The demand for food and foodgrain quality in India

DISCUSSION PAPER Report No.: ARU 6 THE DEMAND FOR FOOD AND FOODGRAIN QUALITY IN INDIA Hans P. Binswanger, Jaime B. Quizon and Gurushri Swamy November 1982 Research Unit Agriculture and Rural Development Department Operational Policy Staff World Bank The views presented here are those of the authors, and they should not be interpreted as reflecting those of the World Bank. The Demand for Food and Foodgrain Quality in India Hans P. Binswanger, Jaime B. Quizon and Gurushri Swamy A recent study by Swamy and Binswanger (S&B) used advances in duality theory and index number theory to formulate flexible consumer demand systems that are linear in estimation equations they estimated such systems for food demand using data on per capita food availability for 20 years and 10 states of India. That data set was rich in price variation but had several weaknesses, particularly in the estimates of certain foodgrain availabilities and of per capita expenditures. Furthermore, the use of per capita expenditure levels restricted the range of expenditure variability to variation across states and over time. The present study attempts to overcome the limitations of the above data set by re-estimating the consumer demand equations developed by S&B using National Sample Survey CISS) data for 15 states and two years, 1961 and 1974. The individual observations refer to the per capita consump- tion and expenditure levels of eight expenditure classes. Income therefore varies considerably both across and within states. On the other hand, since only two years of data are used, the price responses have to be estimated essentially from the change in relative prices between these two years which, fortunately, are over a decade apart. In addition to the demand for foodgrains, this study also estimates the demand for foodgrain quality. Initial inspection of the implicit prices paid for foodgrains by different expenditure groups revealed that poorer consumers tend to pay lower prices for the same conmodities than richer groups. It is likely that these differentials largely reflect genuine quality differences, but they may also reflect differences in the accessibility and convenience of the source from which foodgrains are purchased. The measured price differentials were used to define an expenditure variable, foodgrain quality, which was used to estimate consumer demand for quality with respect to total expenditures. This also allowed the substitution of average prices across expenditure classes for actual prices to "clean out" spurious price differences associated with quality while maintaining consis- tency of the demand systems with demand theory, i.e., expenditures on commodities plus foodgrain quality exhaust the budget. -3- I. THE CONSUMER DEMAND SYSTEMS AND THE DEKAND FOR FOODGRAIN QUALITY The flexible consumer demand systems of the S&B study are derived from three cost (or expenditure) functions, i.e. the Generalized Leontief (GL), the Normalized Quadatic (NQ) and the Transcendental Logarithmic (TL) functional forms. The estimation equations are given below (equations (7), (8) and (9)) and the elasticity formulae are given in Table 1. The justifica- tion for the use of these equation systems are discussed in S&B. In brief, these demand equation are all linear in paameters. Also, given appropriate restrictions, these equation forms (a) are homogeneous of degree one in prices and nominal income, (b) have symmetric cross-price derivatives and (c) satisfy the adding up constraint, i.e., the weighted sum of income elasticities across commodities equals one. These conditions are those imposed by neo- classical consumer demand theory. Finally, each equation contains a linear and a squared income term. As discussed in the S&B paper, this allows much more flexibility in the response of food consumption to changes in income. Income elasticities may be positive, negative, increasing or decreasing and they may first be increasing and then decreasing. Similar flexibility is possible for expenditure shares. The way this paper departs from the earlier S&B study arises from the nature of the data used. For each region and time period, per capita expenditures on and per capita quantities of various food items of eight expenditure groups are given with each data point representing an expenditure- group-specific mean. Prices of various foodgrains are computed implicitly as expenditure divided by quantity. Computations of these prices by expendi- ture group showed that these implicit prices are substantially higher for the higher expenditure groups than the lower ones. Richer people appear to -4- consume higher quality foodgrains and/or to buy them from higher priced sources. Since we cannot distinguish between these two likely sources of higher observed prices, we subsume them both under the term "quality". Total expenditures on foodgrain quality can be computed as presented below. Let Eikrt' ikrt and Pikt be the per capita expenditure, per capita quantity and implicit price of commodity i in expenditure group k, region r and time t. Supressing time and region subscripts, ik ' Eik /Qik A weighted average price across all expenditure classes is then -- k ik (2) I nk Qik where ak is the number of persons in expenditure class k. The quality premium paid for commodity k is thus Pik i and can be positve o nev 2/ positive or negative.- Expanditures on quality for commodity i by expenditure class k therefore is Qik ik -i). Total expenditure on foodgrain quality is the sum of expenditures on quality of each foodgrain and becomes F-Qk ' ik ~ik F- (3) where ieF indicates that i belongs to the foodgrains. 