several food groups, ncludng meat and fsh, fruts and vegetables and dary products. Among the factors sgnfcantly and postvely affectng consumpton are household ncome/expendture, and the educaton and age of the household head. Although the total 21
calores consumed dd not change over the ten year perod, obesty has ncreased. Change n the composton of the det such as shftng away form healthy food as fruts and vegetables toward fatty and sugary products, or unhealthy lfestyle such as alcohol consumpton, as well as the change n lfestyle wth the technologcal progress, are among the possble explanatons. Understandng obesty n Russa s mportant n order to defne what strateges are most lkely to be effectve n preventng and reducng obesty. Ths study ndcates that hgher educaton, other thngs equal, has a sgnfcant and negatve effect on obesty, and contrbutes to the producton and supply of good health. Educaton not only provdes economc returns such as ncreasng earnngs and employment, but also mproves health and well beng. Therefore, nterventons whch enhance educaton could play a vtal role n preventng obesty n Russa. People should be educated about healthy lfestyles and healthy det. 22
Notes 1. We use the sample weghts provded wth the data to correct for sample desgn and samplng procedure. 2. Ths s analogous to the dervaton of the supply functon for farm output n an agrcultural household model (see Huffman, 1991). 3. Ths s not a true panel survey where sample households and ndvduals are followed and ntervewed n each round. However, after 1999 the orgnal desgn was modfed and some households and ndvduals who moved were surveyed at ther new locatons. Most mportantly, the analyses of the RLMS data for attrton, carred out by the Insttute for Socal Research at the Unversty of Mchgan, show that the exts can be characterzed as random and that the sample dstrbutons reman unchanged (Heernga, 1997). 4. In the sample 19 percent of the observatons had mssng values of BMI, 7 percent had mssng values of total calores, and 4 percent had mssng values of ncome. 5. In the alternatve estmaton, where the dependent varable s Weght, lnheght s ncluded as an explanatory varable. We nclude the ndvdual s mature heght n ths equaton as a summary ndcator of an ndvdual s genetc potental and early nvestments n good health. 6. Although the RLMS collects detaled communty level prce data for 58 tems each year, we do not have access to these data and cannot nclude the prces n our analyss. 7. The results from the estmaton of the Heckman selecton model to predct the ndvdual wage are avalable from the authors upon request. 23
8. We use OLS estmates to explot the between and wthn dmensons of the data even though not effcently. To mprove the effcency of the estmates we use cluster opton whch produces cluster robust covarance matrx. Such estmator allows for general forms of heteroskedastcty as well as autocorrelaton wthn a gven ndvdual, and has been recently argued to be preferred or at least as good as random effects approach but wthout the need to make restrctve assumptons about correlatons of the error terms (Verbeek and Vella 2005). 9. Results from estmaton of unbalanced panel are smlar to the reported results for the balanced panel and are avalable from the authors. 10. We appled Wald tests for coeffcent dfferences between the male and female, and urban and rural subsamples. The results are avalable from the authors upon request. 11. Due to data lmtatons, we estmated demands for meat and fsh, fruts and vegetables, and dary at the household rather than the ndvdual level. 12. The results are avalable from the authors upon request. 24
