1
The impact of rising food prices on the poor in
developing countries
ABSTRACT
The aim of this research is to provide insight into the effects of rising food prices on the poor. It shows that rising food prices can cause severe real income losses for the poor. Irrespective of actual food price rises, I have shown that the South and East Asian poor bear by far the largest exposure to rising food prices, followed by the East European, Central Asian and urban African poor. The rural African and Latin and Central American poor will be least severely affected by rising food prices. When looking at the 2007-2008 food price rise, it appears that it had the largest impact on the Asian poor as they experienced the highest food price inflation. The African and Latin and Central
American poor experienced the lowest food price inflation and hence were better off than their Asian counterparts.
The costs of alleviating the effects that the 2007-2008 food price rise had on 240 million poor people in 43 developing countries amount to 12 billion dollars. This amount 45% higher than the amount that would be needed to compensate these poor for the 2005-2006 price increase and 95% higher than the amount that would be needed for the 2009-2010 price increase. Although I have demonstrated that food prices have a large negative effect on poverty, I have shown that there is no correlation between food prices and poverty rates.
Student:
Leonie Broeders
Student number: S1627732
Supervisor:
G.H. Kuper
Date:
August 30, 2012
University of Groningen
Faculty of Economics and Business
Master Thesis M.Sc. Economics
2 TABLE OF CONTENTS
I. Introduction
II. Literature Background 5
II.I Overview of current literature 8
II.II Conclusions and limitations of current literature 10
III. Theoretical Framework 13
III.I The base model 13
III.II Including substitution effects 15
III.III Including wage effects 16
IV. Research Objectives 18
V. Data Collection and Data Construction 21
V.I The average income per poor capita 21
V.II Food consumption budget shares 22
V.III Food production budget shares 24
V.IV. Food price inflation 25
V.V Aggregating per capita compensating variations 25
V.VI Final sample sizes 26
VI. Results 28
VI.I The impact of a 15 percent food price increase on the poor 28 VI.II The impact of the 2007-2008 food price increase on the poor 31 VI.III Incomplete pass-through of consumer food price increases to poor producers 33 VI.IV The empirical relationship between food prices and poverty rates 37
VII. Discussion of results 38
VIII. Concluding Remarks and Recommendations for Future Research 41
References 43
3 I.INTRODUCTION
Rapidly rising food prices have been a recurring concern over the past few years. In 2005 the first food price started to rise, reaching their record height in 2008. After a period of decreasing prices, food prices started rising again in June 2010 and in February 2011 food prices were even higher than their previous record level in 2008. After a short period of declining prices, food prices started rising again in the past few months, amongst others caused by extreme drought in the U.S.1 The 2007-2008 food price crisis has had serious consequences. Numerous countries banned exports of specific food products. However, many developing countries had to deal with much more serious consequences; the rapid rise of food prices led to food riots, violent protests and social unrest. In Haiti food riots even caused the government to fall .
Plenty of research has been done on the causes of rising food prices. The literature on the
consequences of rising food prices is also rapidly growing. This paper belongs to the latter strand of literature and tries to estimate the welfare effects of rising food prices faced by the poor population in developing countries. After the 2007-2008 food price crisis a lot of research on the effect of this crisis on the poor was published. Hence, this literature is looking backward. This paper is no different. However, most of the previous literature base their estimations of the welfare effects of rising food prices on household surveys, of which the use is extremely time consuming and limited to countries that undertake large and reliable household surveys often. I hope to make my contribution to the research by showing that there is a much quicker manner to estimate the effects of rising food prices on poverty. Moreover, this method can be used for all developing countries. I believe that forecasting is more useful than looking back and I hope that my research shows that is possible to make a fairly accurate estimate of the impact of an expected rise in food prices in advance. Accurate forecasts will help governments and organizations as the World Bank, IMF, FAO and WHO to respond to rising food prices fast and effectively. This will hopefully prevent future food riots.
In this paper I will address the following four different research objectives. The first is to identify the countries that are most and least vulnerable to food price increases, irrespective of the actual food price increase. The second objective is to provide an estimate of the amount of money that would be needed to compensate all poor people living in developing countries for the 2007-2008 food price rise. The next objective is to gain insight into which countries are hurt most and least severely by the 2007-2008 food price crisis. The final objective is to gain insight into the empirical relationship between food prices and poverty rates
This paper proceeds as follows. The next chapter will provide an overview of the current literature on this subject; it will discuss assumptions and characteristics of the most important papers, it will conclude on the most important and remarkable results that have been found and it will elaborate on the limitations of the current literature. The third chapter of this paper describes the theoretical framework on which I will base my analysis. I will discuss a base model and will provide the reader
4 with more advanced extensions of this model. The fourth chapter goes deeper into the research
objectives of this paper; it provides a more formal and extensive overview of the research objectives and it explains how the theoretical framework will be used to address these objectives. Furthermore it will explain how this paper adds to the previous literature and provide an overview of important definitions that are used throughout the rest of the paper. The fifth chapter is dedicated to the data collection. I will describe the data that are required to address the research objectives with the
5 II. LITERATURE BACKGROUND
There is a vast amount of literature on the impact of rising food prices on a country’s distribution of income and on its poverty rates. One of the first to research this subject was Deaton (1989). He examined the effect of rise price increases on the income distribution in Thailand. In order to do so he developed a partial equilibrium framework that has been utilized by many other researchers that analyze the relationship between rising food prices and poverty or the income distribution. The main thought of this framework is as follows: food prices have an effect on households via two different channels, one is the household’s consumption expenditure and the other one is the household’s production of food. In order to keep the amount of food consumed at the same level as before a price rise, consumption expenditure must increase. Therefore, food price rises reduces the purchasing power of a household; this is the so-called consumption expenditure effect. If a household produces food, rising food prices result in a higher income from selling food; this is the so-called production income effect. Hence, the impact of rising food prices on a household’s real income depends on whether the household is a net consumer or a net producer of food. Deaton’s model is a partial equilibrium model because only the direct impact of rising food prices on poverty is considered, second-order effects via wages, substitution, policy responses and transmission to general inflation are ignored. Throughout the years some researchers have extended Deaton’s model to incorporate these secondary effects.