1/ The high income groups may purchase from higher priced shops and implicitly purchase better retail services. Another component of the price difference could be that the poorer groups purc6ae a higher proportion of their consumption from government subsidizii"fair price shops. aowever, in this case the price rises should be confined to the urban areas since few fair price shops exist in rural areas. They should also be confined to wheat, rice sugar and cooking oil since other commoditi'es are rarely traded in fair price.sho _9ote also, that it has usually been assumed that the high income consumers would purchase in the free market at lower prices than the poorer groups because their higher incomes would allow them to buy at lower post-harvest prices and store over longer periods of time while poorer groups would not have the savings co do so. 2/ It is therefore not meaningful or possible to compute elasticities of demand 1oi quity. T d the derivatives of quaity demand with respect to income have to be inspected directly. This quality variable is usually positive for richer consumers and negative for poorer ones. Its population-weighted sum in each region and time period is zero. Total expenditures of a consumer group are defined as N k El QikP i + EQk (4) where the commodity index i = 1, ... N, excludes quality which is located as the (N + 1)st commodity. For the commodities i which are not foodgrains, prices come from independent sources which do not allow measurement of price differentials across expenditure groups. For those comodities, = P . We include quality as a dependent variable in the system of demand equations. Expenditures on foodgrain quality are assumed to depend only on income and family size, but not on the prices, i.e. the quality equation contains no price terms. Given the budget constraint in (4), the adding up constraint on the income effects is assumed to hold over both the commodity demands and the quality demand. As in the earlier papers,this income constraint can be imposed directly on the GL and NQ systems (at sample means of the variables only). In the TL system, the fact that expenditure shares on real commodities no longer add up to one allows the joint estimation of all comodity equations, leaving out the quality equation. Furthermore, for the TL system the adding-up constraint implies N+l E (b + 2 b log M).= 0. (5) i=1 41 12 If we have independent estimates of bn+,l1 and bn+1,2' the income coefficients for a quality equation, we can maintain the following constraint at the sample mean of m. N E1 (bil + 2 bi2 log m) bN+,l + 2 bn+1,2 log m (6) -6- We have estimated the quality equation independently as a single equation and used the resulting bN+I,1 and b+1,2 terms to impose constraint (6) in the systems estimation for the real commodities. In addition to estimating the demand for foodgrain quality,we include the number of persons per family (H) into the estimation equatious of per capita consumption. This will enable us to test questions about economies of scale of larger families in the demand equations. Thus,for the three functional forms,the demand equations have the following form Transcendental Logarithmic (TL) S =a i + bil log m + bi2 (log M)2 + CLE log H - N-1 . . (7) + E C log (P i = 1, N SN+1 q aN + bN+,1, log m + N+ l2 (0gm log Normalized Quadratic (NQ) 2 I ai+ b m i+ b + C H N-1 + E C (P /P N-1 j ij i N aN+ bm+bN m + iH N-1 N-1 2l(8) + 1/2 1Z1E 1 C ijP i j PN N+ N1bN- , m + bN,+1, 2 m + CT+1,Ra Generalized Leontief (GL) b = a + b + bm + Cij (P ) i 1, ..., N (9) XN+1 as in (8) The restrictions and elasticity formulae are discussed in S& (1981). -7- II. DATA, DATA TRANSFORMATIONS AND ECONOMETRIC PROCEDURES This study uses data from two rounds of the National Sample Survey (NSS) of India, the 17th for 1961 and the 28th for 1974. Unlike most NSS surveys these rounds contain data on the cuantities consumed of the following foodgrains: rice, wheat, sorghum (jowar), pearl millet (bajra), maize, finger millet (ragi), other small millets, chickpea (bengal gram), cereal substitutes (primarily tapioca). In addition, both surveys contain the following relevant variables: expenditures on each of the foodgrains, expenditure on other food,-/ expendi- ture on nonfoods, total consumer expenditure, and the number of persons per family. Data have been published separately for the urban and the rural regions of 15 states,-/ i.e. are thus available for 30 subregions of the country. In each of these subregions the published data are given separately for a number of expenditure classes and refer to the average per capita consumption over the past 30 days in each of the expenditure classes. (This poses a heteroscedas- ticity problem discussed below.) The expenditure class averages are given for 13 and 14 expenditure classes respectively in the 1961 and the 1974 surveys, and they form the basic data of this enquiry. In addition we made use of both the state-specific consumer price indices far the two' years and the food price subindices. These state-specific indices are published by the Government of India and refer to rural workers (Agricultural Laborer's Index) and to urban workers (Industrial Worker's Index). This data base posed the follow-ing data and econometric problems: 1/ Other foods include the following categories: pulse products; milk; edible oils; meat, fish and egg; vegetables; sugar; salt; spices; beverages. 2/ Andhra Pradesh, Assam, Bihar, Gujarat, Jammu and Kashmir, Mysore (Karnataka), Kerala, Madhya Pradesh, Maharastra, Orissa, Punjab Rajhastan, Madras (Tamil Nadu), Uttar Pradesh, West Bengal. -8- 1. Reconciliation of expenditure classes between 1961 and 1974: The 1974 expenditure class intervals were first deflated to th.e 1961 price level using the national price index for urban workers (Industrial worker index). The 13 classes of the 1961 survey were then reduced into 8 classes by aggregating them as shown below. The 14 classes of the 1974 survey were then matched to the newly formed 1961 classes as closely as possible. Only a single national price deflator was used. The class allocations are as follow: Reconciled 1961 1974 Class No. Class Numbers Class Numbers 1 1 1, 2, 3, 4 2 2 5, 6 3 3 7 4 4, 5, 6 8, 9 5 7, 8 10 6- 9, 10, 11 11, 12 7 12 13 8 13 14 2. Generating price series for foodgrains, foodgrain quality, other foods and other commodities: A foodgrain Drice index for each expenditure class x subregion was computed directly from the NSS data as a Fisher price index. It was assumed that the price of foodgrain ouality would change in proportion to the price of foodgrains, i.e. the 1974 foodgrain quality expenditures were deflated to the 1961 level using the corresponding foodgrain price index. The price of other foods was obtained residually from the computed price index-of fo6dgrains and- tefoodsubind, using the7tatio of foodgrain expenditures to total food expenditures in the NSS data to weight tHe faod- grain and nonfoodgrain component in the food-saindex. -9 Similarly, the price index of other commodities (primarily nonfoods) was computed residually from corresponding price index for all commodities and the food subindex, using the ratio of food expenditures to total expenditures in the NSS surveys. 