5. Food Demand Analysis

Transcription

1 M Food and Nutrton Securty Unversty of Hohenhem Olver Ecker 5. Food Demand Analyss 5.1 Data Collecton at Household Level 5.2 Determnants of Food Demand 5.3 Nutrtonal Economcs

2 Theoretcal framework of food demand Consumer Theory: Basc assumpton: Consumers choose a consumpton bundle x, so as to maxmze utlty u(x) subect to ther budget constrant m = px. Maxmzaton of overall utlty! The optmal consumpton bundle x* depends on the prces of goods p and the avalable ncome m. Marshallan demand functon for good : x * = D (p,m) Consumers expendture (cost) functon: m(u,p) Mnmum expendture requred to attan a specfc utlty level at gven prces 2

3 What does food demand analyss do? It descrbes and explans the level of demand for the food commodtes an ndvdual consumes, gven the structure of relatve prces faced, real ncome, and a set of ndvdual characterstcs. A set of elastctes s an mportant result of food demand analyss. Why s ths mportant? Utlty s determned by the quantty of goods an ndvdual consumes. For the poor, food accounts for the largest share of the consumpton bundle. The nutrton status of ndvduals s to a large extent determned by the level of food consumpton. Therefore, knowledge about consumpton patterns and how they change through exogenous shocks s mportant for food and nutrton securty. 3

4 5.1 Data Collecton at Household Level Data collecton approaches 1. Food expendture data Households ntervewed report on ther food expendture (purchased and from other sources) for the past 7 days (perods can also be longer) hour recall data Households ntervewed report on ther food ntake for the last 24 hours, usually meal by meal. 3. Food weghng (detary record) A surveyor s present n the household durng meal preparaton tmes and weghs all ngredents used. 4

5 Example: food expendture survey form How much of the followng food tems dd your household consume n the past one week (purchases, own producton, receved n-knd)? Code Item Quantty (kg) Value ($) 1 Rce 2 Wheat 8 Red lentls 9 Chckpeas 20 Mlk 21 Cheese 30 Beef 31 Chcken meat The dsaggregaton of food tems vares by survey (typcally ). 5

6 Example: 24h-recall survey form What type of food and how much dd your household consume durng the past 24 hours? Item Code a Quantty (kg) Value ($) Breakfast Lunch Dnner Snacks a 1 = rce, 2 = wheat,., 8 = red lentls etc. 6

7 Pros and cons of survey approaches Food expendture data Food expendtures are usually part of lvng standard measurement studes (LSMS). Together wth non-food expendtures, they are preferred over monthly ncome as a measure of permanent ncome and thus lvng standard. But, households mght forget to menton certan tems because of relatvely long recall perod. It s not really food ntake what s assessed, because food waste, spolage, portons gven to guests or fed to anmals etc. can hardly be captured. Shorter recall perods can mprove on some dsadvantages (e.g., 3- or 7-day recall). 7

8 pros and cons of survey approaches 24-hour recall data For an assessment of household food ntake, 24-h recall data are better and often nclude more detals. Recall perod s short, so that more accuracy. Dsaggregaton by meal helps determne nhbtng and enhancng factors for mcronutrent absorpton. But, 24-h recall s only snapshot; day-to-day dversty neglected. Food weghng In terms of recordng, weghng s stll more accurate than 24- h recall, but the presence of the surveyor mght cause a bas. Problems of all approaches Sngle round of data collecton cannot capture seasonalty effects (optmum: two or more rounds n dfferent seasons). Mostly no nformaton on ntra-household dstrbuton (surveys at ndvdual level are possble but rare). 8

9 5.2 Determnants of Food Demand What determnes household food consumpton? Income ( expendture data) Prces: own- & cross-prces Household characterstcs: Household sze, sex & age structure, decson maker, Educaton & nutrtonal knowledge, Taste, food habts, relgous taboos Food avalablty & accessblty: Farmng & self-suffcency, cash lqudty Market ntegraton & nfrastructural endowment Agro-ecologcal producton condtons Seasonalty of food crops Whch varables reflect these determnants best? Whch varables are avalable from the data? 9

