WPS6858


Policy Research Working Paper                     6858




              The Price Is Not Always Right
   On the Impacts of (Commodity) Prices on Households
                      (and Countries)

                                Daniel Lederman
                                  Guido Porto




The World Bank
Latin America and the Caribbean Region
Office of the Chief Economist
May 2014
Policy Research Working Paper 6858


  Abstract
  This paper provides an overview of the impact of once-                             expenditure dependency suggests sizable impacts and
  and-for-all changes in commodity prices and other                                  adjustments following commodity-price shocks. The
  prices on household welfare. It begins with a collection                           paper explores these effects with a review of the literature.
  of stylized facts related to commodities based on                                  It studies consumption and income responses, labor-
  household survey data from Latin America and Africa.                               market responses, and spillovers across sectors. It ends
  The data uncover strong commodity dependence in both                               up providing evidence on the relative magnitudes of
  continents: households typically allocate a large fraction                         various mechanisms through which commodity prices
  of their budget to commodities and they often depend                               affect household (and national) welfare in developing
  on commodities to earn their income. This income and                               economies.




  This paper is a product of the Office of the Chief Economist, Latin America and the Caribbean Region. It is part of a larger
  effort by the World Bank to provide open access to its research and make a contribution to development policy discussions
  around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors
  may be contacted at dlederman@worldbank.org and guido.porto@depeco.econo.unlp.edu.ar.




          The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
          issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
          names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
          of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
          its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                        Produced by the Research Support Team
        The Price Is Not Always Right: On the Impacts of
       (Commodity) Prices on Households (and Countries)∗
                    Daniel Lederman†                       Guido Porto‡
                            LCRCE                        Dept. of Economics
                       The World Bank                 Universidad de La Plata




       JEL Classification: O1, F1, D1, D6
       Keywords: Commodity prices, household welfare, price transmission




   ∗
      Laura Chioda, Louise Cord, Irene Brambilla, and Emily Sinnott provided insightful comments during
the early stages of this research project. We are grateful to Emmanuel Jiminez and three anonymous referees
for detailed comments on a previous version of this article. We are especially grateful to Cecilia Peluffo and
German Bet for excellent research assistance. All remaining errors are our responsibility.
    †
      World Bank, email: dlederman@worldbank.org
    ‡
      Universidad Nacional de La Plata, Calle 6 e 47 y 48, Oficina 521, La Plata 1900, Argentina; email:
guido.porto@depeco.econo.unlp.edu.ar
1         Introduction
Energy prices rose more than 140% between 2000 and 2013, according to the World Bank’s
energy-price index, while the food-price index rose by more than 80%. These increases in
commodity prices motivated research to assess the implications for households and workers
in developing countries, as well as policy responses. An example is the World Bank’s 4
billion dollar contribution to a joint fund with J.P. Morgan to help developing countries
invest in commodity-price hedging instruments launched in June 21, 2011.1 Yet commodity
prices have stalled since 2012, and observers are once again concerned about the social and
economic consequences of declining commodity prices2 . Clearly, part of the concern is about
the volatility of commodity prices, as much as the medium-term trends per se.
        Whether it is the trends or the volatility of commodity prices, knowledge about the
effects of commodity-price fluctuations on household welfare is useful for policy as much as
it is useful for understanding economic behavior. It is also important to have a sense of the
magnitudes of the impacts involved.
        The objective of this article is to put forth an economic framework for understanding how
people in developing countries are affected by relatively permanent changes in commodity
and other prices. The approach is to review the literature framed by a toy model that
highlights issues relevant in the discussion of prices and welfare. The review of the literature
canvasses research from various branches, including shocks to exchange rates, financial crises,
and trade reforms. In all these cases, the shocks manifest themselves as price changes, and
thus the findings are relevant for the discussion of commodity prices. We complement the
literature review with a detailed analysis of household-survey data on commodity dependence
in Latin America and Africa.
        Since we are interested in the impact of commodity prices on household welfare and
behavior, our framework builds on agricultural household models. This framework analyses
how households choose consumption bundles and how they earn their living. In some cases,
especially in rural areas, income is generated by agricultural production (profits). In other
    1
        World Bank, Press Release No:2011/559/EXT, Washington, DC.
    2
        See, for example, “Life after the Commodity Boom,” The Economist, March 29, 2014