3. Missing observations; In some states, no households were sampled in some of the expenditure classes for one of the two surveys. For the data transformations discussed under 7 and 8 below, however, we required a fully balanced data set. There- fore, when data was missing for a rural expenditure class, we borrowed the per capita data of the matching expenditure class for the urban sample from the same state, and vice versa if urban data were missing. We assigned a sample size of 0.1 households to these missing data, which ensured that these obser- vation would have virtually no weight in all computations of indices and weighted averages. These observations were dropped after data transformation, but prior to econometric estimation. Whenever information was missing for one round, the corresponding expenditure class observation from the other round was also dropped. Ten missing data points thus resulted in 20 dropped observations Cout of a possible 480 observations). 4. Missing prices: For some commodities and some observations quantities consumed were zero. Therefore implicit prices cannot be computed. In those cases, prices were substituted from the corresponding urban or rural observation, and in some cases from a neighboring state. 5. Forming of commodity aggregates and real total expenditure: Fisher indexes were used whenever commodities had to be aggregated from individual price and quantity series as in the case of "inferior cereals". A Fisher index was also used in computing an overall price deflator for total expenditures. - 10 - 6. The price and quantity data were then transformed into the proper dependent and independent variables for the GL, NQ and TL functional forms. 7. Combining cross section and time series:2 The following covariance transformation is equivalent to the intro- duction of a dummy variable for each of the subregions and one of the two years: Ikt ikt *kt Xik +X... (10) where X stands for all variables in the data set and X*kt' ik and X... are weighted means of the variables across the dotted dimension. The weights used are the number of persons in each of the expenditure group, observation ikt. This procedure does not eliminate any variation in the data set along the expenditure class dimension k. In the earlier paper, S&B had used an arror camp nents procedure to handle the time series - cross section problem. Since the data set for this study includes only two years, it appeared impossible to estimate the error component of time for such a model and the simpler fixed-effect model was chosen instead. 8. Retaroscedasticity: The use of group means for expenditure classes (instead of individual observations) leads to a familiar heteroscedasticity problem which was handled by the usual data transformation g* */ V7 (11) where H is the number of households per expenditure class observation. Note that here the number of households is used, not the number of persons, because the sampling error is related to the number of primary sampling units (i.e. households) and not to number of persons. The above equation also indicates - 11 - that the heteroscedasticity transformation is applied after the covariance transformation. 9. Systems estimation: The transformed data set x forms a system of equations which was estimated with and without restrictions using Zellner's method. III. THE SYSTEMS ESTIMATED AND THE RESULTS Three systems with different geographic coverages were estimated. The comodities included in each system are indicated in panel I of Table 2 by the presence of a mean quantity. System A used data from all 15 states. It is focused on wheat, rice and inferior cereals as a group (sorghum, pearl millet, maize, finger millet, barley and small millet). It also contains- equations for other foods, other commodities and foodgrain quality. System C is basically similar, except that inferior cereals are split up. Separate equations are included for the major inferior cereals sorghum and pearl millet. The remaining inferior cereals are aggregated with "other foods 3". System C is estimated using data from only those states where consumption of sorghum and pearl millet is important. System E is again similar to system A in that an equation for all inferior cereals is estimated. In addition, however, it separately estimates an equation for chick-pea. Therefore "other foods 2" does not contain chick-pea. Other foods are thus a category which differs from system to system to include all those foods for which no separate equation is estimated. Restrictions Tests for Functional Form In order to decide which of the three equation systems to use the GL, NQ and TL functional forms were tested to see how well each conforms to the - 12 - symmetry, homogeneity and-adding-up constraints imposed by demand theory. These tests were performed only on equation system A. F-test results are as follows: Summary of Restriction Tests for System A. 1. Generalized Leontief (GL)/. (a) test for symmetry, intercept = 0 and adding-up constraints NUMERATOR : 18.9396 DF: 17 -F Value : 20.4073 DENOMINATOR: 0.9281 DF: 2716 Prob>F : 0.0001 2. Normalized Quadratic (NQ)aE/ (a) test for symmetry, intercept = 0 and adding-up constraints NUMERATOR : 6.4782 DF: 23 F Value 6.7732 DENOMINATOR: 0.9565 DF: 2710 Prob>F : 0.0001 3. Transcendental Logarithmic (TL) (a) test for symmetry, intercept = 0, homogeneity in prices and adding-up constraints NUMERATOR :12.3424 DF: 21 F Value : 12.1009 DNOMINATOR: 1.0200 DF: 2255 Prob>F : 0.0001 a/ The GL and NQ functional forms are automatically homogeneous of degree one in prices and nominal income. For the GL and NQ systems, the reported F-results are tests of the joint hypothesis of symmetry of cross price terms, of the adding up constraints on the income elasticities, and of all intercept terms being equal to zero. In both the GL and NQ systems, homogeneity of degree zero in prices and nominal income of each demand equation is automatically