As the goal of this paper is to examine the effect of the recent food price rise on poverty, this literature review will focus on the more recent papers. Most of the recent literature is aimed at determining the impact of food price increases that took place around the peak years 2008 and 2011 on poverty or the income distribution. This effect had been researched for several (groups of) countries, several food products, different time periods, making use of different definitions of poverty2 and by making different assumptions (e.g. on the level of transmission of world food prices to: consumer prices, food producers’ prices and agricultural wages). On the next page the reader can find a table that provides an overview of the characteristics of all relevant recent papers. The section below the table will extend briefly on each paper by pointing out important assumptions and characteristics. The final section of this literature review will conclude on the most important results and will elaborate on the limitations of the current literature.
2 There are several definitions of poverty. The most common ones are established by the World Bank. The World Bank defines an extreme
6
Authors (Number of)
Countries included
Products included Time period for which inflation is measured Measurement of inflation Full transmission to producer prices Substitution effect included Wage effect included Use of household surveys Zezza et al. (2008) 11; all developing regions included
Three main food staples per country
N.a.; the effect of a simulated 10% price increase for all food items is analyzed. Simulated inflation Yes N N Yes Rios, Shively and Masters (2009) Tanzania, Vietnam, Guatemala Perform analysis for all crops produced by the household as well as individual crops.
N.a; research which farmers (poor/rich) are net food buyers (sellers) and to what extent they are vulnerable to (gain from) price shocks.
N.a. N.a. N N Yes
Aksoy and Dismelik (2008)
9; all developing regions included
Three main tradable staples per country
N.a; research the characteristics and the proportion of the households that are net food buyers (sellers) and to what extent these households are vulnerable to (gain from) price shocks.
N.a. N.a. N Y Y
Arndt et al. (2009)
Mozambique All consumed food items
N.a; research the characteristics and the proportion of the households that are net food buyers (sellers) and to what extent these households are vulnerable to (gain from) price shocks.
N.a. N.a. N N Y
DeHoyos and Medvedev (2009)
73 All consumed food
items 2005-2007 Domestic food CPI relative to domestic non-food CPI Yes N N N Valero and Valero-Gill (2008)
Mexico 11 most consumed food products
2006-2007 and 2006-2008 World prices Yes N N Y
Ferreira, Fruttero, Leite and Luchetti (2011)
Brazil 16 June 2007 – June 2008 Regional prices (they take the maximum price increase observed during the period)
7
Authors (Number of)
Countries included
Products included Time period for which inflation is measured Measurement of inflation Full transmission to producer prices Substitution effect included Wage effect included Use of household surveys Ivanic and Martin (2008) 9; all developing countries included
7, the same for each country
2005-2007 + simulated price increase of 10% for all items simultaneously
World prices Yes N Y Y
Ivanic, Martin and Zaman (2011) 28 countries; all developing regions included
38, the same for each country
June – December 2010 Domestic prices Yes Y N Y
Robles and Torero (2010)
4; only LAC 6, same for each country
2006-2008+ simulated price increase of 10% for all items simultaneously
Domestic prices 3 scenarios: full, no and 50% transmission Y N Y Dessus, Herrera and DeHoyos (2008) 73; all developing regions included
All consumed food items
N.a.: simulated price increase of 10%
N.a. N.a.; only urban poverty is measured. Y N Y Ul Haq, Nazli and Meilke (2008)
Pakistan All consumed food items 2007-2008 Domestic unexpected inflation (average yearly domestic food price inflation 1991-2006 in defined as expected inflation) No transmission Y N Y Wodon et al. (2008)
12; only African Different (numbers of) products per country
N.a.; simulated price increase of 25% and 50%
N.a. Both full and no transmission
N N Y
Vu and Glewwe (2011)
Vietnam 11 items together and rice separately
Jan. 2007 – Sept. 2008 and simulated price increase of 20%
Domestic prices 80%, 100% and 120%
transmission
8 II.I Overview of the current literature
Aksoy and Isik-Dismelik (2008), Arndt et al. (2008) and Rios, Shively and Masters (2009) try to gain insight into which population groups (based on income, region or country) are most vulnerable to food price increases. They do this by calculating the net benefit ratios across different income and regional groups. The net benefit ratio is equal to the share of the value of total production of food (production for own-consumption and for sales) in the household’s total expenditure minus the share of the value of total consumption (both from own-production and purchased items) in the household’s total expenditure. The three studies perform their analysis for a different set of countries and for different food products so their results are not comparable. A similar goal is addressed by Zezza et al. (2009). They investigate the welfare effects of a simulated 10 percent increase in the prices of the three main food staples per country across income quintiles for eleven countries. Another purpose of this study is to develop a household profile for the ‘extreme losers’. Hereby they include characteristics such as the households’ education level, whether or not the household is leaded my a female and the use of resources to increase farm productivity (e.g. fertilizers). De Hoyos and Medvedev (2011) calculate the impact of the food price inflation experienced between 2005 and 2007 on poverty rates in 73
9 incorporate substitution effects when estimating the welfare effects of the food price increases
observed between 2006 and 2008 in Guatemala, Nicaragua, Peru and Honduras. Furthermore, Robles and Torero question whether the price that producers receive for their food products increases with the same pace as food consumer prices do. Therefore, they estimate the welfare effects under several scenarios of pass-through of food consumer price increases to food producer prices. Wodon et al (2008) and Vu and Glewwe (2011) also allow for incomplete transmission of consumer price increases to producer prices. Wodon et al. provide two explanation for incomplete transmission. First, the recent food price increase was accompanied by increasing oil prices; higher oil prices will increase the production costs and transportation costs for producers of food. Second, market intermediaries might keep the profit from higher consumer prices to themselves. Wodon et al. analyze the impact of a simulated 10 percent food price increases on poverty in twelve West and Central African countries and Vu and Glewwe (2011) estimate both the welfare effect of simulated price increases and the welfare effect of domestic food price changes observed over the period 2007-2008 on poverty in Vietnam. Ul Haq, Nazli and Meilke (2008) ignore the effect that higher food prices might have on the production income of farmers and therefore implicitly assume that consumer food price increases are not passed through to producer prices. Ul Haq, Nazli and Meilke investigate the impact of the unexpected food price increase experienced in 2007 on poverty in Pakistan. In order to obtain unexpected food price inflation they define expected food price inflation as the average yearly food price inflation
experienced between 1991-1992 and 2006-2007 and subtract this number from the 2007-2008 food price inflation.
10 II.II Conclusion and limitations of current literature
It is difficult to draw a general conclusion about the size of the effect that rising food prices have on poverty measures or welfare. This is due to the fact that the papers concerning this subject base their analysis on different measures for food price increases; measure food price increases over different time periods; incorporate different (numbers of) food products and use different partial equilibrium models (e.g. incorporation of substitution effects or wage effects). Hence, the only conclusions that can be provided are on the direction of the effect that rising food prices have on poverty. Furthermore, differences between countries and regions, and distributional implications can be addressed.