10 Food consumpton n Russa Study by Qam, von Braun, tho Seeth (1997) 712 households were ntervewed n 1995 n three oblasts. Summer and wnter ntervew rounds to account for seasonalty effects. Food expendture over last 30 days was recorded. Consumpton from own producton s ncluded n total expendture. Household consumpton was converted nto per-capta consumpton, assumng that food dstrbuton follows standard adult-equvalent weghts. Households were stratfed n four quartles accordng to percapta total expendture. 10

11 Food expendture patterns n Russa Expendture quartle Total Tot. pc expend. (Thsd. Rbl) 3,267 1,399 2,410 3,411 5,847 Gn coeffcent Food expend. (Thsd. Rbl.) 2,277 1,009 1,685 2,354 3,964 Food expend. share (%)

12 Food consumpton patterns n Russa Expendture quartle Total Bread (kg/year) Bread exp. share (%) a Potatoes (kg/year) Pot. exp. share (%) a Vegetables (kg/year) 77.9 Veg. exp. share (%) a Meat (kg/year) 45.6 Meat exp. share (%) a a Share n total household food expendture. 12

13 Income as determnant of demand Income elastcty of demand η = q y y q Can be estmated wth Engel functons: Lnear functonal form q = a + b y η = ( y z) q = q, b y q Double-log functonal form ln q = a + b ln y η = b 13

14 Values of ncome/expendture elastctes Normal goods: η > 0 Inferor goods: η < 0 Luxury goods: Essental goods: (most food tems) η >1 η <1 and η > η η staples <η non staples poor rch Engel s law: Wth ncreasng ncome levels, people tend to spend a decreasng share of ther ncome on food. 14

15 Food expendture shares (examples) Germany 11% Czech Republc 25% Argentna 33% Indonesa 50% Sr Lanka 63% Ghana 73% Rwanda 79% Burkna Faso 84% 15

16 Income/expendture elastctes Average Poor Non-poor Food, Serra Leone Sorghum, Nger Rce, Mal Cassava, Ngera Cereals, Brazl Legumes, Brazl Cassava, Brazl Source: Teklu (1996), Gray (1982). 16

17 Prces as determnants of demand Own-prce elastcty: (usually negatve) E = q p p q Cross-prce elastcty: (can be negatve or postve) E = q p p q Can be estmated econometrcally from demand equaton: ( y, p1... p z) q = q n, Prce elastctes are more dffcult to estmate from crosssecton data than ncome elastctes (why?). 17

18 Selected demand equatons n Russa ln ( pccons) = α + β ln( pcexpend) + γ ln( prce) + δ Z + ε Bread Potato Mlk Vegetable Meat Constant 3.27** 2.70** 3.50** Expendture 0.20** 0.39** 0.33** 0.61** 0.84** Own prce -0.25** ** -0.37** -0.83** HH members -0.05** -0.09** -0.15** -0.10** 0.01 Educaton -0.11* * Urban -0.14** -0.77** -1.44** -0.70** -0.21** Adusted R *, ** Coeffcent s statstcally sgnfcant at 10% and 5% level, respectvely. Source: Qam et al. (1997). 18

19 Theory of consumer behavor Estmaton of sngle demand equatons s not fully consstent wth economc theory. ( q z) subect to p q y Max u =, (budget constrant) Lagrangean ( q, z) + ( y p q) = u λ Soluton s a set of n demand equatons, q (p, y, z), wth n ncome elastctes and n 2 prce elastctes. That s, n + n 2 parameters to be estmated. Number of ndependent parameters s reduced through dfferent constrants. 19

20 Constrants from frst-order condtons 1. The Engel equaton q y p = 1 or wη = 1 where w = pq y 2. The n Cournot equatons q p p = q or w E = w for = 1,..., n 2. The n Euler equatons E + η = 0 = 1,..., n 3. The n(n-1)/2 Slutsky equatons w E E w ( ) = + η η for = 1,..., n. w 2 n n Independent parameters n + n 1 n = n + n ( ) ( ) ( ) 20