satisfied and cannot be tested. For the TL functional form, however, an additional restriction in - 13 the overall F-tast is the constraint of homogeneity of degree zero in prices. As discussed in section I the separate estimation of the quality demand equation allows us to impose the adding-up constraint for the TL form as well. For all three functional forms, F-test results show that the restrictions taken together are rejected. This could have resulted because of a number of possible reasons, e.g., from the wrong choice of functional form or from errors in variables. Nonetheless, the systems of demand equations derived from these functional forms remain useful, so long as they are sufficiently stable and able to replicate underlying responses of consumers. Since the NQ and TL functional forms give us the lowest F values, we now investigate which of these two systems provides estimates that are closest to a_riori expectations. Table 3 compares, at the mean of the sample, the actual weighted commodity shares from the sample households, with the implied shares from the NQ system- and the predicted shares f =m the TL system, respectively. Quite clearly, the TL predicted expenditure shares lie much closer to the actual mean shares for all commodities. Moreover, for the NQ system, a negative mean quantity is predicted for inferior cereals. Hence a negative share is estimated, an undesirable characteristic of the estimated NQ system. ComDensated Price Elasticities Table 4 lists the compensated price elasticities from the NQ and IL demand systems as computed from mean prices and NQ predicted quantities and TL predicted shares, respectively. Both the NQ and the TL systems estimate 1/ These implied shares are computed .from the NQ predicted quantities and the actual weighted sample mean prices and expenditures reported from System A. - 14 an unexpected positive response of inferior cereals to its own price. In the NQ system the price derivative is positive and highly significant. (An elasticity is not computed since the predicted quantity is negative). In the TL system, a small positive and non-significant elasticity is estimated. The NQ system shows very low own price elasticities for rice and wheat, both with respect to the TL system estimates of the present study and with those of the S&B study,which used more aggregative data. While it is clear that our preference is based on our a priori judgment of reasonableness, we prefer the TL estimates Appendix Tables 1, 2 and 3 report both the restricted and unrestricted demand equations for TL systems A, C and E, respectively, It is the coefficient estimates of these restricted equation systems that are used to compute the elastTcities reported in the remainder of this'taxt. Elasticities, of course, are not constant and would depend on the saple points at which they are evaluated. The coefficient estimates reported in Appendix Tables 1, 2 and 3 therefore are useful for any later simulation exercise that takes changing commodity prices, incomes and commodity shares into account. Tables 4 and 5 report compensated price elasticities for the TL systems A, C and E. Of the 17 compensated awn price elasticities reported in these tables, only three do not have the expected signs. Two of these are for aggregate inferior cereals and are not statistically significant (systems A and E). The only significant positive own price elasticity is that for pearl millet in system C. Out of the 14 own price elasticities with the expected signs, 13 are statistically significant at the .01 level. The compensated price elasticities for each commodity are discussed below. - 15-- Rice. The own compensated price elasticity for rice ranges from a low of -.836 (system A) to -1.035 (system C) and shows the relatively high sensitivity of rice demand to changes in its own price. These estimates are higher than the S&B estimates of -0.582. Rice appears to be a complement to inferior cereals and its commodity components, sorghum and pearl millet. It is also complementary to chick-pea. For each of the remaining commodities considered in systems A, C and E,rice appears to be a substitute. S&B found rice and inferior cereals to be substitutes, not complements as the present estimates imply. Wheat. The own price elasticity for wheat range from -.667 (system A) to -1.263 (system E), as against -0.226 in S&B. As with rice, wheat demand appears to be quite price elastic. In all systems wheat appears to be a substitute for rice, chick- pea, and the catch-all aggregate other crops. It shows significant comple- mentarities with other commodities in both systems A and C. Inferior cereals and its comonents. In syst.ems A and E the compensated own price elasticity of inferior cereals is positive, small and not significant. A priori, a near zero elasticity would not be surprising since the main consumer groups are poor and might have few substitution possibilities as prices change. However, in S&B, a substantial own price elasticity of -0,63 was estimated. Breaking out sorghum and millet in system C results in an insignificant negative own elasticity for sorghum, but a large positive and significant elasticity for pearl millet. The latter is the most troublesome estimate. Chick-pea. This is the only pulse for which quantity data was available, allowing for estimation of a demand equation (system E), This crop - 16 constitutes only a fourth of all pulse consumption, the rest of which is aggregated with certain other commodities to form the other foods aggre- gate of system E. Chick-pea is one of the main sources of protein in the Indian diet, and appears to have a strikingly large own price elasticity (-2.540). It also appears to be a significant complement to rice, and a substitute for wheat. The S&B study was unable to estimate a demand equation for pulses. Other foods. This aggregate performs fairly well with regard to the expected signs and the significance levels of its elasticity estimates. Own price elasticities are nearly constant and range from -.700 (system C) to -772 (system A). This aggregate appears as a substitute for all other commodities. Its demand elasticities with respect to the prices of these commodities are also rather stable across systems. Other commodities. Other commodities appear to have