66305 H N P D i s c u s s i o N P a P e R THE IMPACT OF THE FOOD PRICE CRISIS ON CONSMPTION AND CALORIC AVAILABILITY IN PAKISTAN: Evidence from Repeated Cross-sectional and Panel Data Jed Friedman, Seo Yeon Hong, Xiaohui Hou November 2011 THE IMPACT OF THE FOOD PRICE CRISIS ON CONSUMPTION AND CALORIC AVAILABILITY IN PAKISTAN: Evidence from Repeated Cross-sectional and Panel Data Jed Friedman, Seo Yeon Hong, Xiaohui Hou November 2011 Health, Nutrition and Population (HNP) Discussion Paper This series is produced by the Health, Nutrition, and Population Family (HNP) of the World Bank's Human Development Network (HDN). The papers in this series aim to provide a vehicle for publishing preliminary and unpolished results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For information regarding the HNP Discussion Paper Series, please contact the Editor Martin Lutalo at mlutalo@worldbank.org or go to www.worldbank.org/hnppublications. © 2011 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 All rights reserved. ii Health, Nutrition and Population (HNP) Discussion Paper The Impact of the Food Price Crisis on Consumption and Caloric Availability in Pakistan: Evidence from Repeated Cross-sectional and Panel Data Jed Friedmana, Seo Yeon Hongb, Xiaohui Houc a DECRG, The World Bank, Washington D.C., USA b SASHD, The Word Bank, Washington D.C., USA c SASSP and WBIHS, The World Bank, Washington D.C., USA Paper prepared for the Pakistan Social Protection Programmatic Policy Notes The World Bank, Washington D.C., USA, November 2011 Abstract: Welfare losses from the 2008 food price crisis in Pakistan are deepening the gap between poor and nonpoor populations and further increasing inequality between the provinces. To estimate welfare losses, the reduction in caloric availability at household level is measured. The analysis of calorie intake by source supports the notion that rural households were shielded from the worst effects of the crisis by their capacity to grow their own food. Compensating variation estimates suggest that the average household would need 38 percent of its total precrisis expenditure to maintain precrisis consumption levels. The impact of the food price crisis (measured as the percentage of total expenditure required to restore consumption to the precrisis level) peaked at the end of 2008 to twice as high as at the start of the year. Average household caloric availability fell by almost 8 percent between 2006 and first half of 2008. Urban households were relatively worse off than rural households during the crisis. Income gains from sales of agricultural commodities produced by rural households presumably offset the negative impact of the food crisis to some degree. The drawdown of assets over 2008–10 was another important coping mechanism, especially for households without access to land. Keywords: Food Crisis, Compensating Variations Disclaimer: The findings, interpretations and conclusions expressed in the paper are entirely those of the authors, and do not represent the views of the World Bank, its Executive Directors, or the countries they represent. Correspondence Details: Xiaohui Hou, World Bank (1818 H. St. NW, Washington DC), 202-473-7773, xhou@worldbank.org 1 Table of Contents FOREWORD..................................................................................................................... 3 ACKNOWLEDGMENTS ................................................................................................ 4 1. INTRODUCTION......................................................................................................... 5 2. DATA AND METHODS .............................................................................................. 7 2.1. DATA ......................................................................................................................... 7 2.2. METHODS .................................................................................................................. 8 2.2.1. Compensating Variation .................................................................................... 8 2.2.2. Changes in Caloric Availability Prior to and During the Onset of the Crisis 10 2.2.3. Impact of Price Changes on Caloric Availability, Panel Household Analysis 11 3. COMPENSATING VARIATION ............................................................................. 12 3.1. BUDGET SHARES AND FOOD PRICE INCREASE ........................................................... 12 3.2. COMPENSATING VARIATIONS .................................................................................. 14 3.2.1. Compensating Variations for Each Composite Food Catogory...................... 14 3.2.2. Compensating Variations with Substitution Effects ........................................ 17 3.2.3. Compensating Variations with Consumption from Self-produced Food ........ 18 4. CHANGES IN CALORIC AVAILABILITY........................................................... 21 4.1. CHANGES IN CALORIC AVAILABILITY, 2005-08 ....................................................... 21 4.1.1. The Changes in Caloric Availability ............................................................... 21 4.1.2. Urban and Rural Differences .......................................................................... 22 4.2. THE WHEAT PRICE AND CHANGES IN CALORIC AVAILABILITY OVER 2008-10 ....... 28 4.1.1. The Changes in Caloric Availability ............................................................... 28 5. CONCLUSIONS AND POLICY IMPLICATIONS ................................................ 32 REFERENCES ................................................................................................................ 34 APPENDIX ...................................................................................................................... 37 2 FOREWORD As part of its technical assistance to the Government of Pakistan, the World Bank initiated the Pakistan Programmatic Social Protection Policy Notes in 2009. The main objective of these notes was to facilitate effective design, development, and implementation of a national safety net system under the Benazir Income Support Program (BISP). In addition, these notes were meant to inform overall social protection policy in Pakistan. Opinions expressed in these policy notes are entirely those of the authors. They do not represent official views of the Executive Directors of the World Bank, the Trust Fund Donors, or the Government of Pakistan. 3 ACKNOWLEDGMENTS The team warmly thanks the government counterparts, especially the Benazir Income Support Program, Federal Bureau of Statistics, and the Ministry of Finance, for facilitating technical work and inputs into various policy notes. We also thank the UK Department of International Development (DfID), the Trust Fund for Environmentally & Socially Sustainable Development (TFESSD), and the Gender Action Plan (GAP) Trust Fund for their financial contribution to this task. The authors would like to thank Dean Mitchell Jolliffe for his insightful written comments and Ning Ma and David Haddad for their assistance. The authors also thank Andrea Vermehren, Amjad Zafar Khan, Cem Mete, Iftikhar Malik, Julie McLaughlin, Khalid Mahmud, Mansoora Rashid, Pablo Gottret, Shumaila Rifaqat, and other members of the Pakistan Human Development team for their comments and support in the overall context of Social Protection work in Pakistan. 4 1. INTRODUCTION Global food prices have almost doubled over the past seven years (FAO Food Price Index). From 2007 to 2008 alone, the price of wheat and rice rose by 121 percent and 76 percent, respectively (World Bank 2010a). In many cases, these price rises have persisted. Domestic food prices in 45 developing countries remained higher in 2009 than before the crisis (FAO 2009). The rising cost of food is expected to worsen poverty and further limit human development in many developing countries. The 2008 food price crisis is projected to worsen poverty for a significant portion of the 2.3 billion people living on less than US$ 2 per day, and another 100 million are likely to join the numbers of the poor (Ivanic and Martin 2008). In addition to the direct impact on monetary measures of poverty, the reductions in real income caused by rising food prices could worsen nutrition, limit the use of educational and health services, and cause the poor to deplete their few productive assets. A recent study found that calorie intake and household dietary diversity declined in Afghanistan during the 2008 food crisis (D’Souza and Jolliffe 2010). Evidence from past crises clearly indicates that children suffer long-term health consequences from short-term shocks (Alderman, Hoddinott, and Kinsey 2006; Friedman and Sturdy 2010). Although the evidence is mixed, aggregate economic shocks may reduce educational attainment because the opportunity cost of keeping children in school may increase (Ferreira and Schady 2009). Rising food prices may also force poor households with low coping capacity (because of inadequate credit markets, for example) to invest less in their livelihoods or sell their livestock and other productive assets (Carter and others 2004; Lokshin and Ravallion 2000; Cater and others 2007). Pakistan, like much of the developing world, was a victim of the global food crisis. From 2005 to the end of 2008, the domestic wheat price rose by 106 percent. Prices of other staple foods rose between 20 and 120 percent, while nonfood prices rose only 17 percent on average. 1 The price of wheat in Pakistan was affected not only by the global food crisis but by domestic hoarding and smuggling of wheat, caused by the relatively low wheat price in Pakistan compared to neighboring countries during 2007. To prevent farmers from hoarding and smuggling, Pakistan increased the procurement price of wheat. However, by 2009 this practice had caused the domestic price of wheat to rise beyond the international wheat price (Ministry of Finance 2008). The floods that coursed through Pakistan in 2010 are expected to reduce wheat production and further raise prices. For Pakistanis, high wheat prices are particularly critical because wheat accounts for nearly 50 percent of the mean daily calorie intake. 2 A large relative change in the wheat price has substantial implications for levels of poverty and human development in Pakistan. 1. Based on the price change calculation described in the results section. 2. Based on PSLM 2005–06. 