All papers discussed in the literature review confirm that higher food prices have a negative impact on both poverty incidence and poverty depth3 at a national and at a global level (except for Vietnam and maybe Peru). However, papers that analyze multiple countries show that the severeness of this impact differs widely across countries (Ivanic, Martin and Zaman (2011), Dessus, Herrera and deHoyos (2010), De Hoyos and Medvedev (2011)). Furthermore, Ivanic, Martin and Zaman and DeHoyos and show that the increase in poverty incidence is small for many countries.
An interesting result that is found by several papers is that Vietnam seems to benefit from increasing prices (Vu and Glewwe (2011), Zezza et al. (2009), Ivanic and Martin (2008), Ivanic, Martin and Zaman (2011)). Furthermore, there is some evidence that shows that Peru benefits from a price increase or at least is relatively immune to food price increases (Ivanic, Martin and Zaman (2011), Ivanic and Martin (2008) and Robles and Torero (2010)). Nicaragua seems to be among the countries that will be hurt most severely by high food prices (Robles and Torero (2010), Dessus, Herrera and deHoyos (2010) and Ivanic and Martin (2008). This also seems to be true for Bangladesh (Ivanic, Martin and Zaman (2011), Zezza et al. (2009), Aksoy and Isik-Dismelik (2008) and De Hoyos and Medvedev (2011)) and Pakistan (De Hoyos and Medvedev (2011) and Ivanic, Martin and Zaman (2011)). On a more aggregate level it seems that Asian countries suffer more from higher food prices than Latin American countries (De Hoyos and Medvedev (2011), Aksoy and Isik-Dismelik (2008)). Explanations could be that Asia experienced higher food price inflation from 2005 till 2008, the Asian poor have higher food consumption budget shares and that initial poverty rates were higher than in Latin America.
Most papers show that the poorest households suffer the most from an increase in food prices (e.g. Zezza et al. (2009), Ferreira, Fruttero, Leite and Luchetti (2012), Rios, Shively and Masters (2009), Robles and Torero (2010)). A logical conclusion, since poor households spend a larger share of their income on food than less poor households. This result implies that rising food price cause income distribution to become more uneven. Furthermore, the largest part of the total welfare loss that is experienced by the poor is attributable to the losses of the old poor. Only a small part can be attributed to the people that fall into poverty only after the increase in food prices. This result is supported by
3 Poverty incidence is equal to the number of people living below the established poverty line. Poverty depth is the income gap between the
11 Robles and Torero (2010), Wodon et al. (2008) and Dessus, Herrera and deHoyos (2008).
Another recurring result is that urban households suffer more than rural households. This is quite comprehensible as urban households are least likely to benefit from the positive effects that higher food prices might have via increased profits from food production or increased agricultural wages. Even more, some papers show that rural middle-income households benefit from food price increases (Vu and Glewwe (2011), Arndt et al. (2008), Deaton (1989), Ivanic, Martin and Zaman (2011), Rios, Shively and Masters (2009)). The positive effect that increasing food prices have on agricultural wages is able to limit the increase in poverty following a food price increase (Ivanic and Martin (2008), Ferreira, Fruttero, Leite and Luchetti (2012)). However, none of the papers is able to show that this effect is large enough to actually decrease poverty incidence and poverty depth. The only two
exceptions seem to be Vietnam and Peru where the poverty gap decreases once wage effects are taken into account (Ivanic and Martin, 2008).
Robles and Torero (2010) show that omitting substitution effects results in an overestimation of welfare losses of up to 12 percent. Vu and Glewwe (2011) and Wodon et al (2008) show that incomplete pass-through of the increase in food prices to producer prices might significantly worsen the effect that increasing food prices have on poverty as compared to full pass-through. However, Robles and Torero (2010) disagree on this. This different in results might be caused by the fact that Latin American countries have relatively smaller rural populations (Robles and Torero’s analysis only includes Latin American countries) (Aksoy and Isik-Dismelik (2008 and De Hoyos and Medvedev (2011).
Almost all papers discussed in the literature section use household surveys to compute total expenditure (or income levels) and food consumption and production budget shares. Furthermore, some papers compute the poverty headcount ratio and the poverty gap from the household surveys by verifying whether a household falls below the poverty line based on their reported income level (e.g. Wodon et al. (2008), Ivanic and Martin (2008), Robles and Torero (2010) and Dessus, Herrera and DeHoyos (2008)). For none of the papers these inputs are adjusted for the gap between the survey year and the year of the food price rise under consideration.4 In some papers this time gap is quite
substantial. E.g. Ferreira, Fruttero, Leite and Luchetti (2012) make use of a 2002-2003 household survey to measure the impact of the 2007-2008 price rise, Ivanic, Martin and Zaman (2011) make use of household surveys ranging from 2000 till 2009 to estimate the effects of the 2010-2011 price increase and Ivanic and Martin (2008) make use of household surveys ranging from 1997 till 2005 to estimate the impact of the 2005-2007 price rise. A large gap between the survey year and the year of the food price increase results in incorrect outcomes as expenditure levels and budget shares (and poverty rates if computed from household surveys) might have changed throughout time. Ivanic and
4 Dessus, Herrera and DeHoyos are an exception. They adjust the reported income level to 2005 with the growth rate of private consumption.
12 Martin (2008) estimate the impact of a hypothesized price increase of 10 percent and compare results across countries. However, as expenditure levels and budget shares are computed based on different survey years one might argue whether these results are indeed comparable across countries. The argument holds for Zezza et al. (2009) with a survey year range from 1997 till 2005 and Rios, Shively and Masters (2009) with a survey year range from 1991 till 2000. A similar problem can be observed concerning the use of initial poverty rates. De Hoyos and Medvedev (2011) simply use the latest numbers available at PovCal as initial poverty rates to analyze the effect of the 2005-2007 price increase on the poverty headcount ratio and the poverty gap of 73 countries. Ivanic and Martin’s (2008) initial values of the poverty headcount ratio and poverty gap are based on different years ranging from 1997 till 2004. De Hoyos and Medvedev (2011) and Ivanic, Martin and Zaman (2011) explain differences in the impact that increasing food prices have across countries by differences in initial poverty rates. Hence, it is important that poverty rates of the same year are used when doing a cross-country comparison.
13 III: THEORETICAL FRAMEWORK
Most of the papers discussed in the literature review are based on a model developed by Deaton (1989). This model only incorporates the loss in real income caused by increasing food prices.