                                                     1
cases, where the economy is more industrialized, wages and employment are the predominant
sources of income, which can be affected by commodity prices. Since prices can affect
both consumption and income, the simple framework opens the door to a discussion of how
household welfare is affected by changes in commodity prices, both in the short- and long-run.
The literature provides numerous empirical findings that can bring together various pieces
of the puzzle illustrated by the framework.
   First, we review the literature on the first-order impacts of price changes. These are the
impacts that would take place if households adjust neither the quantities of commodities
that they consume nor change the sources of income they receive in response to the price
change.   The article derives the commonplace net-consumer-versus-net-producer result,
which establishes that a price increase hurts net-consumers and benefits net-producers.
   Second, we discuss spillovers from commodity prices to other sectors. This refers to,
for instance, whether households adjust all consumption choices in response to higher food
prices, or whether only food expenditures are affected. Also, we report evidence on spillovers
in employment and wages; that is, changes in food prices can affect wages and employment
not only in the food sectors (primary, agro-manufactures) but also in non-food sectors.
   Third, the paper reviews how prices are transmitted to the local economy. Because
commodities are traded in international markets, most of the shocks associated with
commodity prices are global in nature. Consequently, a relevant question is the extent
to which changes in international prices pass-through to domestic markets.
   Fourth, the paper explores available evidence on consumption and production responses.
Here, we review how consumers adjust the quantity and quality of the consumption basket,
as well as other behavioral responses associated with consumption. We also explore some of
the patterns of adjustment in production, particularly supply responses in agriculture.
   It is important to mention at the outset that there are at least five related literature
surveys. These are Winters, McCulloch and McKay (2004) and Goldberg and Pavcnik
(2004), which cover trade liberalization and poverty; Goldberg and Pavcnik (2007) and
Harrison, McLaren and McMillan (2011), which cover trade and the distribution of income;
and Lederman (2013), which covers trade and inclusive growth. Our review complements


                                              2
this work by merging issues of poverty with issues of inequality. Furthermore, we have a
somewhat narrower focus, as we are mainly interested in commodity prices, but we take a
somewhat deeper look at some of these issues, all in the context of a general framework.
    The rest of this paper is organized as follows.      Section 2 presents an overview of
household commodity dependence in Latin America and Africa. Using various household
income and expenditure surveys from Latin American and African countries, we estimate the
average budget shares spent on commodity-related sectors (mostly food and energy), and the
average income share derived from commodity-related sectors (agriculture, forestry, fishing,
mining, food manufactures, and energy). The data reveal strong commodity dependence in
both continents. Section 3 introduces the analytical framework and reviews the literature
that measures those effects using data from developing countries.         The implications of
price changes for net-consumers and net-producers are central in this analysis, with labor
markets, various types of spillovers across markets, and imperfect price transmission of border
prices playing fundamental supporting roles in the story. Section 4 turns to departures
from first-order effects and studies the role of household adjustments in production and
consumption. Section 5 presents a simple example of the magnitudes of the different channels
identified in the paper using Mexican data. Section 6 concludes.



2     Commodity Dependence: Stylized Facts
This section characterizes the dependence on commodities of households in developing
countries. To study dependence on consumption, we focus on expenditure shares spent
on items related to commodities, mainly food and energy. Households’ dependence on
commodities on the income side is then measured by income shares in agriculture (including
forestry and fishing), mining, food manufactures (beverages, food, and tobacco), and energy.
An important aspect of the analysis is related to poverty and inequality, because dependence
on commodities either on the consumption or the income sides can vary across quintiles of
the income distribution in a given point in time.
    Regarding commodity-dependence on the consumption side, Table 1 presents the average


                                              3
expenditure shares on commodities for eight Latin American countries: Argentina, Bolivia,
Colombia, El Salvador, Mexico, Nicaragua, Panama, and Peru.              There is significant
cross-country variation in the average share spent on commodities, ranging from an average
share of 71.3 percent in Bolivia to 31.7 percent in Mexico. A common feature is that, in
all these countries, the share of commodity expenditures decreases across quintiles. This
is expected due to Engel’s Law: richer households typically spend a smaller share of their
total expenditures on food. In Bolivia, for example, the average commodity share is 82.9
percent in the first quintile and 54.4 percent at the top quintile. In Mexico, the average share
of the bottom quintile is 40.6 percent and it is 20.8 percent at the top. In Colombia, the
poorest households spend, on average, 80.1 percent on commodities; the richest households
spend 41.6 percent. Clearly, differences in the populations covered by the surveys across
countries (i.e., urban versus rural populations) can account for part of the differences in the
average commodity expenditure shares. But, leaving this aside, the data reveal a very high
commodity dependence on the consumption side, especially for the poorest households.
   Table 2 reports commodity-dependence of household expenditures for a set of African
countries. There is also significant variation across countries, from a minimum average
share of 42.7 percent in South Africa to a maximum of 89.9 percent in Madagascar and
84.4 in Tanzania. The commodity dependence on consumption seems higher in Africa than
in Latin America, however. In most countries, commodity dependence in consumption is
more prevalent at the bottom of the quintile distribution. This means that Africa and Latin
America are both highly dependent on commodities and, moreover, this dependence declines
with household income. This pattern explains the literature’s concerns about the poverty
impacts of commodity price hikes.
   Tables 3 and 4 turn to dependence on the income side. The aim is to measure the
share of household income derived from commodities, including income from employment
(wages) and self-employment (for instance, agricultural home production). The following
sectors are included in our definition of “commodities:” agriculture, forestry, and fishing;
mining; food manufactures (including beverages and tobacco); and energy. Starting with
Latin American data for 2004, Table 3 reports the total combined share of income derived