5 To respond to rising food prices, in 2008 the Government of Pakistan initiated the Benazir Income Support Program (BISP), which provides cash transfers of Pakistan Rupees 1,000 to poor families identified by the government on the basis of its poverty scorecard census. The program’s primary objective is to mitigate the impact of the food crisis on consumption. Given the continuing volatility of wheat prices in Pakistan, a greater understanding of the past and continuing impact of the food crisis is valuable for informing the discussion on whether and how the country should scale up its safety net programs. We focus on variations in the wheat price, because wheat is the single most important commodity for a typical Pakistani household. Using three rounds of the Pakistan Social and Living Standard Measurement Survey (PSLM), from 2005–06, 2007–08, and 2010, this note examines the distributional and welfare impact of the food crisis through the following measures: (1) The compensation required following the 2008 price increases to return households to consumption levels prevailing before the crisis (2005–06) 3 Estimated impact of the food crisis varies depending on how the analysis accounts for households’ capacity to substitute less expensive commodities for those previously consumed, and households’ capacity to produce selected food commodities themselves. For that reason, this note presents a range of estimates. It also attempts to trace welfare changes across the entire period of food price increases in 2008 rather than at a particular point in time. (2) The loss of caloric availability over the course of the initial food price crisis between 2006 and 2008. The change in available calories at the household level will vary by location and by the availability of productive assets such as agricultural land. (3) The changes in caloric availability in relation to changes in the wheat price between 2008 and 2010. The analysis benefited from recently collected panel data to trace the behavior of households as they coped with sustained price increases over that period. The analysis helps to identify which groups have been most seriously affected by Pakistan’s food price crisis. Studying the impact of rising food prices over time rather than at one particular moment provides a more complete understanding of how the impact of the crisis has unfolded among different groups. In addition to analyzing the welfare impact in monetary terms, this note looks at changes in a simple yet critical dimension of welfare: household calorie consumption. Findings from the caloric availability analysis highlight the differential ability of urban and rural households to cope with the food crisis. They emphasize the protective effect of households’ capacity to produce their own food when prices are rising rapidly. 3. The compensating variation is expressed as a percentage of total household expenditure 6 2. DATA AND METHODS 2.1. DATA The main data sources used for the analysis are three rounds of the PSLM, a nationally representative survey covering 14 large cities and 81 districts, including urban and rural areas. The survey records information on a range of social issues, including education, employment, immunization, women’s decision making, and household consumption. In the consumption module, in addition to 79 nonfood items, the survey collects data on the quantities and values of 58 consumed staple food items while distinguishing between purchased, self-produced, and gifted, and in-kind quantities of staple foods. The three survey rounds used in the analysis were fielded in July 2005–June 2006 (the precrisis period); July 2007–June 2008 (the onset of the food price crisis); and January– June 2010 (after a period of sustained price increases). 4 Households interviewed in January–June 2008 were interviewed again in January–June 2010 within the same calendar quarter. The first two rounds of data are relevant for analyzing welfare before the crisis and during its onset, since food prices began increasing in late 2007 and continued to rise throughout 2008. On the other hand, the panel nature of the last two rounds of data (January–June 2008 and 2010) from PSLM allows us to identify how price changes affected welfare at the household level, and provides detailed insight into households’ coping mechanisms. To measure the effect of the price shocks, individual food and nonfood items in the PSLM are matched to the retail prices of individual commodities obtained from Pakistan’s Federal Bureau of Statistics (FBS). Since 1991, FBS has collected retail price data on 374 goods and services covering 10 fairly exhaustive consumption categories from 71 markets in 35 cities. 5 The 58 staple food items in the PSLM were matched to the retail price data from FBS (the food items are listed in Appendix Table 4). The present analysis aggregated the individual food items into 11 food groups: dairy products; meat; fruit; vegetables; spices; wheat; rice; other cereals; pulses; oil; and other food. Price changes for each item were aggregated using the budget share of each item within its group as a weight. For example, the dairy product category includes five different food items, such as milk and yogurt, each with its own price change. The price change of the dairy product was calculated as 4. Questionnaires and sample design in all rounds are consistent. Summary statistics in both rounds are shown in Appendix Table 1. 5. The 10 nonfood groups are: food and beverages; apparel, textiles, and footwear; house rent; fuel and lighting; household furniture and equipment; transport and communication; recreation and entertainment; education; cleaning, laundry, and personal appearance; and medical care. 7 the weighted average of the price change of each constituent item. The price of the aggregate nonfood category was the weighted price of the corresponding nonfood items. To reflect the geographic variation in food prices, each district in the PSLM data was matched to the nearest urban center where the retail price data for both food and nonfood items were collected. Given that retail prices in rural areas are matched to prices in urban centers, caution should be used when interpreting the results, because price changes in rural and urban areas can differ. Rural impacts may be overestimated if price changes in a rural area were smaller than those in the nearest urban area (the opposite is also true; rural impacts can be underestimated). Based on the matched information about price changes and household consumption of the 12 aggregate goods (11 food and 1 nonfood), we estimate the compensating variation and changes in caloric availability between the two study periods. Lastly, self-reported consumed quantities of 58 staple food items in the PSLM were converted to caloric measures based on nutritional values in Pakistan to represent household-level caloric availability. The food consumption module in the PSLM utilizes either a fortnightly or monthly recall, depending on the item; fortnightly measures were converted to monthly consumption. 6 Because PSLM separately collects information on consumed quantities of purchased foods, self-produced foods, and foods obtained as a form of wage, the caloric availability from purchased foods and self-produced foods could be identified. Caloric information was also disaggregated by source for the analysis into calories obtained from grain, vegetables and fruit, animal products, and other sources. 7 2.2. METHODS We use two methods to understand the impact of the food crisis on household welfare. The first method estimates the compensating variation, as suggested by Friedman and Levinsohn (2002), to understand the magnitude of the food price crisis in terms of the additional expenditure necessary to maintain a reference level of welfare. The second method is a regression analysis using pooled cross-sectional and panel data to capture the effects of the food price crisis on household caloric availability. 2.2.1. Compensating Variation Welfare changes arising from the food price crisis are proxied by the estimated compensating variation (CV), which is defined as the income necessary to compensate the household in order to maintain its precrisis utility level. Following Friedman and Levinsohn (2002), we use precrisis information on household consumption along with price data over the crisis period to forecast the distributional consequences of relative 6. See Appendix Table 4. 7. Grains include wheat, rice, other cereals, and pulses; animal products include dairy products and meat;and other food includes spices and oils. 8 price changes as captured by the CV. These projections can then be assessed for accuracy and updated as post crisis household data become available. The main method is summarized below. Compensating variation (CV) is derived using the first-order Taylor series expansion around the minimum expenditure function, e( p, u ) , with respect to price, p, at a fixed baseline utility level, u0. ∂e( p, u 0 ) CV = e( p 0 , u 0 ) − e( p1 , u 0 ) = ( p1 − p 0 ) ∂p ∂e( p, u 0 ) Simply replace with q , the consumption quantity, and the CV can be expressed ∂p as a price change and baseline consumption quantity. CV = q∆p Reformulating this equation, the CV can also be expressed as a function of the budget share ( wi ) and proportional price changes. Equation 1 CV n ∆pi q p n = ∑ wi , where wi = n i i ∑q p ∑ qi pi i =1 pi i i i =1 i =1 The summation function is indexed over the n consumption goods available to the households. This first-order Taylor expansion does not take into account the possibility that a household can substitute one good for another, depending on the relationship between the two goods and the magnitude of the relative price change. However, the ability to substitute particularly costly goods with less costly items could mitigate the negative impact of the food price crisis. To capture the substitution effect in the CV calculation, the second-order Taylor expansion with some algebraic manipulation can be expressed as: Equation 2 CV n ∆pi 1 n n ∆pi ∆p j n = ∑ wi + ∑∑ wi ε ij ∑q p i =1 pi 2 i =1 j =1 pi p j i i i =1 This expression includes the own-and cross-price elasticity terms, ε ij , which is the price elasticity of good i with respect to good j for each of the n goods, and incorporates the substitution behavior of households attributable to price increases in one good over another. 9 The own and cross-elasticity of each good is estimated using the methodology developed by Deaton (1997), which utilizes the geographic variation in the unit value of a good. As only spatial variation is utilized to identify the elasticity terms, elasticities are estimated using the second-round PSLM (2007–08), which is the more relevant period for estimating the demand changes, given the large spatial divergence in prices after 2007. 8 Additionally, a households’ ability to produce certain important foods may lessen the negative effect of the food crisis on the household’s welfare. We allow for this possibility in robustness exercises that treat consumption from self-production as a negative expenditure and re-estimate the compensating variations, following Friedman and Levinsohn (2002). 