a compensated price elasticity of about -0.5. For all systems, the demand for this aggre- gate appears sensitive to changes in the price of rice and to the price of the mixed aggregate other crops. Cross,elasticities with other commodities are either close to zero or not statistically significant. The Income-Consumvtion Relationships Table 6 reports the predicted expenditure shares by income class. The results are remarkably similar across the three systems. The share of rice consumed declines sharply from between 20 and 30% for the lowest expenditure group to near zero for the highest expenditure group. In the S&B study, the share of rice dropped only at higher incomes, after an initial rise. The share of wheat is substantially more stable and remains at between 5% and 8% for the highest expenditure groups. As expected, the demand for inferior cereals drops from between 25% and 30% for the low income group to near zero - 17 - or slightly negative values for the highest expenditure groups. The share for chick-pea declines from 2% to near zero for the highest expenditure class. In systems A and E the predicted share of other foods starts at around 24% and rises by about 10%. In system C, where other foods contain some inferior cereals, the share stays nearly constant. The strongest positive income effect arises for other comodities which increase from about 13% to about 58%. For each of our TL comodity systems a quality equation was estimated independently, relating quality demand to incomes and family size. This independently estimated equation (see Table 7) was then used to impose the adding-up constraint (at sample means) on the income coefficients of equations for all other commodities. As expected, the share of quality is negative at about -8% for the poorest consumers and goes to roughly +10% for expenditure classes 6 and 7. The predicted share is negative and near zero for the highest expenditure class, an anomaly which arises from the statis- tically significant negative quadratic term in the quality demand relation- ships (Table 8). All in all, the expenditure consumption relationships seem remarkably stable and well estimated. They are qualitatively similar to the S&B estimates, except for the income elasticity of coarse cereals (Table 7). We prefer the NSS estimates, however, because the NSS data is better suited for estimating these relationships. The NSS expenditure estimates are more precise and expenditures vary much more in the data set than in the state level per capita data used by S&B. Table 7 provides the income elasticities across expenditure classes. For rice, inferior cereals and their components, and for chick-pea, negative income elasticities are predicted at higher expenditure ranges. This is a consequence of the quadratic income terms - 18 - which are often statistically significant and cannot simply be suppressed (Appendix Tables 1, 2 and 3). Family Size In Table 8, we see that family size appears to have 'no influence on the demand for foodgrain quality. For the real commodities (Appendix Tables 1, 2 and 3) family size tends to systematically increase the demand for wheat whereas its effect on rice is ambiguous. It tends to decrease the demand for coarse cereals in the all India system A, and sorghum in system C. In the chick pea growing area of system E, demand for inferior cereals also declines with family size, but the effect is not significant. Family size systematically decreases the demand for other foods, but increases the demand for other commodities (except in system E where it appears to have little effect). It is tempting to read economies of scale into these results as the commodities with high income elasticities such as wheat and other commodities appeir to be favored in large families, while the inferior cereals and perhaps rice appear to be reduced. Other foods, however, have a fairly high income elasticity and are nevertheless reduced in larger families, so the issue of economies of scale is not as clear cut. The Uncomnensated Price Effects These reflect both the compensated price elasticities and the income effects (Table 9). The absolute size of virtually all uncompensatad own price effects exceeds that of the compensated ones. High uncompensated price elasticities result for wheat and rice (0.76 to 1.47). The own price elasticity for pearl millet remains problematic. - 19 - IV. DISCUSSION The results of the present study are less encouraging than hoped for as far as the price elasticities are concerned. Restrictions suggested by theory do not appear to hold. Positive own elasticities are estimated for inferior cereals, although they are rarely significant. Elasticity estimates from alternative functional forms differ widely and the test results give only very weak guidance on how to choose among functional forms. Elasticity estimates differ substantially between the S&B and the present study. Given these results it may be advisable to place greater reliance on the price elasticity estimates from the S&B study, where theoretical restrictions were not rejected for the TL form. Furthermore, the state data set used by S&B had greater price variability than the NSS data. The results with respect to the expenditura-consumption relation- ships are fortunately much more encouraging. They already appeared to be well estimated in the S&B study. They also agree closely with a priori expectations, including the demand for foodgrain quality. The demand for foodgrain quality is shown to be sharply increasing with expenditures at low income levels but the income response of foodgrain quality becomes very small for the high expenditure groups. Table 1: orulae for Frtce and Income EastjcitIes NQ System .L system . ... .TL Svltez .. -, pj/2 c - p C c 1. Comvensated n . uC - i . u- . . Cwn Prce x X N Elasticity for , for l for ,.,N... N-1 N-i N-i N- t NN - N 2.'Comensated 0 P.1/2 cc ip Cr0s3 Price i j x p i '27 i Elast:icityiN for im 1 for i 1 . for ±-1,..,T-1 -w ., N, i i j 4z c. N-- • 13 N-1 P. jul iN b +2.l.,1og 2 3.