However, higher food prices can also have an indirect effect on households. An increase in food prices might be passed through to agricultural wages. Furthermore, rising food prices might induce
substitution to (non-food) products with relatively lower prices. In the next section I will discuss the derivation of the expression that allows for direct effects only . The two subsequent sections will derive expressions that include substitution effects and wage effects.
III.I The base model
Deaton’s (1989) model is based on standard economic theory (an indirect utility function and a profit function). He derived an expression that allows one to calculate the amount of money that would be needed to compensate a household for an increase in (food) prices. However, the two extensions on this base model are derived based on a profit function and an expenditure function, rather than an indirect utility function. For the sake of consistency this section will derive the standard expression following this latter method. The expenditure function describes the minimum amount of money a household needs to achieve a given level of utility, u∗ . The expenditure function of household h is described as follows:
(1) Eh =Eh(P,w,u*)
Where P is a vector that contains prices of all the goods that are consumed by the household, w is a vector containing the prices of supplied production factors, e.g. the wage rate for a specific type of labor, and u∗is the constant utility level of the household h . The envelope theorem states that the first derivative of expenditure with respect to the price of a good is equal to the consumed quantity of that good. Hence, i h p E ∂ ∂
equalsqih; the demanded quantity for good i by household h . pi is price of good i
and is included in the vector P .
The profit function describes the profit that the household obtains with production activities through unincorporated enterprises:
(2) πh =πh(P,w)
P is in this context a vector of prices of the products that are produced and w is a vector containing the prices of purchased production factors. From standard economic theory we know that the derivative of a profit function with respect to the price of the produced good i is equal to the sales volume of good
i (xi). I.e. ih i h x p = ∂ ∂π .
The net expenditure of household h , Bh is equal to:
14 Equation (3) is used to derive the compensating variation. The compensating variation (CV) is a concept developed by John Hicks (1939). It represents the extra amount of income that a household would need to achieve the same level of utility *u after an increase in consumer good prices. Hence, in mathematical terms this concept can be written as: CV =Eh(p1,u*)−Eh(po,u*). However, this paper is also concerned with the impact that an increase in food prices might have on a household’s income via the production of food. Therefore the compensating variation is redefined as
*) , ( *) , (p1 u B p u B
CV = h − h o . Note that by using this expression I implicitly assume production
prices are equal to consumption prices. In the last paragraph of this section an expression allowing for unequal consumer and producer prices will be derived. Expressing the compensating variation in continuous time yields CV =dBh given dpi (the change in the price of good i ). This is equal to: (4) CV =dBh =(qih −xih)dpi
Expressing equation (4) in terms of the household’s total expenditure level and expressing the price change in percentages rather than in absolute terms yields the following equation (expressed in discrete time): (5) i i h i h i h h p p E B CV = ∆ =(α −λ )∆ Where α equals ih h i ih E p q
and is the share of the total expenditure of household h that is spent on
commodity i and λih equals
h i ih E p x
and is the value of the production of the same commodity i expressed in terms of the household’s total expenditure. Note that the compensating variation is now expressed in terms of expenditure, by multiplying the right hand side of equation (6), one can easily compute the absolute level of the compensating variation. The compensating variation can be seen as a proxy for a household’s loss in real income or real expenditure that results from a price increase. Hence, by subtracting the compensating variation from a household’s initial income one can find the household’s real income level after price changes.
To allow for unequal consumer prices and producer prices I define a consumer price, pci and a
producers price,ppi. Hence, dEas dE=qihdpciand d
π
asdπ =xihdppi. Combining results to obtainthe compensating variation yields: (6) CV =dBh =qihdpci −xihdppi
15 III.II Including substitution effects – An extension by Robles and Torero (2010)
Households can substitute away from goods of which prices have risen relatively a lot to less costly goods. Compared to the base model, this mechanism will result in lower a lower compensating variation and hence lower costs to maintain utility levels at the same height as before price increases. Second-order Taylor expansion of the net expenditure function (3) allows for substitution effects. Second-order Taylor expansion of the net expenditure function yields the same results as subtracting the expression for the second-order Taylor expansion of the profit function from the expression for the second-order Taylor expansion of the expenditure function so I proceed accordingly. Second-order Taylor expansion of the expenditure function (1) yields:
(8)
∑
∑∑
= = = ∆ ∆ + ∆ = ∆ n i n i j i ij n i i ih h q p s p p E 1 1 1 2 1Where sijis a n× matrix with the derivatives of the Slutsky compensated demand curves. This n
equation can be rewritten as follows:5
(9)
∑
∑∑
= = = ∆ ∆ + ∆ × = ∆ n i n i j j i i ij ih n i i i ih h h p p p p p p E E 1 1 1 2 1 ε α α ijε represents a matrix with compensated price elasticities of demand for good i with respect to the price of good jand hence accounts for the substitution effect. Note that the diagonal of the matrix contains own-price elasticities of demand while the rest contains cross-price elasticities of demand. As this paper is only concerned with the effect of food prices on poverty, i represents all food items while
jrepresents either all food items or all non-food items. Now equation (9) can be simplified to:
(10)
∑
= ∆ ∆ + ∆ = ∆ 2 1 2 1 j j j i i ij ih i i ih h h p p p p p p E E ε α αHere good i represents all food items and good jrepresents all food items if it takes on a value of one and represents all non-food items if it takes on a value of two. Second-order Taylor expansion of the profit function yields an expression similar to (10) including the food production budget shares and a matrix containing the compensated cross-price elasticities of supply of food with respect to the price of all non-food items. In line with Robles and Torero (2010) I choose to suppress supply elasticities since I believe the compensated cross-price elasticity of supply of food with respect to the price of non-food equals zero. Reason for this is that it is extremely difficult for an agricultural household to
5 The step from equation (6) to (7) can be shown using simple algebra. Remember that the Slutsky derivative (
ij s ) is defined as j ih p q ∆ ∆
, the compensated elasticity of demand (εij)as i j j ih q p p q ∆ ∆
and the budget share (αih)as h ih i C q p . By noting that h j i ij i j j ih h ih i ij ih E p p s q p p q E q p × = ∆ ∆ × = ε
α one can see thatsijcan be replaced by
j i h ij ih p p E × ε
α . This yields the following expression:
h n i n i j j i i ij ih n i i ih h E p p p p p q E
∑
∑∑
= = = ∆ ∆ + ∆ × = ∆ 1 1 1 2 1 ε16 switch from producing food to producing non-food items as they lack resources and experience to do so. As a result, second-order Taylor expansion of the profit function yields the following expression: (11) i i ih h h p p E ∆ = ∆ λ π
Finally combining (11) and (10) gives the following expression for the compensating variation including substitution effects:
(12) j j i i ij ih i i ih ih h h p p p p p p E B ∆ ∆ + ∆ − = ∆ ε α λ α 2 1 ) (
III.III Including wage rate effects – An extension by Ivanic and Martin
Ivanic and Martin (2008) have extended Deaton’s model in order to incorporate the effect of food price rises on agricultural wages. Note that agricultural wages are incorporated in the vector w in expenditure function (1) (as this vector contains the prices of supplied production factors) and in the vector w in the profit function (2) (as this vector contains the prices of purchased production factors) The envelope theorem tells us that
i h w E ∂ ∂
is equal to the negative of the quantity of the supplied factor i . This amount will be called lih. I.e.:
(13) ih i h l w E = ∂ ∂
Applying the same logic to the profit function provides
i h
w
∂ ∂π
equals the negative of the demanded quantity of variable production input i , which I call nih . Hence:
(14) ih i h n w = ∂ ∂π
Combining (13) and the result that ih i h q p E = ∂ ∂
and applying the chain rule yields the following equation: (15) i h ih i h p w l q p E ∂ ∂ + = ∂ ∂
Similarly, combining (14) and the result that ih i h x p = ∂ ∂π
17 Where i p w ∂ ∂
represents a Stolper Samuelson vector relating factor prices to changes in the price of commodity i , this vector will be called ηh from here on. lhandbh are vectors of the quantity of
supplied factors and the quantity of demanded production factors, respectively. Now the compensating variation can be defined by subtracting (14) from (13):
(17) CV=dBh =(qih −xih)dpi +(lh −bh)ηhdpi
Note that (lh −bh)equals the vector of household’s net purchases of the variable factors, since l ih
indicates the negative amount of the supplied factor i and bih indicate the negative amounts of the demanded input factor i . Expressing equation (17) in terms of the household’s total expenditure level and expressing the price change in percentages rather than in absolute terms yields the following equation (expressed in discrete time):
(18) i i h h i i h i h i h p p L p p E B CV =∆ =(α −λ )∆ +η ∆ h
L is the vector of the shares of the net costs from net factor demand in the total household expenditure.
By assuming that the agricultural wage rate is the only factor that is influenced by commodity price changes and by noting that i represents all food items together (as this paper is concerned with the effect of rising food prices only) it is possible redefine Lh and ηh : Lh becomes the agricultural labor
costs share in total household expenditure and ηhthe elasticity of agricultural labor wage rates with
respect to food price changes. Furthermore, I assume that an increase in food prices will result in higher agricultural wages, i.e. that ηhis positive. It is easy to observe that the compensating variation
is lower (higher) than in the base scenario if Lhtakes on a negative (positive) value. This makes sense
because a negative value for Lh means that the household supplies agricultural labor and hence
receives compensation for the higher food prices in the form of higher wages.
Combining (7), (12) and (18) yields an expression for the compensating variation that allows for substitution effects, wage effects and differing consumer and producer prices.
18 IV: RESEARCH OBJECTIVES AND RESEARCH DESIGN
This paper has the following three research objectives:
- Identify which developing countries are most and least vulnerable to food price increases, irrespective of actual food price increases.
This objective will be addressed by estimating the yearly costs of compensating:
i) the average poor person within a country for a hypothetical food price increase of 15 percent taking place in the year 2008
ii) all poor people within that same country for a hypothetical price increase of 15 percent taking place in the year 2008, expressed as a percentage of total government expenditure
This analysis will be performed for as many countries as is possible with the available data.
- Provide an estimate of the amount of money that would be needed to compensate all poor people
living in developing countries for the 2007-2008 food price rise and compare these costs with the amount of money that would be needed in less extreme periods. Additionally, I want to provide an overview of the poor that are hurt most and least severely by the 2007-2008 food price crisis.
The first objective will be addressed by estimating the yearly costs of compensating all poor people within a country for the food price changes that has been observed in the year under consideration. This will be done for each year within the period 2005-2010 separately. After that I will
sum up the resulting costs for 2005 and 2006, 2007 and 2008, and 2009 and 2010 in order to compare the costs associated with the 2007-2008 food price increase with the costs associated with less extreme periods.
The second objective will be addressed by estimating the yearly costs of compensating the average poor person within a country for the food price changes that has been observed in the year under consideration. This will be done for the year 2007 and 2008 separately after which the results will be summed for each country. This results in the total costs needed to compensate the average poor person within a country for the 2007-2008 price increase. The total costs per country will be compared across countries.
The analyses will be performed for as many countries as is possible with the available data.
- Gain an insight into the empirical relationship between food prices and poverty rates.
19 increase. Therefore, the objective will be addressed by testing whether the compensating variation per poor person is positively correlated to the changes in the poverty headcount ratio and the change in poverty depth. The change in the poverty headcount is defined as the poverty headcount ratio in the next year minus the poverty headcount ratio in the current year. The same logic applies to the change in poverty depth.
This research contributes to the current literature in the following ways. First, opposed to all other papers, I only estimate the compensating variation if a country’s inputs that are needed for making the estimation are available for the same year as the year of the food price increase under consideration. Therefore, my estimations are more accurate as initial conditions cannot have changed throughout time. Furthermore, my results are more suitable for cross-country comparisons since estimations are based on inputs representing the same year. Second, I allow for incomplete pass-through of consumer prices to producer prices by making estimations under scenarios with different pass-through rates. Third and finally, after De Hoyos and Medvedev (2011), this is the first paper that does not make use of household surveys to estimate the impact of rising food prices on poverty. Although the use of household surveys will produce more accurate results, there are a few disadvantages. First, household surveys are difficult to access. Second, different household surveys contain different questions and rely on different assumptions, so results might not be comparable. Third, and most important, using household surveys is extremely time consuming. A quick-and-dirty approach can be useful when trying to develop policy responses to rising food prices within in a short time-frame.
Within this paper I define a person as poor is he or she has an income of $1.25 per day at 2005 PPP or less. This definition is similar to the World Bank’s definition of an extreme poor person. Furthermore, I define food price as the price of the basket of food that is used by the national bureau of statistics to measure the food consumer price index (food CPI). I assume that this basket is the same as the basket of food products that is consumed by the average poor person.