2.2.2. Changes in Caloric Availability Prior to and During the Onset of the Crisis The method just described leverages information on consumption patterns measured before the crisis. This second method employs information collected during the crisis to examine information on one key welfare indicator before and during the crisis. We use pooled cross-sectional data—incorporating 2005–06 (round 1) and 2007–08 (round 2)— to measure changes in caloric availability between the two rounds. Caloric availability is one indicator of household welfare that is presumably directly affected by the food price crisis. As the relative price of food increases, households substitute, at least partially, the foods experiencing the most rapid price increases, with corresponding changes in caloric availability for the household. In addition, a price increase may cause households that consume but do not produce food to experience a negative real income shock, whereas food-producing households may experience a positive income shock, which may in turn affect total calories consumed. Since we do not observe the same households over the first two rounds of the PSLM, it is a challenge to determine the exact food consumption changes between the two periods; food prices vary by region and season within Pakistan. However, food prices should be substantially more stable within individual districts and quarters of the year. By controlling for the district and quarter of the observation through fixed effects, however, we can compare changes in average household caloric availability between the two rounds. Based on the PSLM survey design, we compare the two observed quarters in 2005 with the corresponding quarters in 2007 (July–September and October–December), and the two quarters in 2006 with the corresponding quarters in 2008 (January–March and April–June). The following regression equation is estimated: Equation 3 ln(Calorie hqtd ) = γ 0 + γ t T + γ q Q + γ tq (T * Q) + γ x X hqtd + γ d D + ehqtd 8. Results are qualitatively the same if the first-round PSLM is used to estimate the elasticity terms (results available upon request). 10 The log calories consumed is indexed by household h measured in quarter q, survey round t, and residing in district d. A complete set of dummy indicators for survey round, T, quarter, Q, and district, D, are included in the specification. The coefficient of the interaction term between T and Q, γtq, captures the changes in caloric availability over the crisis period. In the absence of a food price crisis, we would expect γtq either not to be significantly different from zero, reflecting no change in caloric availability, or to be slightly positive if the caloric consumption within the season and district increases as a result of economic growth. We include quarter dummies to control for the seasonality of food consumption and possible variation in prices. The regression also includes household-level characteristics, X htq , to better balance the cross-round comparison by controlling for observable household demographic and socioeconomic information. Since the food price crisis started at the beginning of the year 2008, we would expect γtq to be the most negative in the first and second quarters of 2008 as opposed to the last two quarters of 2007. In addition, we examine changes in calories available from purchased goods compared to self-produced goods. Since the decision to consume self-produced goods is due to economic considerations in producing households, which can in turn be affected by the food crisis, we model the changes in calories available from self-produced goods using a two-stage model. First, the probability of consuming from self-production is estimated using a Probit specification; then, based on consumption from self-production, the change in caloric availability is estimated using Ordinary Least Squares (OLS). If indeed the capacity to produce food partially protects households from the price increases, access to agricultural land can be an important determinant of welfare. For this reason, the analysis distinguishes between urban and rural households and, within rural areas, between landed and landless households. Lastly, we investigate changes in dietary composition between the two periods. Since the price of grain increased more dramatically than the prices of other goods, we would expect calories obtained from grain to decrease the most during the crisis. We estimate the caloric availability from different sources: grain, animal products, vegetables and fruit, and others. 2.2.3. Impact of Price Changes on Caloric Availability, Panel Household Analysis This method benefits from the panel aspect of the PSLM and makes use of household- level information collected in the shock and post-shock periods. As in the previous method, in this approach the measure of welfare is caloric availability, and geographic variations in price changes between two periods help to identify household-level responses to food shocks. We focus on variations in the wheat price, because wheat is the single most important commodity for a typical Pakistani household (on average, 48 percent of total caloric availability is from wheat), and the wheat price data show a great deal of spatial and temporal variation. 11 Using PSLM household-level panel data in 2008 (January–June) and 2010 (January– June), a household-level fixed effect model is employed to identify the impact of wheat price changes on household caloric availability. In addition to investigating the average impact of wheat price changes, we investigate the potential mitigating role of land accessibility using the following specification. Equation 4 ln(Calorie ht ) = β 0 + β1 ln(Wheatpricedt ) + β 2 ln(wheatpricedt ) * Land t −1 + ϕ ′X t + α h + ε ht Since the same households (h) are observed in two periods (t and t-1) using the household fixed effect model, the effect of changes in the district-level (d) wheat price on changes in caloric availability at the household level can be identified. β1 captures the impact of changes in the district-level wheat price on changes in caloric availability for households with no access to land in 2008. β 2 shows the differential impact of wheat price changes on the caloric availability in households with and without access to land. The same outcome measures investigated in the preceding analysis are investigated here: total caloric availability; caloric availability from purchased foods; caloric availability from self-produced goods; and the probability of consuming self-produced goods. The linear probability model is used to estimate the probability of consuming self-produced goods. 3. COMPENSATING VARIATION 3.1. BUDGET SHARES AND FOOD PRICE INCREASE Among food categories, dairy products have the highest mean household budget share (table 1): On average, 16.5 percent of household expenditure went to dairy products before the food crisis. The second most important food good in terms of budget share was wheat, at 13 percent of household expenditure. The next most important aggregate goods are spices (including sugar) at 9.15 percent of the household budget, meat at 7.17 percent, vegetables at 6.98 percent, and oil at 6.79 percent. The budget share of each aggregate food category varies across the income distribution. The poor spend proportionally more on wheat, oil, spices, and vegetables, but less on meat, milk, and fruit. Wheat consumption accounts for about 20 percent of the total household expenditure for the poorest 10 percent (as defined by per capita household expenditure) and only 4.7 percent for the top 10 percent of households. The top 10 percent of households spend a significant share on nonfood items: 43 percent compared to 22 percent for the average household; the difference is statistically significant. 9 9. Standard errors are reported for the comparison between the bottom 10 percent and the top 10 percent of households in the per capita household expenditure distribution. 12 Table 1. Budget Share of Individual Aggregate Goods and Their Price Changes Budget Share (%), 2005–06 Price Change (%) All Bottom 10% Top 10% All households households Dairy 16.47 12.22 (0.29) 14.68 (0.24) 42.01 Meat 7.17 5.09 (0.20) 9.43 (0.22) 24.52 Fruit 2.50 1.38 (0.06) 3.41 (0.08) 42.11 Vegetables 6.98 9.06 (0.14) 4.10 (0.08) 9.38 Spices 9.15 11.16 (0.13) 5.66 (0.11) 13.10 Wheat 12.99 20.83 (0.24) 4.67 (0.11) 79.74 Rice 2.59 2.93 (0.12) 1.58 (0.05) 173.19 Cereals 0.17 0.13 (0.01) 0.17 (0.01) 5.99 Pulses 1.89 2.37 (0.06) 1.09 (0.03) 43.24 Oil 6.79 8.12 (0.11) 4.31 (0.09) 86.88 Other foods 1.45 0.73 (0.05) 2.74 (0.08) 12.14 Nonfood 27.88 21.80 (0.32) 43.33 (0.48) 16.55 Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS Note: Standard errors in parentheses for the comparison between bottom 10 percent and top 10 percent; the price change is between the month that the household was surveyed in 2005-2006 and June, 2008. This price change is used in the CV calculation. Most staple food prices increased gradually until early 2008, started to rise dramatically in early to mid-2008, and remained high until the end of the year. Figure 1 shows proportional price changes (in percent) between 2005–06 and 2007–08 10 for six aggregate staple goods. 11 Among the six staple goods, the price of wheat increased most dramatically; by December 2008, it was 106 percent over the 2005–06 price. 10. In price change calculations, the baseline price is the retail price prevailing in the month when the PSLM 2005–06 respondents were interviewed about their food consumption. 11. Dairy products, wheat, spices, meat, vegetables, and oil are the six most important aggregate food goods in terms of budget share (from high to low). 13 Figure 1. Proportional Price Changes (%) Compared to 2005–06 for Six Selected Aggregate Goods 120 100 80 Percent 60 40 20 0 July, Oct, Jan, Apr, Jul, Oct, 07 07 08 08 08 08 dairy product meat spices wheat oil vege Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS The distributional implications for the price changes depend on precrisis consumption patterns as well as the magnitude of the relative price changes between goods. For example, as household expenditure increases, the budget share of both wheat and oil decreases, suggesting that increased prices for these two goods will affect the poor disproportionately. The price of rice increased 173.2 percent from 2005–06 to June 2008, but the impact may not have been particularly significant, given that rice accounts for only 2.6 percent of Pakistani household budgets. Dairy products and meat are more important goods for the nonpoor than for the poor, and their prices rose modestly, implying a relatively smaller impact on richer households. 