                                              4
from commodities. Income commodity dependence varies widely across countries. This
comparison, however, is unlikely to be very useful because of differences in coverage. An
example of this is Argentina, which in spite of being a country with a relatively important
agricultural sector, the household survey suggests that only 3.2 percent of household income
is, on average, derived from commodities. Clearly, this is more a manifestation of the urban
focus of the survey than of the economy. In the discussion that follows, it is thus convenient
to focus more on within country comparisons across quintiles. A regularity in most of our
Latin American data is that lower quintiles exhibit higher shares of income derived from
“commodities.” The differences are sometimes important. In Paraguay, for instance, while
the bottom quintile earns 44.4 percent of income from commodities, the top quintile earns
only 14.7 percent.
   Table 4 uncovers similar patterns in the African data. The income commodity dependence
in Africa is very high (higher than in Latin America) and varies across countries. The lower
quintiles also tend to show higher commodity dependence. In Ghana, for instance, the share
of income derived from commodities is 71.4 in the first quintile, and 27.1 in the top quintile.
Table 4 shows that, in the data, most of the dependence from commodities in Africa is
related to dependence on agriculture. Since the data from both continents indicates that
poor households are on average highly dependent on commodities on both the consumption
and income sides, the distinction between net consumers and net producers will become
particularly relevant for the literature that attempts to assess the distributional and poverty
impacts of commodity-price fluctuations.
   As a preliminary step in this direction, it is worth analyzing the evolution of consumption
and income dependence on commodities during a period of time when global commodity
prices rose. The purpose of the empirical exercise is to describe the changes in income
shares observed in the data from 2004 to 2008. Transitions in and out of dependency can
be studied with panel data, which is unavailable in our sample countries. We can, however,
build pseudo-panels of households using repeated cross-section with the method proposed
by Dang, Lanjouw, Luoto and McKenzie (2011). Since only the income surveys in Latin
America are repeated cross-sections, we focus here solely on income commodity dependence


                                              5
for Latin America. It suffices to say that due to sampling and measurement issues, one can
get a distorted view of over-time trends by looking just at the household expenditures and
incomes between cross sections. What is needed is a methodology that removes the effect
of changes in household samples on the average household consumption and income shares
derived from commodities.
    To build the 2004-2008 pseudo-panel, we focus on households observed in 2004 and
estimate commodity income shares for these households in 2008. This procedure requires the
estimation of a regression model for commodity income shares separately for the 2004 and
2008 samples. We then use the estimates from the 2008 sample, together with the residuals
from the 2004 regression to produce counter-factual panels. Intuitively, we predict average
income shares using the observed covariates in 2004 and the estimated parameters for 2008.3
    The results are reported in Table 5. The first row reproduces the observed income
shares in 2004 that were reported in Table 3, while the second row shows the predicted
2008 shares. A key result is that income dependence on commodities increased in Chile, El
Salvador, Honduras, Peru, and Uruguay, which is expected given the rise in global commodity
prices. However, dependency remained roughly constant in (urban) Argentina and in the
Dominican Republic; Brazil and Mexico experienced a slight decline in commodity income
dependence; while in Costa Rica, Ecuador, Panama, and Paraguay, the decline was large.
This international heterogeneity might be surprising – given the commodity price boom –
but differences are expected due to various factors discussed below.
    It is also noteworthy that the dynamics of income dependence on commodities varies
across quintiles within countries.              In El Salvador, Honduras, and Uruguay, income
dependence increased not only on average, but also for all quintiles. In Chile, average
dependence rose, but it decreased for lower quintiles and increased for higher quintiles.
    3
      Let si,t be the commodity income share of household i at time t. The cross-section model for t is
si,t = xi,t βt + εi,t . For t + 1, the model for household j is sj,t+1 = xj,t+1 βt+1 + εj,t+1 , where we use
j instead of i to highlight the fact that these are repeated cross-sections, not panels. Denote βt+1 the
OLS estimates of βt+1 . From the empirical distribution of the residuals, let εi,t+1 be the draw attached to
household i. With these estimates, the prediction for the income commodity shares for household j at time t
is si,t+1 = xi,t βt+1 + εi,t+1 . Note that, since we are taking averages, the residuals vanish (or are very small).
Consequently, the lower and upper bounds in Dang, Lanjouw, Luoto and McKenzie (2011) are very close in
our experiment. We report below the lower bound.