21 Estmaton of complete demand systems Complete systems are needed to be consstent wth theory, but trade-off between cost of estmaton and theoretcal foundaton. Complete systems are partcularly mportant when used n general equlbrum models, and complex smulatons exercses. Dfferent systems approaches can be used that dffer n ther specfcaton of the utlty functon and addtonal assumptons: Lnear expendture system (LES) (restrctve assumptons) Almost deal demand system (AIDS) (more flexble) Polcy mplcatons Income and prce elastctes of food demand, whether derved through estmaton of sngle equatons or complete systems, are key to assess food consumpton mpacts of polces and other external shocks. 21

22 5.3 Nutrtonal Economcs From food consumpton to calore ntake For food and nutrton securty analyss, not only the quantty of food tems consumed, but also the actual calore and nutrent ntake s of nterest. Knowng the daly per-capta consumpton of food tems, calore and nutrent ntakes can be calculated usng food composton tables. Food composton tables should reflect local food characterstcs, whenever possble, and are generally preferable over nternatonal standard tables. Especally for mcronutrent ntakes, ssues of boavalablty need to be accounted for (e.g., beta-carotene can only be converted to vtamn A f there s some fat n the det; ron from plant sources s less boavalable than from meat). 22

23 Calore ntake n Brazl Total calores per capta Poorest 30% Mddle 50% Rchest 20% 1,876 2,162 2,369 Of whch n % Cereals Roots and tubers Legumes Sugar Vegetables Fruts Meat and fsh Dary products Ols and fats Beverages Source: Gray (1982). 23

24 Calore elastctes Calore elastctes can be estmated drectly, ust as demand elastctes, when consumpton as dependent varable s expressed n terms of calores. Alternatvely, calore elastctes can be derved as follows: Total calore elastcty wth respect to prce of good E c = c p p c = where a s a set of techncal coeffcents measurng calore contents of food tems (from composton table), and E s the set of prce elastctes of food demand. Interpretaton: How does overall calore ntake change, when the prce of good changes by one percent? a c a q c q E 24

25 calore elastctes Calore elastcty wth respect to ncome: η cy c y = = y c a c a q c q η There s no consensus whether ths calore-ncome elastcty s sgnfcantly dfferent from zero across ncome groups. But dsaggregate analyss suggests that, at least for the poor, t s postve (.e., rsng ncomes reduce under-nutrton). 25

26 Are food consumpton databases gvng relable estmates? Comparatve study by Bous (1994) for Kenya and the Phlppnes Kenya: per-capta calore consumpton Expendture quartle 24-h recall data 7-day food expendture data Household Adusted for guests Household 1 1,706 1,651 1, ,948 1,839 1, ,026 1,912 2, ,232 2,045 2,293 η cy η cy s the total calore-ncome elastcty. 26

27 relable estmates? Phlppnes: per-capta calore consumpton Expendture quartle 24-h recall data 30-day food expendture data Household Adusted for guests Household 1 1,726 1,625 1, ,877 1,762 1, ,035 1,876 2, ,196 1,983 2,540 η cy Elastctes from food expendture data are often overestmated. People s weght ncrease across groups s less than mpled by elastctes. Reasons: Food expendture data have bgger problems wth waste, spolage, etc., and transfers across ncome groups are more dffcult to capture (e.g., laborers recevng meals). 27

28 Nutrton effects of polces: example of settng research prortes n Colomba How should a gven agrcultural research budget be allocated to dfferent commodtes when the man obectve s an mprovement n the calore status of the urban poor? Study by Pnstrup-Andersen et al. (1976) cted n Sadoulet and de Janvry (1995). Research n a gven commodty wll eventually ncrease the market supply of that commodty and lower the market prce. Accordng to prce responsveness of supply and demand, there wll be drect and cross prce effects. Consumpton patterns wll be affected. How does calore ntake of the poor change when research money s allocated to commodty? 28

29 Impact of 10% ncrease n supply on daly per capta calore ntake (Colomban example) Beef Mlk Vegetables Peas Beans Rce Maze Cassava a n kcal. Drect effect a Defcent households Indr. effect a Net effect a % of defcency Drect effect a Non-defcent hh Indr. effect a Net effect a