Rising Food Prices and Household Welfare: Evidence from Brazil in 2008

Rising Food Prices and Household Welfare: Evidence from Brazil in 2008 Francisco H. G. Ferreira, Anna Fruttero*, Phillippe Leite* and Leonardo Lucche The World Bank and IZA * The World Bank University of Illinois at Urbana Champaign

Plan of the talk 1. Introduction 2. Data 3. Analytical framework 4. Empirical approach 5. Results 6. Conclusions

1. Introduction The 2006 2008 period saw substantial increases in the world prices of a number of staple foods. 2005 07 rises of 80% for maize, 70% for wheat, 90% for dairy. After falling markedly during the Great Recession, food prices rose again in 2010 2011. Some now worry that higher and more volatile food prices may be a new normal (e.g. World Bank, 2011).

1. Introduction Widespread concern with the impact of dearer food on poverty: Friedman and Levinsohn (2002): find very large welfare losses associated with rising food prices during the Indonesian currency crisis of 1997. Ivanic and Martin (2008): Look at nine developing countries and find poverty increases in most of them. Average increase in P 0 : 3.0 pp and as much as 7.8 pp in Nicaragua. Robles and Torero (2010): Average impacts on welfare around 1.5 2.5% of baseline expenditure in Guatemala, Honduras & Peru. 7% in Nicaragua. Rising food prices disproportionately affect the poor, given the higher share of food in their overall expenditures. (World Bank, 2011, p.11)

1. Introduction But (i): Many developing countries are net food exporters. Argentina (13.1% of GDP); Brazil (5.7% of GDP) in 2008 In many cases, most of the agricultural labor is on wage contracts (rather than in traditional family farms). Studies based on estimates of the compensating variation from net household expenditure functions miss any such general equilibrium effects. They only incorporate income gains from food sold directly by sampled households or individuals.

1. Introduction This limitation has long been recognized: A more serious deficiency [of a first order approximation based on net food purchases is] its neglect of repercussions in the labor market. Changes in the price of the basic staple will affect both supply and demand for labor, and these effects can cause first order modifications to the results. (Deaton, 1997, p.187) SUSENAS contains much less detail on sources of household income, and the available information on changes in factor incomes is also much less detailed. Our inability to accurately forecast changes on the income side renders us mute in terms of the real impacts of the crisis. (Friedman and Levinsohn, 2002, p. 420)

1. Introduction But (ii): During the 2007 2008 food price crisis, some governments did use social policy instruments to mitigate the welfare effects. Despite policy interest, no studies that we are aware of investigate their efficacy. A number of studies uses relatively aggregated prices sometimes assuming a 100% pass through from international prices, and a single price within the country. Very strong assumption (see Brambilla and Porto, 2009; Friedman and Levinsohn, 2002).

1. Introduction This paper: Uses monthly consumer prices for 156 food items (then grouped into 16 categories), collected in the main 11 large urban areas of Brazil during 2007 2008. Estimates the net effect of rising prices as the sum of: Expenditure effect: first order approximation to the change in consumer surplus Market income effect: estimated under two scenarios for the effect of higher food prices on agricultural wages. Transfer income effect: incorporates changes made to benefit levels of two main social assistance programs in 2008. Separately for large urban areas, rural areas, and whole country. Results presented by means of price change incidence curves.

2. Data Three sources of data: 1. Índice Nacional de Preços ao Consumidor (INPC IBGE). Monthly price series from January 2007 to May 2009. 156 food items, grouped into 16 categories (97% of food consumption in POF) Representative of each of 9 metropolitan regions and two large cities. 2. Pesquisa de Orçamentos Familiares (POF IBGE) Nationally representative, pre shock: 2002/3 wave Detailed consumption expenditure data for 48,568 households Used to estimate the expenditure effects of the price shock. 3. Pesquisa Nacional por Amostra de Domicílios (PNAD IBGE) Nationally representative, pre shock: 2006 wave Contains no consumption information, but detailed occupational, income, and social assistance receipt information. Used to estimate (market and transfer) income effects of the price shock.

2. Data INPC: The relative price of food did rise substantially during 2007 and 2008 in Brazil

2. Data POF: Food is a necessity, with a textbook declining Engel curve throughout. 35.00 Total food consumption as a share of total consumption (%) 30.00 25.00 23.21 20.00 15.00 10.00 0 25 50 75 100 Percentile of per capita household total expenditure

3. Analytical Framework Question: What is the effect of the changes in a set of prices on household welfare? Starting point: Change in consumer surplus, estimated by the compensating variation: Using Shephard s Lemma and moving to proportional changes: Allowing for household production and substitution effects:

3. Analytical Framework Allow for general equilibrium effects on household income, and define the proportional change in money metric welfare as: Data limitations: No information on household s own agricultural production ( ) Insufficient spatial variation to estimate substitution matrix ( ) These lead to our estimating equation:

3. Analytical Framework Allow for general equilibrium effects on household income, and define the proportional change in money metric welfare as: Data limitations: No information on household s own agricultural production ( ) Insufficient spatial variation to estimate substitution matrix ( ) These lead to our estimating equation: Expenditure effect Market income effect Transfer income effect

4. Empirical Approach Vector of price changes used in the analysis (16 category INPC POF concordance) Table 1: Average and Maximum Price Change by Food Items Food Items Average Price Maximum Price Increase over 12 consecutive months Increase Value Peak month Cereals 30.20% 76.60% July 2008 Flour, starches, and pasta 11.80% 20.90% June 2008 Tuber and roots 15.50% 49.10% July 2008 Sugars and derivatives 9.20% 9.80% November 2008 Vegetables 9.60% 21.00% June 2007 Fruit 6.90% 16.20% August 2008 Meat 20.30% 42.30% July 2008 Poultry and eggs 15.90% 30.10% April 2007 Milk and derivatives 10.70% 33.70% August 2007 Baked 9.90% 22.50% June 2008 Oils and fats 17.80% 39.80% May 2008 Drinks and teas 4.90% 7.50% October 2007 Canned and preserved 3.80% 8.70% December 2008 Salt and condiments 2.90% 7.50% November 2008 Food away from home 8.30% 12.00% October 2008 Industrialized fish and meat 9.40% 20.40% November 2008 Source: IBGE - National Consumer Price Index (INPC).

4. Empirical Approach Table 2: Assignment of price changes from large urban area in the INPC to state in the POF 2002/03 expenditure survey Region (1) POF 2002/03 (2) PRICES - INPC Assignment of (2) to (1) North Northeast Southeast South Center-West Rondônia - Belem Acre - Belem Amazonas - Belem Roraima - Belem Pará Belem Belem Amapá - Belem Tocantins - Belem Maranhão - Fortaleza Piauí - Fortaleza Ceará Fortaleza Fortaleza Rio Grande do Norte - Recife Paraíba - Recife Pernambuco Recife Recife Alagoas - Recife Sergipe - Recife Bahia Salvador Salvador Minas Gerais Belo Horizonte Belo Horizonte Espírito Santo - Belo Horizonte Rio de Janeiro Rio de Janeiro Rio de Janeiro São Paulo Sao Paulo Sao Paulo Paraná Curitiba Curitiba Santa Catarina - Porto Alegre Rio Grande do Sul Porto Alegre Porto Alegre Mato Grosso do Sul - Goiania Mato Grosso - Goiania Goiás Goiania Goiania Distrito Federal Brasilia Brasilia

4. Empirical Approach Expenditure effect estimated for each household in POF, using their observed baseline budget shares, and the INPC price changes. Market income effect estimated for each household in the PNAD, as follows: if household contains no worker in any agricultural activity if household contains a worker in any agricultural activity, producing good i. 0.5,1.0 Transfer income effect set to the exact transfer increases announced for Bolsa Família and BPC in 2008, for each family with at least one recipient of the benefit in the PNAD 2006. Increases followed the benefit formulae for each program.

4. Empirical Approach Bolsa Família s benefit increase took place in July 2008, and appears to have been motivated explicitly by rising food prices. The BPC was raised by 10% in March 2008, in line with the annual increase in the minimum wage. Program 2006 benefit value 2008 increase Bolsa Família basic benefit R$50 R$4 add. per child R$ 15 R$2 BPC R$350 R$35 Minister Patrus Ananias (Social Development) said on Wednesday that the average Bolsa Família benefit will rise from R$78.70 to R$80.00. He added that the 8% value of the increase [in the basic benefit] was determined on the basis of the INPC (Índice Nacional de Preços ao Consumidor), with the objective of improving the purchasing power of low income families in the midst of the world food crisis (Folha de São Paulo newspaper, online, 25 June 2008, our translation and emphasis).

4. Empirical Approach Once these are estimated, compute Price Change Incidence Curves (PICs), for each component and for the overall effect, as follows: Each of these can be graphed (against π). Also, they can be applied to baseline (PNAD) distribution, to obtain counterfactual income distributions: Estimated poverty and inequality changes are computed from these distributions.

4. Empirical Approach (caveats) There are four main caveats to this approach: 1. No large representative Brazilian survey contains reliable information on household own agricultural production. So we cannot estimate 2. Insufficient spatial price variation prevented us from estimating the substitution matrix for the POF. So we cannot estimate 3. Our estimates of the market income effect are coarse. Future work should try to obtain causal estimates of the pass through from food prices to agricultural wages. 4. Estimating compensating variations requires pre shock budget shares. However, one would ideally have liked budget share information nearer the date of the price increases.