References Branerd, E. and D. Cutler, 2005. Autopsy on an Empre: Understandng Mortalty n Russa and the Former Sovet Unon. Journal of Economc Perspectves 19 (1): 107 130. Chou, S., M. Grossman and H. Saffer, 2004. An Economc Analyss of Adult Obesty: Results from the Behavoral Rsk Factor Survellance System. Journal of Health Economcs 23(3): 565 587. Cockerham, W., 2000. Health Lfestyles n Russa. Socal Scence and Medcne 51: 1313 1324. Gardner, B. and K. Brooks, 1994. Food Prces and Market Integraton n Russa: 1992 93. Amercan Journal of Agrcultural Economcs 76: 641 646. Gnter, E., 1995. Cardovascular Rsk Factors n the Former Communst Countres. European Journal of Epdemology 11 (2): 199 205. Grossman, M., 2000. The Human Captal Model. In A.J. Cutler and J.P Newhouse, Handbook of Health Economcs, Vol 1A, New York, NY: Elsever, p. 248 408. Heernga, 1997. Russa Longtudnal Montorng Survey (RLMS) Sample Attrton, Replenshment, and Weghtng n Rounds V VII. Avalable at: http://www.cpc.unc.edu/projects/rlms/project/samprep.pdf Hossan, F. and H.H. Jensen, 2000. Lthuana's Food Demand durng Economc Transton. Agrcultural Economcs 23: 31 40. Huffman, S. and S. Johnson, 2004. Impacts of Economc Reform n Poland: Incdence and Welfare Changes wthn a Consstent Framework. The Revew of Economcs and Statstcs 86(2): 626 636. Huffman S. and M. Rzov, 2007. Determnants of Obesty n Transton Economes: The Case of Russa. Economcs and Human Bology, Vol. 5 (3): 379 391. Huffman, W.E., 1991. Agrcultural Household Models: Survey and Crtque. In M.C. Hallberg, J.L Fndes, and D.A. Lass, (Eds.), Multple Job Holdng among Farm Famles. Ames, IA: Iowa State Unversty Press, p. 79 111. Huffman, W., S.K. Huffman, A. Tegene and K. Rckertsen, 2006. The Economcs of Obesty Related Mortalty among Hgh Income Countres. Iowa State Unversty, Department of Economcs Workng Paper #06021 Komlos, J. and M. Baur, 2004. From the Tallest to (one of) the Fattest: The Engmatc Fate of the Amercan Populaton n the 20 th Century. Economcs and Human Bology 2, 57 74. 25
Lakdawalla, D. and T. Phlpson, 2002. The Growth of Obesty and Technologcal Change: A Theoretcal and Emprcal Examnaton. Natonal Bureau of Economc Research Workng Paper Seres, Workng Paper 8946. Lakdawalla, D., T. Phlpson, and J. Bhattacharya, 2005. Welfare Enhancng Technologcal Change and the Growth of Obesty. Amercan Economc Revew 95:253 257. Lefert, W., 2004. Food Securty n Russa: Economc Growth and Rsng Incomes are Reducng Insecurty. Food Securty Assessment/ GFA 15/ May 2004. Maheshwar, N., J.G. Robnson, N. Kohatsu, and B. Zmmerman, 2005. Obesty Prevalence Increasng 3 Tmes Faster n Hgh than Low Income Groups: Natonal Health and Nutrton Examnaton Surveys 1971 to 2002. Crculaton: 111:190. Mendez, M.A. and B. M Popkn, 2004. Globalzaton, Urbanzaton and Nutrtonal Change n the Developng World. Journal of Agrcultural Development Economcs 1:220 241. Natonal Insttute of Dabetes and Dgestve and Kdney Dseases (NIDDKD), 1996. Statstcs Related to Overweght and Obesty. Washngton DC: US Government Prntng Offce. Avalable at http://www.medhelp.org/nihlb/gf 367.html Oglobln C. and G. Brock, 2003. Smokng n Russa: The Marlboro Man Rdes but Wthout Vrgna Slms for Now. Comparatve Economc Studes, 45: 87 103. Rashad, I., 2006. Structural Estmaton of Calorc Intake, Exercse, Smokng, and Obesty. Quarterly Revew of Economcs and Fnance, 46(2): 268 283. Rashad, I., M. Grossman and S. Chou, 2006. The Super Sze of Amerca: An Economc Estmaton of Body Mass Index and Obesty n Adults. Eastern