In my analyses I ignore the income effects that food prices might have via substitution effects or agricultural wages. The goal of this research is to estimate the impact of an increase in the price of the basket of all consumed food items rather than the impact of price increases of specific food items. Therefore, substitution between different food items cannot occur because all food items belong to the same basket that has one single price. Substitution could only occur between food items and non-food items. Data on the poor’s demand elasticity of non-food with respect to food do not exist. However, I expect it to be close to zero as it is unlikely that poor people that struggle to survive will reduce their consumption of food after a food price increase only to increase their consumption of non-food items. In order to use the model that includes wage effects I would need information on the average
21 V DATA REQUIREMENTS AND DATA CONSTRUCTION
The inputs that are needed to compute the compensating variation for a food price increase, ignoring wage and substitution effects, are: data on food prices, the household’s expenditure level, the share of the household’s expenditure that is spend on food and the profit the household obtains from the production of food, expressed in terms of expenditure. Note that this would require very detailed information per household that can only be obtained from household surveys. As explained in the previous chapter, household surveys are not easily accessible and also very time-consuming to
employ. Moreover, this research is concerned with the effects of rising food prices on the average poor person rather than the effect on all households. That is why I will estimate the average compensating variation per poor capita within a country rather than the compensating variation per household. Furthermore, I approximate expenditure levels with income levels. Reason for this is that data on the average expenditure level among the poor are not available. The average income level among the poor can be constructed using World Bank data; the procedure for this will be explained in the next section. Next, this paper is concerned with the effects of a rise in the price of the basket of all consumed food items rather than the price increase of specific food items. Taking into account these limitations and objectives the inputs needed for computing the average compensating variation per poor capita are: the average income level per poor capita, the share of income that the average poor person spends on food (food consumption budget share) and the average profit that a poor person makes with the production of food, expressed in terms of income (food production budget share) and data on food price inflation. The former three inputs are not readily available but can be constructed making use of existing data. The procedures for doing so will be explained in the next three sections. The fourth section will discuss considerations concerning the measurement of food price increases and the final section explains how individual compensating variations are aggregated to obtain the monetary costs of compensating all poor individuals within a country for a food price increase. In the appendix one can find descriptive statistics for all inputs.
V.I The average income per poor capita
The average income level of the poor is constructed with the poverty headcount ratio and the poverty gap. These data are acquired from the WDI database. The World Bank defines poverty headcount as the proportion of a population whose consumption or income is less than the established poverty line, which is $1.25 per day in our case. In mathematical terms:
∑
= < = N i i y I N PH 1 ) 25 . 1 ( 122 line of the non-poor (persons with an income above $1.25 a day) is zero. Writing the poverty gap in mathematical terms yields:
∑
= < − = N i i i I y y N PG 1 1.25 ) 25 . 1 ( * ) 25 . 1 ( 1Substituting the expression for the poverty headcount into this expression yield and recognizing that the average income of the poor equals
∑
= < = N i i i I y y N y 1 ) 25 . 1 ( * 1yields the following expression:
PH y PG * 25 . 1 ) 25 . 1 ( − =
Rearranging terms yields the formula for the average income per poor capita: (20)
PH PG y=1.25−1.25
The compensating variation will be calculated for each year within the period 2005-2010 separately. For reasons that will be explained in the next section, the analysis is limited to the 73 countries included in the GIDD database. Hence, I need the poverty headcount ratio and the poverty gap for all countries in the GIDD database for every year from 2005 till 2010. For eight countries these numbers are not available for any of the six years. Thus, these countries are left out of the analysis. For most other countries, the indicators are only available for one or a few years within the period 2005-2010. In order to increase the sample size I make the assumption that poverty headcount ratios and poverty gap rates do not change significantly within one year; if the poverty headcount ratio and the poverty gap are not available for a year, I replace the average income level per poor capita by its value of the previous or next year (if available).Hereby, I assume that the previous year is more representative than the next year. E.g. if the poverty headcount ratio and poverty gap of a country are available for 2005 and 2007 only, the average income level per poor capita of 2005 is used to calculate the compensating variation for the 2006 food price increase. Unfortunately, even after applying this method, I still do not have an average income level per poor capita for all six years for all countries in the GIDD database.
V.II Food consumption budget shares
23 expenditure per household is available. The authors use this information to estimate the Engel curve for developing countries. They have estimated the following equation:
(21) ECA SAS LAC EAP Urban H y yi i i i * 055 . 0 * 014 . 0 * 11 . 0 * 042 . 0 * 04 . 0 ln * 017 . 0 ) (ln * 0001 . 0 ln * 109 . 0 044 . 1 2 − − − + − − + − = α
I will use equation (21) to estimate the average food consumption budget shares per capita among the poor rather than the food consumption budget share per household i. In equation (21) yi represents the monthly household income per capita of household i . To be able to estimate the average food consumption budget share per poor capita I replace yi by y , the average yearly income per capita
among the poor divided through 12. Hiis the household size of household i I replace this parameter
by H , the average household size of households living at less than $1.25 per day. To obtain this input I make use of the GIDD database6. The GIDD database contains distributional data collected from household surveys for 121 countries and covers 90 percent of the world population. All data in the GIDD database are standardized to the year 2000. Data on household sizes are available for 73 of the 121 countries. To my knowledge, there are no other sources that publish data on average household sizes among the poor. Hence, the dependence on the GIDD database limits the analysis to 73
countries. The GIDD database publishes the average household size for each vintile of the population, ranked from the poorest five percent till the richest five percent. In order to calculate the average household size for the population living at less than $1.25 a day I match the poverty headcount ratio to the corresponding number of vintiles and take the average of the average household sizes within these vintiles. If the poverty headcount percentage is in the lower half of a vintile I round the number of vintiles off downwards and if the poverty headcount percentage is in the upper half of a vintile I round the number of vintiles off upward. This approach yields the average household size among poor households assuming that the average household size per income vintile has not changed since 2000. For instance, if the poverty headcount is 17.12% the average household size of the poor is the average of the average household sizes of the first three income vintiles. The term Urban is an urban identifier, which takes on a value of one if the household resides in an urban area, and zero is it resides in a rural area. Finally, the previous steps allow me to calculate the average food consumption budget share per poor household. By assuming that the food consumption budget share of the average poor individual is equal to the food consumption budget share of the average poor household I am able to compute the average food consumption budget share per poor capita.