~2 In=e· n ba+ 2 n g EIlastect1 for i .l 1,...,3-1 for i u , - S. N s . j1 im 4. Unco=ensated 3 ,- Own Price ii i. - - Elasti.city Uncompensated fl. Cross Price i Elasticity 1/ This table ts from S& (1981. t oiåUEus are iiLned as folibla (a) na elasticity (compensated) of god : with respect to the prica of good j, (b) C - the (t )h estimated regression coefficent of the system (see equations (7), (8), and (9) for exact particujars), (c) ? price of good L, (d) m per capita real income, (e) si expendiLture share on good i, (f) b and bC - esti=ated regression coe.fficients for income (see equations (7), (8) and (9) for exact particulars), (g) ni - elasricity of good i with respect to real income n, (h) - elasttc±ty (=ncompensated) of good 1 with respect to the price of i good j. Tle 2:_Sste~stimted, Variable Defiotio.s, nits of Measur ment and Weighted vean Vlaues We:ghted Meansl/ Variable Variable DefinitIoa Abbreviation System A (All Indla) System c (So~r~hum- System E7 Pearl Mille t) ~(Ch~ckeaT~ I. Quanrities 1. Rice FISQAGGR 4.1781 2.8023 2.6864 2. Wheat UISQAGG 1.8710 2.0470 3.5749 3. Jowar (Sorghum) FISQAGGJ 0.9853 4. Bajra (Pearl millet) CISQAGGB 0.5456 5. Gram (chickpea) :ISQAGGG 0.1809 6. Inferior Cereals FISQAGGI 1.5047 1.5527 7. Other foods (1) FISQAGG1 8.4422 8. Other 6å:ds. (2) FISQAGG2 7.9084 9. Other i'~oids (3) FISQAGG3 8.8119 10. Odther Cädities FMSQAGGN 8.7325 9.0375 8.6777 11. 7ualiy ~ FISQAGGQ -0.0835 -0.0410 -0.1955 II. Prices 1. Rice FISPADJR 2.4699 2.5340 2.4744 2. Wheat FISPADJW- 2.4060 2.5651 2.3099 3. Jowar (Sorghum) FISTADJJ 2.4412 4. Baj ra (Pearl millet) FISPADJ3 2.0981 5. Gra (Chickpea) FISIADJG 3.5380 6. Inferior Cereals FISIADJI 2.4513 2.5061 7. Other Foods (1) FISPADJI 2.2187 8. Other Foods (2) FISPADJ2 2.5919 9. Other Fcods (3) FISPADJ3 2.2137 10. Oter Commodities FISPADJN 1.5806 1.5334 1.6022 11. Quailty ~nFSPADJQ 2.3672 2.4061 2.3818 III. Other Var:iables 1. Tcome lCOM 24.2512 23.8668 23.6099 2. Number of persons/ family 5.3491 5.3016 5.4972 1/ Weigftts usåd are number oi .tal paesns for each observation. .States.covered: System A:- All 15 Statecr-System-C: Å:;.--ujrat- Mysore, M.P., Yaharastra, Rajhastan, Uttar ?rades. -.System: Bihar, -M.P., Punjab, ttar-Pradesh. Deinitions: Inferior Careals - Sorghum, Pearl Mllet, Haize, Barley, Finger millet, Small miletzs. Other Foods - All three other food aggregates contain pulse products; milk edible oils; meat, fush and eggs; vegerables; sugar; salt; spices; beverages; cereal, subst1.tues. In additton "other foods (1)" contains chick paa whSila "other foods (3)" contains chick pca, maze, finger millet and small millets. 'Tale 3: Actual Weighted Maan Values of Variables and the Predicted - Quantities and Shares Used to Compute Compensated Price Elasticities in the NQ and TL Forms, System A Actual Weighted Predicted Quantities Mean Values and Implied Shares Predicted Shares Variable Name of Variables1/ NQ TL Price Indices Rice 2.4699 Wheat 2.4060 Iniofidr Cereals 2.4513 Otherfbods 2.2187 i-tommodities 1.5806 _a_3___ 2.3672 Ouantitv - ice 4.1781 2.0564 Indices ha't 1.8710 1.4170 InfWid Cereals 1.5047 -0.1889 Other foods 8.4422 9.2441 Otahe mmodities 8.7325 12.1530 quality -0.0835 a.5441 Shares Rice -0.2030 0.1036 .0.2296 Wheat 0.0885 .0.0695 0.0933 Inferior Cereals 0.0725 -0.0094 0.0608 . Other Crops 0.3684 0.4183 0.3507 Other commodities 0.2715 0.3918 0.2656 S6 Quality -0.0039 0.0263 0.0002 INCOME (m) 24.2512 H_ (persons/family) 5.3491 1/ Weights used are number of total persons for each observation. Table 4 : Compensated Price Elasticities for the Normalized Quadratic (NQ) and the Transcendental Logarithmic (TL) Equation Systemse., System A Rice Wheat Inferior Other Other Cereals Foods Commodities A. NQ Equation System A Rice -0.208 0.032 -0.483*** 0.634*** 0.005 Wheat 0.047 -0.066 0.086 0.057 -0.124 Inferior Cereals b/ b/t b/ b/ b/ er foas 0.162*** 0.010 0.098*** -0.386*** 0.116*** Other commodities 0.001 -0.022 -0.058*** 0.124*** -0.038 B. TL Ecuation System A Rice -0.836*** 0.337*** -0.282*** 0.537*** 0.245*** Wheat 0.829*** -0.667*** -0.088 0.102* -0.176** Inferior Cereals -1.068*** -0.136 0.068 0.861*** 0.274* Other foods 0.352*** 0.027* 0.149*** -0.772*** 0.244*** Other commodities 0.212*** -0.061** 0.063* .32*** -0.535*** C. TL Eauations,.Swamv - Binswan&er Study 6 Rice -0.582*** 0.128*** 0.144 0.310 Wheat 0.227*** -0.226* 0.052 0.052 Inferior Cereals 0,243*** 0.049 -0.630*** 0.338*** Other Comodities 0,059*** -0.006 0.038*** -0.091 a/ Elasticities are computed at weighted mean prices and predicted quantities and shares. These elasticities are significant at these defined levels. b/ Coefficient is positive, i.e. of wrong sign. Predicted quantity is negative, therefore elasticity not meaningful. c/ Includes all other foods not estimated in separate equations. Significant at the .01 level Significant at the .05 level * = Significant at the .10 level Table 5 Compenanted Price Etastlctilr for the Transcendentl Logarithmic Equation Syatem ayatems C and E Pearl . Inferior Other Other Rice Wheat Sorglum Mllet Chlckpea Cereala Foods (3) Commod"u StemopuMPearl Millet System Rice -jQUAA 0.652A** -0.016 -0.647A*k 0.825*** 0.224 Wheat 0.881*** -1.08i^* -0.146 -0.038 0.423*** -0.031 Sorghum -0.037 -0.255 -0.18 0.287 0.500*** -0.305 Pearl Hillet-5.283A** -0.232 0.992 2.820A** -1.146** 2.852* Other Foode 0.310*** 0.118*** 0.080"k -0.053** -0.700**t 0.247** Other 0.117* -0.012 -0.068 0.182*** 0.342*A -___* Contmod iti e4 Other Syatem E L Foods (1) Rice 1-.L86*** 1.207*** -0.268*** -1.170*** 0.614*A* 0.480*** Wheat 0.686*** -1jA*** 0.227*** 0.170* 0.376*** -0.196* Chickpea -3.366A** 5.020*** -2j5AQh** 0.529 0.451* -0.094 Inf.Cerealu -1.893*** 0.483* . 0.068 0,2 0.503 0.544** Other Foodo 0.207*** 0.222*k* . 0.012* 0.104*** -f.12L*** 0.179*** Other 0.225*** -0.162* -0.004 0.158** 0.250*** -0.467*** Comnodities a/ Elautlcities computed at weighted mean price and predicted aharea. Theae elasticitleu are aignificant at these defined levels. * Significant at the .01 level. ** SIgnificant at the .05 level. * Significant at the .10 level. Table 6: Predicted Expenditure Shares and Weighted Incomes, by Commodity, by Expenditure Class and by System (TL Equation Estimates) Z-oenditure Grou Mean Expend- 1 237 8 ture Shares Svste= A, All India Predicted Shares s1(rice) 0.3156 0.3034 0.2848 0.2618 0.2257 0.1690 0.0901 0.0239 0.2323 S2(wheat) 0.0924 0.0983 0.0991 0.0999 0.0984 0.0847 0.0618 0.0403 0.0978 S3(inierior cereals) 0.2665 0.1945 0.1599 0.0996 0.0516 0.0215 -0.0036 -0.0125 0.0619 S4(other foods 1) 0.2593 0.2872 0.3073 0.3320 0.3545 0.3768 0.3903 0.3751 0.3502 S5(ather comodities) 0.1410 0.1637 0.1812 0.2196 0.2682 0.3374 0.4527 0.5808 0.2584 S6(quality) -0.0749 -0.0472 -0.0324 -0.0128 0.0015 0.0106 0.0087 -0.0076 -0.0006 DICOMES (M) 6.5914 9.4537 111.8577 17.1725 25.4213 40.2200 74.3563 127.4000 23.6788 Syscem C, Sorghum - pearl millet states Predicted Shares S1(rice) 0.2049 0.1928 0.1853 0.1687 0.1445 0.1142 0.0687 0.0097 0.1505 S2(wheat) 0.0911. 