3.2. COMPENSATING VARIATIONS 3.2.1. Compensating Variations for Each Composite Food Catogory Table 2 presents estimates of the CV for each aggregate food good as a percentage of household expenditure based on budget share and price changes of aggregate goods estimated in June 2008. The most prominent impact of the food crisis was the increased wheat price. The average household needed to spend an additional 10.5 percent of its total expenditure on wheat prior to the crisis to maintain its precrisis level of wheat consumption. Poor households were affected much more than others by rising wheat prices: The bottom 10 percent of households would need compensation of 17.9 percent of household expenditure to maintain their precrisis wheat consumption level, whereas the top 10 percent of households would need only 3.8 percent. Rice and oil price increases also affected poor and rich households differently, primarily because of their relative 14 importance in the household budgets of the poor. Price changes of other aggregate goods, on the other hand, affected the poor and the rich to a similar extent. Table 2. Compensating Variation for Each Aggregate Good (Percentage of total household expenditures) All Poor Bottom 10% Top 10% households households Dairy 7.05 5.25 (0.15) 6.18 (0.12) 5.70 Meat 1.86 1.26 (0.05) 2.34 (0.06) 1.35 Fruit 1.33 0.71 (0.06) 1.99 (0.09) 0.75 Vegetables 0.62 0.90 (0.05) 0.25 (0.03) 0.86 Spices 1.10 1.40 (0.06) 0.79 (0.03) 1.38 Wheat 10.52 17.85 (0.34) 3.77 (0.09) 16.30 Rice 4.77 5.14 (0.24) 2.99 (0.01) 5.36 Cereals 0.03 0.01 (0.00) 0.04 (0.00) 0.02 Pulses 0.84 1.00 (0.04) 0.52 (0.02) 0.95 Oil 5.78 7.37 (0.12) 3.35 (0.06) 7.11 Other foods 0.25 0.10 (0.01) 0.59 (0.03) 0.11 Nonfood 4.12 3.67 (0.07) 4.78 (0.07) 3.74 Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS Note: Based on the price changes measured in June 2008 compared to 2005–06. Poor households are defined as being at or below the 100 percent poverty line in 2005 and 2006. Standard error is in parentheses for the comparison between the bottom 10 percent and top 10 percent. The compensating variations incorporating the price changes of all goods are shown in Table 3. Average households need 38 percent of their total expenditures (Rs 3,801 per month) to recover their precrisis utility level, without considering substitution to relatively cheaper goods. Again, poor households generally were affected more than nonpoor households. Poor households would need to be compensated for 44 percent of their total expenditures (an average amount of Rs 2,616 per month), whereas nonpoor households would need only 37 percent of total expenditures. The food price crisis affected rural households more than urban ones, as long as their prospects for producing their own food was not taken into account. Urban nonpoor households would require compensation of 36 percent of their total expenditure and rural nonpoor households 38 percent to recover their precrisis patterns of consumption. Among the poor, urban households were affected to a slightly greater degree by the food crisis; they had an estimated CV of 46 percent versus 43 percent for the rural poor. 15 Table 3. Compensating Variation (%), without Substitution Effects All Nonpoor Poor households All households 38.3 37.2*** (0.11) 43.6*** (0.35) Urban 36.6 35.7*** (0.16) 45.6*** (0.42) Rural 39.1 38.0*** (0.15) 43.2*** (0.41) Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS Note: Measured in percentage of precrisis household expenditure. Based on price changes measured in June 2008 compared to 2005–06. Difference between poor and nonpoor is significant at 0.01 level (***). Standard errors are in parenthesis for the comparison between nonpoor and poor households. Food prices of staple goods rose continuously and did not peak until October 2008. Depending on when we measure the price changes, the welfare effects of the food crisis might differ. Previous results were based on prices in June 2008, the middle of the food crisis. Figure 2 shows the estimated CVs over time. Compensating variation estimates move along with overall price increases; there is a gradual increase in CV estimates to February 2008, and a dramatic increase over the rest of year. The peak, when the highest level of compensation was needed to maintain precrisis consumption levels, was in October 2008. Over time, the food price crisis disproportionately affected the poor. For the poor and nonpoor alike, welfare changes were similar in July 2007, but the gap widened: In June 2008, the estimated CV for poor household is 10 percentage points higher than nonpoor household, and this difference persisted until the end of the year (Figure 2). The distributional gap arising from this cumulative change in welfare might be much larger than the welfare change estimated at one point in time. The CV also exhibits substantial geographic variation, 12 with the widest variation occurring between Punjab and Balochistan (Figure 3). Households in Balochistan were the most adversely affected over the whole period, and KPK was hit hard at the end of the year. The difference in CV changes between Punjab and Balochistan was around 17 percent of total household expenditure in June 2008; the CV for households in Punjab is estimated at 37 percent of total household expenditure and in Balochistan at 54 percent. This regional gap persisted to the end of 2008 and was more severe than the gap between the poor and nonpoor throughout Pakistan. 12. PSLM adopted a two-stage stratified sample design and kept sample size at approximately 17,600 households comprising 1,252 sample villages/enumeration blocks, which is expected to produce reliable results at the provincial level (see http://www.statpak.gov.pk/fbs/content/methodology-4). 16 Figure 2. Welfare Changes During the Crisis Year for the Poor and Nonpoor 70 % of Household's Total Expenditure 60 50 40 30 20 10 0 08 0 7 09 0 7 10 0 7 11 0 7 12 0 7 01 0 7 02 0 8 03 0 8 04 0 8 05 0 8 06 0 8 07 0 8 08 0 8 09 0 8 10 0 8 11 0 8 12 0 8 8 00 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 ,2 , 07 All Households Non Poor Poor Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS Figure 3. Welfare Changes During the Crisis Year, by Region % of Household's Total Expenditure 70 60 50 Punjab 40 Sindh 30 KPK Balochistan 20 10 0 07, 2007 08, 2007 09, 2007 10, 2007 11, 2007 12, 2007 01, 2008 02, 2008 03, 2008 04, 2008 05, 2008 06, 2008 07, 2008 08, 2008 09, 2008 10, 2008 11, 2008 12, 2008 Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS 3.2.2. Compensating Variations with Substitution Effects When prices rise disproportionately among food goods, households substitute cheaper foods for more expensive foods. To capture this substitution effect, the own and cross- price elasticities of each good are estimated using geographic variations in quality and quantity (Deaton 1997). Own and cross-elasticities of composite goods are estimated using the second round of PSLM data, and these elasticities are given in Appendix Table 2. Own-price elasticity estimates are reported on the diagonal of the tables. For example, the own-price elasticity of wheat in 2007–08 is estimated to be –0.430, which is in line with the global estimates of elasticities of key staple commodities. 17 Figure 4. Compensating Variations, with and without Substitution Effect .5 Compensating Variation .2 .3 .1 .4 6 7 8 9 10 Log of per capita HH expenditure w/ sub, Urban w/ sub, Rural w/o sub, Urban w/o sub, Rural Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS When the substitution effect is included in the CV estimates, the compensating variation is substantially smaller than the estimates without the substitution effect. On average, households need 26.9 percent of their total expenditure to recover their previous utility level from consumption. The distributional effect of the food crisis on welfare diminishes in absolute terms when substitution effects are considered, as seen by the changes in slope in Figure 4, which shows nonparametric regression lines of CV as a function of baseline household per capita expenditure. Four lines are presented, two each for urban and rural areas, both with and without substitution effects. When the prices of goods that poor households relied on went up substantially, they substituted cheaper goods for more expensive ones. Substitution lessens the negative effects of high food prices for poor households more than it does for nonpoor households, but the distributional consequences are unchanged: Poorer households still require a greater proportion of their original household expenditure to maintain their consumption. As for urban–rural differences, urban households were affected more across the income distribution, but only marginally. 3.2.3. Compensating Variations with Consumption from Self-produced Food The effects of the food price crisis may have been tempered by the capacity of many rural households and some urban households to produce some of the food they consume. The mean budget share of self-production was only 3.1 percent for urban households but 21 percent for rural households in 2005–06. Table 4 shows the relative importance of household production for a range of commodities. Rural households, on average, produce 47.3 percent of the dairy products they consume and 34.3 percent of the wheat, whereas urban households produce only 5.8 percent of the dairy products they consumer and 3.9 percent of the wheat. Since dairy products and wheat together accounted for approximately 30 percent of the household budget, a household’s ability to produce these goods was likely to lessen the shock of the food crisis. 18 Table 4. Share (%) of Commodities Produced by the Household for Its Own Consumption All Urban Rural households households households Dairy 32.8 5.8 47.3 Meat 9.2 2.0 13.5 Fruit 1.4 0.7 1.8 Vegetables 2.7 0.6 3.7 Spices 1.7 0.6 2.3 Wheat 23.9 3.9 34.3 Rice 9.8 2.0 14.1 Cereals 1.9 0.9 2.5 Pulses 2.1 0.6 2.8 Oil 5.2 1.2 7.4 Other food 8.0 5.0 10.1 Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS To account for the protective role of household food production during the food crisis, an alternative CV was estimated that treats the consumption of self-produced goods as negative expenditures. This change has the greatest implications for welfare effects on rural households (Figure 5). The CV for urban households at any level of household resources is virtually identical regardless of whether their own food production is treated as a negative expenditure. In contrast, the welfare implications for rural households are dramatic. The estimated CVs accounting for the protective effects of household food production in rural areas are approximately 40 percent less at every level of baseline household resources. Although the CV remains higher for poor rural households than for richer rural households, this result shows that self-production buffers many rural households from outside shocks. It indicates that the greatest vulnerability to the food price crisis may be found among poor urban households as well as poor rural households that cannot produce their own food. This result does not mean that rural households would have consumed the same quantity of self-produced goods as they did before the price crisis. Given the opportunity to sell their agricultural goods at higher prices, households might consider selling rather than consuming the food they produce for themselves. The next section explores the changes in caloric availability from purchased goods and self-produced goods during the crisis. The analysis reveals how rural households responded to the price changes in more detail. 