As explained above, De Hoyos and Lessem (2008) estimate the Engel Curve based on the reported consumption budget shares in household surveys of 22 developing countries. In their paper, De Hoyos and Lessem report their estimations of the food consumption budget shares among the extreme poor for the other 51 countries in the GIDD database. They also list the food consumption budget shares among the extreme poor for the 22 countries as derived from the household surveys. I have compared
24 my estimations of the average food consumption budget shares per country7 with the food
consumption budget shares as reported by De Hoyos and Lessem. Differences can arise for several reasons. First, DeHoyos and Lessem calculate the food consumption budget share among the extreme poor by calculating the food consumption budget share for all households within a household survey separately. Then they calculate the average food consumption budget share of all extreme poor households within that survey. I calculated the food consumption budget share based on the average income level of the poor directly. Second, the household surveys within the GIDD database are adjusted to reflect the year 2000. Hence, the food consumption budget shares reported by De Hoyos and Lessem are based on income levels of the year 2000. My estimates are recalculated for each year within the period 2005-2010 and are based on income levels of that same year. Finally, the reported food consumption budget shares of the 22 countries that are used to estimate the Engel Curve are the values taken from household surveys directly and are not calculated using the Engel Curve.
If the difference between the food consumption budget share as calculated by me and the food consumption share as reported by De Hoyos and Lessem differs by more than 0.05 I replace my estimation of the average food consumption budget share by De Hoyos and Lessem’s estimation. Unfortunately it is not possible to adjust De Hoyos and Lessem’s estimates to reflect more recent years.
V.III Food production budget shares
De Hoyos and Medvedev (2011) have estimated the relationship between the share of income that is derived from agricultural production,λi, per capita household income, and regional effects. For this they use the share of income generated by agricultural production activities reported in household surveys in 19 developing countries. These data are available in the Rural Income Generating Activities (RIGA) database from the Food and Agricultural Organization (FAO). This paper will employ this equation to estimate the food production income budget shares of the countries under analysis. The equation takes on the following form:
(22) i 0.76 0.54*yi 0.0002*yi 0.38*EAPi 0.30*ECAi 0.44*LACi 0.49*SASi 2 − − − − + − = λ
De Hoyos and Medvedev define εias the share of income that is derived from agricultural
self-employment activities. It is assumed that εi is zero for households in urban areas. In this equation yi
is the daily income of household i . As this papers works with the average income of the poor living at less than $1.25 a day, I replace yi by y the average yearly income per capita among the poor divided through 365. The other variables are the fixed regional effects, EAP stands for East Asia and the Pacific, ECA for Eastern Europe and Central Asia, LAC for Latin America and SAS for South Asia. These variables take on a value of one if the household resides in the relevant area and zero otherwise.
7 The average food consumption budget share per country is a weighted average of the rural and the urban food consumption budget share.
25 The country effects for the Middle East and North Africa (MNA) and Sub-Saharan Africa (SSA) are zero.
When examining equation (22) more closely one observes that is has a minimum. The intercepts range from 1.41 and 2700 for Middle East and North Africa and Sub-Saharan Africa to 0.50 and 2700 for South Asia. This means that the income of the South Asian rural poor must be smaller than 0.50 in order to receive income from the production of food and that all Middle East and North African and Sub-Saharan African rural poor receive income from the production of food.
V.IV Food price inflation
Another input needed to compute the yearly compensating variation per poor capita is the yearly food price inflation per country. I choose to use national prices rather than world market prices because I agree with De Hoyos and Medvedev (2011) that full transmission of world prices to local prices is not realistic. I do not correct this number for inflation of non-food items or expected inflation because I believe that these adjustments are inconsistent since the poverty line, and hence, the average income level of the poor is also not corrected for inflation. I measure food price increases (decreases) with the food Consumer Price Index (food CPI) that is available in a database from the Food and Agriculture Organization (FAO). From this database I need the yearly food price change for all countries and corresponding years included in this research. Missing data are complemented with data from the International Labour Organization (ILO) and with data from national statistical bureaus. Unfortunately the food price inflation rate for Costa Rica 2006, Georgia 2009, Romania 2010 and the Kyrgyz
Republic 2007, 2008, 2009 and 2010 are not available. Therefore, these countries and corresponding years are removed from the sample.
V.V Aggregating per capita compensating variations
The previous sections explained how this paper proceeds in calculating the compensating variation per poor capita per year. This section explains how this compensating variation per capita can be
transformed to compute the amount of money that would be needed to compensate all poor individuals in a country for the increase in food prices that country experienced in a given year. This number can easily be calculated by multiplying the compensating variation per capita by the number of persons that live at less than $1.25 per day. This number can be computed by multiplying the poverty headcount ratio of a country with its population, a number that can be obtained from the WDI database.
26 this assumption. As the compensating variation per capita differs for rural and urban individuals, data on the number of poor people living in urban versus rural areas is needed in order to calculate the amount of money that would be needed to compensate all poor individuals for an increase in food prices. The GIDD database contains the average proportion of the population that lives in urban areas per income vintile. In order to acquire the average proportion of the poor living in urban areas I employ the same matching procedure as used for calculating the average household size of the poor. This approach yields the average proportion of the poor that lives in urban areas assuming that no urbanization took place since 2000. I adjust these numbers for urbanization effects by letting the computed proportion grow with the urbanization between 2000 and the year under examination. The urbanization rate is the growth rate of the fraction of the total population living in urban areas within each country and is obtained from the WDI database. I assume that the urbanization rate of the poor is the same as the urbanization rate of the entire population. This assumption is made because
urbanization rates for specific groups of the population are not available. The assumption will be challenged in a sensitivity analysis in the appendix. Now I am able to compute the number of poor living in urban (rural) areas by multiplying the proportion of poor living in urban (rural) areas by the number of poor within a country and subsequently the amount of money needed to compensate all poor individuals in urban (rural) areas for an increase in food prices can be computed. The amount of money needed to compensate all poor individuals in a country for an increase in food prices within a given year is the sum of the amount of money needed to compensate all poor individuals living in urban areas and the amount of money needed to compensate all poor individuals living in rural areas. Unfortunately, the household surveys of Argentina and Uruguay within the GIDD database are held among urban citizens only. Therefore it is not possible to compute the proportion of the poor living in urban areas and consequently I cannot compute the amount of money needed to compensate all poor individuals in these two countries for food price increases. The same is true for Peru, where the household survey was held among rural citizens only. However, as it is still possible to calculate the average compensating variation per poor rural capita and per poor urban capita for these countries, they are not removed from the sample.