5. Results (PICs) Figure 5: Price Increase Incidence Curve Expenditure Effect (Large Urban Areas) 0.00-1.00-2.00-3.00-4.00-5.00-6.00-7.00-8.00-9.00-10.00 0 25 50 75 100 Percentile of per capita household total expenditure Source: IBGE Household Budget Survey (POF), 2002/03. Note: Horizontal black solid line represents the average change in per capita expenditure.

5. Results (PICs) Figure 6: Price Increase Incidence Curve Expenditure Effect (Rural Areas) Change in per capitai total expenditure (%) 0.00-2.00-4.00-6.00-8.00-10.00-12.00-14.00-16.00-18.00-20.00 0 25 50 75 100 Percentile of per capita household total expenditure Source: IBGE Household Budget Survey (POF), 2002/03. Note: Horizontal black solid line represents the average change in per capita expenditure.

5. Results (PICs) Figure 7: Price Increase Incidence Curve Expenditure Effect (All Brazil) 0.00-2.00 Change in per capitai total expenditure (%) -4.00-6.00-8.00-10.00-12.00-14.00-16.00 0 25 50 75 100 Percentile of per capita household total expenditure Source: IBGE Household Budget Survey (POF), 2002/03. Note: Horizontal black solid line represents the average change in per capita expenditure.

5. Results (PICs) Figure 9: Price Increase Incidence Curve Net Effect (Large Urban Areas Alpha = 1) 0.00-1.00-2.00 Change in per capitai total expenditure (%) -3.00-4.00-5.00-6.00-7.00-8.00 Exp, Lab income and transfers effect Exp and lab income effect Expenditure effect -9.00 0 25 50 75 100 Percentile of per capita household total expenditure Source: IBGE Household Budget Survey (POF), 2002/03. Note: Horizontal black solid line represents the average change in per capita expenditure. Figure 10: Price Increase Incidence Curve Net Effect (Rural Areas Alpha = 0.5)

5. Results (PICs)

5. Results (PICs)

5. Results (PICs) Figure 13: Price Increase Incidence Curve Net Effect (All Brazil Alpha = 1) 6.00 4.00 2.00 Change in per capitai total expenditure (%) 0.00-2.00-4.00-6.00-8.00-10.00-12.00-14.00-16.00 Exp, Lab income and transfers effect Exp and lab income effect Expenditure effect 0 25 50 75 100 Percentile of per capita household total expenditure Source: IBGE Household Budget Survey (POF), 2002/03. Note: Horizontal black solid line represents the average change in per capita expenditure.

5. Results (Poverty and Inequality) Table 9: Food Price Effects on Poverty and Inequality (All Brazil) Alpha =1 Baseline Expenditure Effect Expenditure and Market Income Expenditure and Bolsa Familia Effect Expenditure and BPC Effect Extreme Poverty Headcount 11.04 13.53 12.43 13.42 13.49 12.30 (0.14) (0.11) (0.14) (0.14) (0.14) (0.14) Poverty gap 3.80 4.78 4.33 4.67 4.77 4.21 (0.05) (0.07) (0.05) (0.06) (0.07) (0.06) Squared poverty gap 1.95 2.47 2.22 2.38 2.47 2.14 (0.04) (0.04) (0.04) (0.03) (0.04) (0.04) Moderate poverty Headcount 31.29 35.10 33.53 35.05 35.00 33.39 (0.14) (0.17) (0.18) (0.17) (0.17) (0.17) Poverty gap 12.34 14.45 13.56 14.34 14.41 13.42 (0.10) (0.10) (0.11) (0.09) (0.12) (0.09) Squared poverty gap 6.66 8.03 7.45 7.92 8.01 7.33 (0.06) (0.07) (0.07) (0.08) (0.08) (0.07) Inequality Gini 55.7 57.0 56.5 56.9 57.0 56.4 (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Source: IBGE National Household Income Survey (PNAD), 2006. Note: Standard errors in parentheses. Extreme and Moderate Poverty Line can be found in table 2 in the the annex. Headcount, poverty gap and squared poverty gap measures are also known as FGT (0, 1 and 2) respectively (Foster et al. (1984)) Total

6. Conclusions 1. Expenditure effects large and regressive. The national average compensating variation was 7.5% of baseline income. But estimates are an upper bound. 2. Market income effects were generally progressive, and were substantial in rural areas. With full pass through from prices to wages, bottom quartile gains 5 10% of baseline income. 3. Transfer income effects were progressive but small. Reflecting Bolsa Família s targeting, and small benefit increases. 4. Nationally, the net effect was U shaped. Reflecting Brazil s urbanization rate (80%), PICs are negative over most of the domain. But middle groups experienced proportionately larger welfare losses than the very poor. Though the unprotected urban poor will have done worst of all. 5. The net effect implies an increase in extreme poverty from 11.0% to 12.3 12.8%. Without the income effects: 13.5%