19 Figure 5. Compensating Variations without Self-Production Compensating Variation, with and without self-production .1 .2 .3 .4 .5 .6 .7 .8 6 7 8 9 10 Log of per capita HH expenditure w/o self-prod, Urban w/o self-prod, Rural w/ self-prod, Urban w/ self-prod, Rural Source: Authors’ Calculation based on PSLM 2005-06 and retail prices from FBS; the vertical line is the poverty line in 2005-2006 20 4. CHANGES IN CALORIC AVAILABILITY 4.1. CHANGES IN CALORIC AVAILABILITY, 2005-08 4.1.1. The Changes in Caloric Availability Turning from projections of impact based on consumption patterns before the food price crisis, we look at estimates based on observed changes in caloric availability. Table 5 depicts the initial regression results for caloric availability. The dependent variable in the first column of results is the log of total calories consumed by the household. The second column looks at calories from market purchases only, while the third investigates log calories from self-produced food. The final column uses a probit specification to explore the probability that a household will obtain any of its calories from food it has produced. The key coefficients in each panel are the final three rows that give the relative change in caloric availability across rounds within the same quarter. Caloric availability in the second round is significantly less and declined over time. In the two quarters of 2008 (January–June, the period most affected by rising food prices), average caloric availability was almost 8 percent less than in the same quarters of 2006. Table 5.: Total Change in Caloric Availability Pr(caloric Log(caloric Log(Caloric Log(Total Caloric availability from availability from Availability from Availability) self-production>0), purchased food) Self-production) Probit Year 07/08 0.042** (0.008) 0.034*** (0.011) 0.225*** (0.051) –0.037*** (0.012) Quarter 2 0.030** (0.008) 0.030*** (0.011) 0.025 (0.050) –0.037*** (0.012) Quarter 3 0.030** (0.008) 0.057*** (0.011) 0.059 (0.051) –0.037*** (0.012) Quarter 4 0.027** (0.008) 0.062*** (0.011) 0.082 (0.053) –0.038*** (0.012) Year 07/08 X –0.044*** (0.011) –0.020 (0.016) –0.141* (0.073) –0.027 (0.017) Quarter 2 Year 07/08 X –0.075*** (0.011) –0.034** (0.016) –0.308*** (0.074) –0.014 (0.017) Quarter 3 Year 07/08 X –0.078*** (0.011) –0.018 (0.016) –0.307*** (0.077) –0.022 (0.017) Quarter 4 Sample size 30,823 30,803 12,779 30,827 Source: Authors’ Calculation based on PSLM 2005-06 and 2007-08 Note: Variables included in the regression but not presented are the household heads’ gender, marital status, age, and education level; household size; number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years or older; and district indicators. Robust standard errors are in parentheses. The incremental effect at the mean is reported for the probability of having positive caloric availability from self-production. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. 21 To better understand the mechanisms through which households reduced average caloric consumption, we analyze caloric availability from purchased food and from self- production. Surprisingly, there is relatively little change in caloric availability from purchased foods, even when the food price increased dramatically in the first two quarters of 2008. Caloric availability from purchases is significantly lower (3 percent lower) in the first quarter of 2008 but not in any other quarter. Strikingly, the caloric availability from self-produced foods decreased dramatically. In the first and second quarters of 2008, households that consumed food they produced reported obtaining about 30 percent fewer calories from this source than in the equivalent period in 2006. This finding is intriguing because the incidence of caloric consumption from self-production did not change; households were no less likely to consume any calories from home production during the crisis quarters, as seen in the probit results in the last column of the table. This result implies that households may have preferred to sell the food they produced rather than to consume it in order to reap the income gains from higher food prices. Further analysis on caloric availability will explore this possibility. 4.1.2. Urban and Rural Differences As suggested by the CV results for self-production, urban and rural households may have different strategies for coping with the food crisis, given that rural households have greater access to resources for producing their own food. If their access indeed has a protective effect, the implications for social protection policy in the wake of the crisis are clear. Table 6 repeats the caloric analysis from Table 5 but stratifies the results by urban or rural location. Total caloric availability (the first panel in the table) fell more significantly among rural households than urban households, particularly in the first two quarters of 2008. During those quarters, urban households consumed an average of 3.3 percent and 4.5 percent fewer calories, while rural households consumed 10.3 percent and 10.1 percent fewer calories. However, the sources of reduced consumption differ across rural and urban households. Calories from market purchases fell significantly in urban households in the last quarter of 2007 and second quarter of 2008, presumably reflecting food price increases. Market-purchased calories were also lower in the first quarter of 2008, but this effect is not a precise estimate. The decline in market-purchased calories likely reflects the reduction in real incomes resulting from the crisis. Rural households purchased fewer calories only in the first quarter of 2008. On the other hand, calories obtained through self-production were dramatically lower in rural areas—more than 35 percent lower in the first two quarters of 2008. The same parameters for urban households are not significantly different from zero. While the probability of any calorie intake from self-production is largely unchanged, these results may suggest that rural households focused on diverting their production from the household to the market. 22 Table 6. Change in Caloric Availability, Urban and Rural Households Pr(Caloric Log(Caloric Log(Caloric Log(Total caloric availability from availability from availability from self- availability) self-production>0), purchased food) production) Probit Urban Year 07/08 0.033** (0.014) 0.029* (0.016) 0.193 (0.171) –0.012* (0.007) Quarter 2 (Oct~Dec) 0.048*** (0.014) 0.027* (0.016) 0.132 (0.154) –0.011* (0.006) Quarter 3 (Jan~Mar) 0.038*** (0.014) 0.021 (0.016) 0.160 (0.159) –0.011* (0.007) Quarter 4 (Apr~Jun) 0.047*** (0.014) 0.046*** (0.016) 0.224 (0.167) –0.012* (0.007) Year 07/08 X Quarter 2 –0.048** (0.019) –0.038* (0.022) –0.093 (0.238) –0.032*** (0.012) Year 07/08 X Quarter 3 –0.033* (0.019) –0.024 (0.022) –0.136 (0.235) –0.006 (0.011) Year 07/08 X Quarter 4 –0.045** (0.019) –0.048** (0.022) –0.045 (0.248) 0.003 (0.011) Sample size 12,413 12,405 1,743 12,417 Rural Year 07/08 0.048*** (0.009) 0.036** (0.016) 0.233*** (0.053) –0.041*** (0.015) Quarter 2 (Oct~Dec) 0.018* (0.009) 0.033** (0.016) 0.008 (0.052) –0.041*** (0.016) Quarter 3 (Jan~Mar) 0.025*** (0.009) 0.083*** (0.016) 0.052 (0.054) –0.041*** (0.015) Quarter 4 (Apr~Jun) 0.014 (0.009) 0.073*** (0.016) 0.063 (0.055) –0.040*** (0.015) Year 07/08 X Quarter 2 –0.041*** (0.013) –0.006 (0.022) –0.148** (0.076) –0.009 (0.022) Year 07/08 X Quarter 3 –0.103*** (0.013) –0.039* (0.022) –0.355*** (0.077) –0.019 (0.021) Year 07/08 X Quarter 4 –0.101*** (0.013) 0.004 (0.022) –0.356*** (0.080) –0.049** (0.021) Sample size 18,410 18,398 11,036 18,410 23 (Continued) Pr(Caloric Log(Caloric Log(Caloric Log(Total caloric availability from availability from availability from self- availability) self-production>0), purchased food) production) Probit Rural and landless Year 07/08 0.050*** (0.012) 0.071*** (0.019) 0.217** (0.092) –0.091*** (0.019) Quarter 2 (Oct~Dec) 0.012 (0.013) 0.035* (0.019) -0.020 (0.090) –0.088*** (0.019) Quarter 3 (Jan~Mar) 0.035*** (0.012) 0.099*** (0.018) 0.044 (0.091) –0.089*** (0.019) Quarter 4 (Apr~Jun) 0.034*** (0.012) 0.087*** (0.018) 0.061 (0.092) –0.090*** (0.019) Year 07/08 X Quarter 2 –0.014 (0.018) –0.014 (0.026) -0.258* (0.134) 0.032 (0.028) Year 07/08 X Quarter 3 –0.086*** (0.017) –0.086*** (0.025) -0.202 (0.131) 0.036 (0.027) Year 07/08 X Quarter 4 –0.082*** (0.017) –0.066*** (0.025) -0.140 (0.133) 0.033 (0.027) Sample size 11,158 11,149 4,230 11,158 Rural with access to land Year 07/08 0.048*** (0.013) 0.002 (0.022) 0.154*** (0.048) 0.026*** (0.009) Quarter 2 (Oct~Dec) 0.023* (0.013) 0.063*** (0.022) -0.057 (0.048) 0.021*** (0.007) Quarter 3 (Jan~Mar) 0.035*** (0.013) 0.037 (0.023) 0.109** (0.050) 0.020*** (0.006) Quarter 4 (Apr~Jun) 0.028** (0.014) –0.024 (0.023) 0.149*** (0.051) 0.022*** (0.007) Year 07,08 X Quarter 2 –0.063*** (0.018) –0.042 (0.032) -0.001 (0.069) –0.018 (0.015) Year 07,08 X Quarter 3 –0.124*** (0.019) –0.062* (0.033) -0.293*** (0.071) –0.002 (0.016) Year 07,08 X Quarter 4 –0.117*** (0.020) –0.002 (0.034) -0.235*** (0.075) –0.039** (0.016) Sample size 7,252 7,249 6,806 6,930 Note: Variables included in the regression but not presented are the household heads’ gender, marital status, age, and education level; household size; number of female and male household members between zero and five years old, six and fifteen years, Sixteen and fifty-five years, and fifty- six years or older; and dummies for districts. Robust standard errors are in parentheses. Marginal effect is reported for the probability of having positive caloric availability from self-production. The incremental effect at the mean is shown. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. Household is defined to have access to land if it owned, rented, or sharecropped agricultural land. 24 During the food price crisis, rural households consumed much less of the food they produced. Although one explanation is that they may have diverted food from consumption to sale, another possible explanation is that wheat production fell, either because of a production constraint or the increasing opportunity cost of producing something else. Table 7 shows cropping patterns during the two survey rounds. It indicates the relative likelihood that a household will report producing a given crop in the second survey round compared to the first. Table 7. Crop Production Pr(Producing crop), Log(Amount Probit harvested, kg) Wheat: Year 07/08 -0.064*** (0.007) 0.088*** (0.028) Cotton: Year 07/08 -0.042*** (0.005) -0.027 (0.056) Sugarcane: Year 0.007** (0.003) 0.080 (0.100) Rice: Year 07/08 -0.011** (0.005) -0.064 (0.051) Source: Authors’ Calculation based on PSLM 2005-06 and 2007-08 Note: Variables included in the regression but not presented are the household head’s gender, marital status, age, and education level; household size; number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years old or older; and dummies for districts. Robust standard errors are in parentheses. Marginal effect is reported for the probability of producing crop. The incremental effect at the mean is shown. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. The results show that rural households were 6 percent less likely to report producing wheat, 4 percent less likely to report producing cotton, and 1 percent less likely to report producing rice. On the other hand, households were 1 percent more likely to report producing sugarcane. This result is consistent with the changes in agricultural production reported in the Pakistan Economic Survey, 2007–08; farmers shifted from growing wheat and cotton to sugarcane because of relatively high sugarcane prices, elevated fertilizer prices, and the lack of irrigation water during the rabi growing season (Ministry of Finance 2008). 