V.VI Final Sample Sizes
In order to address the first research goal of identifying which developing countries are most (least) vulnerable to a 15 percent food price increase I need the above described parameters for the year 2008 only. These data are available for 47 countries of the 73 countries within the GIDD database. Hence, I will calculate the costs of compensating the poor for a 15 percent food price increase for 47 different countries.
27 for the 2007 and 2008 food price increases separately and add them together. Therefore, I can only calculate the compensating variation for the 2007-2008 food price increase for countries for which the needed inputs are available for both the year 2007 and 2008. This requirement is satisfied by 46 countries. Furthermore, I want to compare the costs of compensating all poor individuals in these 46 countries for the 2007-2008 food price increase with the costs that would be needed to compensate all poor individuals within these same countries for the food price increases that have been experienced during 2005-2006 and 2009-2010. Hence, the inputs needed to calculate the compensating variations must also be available for the years 2005, 2006, 2009 and 2010. Unfortunately, this is not the case for all 46 countries. There are only 22 countries for which the inputs are available for all 6 years. There are 33 countries for which the inputs are available for the all years from 2005 till 2008 and 24 countries for which the inputs are available for all years from 2007 till 2010. I will still calculate the compensating variation for the 2007-2008 price increase for the 46 counties as this will address the objective of identifying the countries that were most and least affected by the 2007-2008 food price rise.
28 VI. RESULTS
The first section of this chapter will discuss the impact of a simultaneous 15 percent price increase on the poor in a wide range of countries. The next section will elaborate on the impact that the 2007-2008 food price crisis has had on the poor in 46 countries. The third section tests how the results provided in the first two sections change if food price increases are not passed through completely to poor food producers. The final section addresses the empirical correlation between food prices and poverty rates.
VI.I The impact of a 15 percent food price increase in the poor
This section will discuss the impact of the 15 percent price increase on poverty for the following four regions: Africa and the Middle East (called Africa henceforth), South Asia, East Asia and Pacific Asia (referred to as Asia henceforth), East Europe and Central Asia (referred to as ECA henceforth), and Latin and Central America (referred to as LAC henceforth). I will not discuss the results for each country separately, except for some remarkable outcomes. The entire range of compensating variations per country (average, rural and urban) can be found in the appendix.
The amount needed to compensate the average poor person for a hypothesized price increase of 15 percent taking place in 2008 ranges from $9.76 (Brazil) to $41.68 (Thailand). Table 1 shows the compensating variations per region. These amounts are a weighted average of the national compensating variations, weighted by the number of poor per country. The rural and urban compensating variation are weighted by the number of rural and urban poor per country. Table 2 shows the unweighted compensating variations per region.
Table 1: The weighted compensating variations for a 15 percent price increase, per region
Region Average CV Urban CV Rural CV
LAC 14.45 15.29 12.69
Asia 36.48 34.81 37.36
ECA 26.63 26.76 26.50
Africa 16.08 28.48 15.14
Table 2: The unweighted compensating variations for a 15 percent price increase. per region
Region Average CV Urban CV Rural CV
LAC 19.69 19.03 19.88
Asia 37.64 35.76 37.98
ECA 26.31 26.70 26.09
29 As one can see, the compensating variation per average poor person hides the differences between the compensating variation for the rural and that of the urban poor. When looking at the average
compensating variation, it seems that the African poor are only slightly worse off than the LAC poor with a compensating variation of $16.08 per poor person per year versus $14.45 per poor person per year in LAC. However, this is only true for the rural African poor, who have only limited exposure to food price increases due to a high food production budget share: their food production budget share is by far the highest of the sample. The African urban poor are much worse off than their rural
counterparts. The compensating variation for the urban poor is almost twice the compensating
variation for the rural poor. Among the urban poor the LAC poor suffer the least. This makes sense as the food consumption budget share is by far the lowest for LAC countries. When looking at the weighted rural compensating variation in seems that LAC rural poor are better off in the event of a food price crisis than the African poor. However, the low weighted rural compensating variation of the LAC is mainly the result of 67% of the rural poor being from Brazil, the country with the lowest rural compensating variations of the entire sample.8 Note that the LAC countries have the second lowest food production budget share according to equation 22 (after South Asia). Hence, the low rural compensating variation among the LAC poor indicates that the food consumption budget share among LAC countries is small enough to make up for their low food production share.
Asian countries are most exposed to a food price increase. This is true for both their urban as well as their rural poor and hence on average too. The Asian group consists of 6 East Asian and Pacific countries and 2 South Asian countries. Because there are only two South Asian countries included in this analysis it might be inappropriate to conclude that all South Asian countries will suffer
considerably after a food price increase. Especially since their food consumption budget share is lower than that of East Asian and Pacific and African countries. However, South Asian countries do have the smallest food production share so it might be true that South Asian rural poor are the worst off in the event of a food price rise. East Asian and Pacific countries have by far the highest food consumption budget share. Their food production budget share is low as well, but that of South Asian poor and LAC poor is lower. Hence, the large food consumption budget share of the East Asian and Pacific poor seems to be the most important source of their large exposure.
The ECA region seem to be second worst off in the event of a food price increase. Their average and rural compensating variation is approximately in the middle of the Asian and African compensating variation. Their urban compensating variation is smaller than that of Africa. This makes sense as their food consumption budget share is quite a bit lower. However, more than 90 percent of the African poor population lives in rural areas, causing their average compensating variation to be lower than that
8 This fact also explains the large gap between the weighted average compensating variation and the unweighted average compensating
30 of ECA.
Note that the compensating variation is heavily dependent on the initial average income level. Expressing the average compensating variation as a percentage of initial income yields a range from 4.85% for Burkina Faso to 10.43% for Lao and Cambodia. Table 3 shows the weighted compensating variations per region expressed as a percentage of initial income. When comparing these results to table 1 and 2 one can see that the range of compensating variations is not very different. However, in relative terms, the African poor seem to be better off than the LAC poor. This is caused by the
relatively low weighted average income level among the LAC poor; $0.60 a day compared to $1.00 in Asia, $0.86 in ECA and $0.83 in Africa.
Table 3: The weighted compensating variations for a 15 percent price increase expressed as a percentage of initial income, per region
Region Average CV Urban CV Rural CV
LAC 6,51% 6,77% 5,94%
Asia 9,97% 9,51% 10,25%
ECA 8,50% 8,55% 8,45%
Africa 5,28% 9,54% 4,78%
Next, I examine what the cost would be to compensate all poor individuals in a country for the simulated 15 percent food price increase taking place in the year 2008. Since these costs are linear to the population size of a country I have expressed these costs as a percentage of the countries’