0.1075 0.1100 0.1110 0.1120 0.0086 0.0664 0.0506 0.1125 S3(forgham) 0.2060 0.1580 0.1312 0.0976 0.0616 0.0343 0.0160 -0.0042 0.0672 - -S(pearl. i±llac)- . 0. a-a0 Q.0344 Q.221 Q.2201. .Q291 -0.0007 -0.0134 -0.0219 0.0122 Ss(other foods 3) 0.4121 0.4014 0.3984 0.3952 0.3919 0.3918 0.4004 0.3941 0.3912 S6(other commodities) 0.1314 0.1649 0.187.Q 0.2262 0.2815 0.3529 0.4565 0.5876 0.2706 S7(qualizy)- -0.0895 -0.0590 -0.0413 -0.0189 -0.0014 0.0089 0.0054 -0.0160 -0.0042 INCOZIES (2) 6.6945 9.3651 11.7613 16.7298 25.1207 40.4805 74.3679 128.7790 23,2287 ystem E, Chickpea states Predicted Shares Sl(rice) 0.1958 0.1788 0.1657 0.1412 0.1104 0.0723 0.0320 -0.0154 0.1143 S2(wheat) 0.1895 0.2111 0.2195 0.2196 0.2176 0.1947 0.1277 0.0886 0.2173 S3(chickpea) 0.0210 0.0171 0.0145 0.0114 0.0082 0.0058 0.0019 0.0008 0.0088 S4(inferior cereals) 0.3163 0.2290 0.1776 0.1159 0.0652 0.0345 0.0062 0.0145 0.0727 S5(other foods 2) 0.2308 0.2715 0.2964 0.3289 0.3506 0.3660 0.3666 0.3450 0.3475 S6(otber comoditles) 0.12-34 0.1406 0.1590 0.1961 0.2474 0.3175 0.447 0.5723 0.Z399 S7(quality) -0.0768 -0.0481 -0.0328 -0.0131 0.0006 0.0092 0.0081 -0.0063 -0.0006 MtCOMES ( 5.7906 8.5804 11.0227 16.2080 24.4107 37.8979 72.1733 122.4760 23.2640 К ,� _ . г• + � • ' � т в Tublu 7 г lпситп 81ди[дсJ[1еи by СотачгJl[у, Ьу Bкeendltura CLaaa und Ьу SyaCea ('l'1, ka[lmnten9 � СлдиывJд[у anJ Sуы[еш �/ Ниреnд д сьь гы ------ Сiаыд R1сл црепг 1пЕегlог Сегеаlы Реаг1 - §1]SцioШL 1jljip.t ChlckNeд _-..d11.iцF+�AAde Z, ОЪ�1ес СотаlоддСlеы _ � g � С Е А G g А 8 С С Е А С @ А С Е 1 f о.Вб12 О.о99о о,7434 1.2175 ].328'! 1.2751 о.22д2 o.28Q.i 0.32в9 0.4915 д.б772 1.3894 о.9Э27 1.5425 1.30з4 1.5967 У.1до4 2 I о,е2з5 о.еэ2а о.бо?.4 1.741s 1.2442 1.1]96 о.аб47 о.►Ы7 U.1855"`^* о.з п s""* о,бг57 1.2854 о.932о 1.збз2 1.4267 1.Ы 19 1.зга4 3 i о.79о5 о.7в2о о.б722 1.1о11 7.2155 1.lз13 -о.о44в О.ФЭ74* 0.0692* о.279Э* o.57S7 1.22)8 О.ц]22 1.2754 1.4794 1.6213 1,4/i9Ч 4 . о.7338 о.бв7о о.5229 1.о]b7 1.17д5 1.0679 -о,4з2Э -о.7941%д�0.148в* -о.о09Эй о.4901 1.152о 0.9]28 1.1688 1,5225 1.6179 1.5057 g � о,б444 0.5зIо о.3278 0.9688 1.1зб£. 1.ооо5 r1.26 9-о.5952 -о.б]2о -о.9824*• о.зЭ95* 1.оВ4о о.9зз5 1.о79з 1.5Э77 d.5971 1.5.53 Ь � о.4515 о.2б8о*а"-о.12ва* о,в]од 1.1о15 o.91e9 -з.о2о4 -о:ц4з�`�'1.s3zo е о.14з1* 1.оl4а о.9]5о o.994b 1.52'I6 1.55з] 1.52з1 7 ' о.а1Э7* -o.525U** -1.8776*�*о.652Э 1.о49о о.7162 9 -1.D554*-3.75lS*д* � -1.Э5Э9* о.9312 0.9784 0.8756 1.4967 1.5249 l.4815 � . в �-4.1т97 , 11.абео 8 о.гзеа" о,ц4з9 о.Э475й 8 3.1е58** Q е -з.чб77* о.е526 о.9э91 о.7641 1.4568 1.й7о4 1.4абе mwn т U.6626 о.5735 U.3581 О.Ч811 1.14Э6 1.0085 -о.9616 -0.4865 -о.5276 -о.Ы14 0.ЭВ45 1.о957 о.9Э71 1.о89.4 1.5з73 1.5971 1.5366 S6t1 теап ов4 • • 1.о65 olвnt1c1t1n® о.В7 1•25 , 1/ Eиtlmutea ьлlth [Ьл Q иlgп дгв poadtdvл ипд not аhоып 4есвиие ргвдде[ед дЬдгед ииад Со comput® t11e1o асе negntdvn. Ы1 rлрог[лд еlди[1е1[1ев ees algndflean[ at [Ье .01 ;evu1 ое Lе[[ве ыаlсид ot�ьerulen notcd. 1.в., • *** + Sdgnlflcank в[ thи .05 1eve1. ** � S1gn1f1cиnt дt [i1p .1о 1uvn1. * � Ии[ 81gn1f1ca11t. 2/ DtЬer &'иид rufлra io Jlffвrent сошшодl[у дggгеgи[еи деrоии ыув[ета. , I Table 8 : The Demand for Quality Equation by Commodity System Family Dependent Variable and System IndeDendent Variables Size ln (Income) Un (Income))? ln (H) SYSTEM A S6 (share of quality) 0.1686 -0.0220 -0-.0080 (7.8646) (-6.2112) C-0.5244) SYSTEM C S7 (shaie of quality) 0.2012 -0.0264 -0.0079 (6.5655) (-5.0838) C-0.3036) SYSTEM E S7 (share of quality) 0.1547 -0.0206 -0.0122 (4.8239) C-3.8146) (-0.4900) I/ t-statistics in parentheses. Table 9 1 Unconepentated Price Elasticities for the Transcendental Logrithic Equation Syifteua-/vy System Pearl Inferior Other Rice M *Thunt r11". 11m - hik ereA -oh -----Commoditiea ALL INDIA System .System_A Rice -Q i0.276 -0.3223 0.307 0.071 Wheat 0.605 -fl259 -0.1477 -0.240 ---0.435 Inferior Carpalm -0.844 -0.045 0.127 1.202 0.532 Other Foode ' 0.101 -0.075 0.0828 -1.155 -0.046 Other CommodItie6-0.139 -0.204 -0.0304 -0.215 8Jortn-m.iu lat. 1gates System C Rice" -1.1161 0.5922 -0.0503 -0.6571 0.609 -0.0014 Wheat - 0.7131 -1 106 -0.2172 -0.0589 -0.022 -0.352 Sorghum I 0.0567 -0.1853 -0,1460 0.2982 0.750 -0.125 Pearl '11let -5.2690 -0.2212 0.9982 2,8212 -1.108 2.879 Other Foods- 0.1740 0.0171 0.0220 -0.0695 -1.062 -0.014 Other Comodi iea-0.1146 -0.1834 -0.1658 0.1540 -0.273 -1L001 Chickpeai System1. Sygtem E Ace ),-02 1.132 -0.2709 -1.1963 0.486 0.388 iheat " 0.569 -11M 0.2177 0.0972 0.027 -0.446 Chickpea -3.415 4.935 -2.5438 0.4985 0,306 -0.197 Inferior Cerealf -1.838 0.579 0.0723 . 0.3293 0.666 0.661 Other Fooda 1 0.080 -0.001 0.0020 0.02611 -1.101 -0.091 Other Commoditioq 0.047 -0.475 -0.0177 0.0478 -0.280 -0.4i a/ Elasticitles computed at weighted mean prices and incomes and predicted sharea. Ap£enIIdix T4ble I t taLricteLd mind Unrcårtrccd Coutteiclnt Eøuttwac ut thi Oulplit Dbman4 quatfona. Ti. Systmul Å (Al[ India) Independntll Vurabla»21 Reatrieced EsLimates Uopundont 1n(FISPADJR)- lo(FISPAD.MH- 1n(FTSPADj1)- ln(FKSPAD11)- jn(FISPADJN)-2 Varlaba ln(FISPABJQ) In(FISPADJQ) In(FISPADJQ) In(FISPADJQ) ln(FISPAD.1Q) 1infINCOME) 1-(1n(WC0HE) 2 In (I) INTERCEPT Sl(rice) -0.0152 0.0558 -0.0788 0.0429 -0.0047 0.0072 -0.0135 0.0388 (-0.6250) ( 4.9853) (-4.5836) ( 4.0252) (-0.3831) ( 0.1875) (-2.1353) ( 1.4118) S2(whca) 0.0559 0.0224 -0.0139 -0.0232 -0.0411 0.0524 -0.0085 0.0478 ( 4.9853) ( 2.3342) (-1.5134) (-3.9176) (-5.1008) ( 2.6580) (-2.6415) ( 3.4194) . 3(intior cereala) -0.0787 -0.0139 0.0612 0.0309 0.0004 -0.3298 0.0329 -0.0551 (-4.5836) (-1.5134) ( 3.2398) ( 3.3861) ( 0.0466) (-9.3025) ( 5.6351) (-2.1915) S4(othcr 0.0428 -0.0232 0.0309 -0.0429 -0.0076 0.2005 -0.0263 -0.0463 f0,14) ( 4.0252) (-3.9176) ( 3.3861) (-4.9987) (-1.0622) ( 7.1902) (-5.7708) (-2.3892) Ss(øiur -0.0047 -0.0411 0.0004 -0.0076 0.0530 -0.0989 0.0375 0.0229 commnditieu)(-0.3831) (-5.1008) ( 0.0466) (-1.0622) ( 3.9161) (-4.8789) (11.2202) ( 1.5691) lInrefitrIcl.L,d Eitimate SI(rice) -0.1388 -0.0215 -0.0697 0.0355 -0.0087 -0.0011 -0.0130 0.0278 -0.0280 (-2.2630) (-0.9814) (-3.1383) ( 2.4116) (.