13 However, the implications of this shift for farm income are unclear. The income for households that continued producing wheat grew 9 percent more, on average, than in the previous period. The increase in wheat production is indicated in the second column of Table 7, which shows results of the regression of the quantity of crops produced (conditional on growing a positive amount) on the period indicated (Year 2007–08). 13. Pakistan has two crop seasons. Kharif begins in April–June (sowing) and ends in October–December (harvest). Rabi begins in October–December (sowing) and ends in April–May (harvest). Kharif crops include rice, sugarcane, and cotton; rabi crops include wheat, gram (pulses), and tobacco (Ministry of Finance 2008). 25 Further analysis is required, but one conclusion is that shifting agricultural patterns alone cannot explain a mean 30 percent reduction in calories from self-produced food. If real incomes were severely affected by changes in crop choice, we would expect to observe a decline in calories through market purchases, but rural households reported no such decline. The analysis of caloric availability from different sources reveals that food price increases affected not only total calorie intake but its composition. Table 8 presents changes in caloric availability by type of commodity. The largest price increases were for grain, especially wheat, and grain saw the largest declines in overall consumption. On average, households consumed 10 or 11 percent fewer calories from grain in 2008 than during the equivalent period in 2006. Urban households consumed less grain, but the declines among rural households were particularly steep. Rural households reported consuming 13 and 14 percent fewer calories from grain during the two crisis quarters of 2008 because they had cut back on consuming their own production. Urban households report consuming fewer calories from vegetables and fruits, indicating that overall food consumption was declining for these households as their real income declined. Rural households do not report a similar decline, which one would expect if they had experienced a decline in real income and reduced consumption across a spectrum of foods. Rural households may not have been as badly off as urban households during the crisis, even though they consumed less of the food they produced. Presumably the income gains from agricultural sales offset the caloric decline to some degree. 26 Table 8. Dietary Changes All Urban Rural Log(caloric availability from grain) Year 07/08 0.012 (0.009) 0.005 (0.015) 0.017 (0.011) Quarter 2 (Oct~Dec) 0.026*** (0.009) 0.056*** (0.015) 0.006 (0.011) Quarter 3 (Jan~Mar) 0.045*** (0.009) 0.082*** (0.015) 0.021* (0.011) Quarter 4 (Apr~Jun) 0.046*** (0.009) 0.079*** (0.015) 0.024** (0.011) Year 07/08 X Quarter 2 –0.039*** (0.013) –0.037* (0.022) –0.045*** (0.015) Year 07/08 X Quarter 3 –0.104*** (0.013) –0.073*** (0.021) –0.128*** (0.015) Year 07/08 X Quarter 4 –0.110*** (0.013) –0.061*** (0.022) –0.144*** (0.015) Log(caloric availability from vegetables and fruit) Year 07/08 0.006 (0.011) 0.044** (0.018) –0.018 (0.014) Quarter 2 (Oct~Dec) 0.074*** (0.011) 0.083*** (0.018) 0.067*** (0.014) Quarter 3 (Jan~Mar) 0.095*** (0.011) 0.088*** (0.018) 0.102*** (0.014) Quarter 4 (Apr~Jun) 0.072*** (0.011) 0.080*** (0.018) 0.068*** (0.014) Year 07/08 X Quarter 2 –0.019 (0.016) –0.068*** (0.025) 0.016 (0.020) Year 07/08 X Quarter 3 0.000 (0.016) –0.011 (0.025) 0.003 (0.020) Year 07/08 X Quarter 4 –0.036** (0.016) –0.068*** (0.025) –0.016 (0.020) Log(caloric availability from other foods) Year 07/08 0.081*** (0.009) 0.080*** (0.014) 0.081*** (0.011) Quarter 2 (Oct~Dec) 0.015* (0.008) 0.011 (0.014) 0.017 (0.011) Quarter 3 (Jan~Mar) 0.006 (0.008) 0.002 (0.014) 0.010 (0.011) Quarter 4 (Apr~Jun) –0.004 (0.008) 0.002 (0.014) –0.008 (0.011) Year 07/08 X Quarter 2 –0.023* (0.012) –0.029 (0.019) –0.019 (0.015) Year 07/08 X Quarter 3 –0.049*** (0.012) –0.018 (0.019) –0.069*** (0.015) Year 07/08 X Quarter 4 –0.022* (0.012) –0.027 (0.019) –0.018 (0.015) Log(caloric availability from wheat) Year 07/08 0.018* (0.009) 0.000 (0.015) 0.029* (0.011) Quarter 2 (Oct~Dec) 0.012 (0.009) 0.033** (0.015) –0.002** (0.011) Quarter 3 (Jan~Mar) 0.040*** (0.009) 0.066*** (0.015) 0.023** (0.011) Quarter 4 (Apr~Jun) 0.046*** (0.009) 0.069*** (0.015) 0.029* (0.011) Year 07/08 X Quarter 2 –0.031** (0.013) –0.026 (0.022) –0.038*** (0.016) Year 07/08 X Quarter 3 –0.110*** (0.013) –0.059*** (0.022) –0.145*** (0.016) Year 07/08 X Quarter 4 –0.096*** (0.013) –0.041* (0.022) –0.132*** (0.016) Note: Variables included in the regression but not presented are the household heads’ gender, marital status, age, and education level; household size; number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years or older; and dummies for districts. Robust standard errors are in parentheses. Marginal effect is reported for the probability of having positive caloric availability from self-production. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. 27 4.2. THE WHEAT PRICE AND CHANGES IN CALORIC AVAILABILITY OVER 2008-10 4.1.1. The Changes in Caloric Availability The previous section showed how caloric availability changed from 2005 through 2008. The analysis in this section explores how caloric availability changed as wheat prices continued to rise between 2008 and 2010. The wheat price increased dramatically during 2008 and continued to increase through much of 2010, but the price changes were much more dramatic in some districts than others. Figure 6 depicts the time path of district wheat prices at the median, at the 10th percentile (p10), and at the 90th percentile (p90). 14 This figure shows how the regional gap in wheat prices has widened since 2008 when the wheat price began its sharp climb. We use the geographic variation in wheat prices over time to identify how rising wheat prices may have affected the welfare of individual households (measured in caloric availability), as well as the coping mechanisms—such as sales of assets—they may have used to smooth consumption. Figure 6. Wheat Price Trend and Geographical Variations, June 2005 to June 2010 340 290 240 190 140 90 2005/12 2006/06 2006/12 2007/06 2007/12 2008/06 2008/12 2009/06 2010/01 P10 Median P90 Source: Authors’ Calculation based on retail price of wheat in 35 urban cities Table 9 contains the elasticity estimates for wheat prices in relation to caloric availability. For households without access to agricultural land in 2008, elasticity is estimated to be 0.02 yet not statistically significant (first row of the first column, Table 9). On the other hand, among households with access to agricultural land in 2008, the elasticity estimate increases by 0.117 and is statistically significant. These point estimates imply that the 40 14. Ten percent of districts are at or below the 10th percentile price level, and 90 percent of districts are at or below the 90th percentile price. 28 percent increase in wheat price observed at the median over 2008–10 resulted in a 5 percent higher caloric availability for households with access to land compared to households without access. Similar to the finding in the cross-sectional analysis, this analysis shows that households with access to land consumed less of their own production and purchased more of their food. A 40 percent rise in the wheat price causes these households to consume 25 percent more calories from the market and 54 percent fewer calories from their own production (if any). Compared to households without access to land, these households are 13 percent less likely to consume any self-produced food. Both urban and rural households exhibit similar patterns, yet the estimates for the rural households are larger and more precise. 15 Results from the cross-sectional analysis and panel analysis consistently indicate that when food prices rise dramatically, agricultural households generate income by selling grain or other staple foods and at the same time buy more food from the market. Table 9. Total Change in Caloric Availability, 2008–10, Household Fixed Effect Pr(Saloric Log(Caloric Log(Caloric availability from Log(Total caloric availability from availability from self-production>0), availability) purchased food) self-production) LPM All Log(Wheat Price) 0.019 (0.016) –0.156*** (0.039) 1.746*** (0.294) 0.136*** (0.035) Log(Wheat Price) X Access to Land in 2008 0.117*** (0.030) 0.616*** (0.043) –1.349*** (0.217) –0.326*** (0.039) Sample size 12,911 12,905 4,641 12,911 Urban Log(Wheat Price) 0.006 (0.025) –0.016 (0.030) 0.217 (0.408) 0.025 (0.024) Log(Wheat Price) X Access to Land in 2008 0.074 (0.105) 0.680*** (0.126) –0.658 (0.582) –0.595*** (0.102) Sample size 4,971 4,967 670 4,971 Rural Log(Wheat Price) 0.032 (0.021) –0.219*** (0.034) 1.126*** (0.206) 0.068** (0.032) Log(Wheat Price) X Access to Land in 2008 0.111*** (0.033) 0.699*** (0.052) –1.488*** (0.239) –0.318*** (0.049) Sample size 7,940 7,938 3,971 7,940 Source: Authors’ Calculation based on PSLM 2008 and 2010 Panel Note: Variables included in the regression but not presented are the household head’s gender, age, and education level and the number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years or older. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. 15. This difference arises largely because of the relative size of the two populations: 32 percent of rural households have access to agricultural land, whereas only 5 percent of urban households have access. 29 Changes in the wheat price have had a differential impact on diets in urban and rural households and on households with and without access to agricultural land (Table 10). Unlike rural households, urban households substituted other foods for grain as the wheat price increased, but rural households with access to land witnessed a significant increase in the calories available from grain. For all households, especially those in urban areas, the higher wheat price significantly reduced consumption of animal products (2.7 percent for a 40 percent increase in the wheat price). In contrast, calories obtained from vegetables, fruit, and other foods (notably cooking oil) increased substantially. Consistent with the findings from the cross-sectional data, as the wheat price rises, households without access to land substitute vegetables 16 or other relatively cheaper foods for wheat or grain. Again, if households had access to agricultural land in 2008, they tended to consume more calories from vegetables and other sources as wheat prices rose. They may not have substituted those foods for grain, given that their grain consumption increased, probably owing to general income gains in regions where wheat price increases were particularly large. 