Economc Journal, 32(1): 133 148. Rosenzweg, M.R. and T.P. Schultz, 1982. The Behavor of Mothers as Inputs to Chld Health: The Determnants of Brth Weght, Gestaton, and Rate of Fetal Growth. In V.R. Fuchs, (Ed.), Economc Aspects of Health, Natonal Bureau of Economc Research, The Unversty of Chcago Press, pp. 53 92. Russan Longtudnal Montorng Survey (RLMS). Avalable at http://www.cpc.unc.edu/projects/rlms Shkolnkov, V., Andreev, E, Leon, D, McKee, M., Mesle, F., and J. Valln, 2004. Mortalty Reversal n Russa: the Story so Far. Hygea Internatonals 4, 29 80. Stllman, S., 2006. Health and Nutrton n Eastern Europe and the Former Sovet Unon durng the Decade of Transton: A Revew of Lterature. Economcs and Human Bology 4, 104 146. 26
Verbeek M. and F. Vella, 2005. Estmatng Dynamc Models from Repeated Cross Sectons. Journal of Econometrcs 127: 83 102. World Health Organzaton, 2006. Avalable at http://www.who.nt/detphyscalactvty/publcatons/facts/obesty/en/ Zohoor, N., T. Mroz, B. Popkn, E. Glnskaya, M. Lokshn, D. Mancn, P. Kozyreva, M. Kosolapov and M. Swafford, 1998. Montorng the Economc Transton n the Russan Federaton and ts Implcatons for the Demographc Crss the Russan Longtudnal Montorng Survey. World Development 26 (11): 1977 1993 27
Table 1 Varable names, defntons, means and standard devatons (weghted data) Symbol Mean (SD) Defnton lnweght 4.272(0.003) Indvdual weght (kg) n logarthm lnheght 5.106(0.001) Indvdual heght (cm) n logarthm lnbmi 3.270(0.002) Indvdual weght dvded by heght (m) squared (kg/m2) n logarthm lncalores 7.493(0.004) Total calores consumed per day n logarthm * Fat 32.994(0.138) Percent of daly calores from fat Proten 12.885 (0.048) Percent of daly calores from proten lnmeat&fsh 5.307(0.051) Real total household expendture per month on meat and fsh (rubles) n logarthm lnfruts&vegges 3.574(0.052) Real total household expendture per month on vegetables and frut (rubles) n logarthm lndary 4.097(0.046) Real total household expendture per month on dary products (rubles) n logarthm Alcohol 0.547 (0.007) Dummy varable equal to 1 f the ndvdual consumes alcohol and 0 otherwse ** Smoker 0.289(0.006) Dummy varable equal to 1 f the ndvdual smokes currently and 0 otherwse Male 0.417(0.007) Dummy varable equal to 1 f the ndvdual s a male and 0 otherwse Educaton1 0.378(0.006) Dummy varable equal to 1 f the ndvdual has educaton level below grade 8 and 0 otherwse Educaton2 Dummy varable equal to 1 f the ndvdual has completed hgh school and 0 otherwse 0.465(0.007) Educaton3 0.157(0.005) Dummy varable equal to 1 f the ndvdual has completed hgher educaton and 0 otherwse Age 48.00(0.208) Age n years Age2 2550.34(20.72) Square of age Work 0.569(0.007) Dummy varable equal to 1 f the ndvdual s employed and 0 otherwse lnexpend 8.718(0.887) Monthly total household expendtures (rubles) n logarthm lnexpend_sq 76.798(15.297) Square of monthly total household expendtures n logarthm lnwage 3.048(0.908) Hourly wage (rubles) n logarthm *** lnpred_wage 2.851(0.465) Predcted value of hourly wage (rubles) n logarthm lnnl_ncome 8.021(1.903) Monthly non labor ndvdual ncome (rubles) n logarthm Marred 0.722(0.448) Dummy varable equal to 1 f the ndvdual s marred and 0 otherwse Kds6 0.248(0.559) Number of chldren age 6 and less Kds18 0.534(0.812) Number of chldren age between 7 and 18 * Snce the data for year 1994 was not avalable, we calculated t by extrapolaton usng nformaton from fve years over the 2000 2004 perod. ** Indvduals are asked drect questons n the health module about alcohol consumpton. ***Number of observatons s 2777. 28