-0.3319) (-0.0277) (-1.9633) ( 0.9815) (-1.0328) 32(wis,at) 0.1677 0.0052 -0.0278 -0.0067 -0.0612 0.0593 -0.0092 0.0566 0.0208 ( 5.3057) ( 0.4645) (-2.4280) (-0.8919) (-4.5226) ( 2.8650) (-2.7094) ( 3.8770) ( 1.4890) 83(inferior -0.3851 0.0550 0.0293 -0.0143 0.0024 -0.3153 0.0289 -0.977 -0.0085 cereale) (-6.7875) ( 2.7116) ( 1.4270) (-1.0532) ( 0.0993) (-8.4865) ( 4.7113) (-3.7264) (-0.3414) 84(ather 0.2326 -0.0713 0.0510 -0.0594 0.0539 0.1828 -0.0232 -0.0313 -0.0057 fOda) ( 5.1978) (-4.4570) ( 3.1449) (-5.5249) ( 2.8093) ( 6.2372) (-4.7989) (-1.5130) (-0.2887) 0.1015 0.0092 0.0099 -0.0136 0.0705 -0.0991 0.0380 0.0308 0.0154 commuodie LI) ( 3.1297) ( 0.7956) ( 0.8470) (-1.7519) ( 5.0742) (-4.6662) (10.8169) ( 2.0574) ( 1.0749) a/! £ - Staitamica in p.aruIena. In Appendix Table 2 1 Rreicted and ULrUricted Coetficient Esimate of the Output Demand Equations. TL System C (Sorghui - Pearl HIlet System) Indepcnden Variables Restricedw Estimates Dependent ln(FISPAFIMS)- -(ISPADJW)- ln(VISPADJJI- ln(VISPADJB)- ln(FISPADJ3)- ln(FISPADJH)- 2 Varlables ln(FISPADJQ) In(FISVADJQ) ln(FISPAnJq) l(FISADJQ) InVISPAD.Q In(FISPADq ln(IICOHE) (In(INCOME)) -In-AR INTERCEPT SI(rice) -0.0265 0.0795 -0.0113 -0.0972 0.0637 -0.0081 0.0456 -0.0174 -0.0450 (-0.7981) (5.2412) (-0.4410) (-5.1586) ( 3.7682) (-0.4081) ( 1.1913) (2.6842) (-1.3626) 82(wheat) 0.0795 -0.0210 -0.0225 -0.0060 0.0038 -0.0336 0.0510 -0.0055 0.1041 ( 5.242) (-1.5309) (-1.5328) (-0.5524) ( 0.3443) (-2.8113) ( 1.9849) (-1.2699) ( 4.6998) 03(Soighum) -0.0113 -0.0225 10466 0.0166 0.0069 -0.0362 -0.1926 0.0142 -*0.0755 (-0.4410) (-1.5328) ( 1.4817) ( 0.9093)- ( 0.3852) (-1.7052) (-4.8180) ( 2.1033) (-2.1748) S4(Pearl Hil1t) -0.0972 -0.0060 0.0166 0.0681 -0.0274 0.0460 -0.0262 0.0010 -.0.0005 (-5.1586) (-0.5524) ( 0.9093) ( 28368 (-2.4500) ( 2.6388) (-1.3638) ( 0.3170) (-0.0306) 65(otker foodu) 0.0637 0.0038 0.0069 -0.0274 -0.0341 -0.0128 -0.03%1 0.0006 -0.0514 ( 3.7682) ( 0.3443) ( 0.3852) (-2.4500) (-1.7535) (-0.8778) (-0.6857) ( 0.0851) (-1.3644) 86(ather -0,0081 -0.0336 -0.0362 0.0460 -0.0128 0.0449 -0.0488 0.0334 0.0763 commodities (-0.4087) (-2.8113) (-1.7052) ( 2.6388) (-0.8778) ( 1.7473) (-1.8904) ( 7.6165) ( 3.4167) Unrstricted Eac lwas S1(rIe) -0.1688 -0.0473 -0.0411 -0.0647 0.0710 -0.0436 0.0546 -0.0204 -.0.0684 -0.0264 (-2.9642) (-1.9221) (-0.6815) (-1.2599) ( 2.6492) (-0.9583) ( 1.3434) (-2.9644) (-2.0301) (-0.9534) 82(llear) 0.2404 -0.0318 -0.1373 -0.0179 0.0418 -0.0920 0.0571 -0.0058 0.1273 0.0207 ( 6.2814) (-1.9249) (-3.3335) (-0.5209) ( 2.3218) (-3.0030) ( 2.0877) (-1.2489) ( 5.6241) ( 1.1148) S3(SorIhum) 0.0084 0.0690 0.1006 0.0183 -0.0969 -0.1783 -0.2132 0.0185 -0.0103 0.0131 ( 0.1425) ( 2.6834) ( 1.5716) ( 0.3427) (-3.4641) ( 3.7481) (-5.0162) ( 2.5744) (-1.9993) ( 0.4534) S4(oarl Hillet) -0.1036 0.0055 0.0320 0.0630 -0.0319 0.0547 -0.0328 0.0021 -0.0006 -0.0126 (-3.6261) 4 0.4514) ( 1.0396) ( 2.4433) (-2.3755) ( 2.3937) (-1.6075) ( 0.6174) (-0.0392) (-0.9062) 65(other foods) -0.0286 -0.0246 0.0534 -0.0373 -0.0082 -0.1430 -0.0202 -0.0026 -0.0777 -0.0158 (-0.4382) (-0.8733) ( 0.7599) (-0.6332) (-0.2676) (-2.7376) (-0.4329) (-03368) (-2.0116) (-0.4998) 56(oLlwr couedicfua) 0.0236 0.0048 -0.0471 0.0550 -0.0417 0.1000 -0.0453 0.0334 0.0820 0.0297 0.6155) ( 0.2909) (-1.1396) 4 1.5903) (-2.3110) 4 3.2563) (-1.6521) 4 7.1780) ( 3.6116) 4 1.5924) А1гl�anJ1B'[иЫе 3t 8иwе[1е[ед дид Ue,raacaleted CuefflelшtL бас[шисиы uf гl�в Out�uC Dewund L'уииг[оnыр TL 5ув[еш Е(L'loic&иав $увtлт) - 1nJг��uaJвnt V;1r4nb1e �/ иаы[гlcceJ 4'ысlта[са Uерепдг:и[ !а�(NISNADJN)- !о(FISPRDJ4t)- lu(PISPAll.1L')- 1n(F151'AD31)- 1п(FISPAD.12)- !п(FISYADJH)- vвriцt�lu !п(Етsело.lg) ]и(F1SPAO.)4) !п(F1st�лn.uгZ In(вlsелиз�) 1n(I:est�nn.1 1п(t�tseлn.гg) 1п(ъисон� (ьп(тисон8D)2 ,�,_,(ц� lиrеис�е•е S1(гlси) O.ooz2 о,1 иа -о.озzз -o.14s1 О.дз1I о.д269 -о.о2о9 -О.оовз -О.огез { О.ОS9з) ( з.зе48) {-4.п ьо) (-ь,2ззs) ( 2•2l54) ( l.18з2) (-О,йз77) {-ь.оэЭS) (-о.7з77) S2(uhь:acD O.1170 -О.Оуб2 O.Ok41 0.02о0 R.0058 -О.д913 0.1d56 -о.01В0 0.1214 ( Э.Эаад) (-2.о228) ( 4.b012) ( 0.д249) ( о.ч4sэ) (-Э.7з92) ( 2.9аг7) (-2.72ьlэ ( з.э74s) &з(�1�lckpcu) 0.03't3 0.04й7 -0.014Э 0.0042 o.00U') -О.о0з1 -О.00В4 Q.д00й -О.Ооз1 (-4.U 6о) t 4.бо12) (-1.9so4) ( O.es22) ( о.з2д7) (-o.s443) (-l.zo10) ( о.4о2з) (-0.5Э20) S4(infnriлr enrnвla> -о.1451 о.ого0 о.о042 О.Оеыз о.0l1i о.а2lз -о,з72д о.ойао -о.оlд5 (-4.2�75) ( О,д249} ( о.д522) ( з.464s) ( l.oo74D ( 1.l4ох) (-9.езз7) ( 6.597з) (-о.ы 2s) S5(other fuudn) 0.03l1 0.0058 О.Ф009 о.0111 -О.о247 -0,24з 0.2441 -0.0338 -0.0470 ( 3.2►54) ( 0.445Э) ( 0.Э2д2) ( 1,0о74} (-2.ЭЭдй) (-1,9403) { 0.4t8o) (-t.24d9) (-1,SЬз0) 86(и[!�¢г слшwлдltЪев) 0.0269 -0.091� -0.0071 9,021з -0.0241 0.0]ОЬ -О,Л 22 О.о3В2 -0.0122 ( 1.18Э2} (-3.7392) (-о.sч4Э) ( 1.1402) (-1.940з) ( 2.7243) (-з.3tбб) ( 6.7222) (-0.4516) � Unrcatrlcted Евtlаш[св � � S1(rice) -0.114д 0.0220 -O,219069 0.01у2 д.д1i! 0.0516 -О.П'160 -о.о07Э -0.02з0 -О.Оз84 � (-1,7979) ( D.1б97) (-2,D587) ( 0.507Л) ( O.S9d4) ( 1.2614) (-0.4980) (-о,8274) (-0.5509) (-Э,0295) S2(uticat) 0.299l92 -0.2695 0.05з4 -0.0671 0.0064 -0.0555 0,1158 -0.о174 O.i377 11.оЭ04 ! ( 5.7256) (-2.5409) ( 1.1dk8) (-2.1571) ( 0.7749) (-1.6608) ( 2.7дй9) (-2.4144) ( й.д2з5) ( O.UD44) i S7(�1�lckpen) -U.0247 D.0285 � -0.01з0 •О.ООз2 -О.RОд8 0.00007 -O.U097 0.0005 -О.дОз4 -0.0019 (-3.1д8]) ( 1.5Dб1) (-1.b222) ( д.587у) (-0.2672) ( 0.01Э1) {-L 2205) ( о.4565) (-д.559з) (-0.7629) i 54(lnfurlur сегепlа) -О.ЬЭ16 д.2tl20 О.Оз61 0.07Z2 -о.02уз O.O149 -0.Э'Ltl3 0.0Э13 -0.1248 -0.о07Э 'j (-1z.ччхд) ( 2,747ь) ( о,ехыд) ( l.дтде) (-l.т944) ( 0.4614) (-7.9274Ъ ( 4.4ьэ4) (-з.7ьд5) (-д,z49э) � S5(othcr елп,lд о,2756 -0.3179 o.1ss9 -а.о47з О.о0sь -о.071т О,z24з -О.озоз! -0.019s -о.о0ы � ( 5.4ЭОг) (-з,ОдS3) ( з.5588) (-I.5654D ( 0,]4Э6) (-2.2063) ( S.ЭЧЬЭ) (-й.Э120) (-0.58В5) (-0.2060) I Sбlnlhвr cuвvuдdlCina) 0.1305 -0.00Э3 -0.0914 o,U21D -О.д075 0.О76В -U.1136 О.ОЗ9Ч О.U1Ьд O.UiSU ( 2.9004) (-0.ЭЬ4) (-2.3szs) ( 0.7845) (-о.s2хь> ( 2.ььб0) (-з.ад11) (�6.4о2г) ( n.s7021� ( o.s7дs) 1 1 ., � н/ С - SСв[tвtli:r 1п r. � � `� � , � � ' � � ' ` • � п �