17 Table 10. Dietary Changes, 2008–10, Household Fixed Effect All Urban Rural Log(caloric availability from grain) Log(Wheat Price) –0.014 (0.021) –0.058* (0.034) 0.028 (0.028) Log(Wheat Price) X Access to Land in 2008 0.122*** (0.040) 0.112 (0.144) 0.084** (0.043) Log(caloric availability from animal products) Log(Wheat Price) –0.068** (0.031) –0.092** (0.043) –0.051 (0.043) Log(Wheat Price) X Access to Land in 2008 –0.033 (0.058) –0.434** (0.183) –0.015 (0.067) Log(caloric availability from vegetables and fruit) Log(Wheat Price) 0.073*** (0.027) 0.094** (0.042) 0.053 (0.035) Log(Wheat Price) X Access to Land in 2008 0.115** (0.050) 0.057 (0.176) 0.142*** (0.054) Log(caloric availability from other) Log(Wheat Price) 0.059*** (0.019) 0.046 (0.030) 0.067*** (0.025) Log(Wheat Price) X Access to Land in 2008 0.167*** (0.036) 0.222* (0.125) 0.159*** (0.039) Note: Variables included in the regression but not presented are the household head’s gender, age, and education level and the number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years or older. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. 16. The grain price increased by 30 percent while the vegetable and fruit price increased by 24 percent from January to December 2008, based on national average prices of each item and weights for the consumer price index (0.2 for grain and 0.13 for vegetables and fruit). 17. Grain consumption accounts for 52 percent of calories consumed, animal products for 12 percent. vegetables and fruit for 5 percent, and other foods for 31 percent (in the 2008 data). 30 Access to agricultural land may mitigate the severity of rising food prices. Other assets, however, may also help households cope. Table 11 explores changes in holdings of these other assets in relation to local wheat price increases (see also Appendix Table 1). Among households without access to agricultural land, ownership of nonproductive asset such as financial assets and precious metals (especially gold) falls as wheat prices rise. 18 This finding suggests that an important coping mechanism for these households, until this point in the food crisis, was to divest themselves of their nonproductive assets. Households with access to agricultural land were more likely to draw down financial assets other than precious metals, suggesting that their access to land may have buffered their losses to some extent. Regardless of their access to agricultural land, however, households in areas with higher wheat prices gave up significantly more assets than households in areas where prices did not rise as much. With respect to productive agricultural assets, rural households without access to land acquired more livestock as wheat prices rose. Households with land sold off their livestock, presumably to dedicate more resources to crop production. Table 11. Changes in Assets, 2008–10 Panel Data, Household Fixed Effect All Urban Rural Pr(Financial Asset>0) Log(Wheat Price) –0.149*** (0.020) –0.168*** (0.032) –0.135*** (0.027) Log(Wheat Price) X Access to Land in 2008 –0.086** (0.038) –0.198 (0.135) –0.086** (0.042) Pr(Precious Metal Own) Log(Wheat Price) –0.468*** (0.025) –0.422*** (0.038) –0.502*** (0.034) Log(Wheat Price) X Access to Land in 2008 0.125*** (0.047) 0.113 (0.159) 0.162*** (0.052) Pr(Livestock Ownership) Log(Wheat Price) 0.130*** (0.017) 0.006 (0.013) 0.233*** (0.027) Log(Wheat Price) X Access to Land in 2008 –0.331*** (0.031) –0.515*** (0.055) –0.408*** (0.042) Source: Authors’ Calculation based on PSLM 2008 and 2010 Panel Note: Variables included in the regression but not presented are the household head’s gender, age, and education level and the number of female and male household members between zero and five years old, six and fifteen years, sixteen and fifty-five years, and fifty-six years or older. (***) significant at 1% level, (**) significant at 5% level, (*) significant at 10 % level. 18. In 2008, 19 percent of households owned financial assets, 69 percent owned precious metals, and 15 percent owned livestock. 31 5. CONCLUSIONS AND POLICY IMPLICATIONS This note has examined how the global and domestic food crisis affected Pakistani household welfare in terms of baseline consumption value and calorie intake. Rising global food prices affected Pakistani households throughout 2008; compensating variations began to rise dramatically at the start of 2008 and had doubled by the end of the year. This outcome demonstrates that ex ante estimates of welfare losses will depend on when the effect is measured (as well as the response of incomes over the period). Over time, the food price crisis expanded the gap between Pakistan’s poor and nonpoor populations, indicating that the effects of such a crisis on income distribution can be underestimated if compensating variations are measured at only one point in time. Second, while differences in welfare effects between rural and urban households were minor, the compensating variation from the food crisis in Pakistan differed greatly across provinces. The large regional difference in welfare effects arising from the food crisis is greater than the mean gap between the poor and nonpoor in Pakistan. One of the limitations of the compensating variation method is that it forecasts distributional effects only on the basis of consumption levels before the crisis. It is a measure of possible impact, but it does not account for potential shifts in income that can mitigate the adverse consequences of rapid price changes. For this reason, we used a regression method to understand how the food crisis affected caloric availability. Caloric availability started to decline as early as the last quarter of 2007, and total calorie intake decreased by 8 percent from April 2008 to June 2008, compared to the period before the crisis. This decline in caloric availability supports the predictive results from the compensating variations. Both the compensating variation and caloric analyses show that urban households are unambiguously worse off because of the crisis. They would require a compensating variation of at least 35 percent of their precrisis expenditures to maintain their baseline level of utility, and poor urban households would require compensated income of 45 percent to maintain consumption. The effects on urban households are also seen in their caloric consumption patterns. Urban households consumed significantly fewer calories—about 5 percent fewer on average. This decline appears to be a direct result of rising food prices, because it derives from fewer market purchases of various categories of food. Analyzing the changes in calorie intake by source gives a better understanding of the protective role played by agricultural production among rural households. The compensating variation calculations that take self-production into account predicted that rural households would not be as adversely affected by the crisis. Indeed, although caloric availability declined for rural households, the calories obtained through market purchases remained largely unchanged over the crisis. This result suggests that the decline in real income was not as severe for rural households compared to urban households. Although calories from food produced by the household declined significantly, the decline does not appear to result from a dramatic reduction in agricultural output. Perhaps agricultural 32 households sought to capitalize on the relatively higher prices for agricultural goods, although more work is needed before this conclusion can be drawn with confidence. An analysis of the panel data from 2008–10 further emphasizes the protective role of access to land in the midst of the food crisis. Regions with higher local prices for wheat, the single most important staple food in Pakistan, exhibited a clear divergence in caloric availability between households having access to agricultural land and those lacking access. The higher wheat prices induced substantial increases in caloric availability among households with access to land, indicating gains in real income for these households. The drawdown of assets also enabled households, especially those with no access to land, to cope with rising food prices during the 2008–10 period. In summary, cumulative welfare losses during the food price crisis in Pakistan widened the gap between the poor and nonpoor and exacerbated inequalities across provinces—a conclusion reinforced by measured reductions in caloric availability. The analysis of changes in calorie intake by source provides a better understanding of how the capacity to engage in food production can protect rural households from the worst impacts of rising food prices. These findings highlight the significant contribution that BISP could make in extending social protection in the event of a new food price crisis and suggest strategies to improve the identification of beneficiaries. The poverty scorecard, BISP’s primary instrument for identifying its beneficiaries, helps to identify poor and vulnerable segments of the population, based on national census data. In the event of a new food price crisis, the government could raise the poverty cut-off score or complement BISP’s current targeting scheme with more categorical targeting strategies. For example, the government could identify households at different poverty thresholds that lack access to land. Simulations using these data could provide information for designing a complementary social protection program, within the allocated budget, to minimize the damage of escalating food prices on the most needy and vulnerable families. Aside from cash transfer programs such as BISP, the government could consider supply interventions that make food more accessible, particularly for households without access to agricultural land. For example, the government maintains networks of utility stores to sell discounted food but places no limitations on who can purchase that food. The government may want to consider targeting eligibility to purchase discounted food to poorer segments of the population, identified on the basis of the poverty scorecard census. 33 REFERENCES Alderman, H., J. Hoddinott, and B. Kinsey. 2006. "Long Term Consequences of Early Childhood Malnutrition." Oxford Economic Papers 58(3): 450–74. Carter, M.R., P.D. Little, T. Mogues, and W. Negatu. 2004. "Shocks, Sensitivity, and Resilience: Tracking the Economic Impacts of Environmental Disaster on Assets in Ethiopia and Honduras." Madison: BASIS Research Program on Poverty, Inequality and Development, Wisconsin. http://www.basis.wisc.edu/basis_crsp/live/assets/shocks,%20sensitivity%20and%20r esilience.pdf, accessed May 2011. ------. 2007. "Poverty Traps and Natural Disasters in Honduras and Ethiopia." World Development 35(5):835–56. Chen, S., and M. Ravallion. 2003. "Household Welfare Impacts of China’s Accession to the World Trade Organization." World Bank Policy Research Working Papers No. 3040. Washington, DC: World Bank. Chumrusphonlert, K., and J. P. Formby. (2004). "A Dominance Analysis of Thailand's Regional Income Distributions, 1992-2000." Studies on Economic Well-Being 12:405–41, 499. Cudjoe, G., C. Breisinger, and X. Diao. 2010. "Local Impacts of a Global Crisis: Food Price Transmission, Consumer Welfare and Poverty in Ghana." Food Policy 35(4): 294–302. Deaton, A. 1997. The Analysis of Household Surveys: A Microeconometric Approach to Development Policy. Baltimore: Johns Hopkins. De Janvry, A., and E. Sadoulet. 2008. "The Global Food Crisis: Identification of the Vulnerable and Policy Responses." University of California at Berkeley, http://are.berkeley.edu/~sadoulet/papers/FoodCrisis_AREUpdate.pdf, accessed May 2011. ------. 2010. "The Global Food Crisis and Guatemala: What Crisis and for Whom?" World Development 38(9):1328–39. De Janvry, A., M. Fafchamps, and E. Sadoulet. 1991. "Peasant Household Behaviour with Missing Markets: Some Paradoxes Explained." The Economic Journal 101(409):1400–17. Dessus, S., S. Herrera, and R. de Hoyos. 2008. "The Impact of Food Inflation on Urban Poverty and Its Monetary Cost: Some Back-of-the-Envelope Calculations." Agricultural Economics 39:417–29. D’Souza, A., and D. Jolliffe. 2010. "Rising Food Prices and Coping Strategies: Household-level Evidence from Afghanistan." Policy Research Working Paper No. 5466. Washington, DC: World Bank. 34 FAO (Food and Agriculture Organization). 2009. The State of Food Insecurity in the World: Economic Crises—Impacts and Lessons Learned. Rome. FAO (Food and Agriculture Organization). FAO Food Price Index, http://www.fao.org/worldfoodsituation/wfs-home/foodpricesindex/en/, accessed May 2011. Ferreira, F., and N. Schady. 2009. "Aggregate Economic Shocks, Child Schooling, and Child Health." World Bank Research Observer 24(2):141–81. Friedman, J., and J. Levinsohn. 2002. "The Distributional Impacts of Indonesia's Financial Crisis on Household Welfare: A "Rapid Response" Methodology." World Bank Economic Review 16(3):397–423. Friedman, J., and J. Sturdy. 2011. “The Influence of Economic Crisis on Early Childhood Development: A Review of Pathways and Measured Impact." Chapter 2 in No Small Matter: The Impact of Poverty, Shocks, and Human Capital Investments in Early Childhood Development, edited by H. Alderman. Washington, DC: World Bank. Pp. 51–84. Hartini, T.N.S., A. Winkvist, L. Lindholm, H. Stenlund, A. Surjono, and M. Hakimi. 2002. "Energy Intake during Economic Crisis Depends on Initial Wealth and Access to Rice Fields: The Case of Pregnant Indonesian Women." Health Policy 61(1):57– 71. Ivanic, M., and W. Martin. 2008. "Implications of Higher Global Food Prices for Poverty in Low-Income Countries." Agricultural Economics 39:405–16. Kang, S.J., and Y. Sawada. 2008. "Credit Crunch and Household Welfare: The Case of the Korean Financial Crisis." Japanese Economic Review 59(4):438–58. Lokshin, M., and M. Ravallion. 2000. "Welfare Impacts of Russia's 1998 Financial Crisis and the Response of the Public Safety Net." Economics of Transition 8(2):269–95. Ministry of Finance, Government of Pakistan. 2008. Pakistan Economic Survey 2007–08. Islamabad. http://finance.gov.pk/survey_0708.html, accessed May 2011. ----- 2009. “Agriculture.� Chapter 2 in Pakistan Economic Survey 2008–09. Islamabad. http://finance.gov.pk/survey/chapters/02-Agriculture09.pdf, accessed May 2011. Razack, A., S. Devadoss, and D. Holland. 2009. "A General Equilibrium Analysis of Production Subsidy in a Harris-Todaro Developing Economy: An Application to India." Applied Economics 41(21):2767–77. Roache, S. 2010. “Wheat Price Jump a Supply Shock That Should Unwind.� IMF Survey Magazine, September 1, 2010. International Monetary Fund, http://www.imf.org/external/pubs/ft/survey/so/2010/new090110a.htm, accessed May 2011. 35 Ruel, M.T., J.L. Garrett, C. Hawkes, and M.J. Cohen. 2010. "The Food, Fuel, and Financial Crises Affect the Urban and Rural Poor Disproportionately: A Review of the Evidence." Journal of Nutrition 140(1):170s–76s. Scharpf, F.W. 2000. "The Viability of Advanced Welfare States in the International Economy: Vulnerabilities and Options." Journal of European Public Policy 7(2):190–228. Timmer, C.P. 2010. "Reflections on Food Crises Past." Food Policy 35(1):1–11. World Bank. 2009. "Commodity Markets Review." Various Issues. Washington, DC. ------. 2010a. “Food Price Increases in South Asia: National Responses and Regional Dimensions.� Washington, DC. ------. 2010b. "Food Price Watch." May. http://siteresources.worldbank.org/INTPOVERTY/Resources/335642- 1210859591030/Food_Price_Watch_May2010.pdf , accessed May 2011. 36 APPENDIX Appendix Table 1: Summary Statistics, Quarter of Interview Jul~Sep Oct~Dec Jan~Mar Apr~Jun 2005 2007 2005 2007 2006 2008 2006 2008 Caloric availability Total caloric availability, per capita 2,422 2,508 2,462 2,462 2,471 2,367 2,593 2,392 Caloric availability from purchased food, per capita 1,899 1,948 1,920 1,956 1,965 1,964 1,990 2,024 Caloric availability from self-produced food, per capita 447 473 462 429 439 353 397 303 Pr(consuming self-produced food) 47% 44% 47% 42% 42% 37% 39% 33% Diet Caloric availability, grain 1,357 1,374 1,383 1,333 1,389 1,268 1,407 1,271 Caloric availability, animal products 293 326 302 321 311 311 295 303 Caloric availability, vegetables and fruit 126 119 135 127 131 134 125 139 Caloric availability, other food 647 689 641 681 640 655 767 678 Caloric availability, wheat 1,150 1,179 1,154 1,144 1,182 1,076 1,191 1,093 Agricultural activity Having access to land 29% 28% 30% 27% 26% 21% 23% 18% Wheat producer 26% 23% 26% 23% 23% 18% 19% 16% Cotton producer 8% 6% 8% 6% 7% 4% 5% 4% Sugarcane producer 3% 4% 3% 3% 3% 4% 2% 2% Sample size 3,888 3,790 3,847 3,708 3,841 4,106 3,835 3,889 Source: PSLM 2005–06 and 2007–08, by quarter of interview 37 Appendix Table 2: Own and Cross-Price Elasticities of Aggregate Goods, 2007–08 Milk Meat Fruit Vegetables Spices Wheat Rice Cereals Pulses Oil Other Non-Food Milk –0.792 –0.155 0.010 0.028 –0.069 –0.106 -0.227 0.055 –0.277 0.221 –0.031 –0.404 Meat –0.017 –0.904 0.050 0.120 0.164 0.033 0.089 0.065 0.533 –0.661 –0.050 –0.275 Fruit 0.081 –0.127 –0.696 0.358 –0.226 0.283 -0.069 0.129 0.115 –0.753 –0.097 –0.084 Vegetables –0.106 –0.021 –0.039 –0.462 0.574 0.052 -0.012 –0.055 –0.147 –0.039 –0.017 –0.087 Spices 0.060 0.110 0.054 –0.021 –1.213 –0.193 0.157 –0.056 0.208 –0.178 0.025 –0.099 Wheat –0.143 0.336 –0.066 –0.017 –0.280 –0.430 0.213 0.056 –0.196 –0.141 –0.001 –0.239 Rice 0.010 –0.050 0.010 –0.496 1.468 0.177 -1.177 –0.410 –0.316 0.353 0.261 –0.100 Cereals 0.521 0.051 0.201 0.507 –0.349 0.133 -0.316 –0.405 0.257 –0.259 –0.117 –0.053 Pulses 0.022 –0.164 0.033 0.213 0.530 0.244 0.233 –0.124 –0.289 –0.028 –0.057 –0.189 Oil –0.094 0.002 0.007 –0.002 0.165 –0.062 -0.044 –0.060 –0.154 –0.520 –0.028 –0.030 Other 0.513 0.244 0.620 0.052 0.422 0.508 0.092 0.480 0.416 –0.062 –0.954 0.072 Nonfood 0.008 –0.015 –0.012 –0.018 –0.037 –0.048 0.006 –0.002 0.007 0.004 0.002 –0.096 Source: Authors’ Calculation based on PSLM 2007-08 38 Appendix Table 3: Share of Caloric Availability from Different Sources, 2005–06 Urban Rural Calories from grain (%) 51.0 (13.0) 58.4 (10.9) Calories from animal products (%) 14.7 (8.3) 12.1 (7.5) Calories from vegetables and fruit (%) 6.2 (5.6) 4.7 (3.5) Calories from other foods (%) 28.1 (7.5) 24.8 (7.0) Sample size 6,213 9,198 Source: Authors’ Calculation based on PSLM 2005-06 Note: Standard deviation is in parentheses. 39 Appendix Table 4: Food Aggregates and Individual Items (PSLM) Recall Food aggregates Individual items period 1. Dairy product 14 days Milk (fresh and boiled) 14 days Milk (packed by milk plants) 14 days Milk, powdered (for adults and children ) 14 days Curd/yoghurt, lassi ( buttermilk) 14 days Butter, margarine, cream, cheese 14 days Other (e.g., ferni, kheer, condensed milk, ice cream, kulfi ) 2. Meat 14 days Beef 14 days Mutton 14 days Chicken meat (fresh, frozen ) / Other poultry birds (e.g., duck, quail, turkey ) 14 days Eggs 14 days Fish (fresh, frozen, dried)/ Prawns, shrimps, or crabs ( fresh, frozen, canned ) 3. Fruit 14 days Banana 14 days Citrus fruit (e.g., mosummi, malta, kinno) 14 days Apple 14 days Dates 14 days Grapes 14 days Mango 14 days Other fresh fruit (e.g., pomegranate, apricot, jamons, lemon, pear, peach, plum, papaya, melon, garma, sarda 14 days Canned fruit 14 days Raisin, dates, apricot (dried), other (e.g., almond, walnut,chilgoza, pistachio, peanuts, aniseed, cashew, coconut, sesame seeds ) 4. Vegetables 14 days Potato 14 days Onion 14 days Tomato 14 days Cabbage, cauliflower 14 days Karaila, ladyfinger, brinjal, cucumber 40 Recall Food aggregates Individual items period 14 days Tinda, pumpkin, bottle gourd 14 days Radish, turnip, carrot 14 days Peas, moongra 14 days Other (e.g., green chillies, tural, lettuce, kulfa ) 14 days Canned vegetables 5. Spices 14 days Salt simple (rock and sea) 14 days Salt (iodized ) 14 days Chillies, red 14 days Turmeric, coriander seed 5. Spices 14 days Ginger Continued 14 days Garlic 14 days Cinnamon, caraway, cardamom, salan masalah /Other spices (e.g., licorice root, cumin seeds, black pepper, cloves, mixed condiments) 14 days Sugar (desi or milled ) 14 days Gur / Shakkar 14 days Honey ( fresh or processed ) 14 days Confectionery (e.g., toffee, chocolate, chewing gum) 14 days Barfi, jaleebi, halwa , and other sweetmeats 14 days Glucose, Energile, and so forth 6. Wheat 1 month Wheat and wheat flour 7. Rice 1 month Rice and rice flour 8. Cereals 1 month Suji, maida, besan 1 month Other cereal products (e.g., vermicelli, corn flakes, noodles, macaroni, spaghetti) 9. Pulses 1 month Gram whole ( black and white) 1 month Dal chana 1 month Mash 1 month Moong 41 Recall Food aggregates Individual items period 1 month Masoor 1 month Other (e.g., arhar, chick/pigeon/garden peas, sunflower, soybean) 10. Oil 1 month Desi ghee 1 month Vegetable ghee 1 month Cooking oils, other oils, and fats 11. Other foods 14 days Carbonated beverages 14 days Squashes and Syrups (not medicated) 14 days Sugarcane juice, other fresh juices, fruit juices (packed), mineral water, and so forth 14 days Readymade meals, snacks, tea, ice cream, drinks, instant foods 1 month Tea (black, green, loose and packed) 1 month Coffee, other (e.g., Ovaltine, Horlick’s, Milo, Complan) 1 month Biscuits ( sweet and savory) 1 month Bread, bun, sheermal 1 month Tandoori roti, nan, kulcha, puri, paratha 1 month Other baked or fried products (e.g., pakoras, samosa, qatlama, popcorn), cake, bakerkhani, pastries, patties) 1 month Jams, marmalades/ tomato ketchup/pulp/ pudding, jelly, pickles, chatni, vinegar, yeast, Ice, and so forth 1 month Food and grain milling/grinding charges 42 about this series... This series is produced by the Health, Nutrition, and Population Family (HNP) of the World Bank’s Human Development Network. The papers in this series aim to provide a vehicle for publishing preliminary and unpolished results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For free copies of papers in this series please contact the individual authors whose name appears on the paper. Enquiries about the series and submissions should be made directly to the Editor Martin Lutalo (mlutalo@worldbank.org) or HNP Advisory Ser- vice (healthpop@worldbank.org, tel 202 473-2256, fax 202 522-3234). For more information, see also www.worldbank.org/hnppublications. THe woRlD baNk 1818 H Street, NW Washington, DC USA 20433 Telephone: 202 473 1000 Facsimile: 202 477 6391 Internet: www.worldbank.org E-mail: feedback@worldbank.org