                                                        6
In contrast, in Peru, average dependence also rose, but it increased for lower quintiles
and decreased for higher quintiles. In Argentina and the Dominican Republic, average
dependence remained constant; but it increased for low quintiles and declined for high
quantiles in Argentina; and declined for low quintiles and increased for high quintiles in
the Dominican Republic. In Brazil, Ecuador, Mexico and Paraguay, average dependence
rose, declined for low quintiles, but it increased for higher quintiles. Finally, in Costa Rica
and Panama, dependence declined for the average as well as for all quintiles.
        These results uncover several observations that are important for the subsequent analysis.
First, the data show that households adjust after price shocks. In fact, the commodity income
shares changed between 2004 and 2008, when global prices rose to record levels. Second,
these adjustments were heterogeneous, not only across countries but also across quintiles of
the distribution of income within countries. In what follows, the literature review aims to
shed light on the impacts of commodity price changes on household welfare and on household
responses and adjustments in response to price signals.



3         Price Changes and Household Welfare
This section explores the impacts of price changes on household welfare. Our analytical
framework builds on standard agricultural household models, as in Singh, Squire and Strauss
(1986). Although this framework has influenced a large body of research spanning various
fields of economics, the ongoing discussion will modify it as needed to illustrate key issues
associated with the welfare impacts of commodity price changes.
        The framework analyzes the consequences of price variations from the viewpoint of
households. Analyses of welfare changes resulting from such price signals tend to follow the
approach pioneered by Dixit and Normal (1980), in which a household’s budget constraint is
(at least initially) assumed to be fixed.4 Simply put, in equilibrium, household expenditures
(including savings) have to be financed with household income (including transfers).
        In turn, total household expenditures are, by definition, the sum of expenditures across
    4
        An alternative method is to start with the indirect utility function,see Deaton (1997).



                                                         7
goods, evaluated at their corresponding prices and quantities (for a given level of household
utility). However, the households decision about how much to spend depends on the prices
of the consumed goods, on the level of utility, and on other household characteristics (such
as household composition).
   A household’s income, on the other hand, comprises the sum of the wages of all working
members of the household plus profits from business or other economic activities. Profits
include, for instance, the net income from agricultural production or farm enterprises.
They depend on prices, technical capacity and household characteristics. Since profits are
defined as sales net of purchases of inputs, some of the costs imposed by taxes on inputs or
intermediate goods can be assessed by their impacts on profits. Income may also comprise
transfers (public or private), savings and other unmeasured returns to factors of production,
such as land holdings.
   This simple framework highlights the fact that household welfare depends on market
equilibrium variables such as prices and wages (that affect household choices) and also
on household endowments. For example, household consumption depends on the prices
of consumer goods, and household income depends on the type of labor endowment (skilled,
unskilled), the market wage rate, and the prices of goods produced by economic enterprises
run by members of the household. It follows that changes in commodity prices affect
welfare directly via consumption and production decisions, and that these impacts are
heterogeneous insofar as they depend on household choices and endowments. In addition,
there are short-run impacts when households do not adjust their consumption, employment
or production choices; medium-run impacts refer to the welfare (or consumption or income)
effects that take into account partial adjustments; and long-run impacts take into account
income growth, investments, and other factors that are related to long-run choices. In other
words, the time horizon associated with the impact of price changes on household welfare is
commonly defined by the extent to which a household has been able to maximize its total
consumption or income in light of the price changes.




                                             8
3.1       First-Order Impacts

Let us consider now the impacts of changes in the price of a commodity. To simplify,
we follow Deaton (1989a) and assume that the principle of “separability” holds.5 Under
this assumption, production decisions are independent of consumption decisions (utility
maximization). This means that we can consider the income level of the household as
exogenous (once optimal production decisions have been made) when utility maximization
takes place as the household adjusts its consumption bundle. The separability assumption
is not just a technicality: it allows us to simplify the welfare analysis, but it is valid only
under certain economic conditions, such as perfect and complete markets for goods, credit,
                         6
insurance and so on.
       Prices affect both expenditures and income, even if separately. On the consumption side,
consumers are worse off if prices go up but are better off if prices go down. In a first order
approximation, these impacts can be measured with budget shares. That is, the size of the
effect of a price change on a household directly depends on how important the good is in the
household’s total expenditures. On the income side, there is also a direct impact on profits, if
the household produces the affected good, which depends on the share of income attributed
to this good. In rural economies, this source of income can account for a large fraction of
total income. In more urbanized economies with more developed labor markets (as in many
places in Latin America), the role of the direct production of (agricultural) goods tends to
be less important.
       Comparisons of budget shares and income shares establish a key result in the literature:
a price increase hurts net consumers (defined by the difference between expenditures and
incomes associated with the commodity whose price changed) and benefits net producers.
The opposite is true for price declines: net consumers are made better off and net producers
worse off. Further, the direction of welfare impacts carries over to the level of a national
economy or any aggregate unit of analysis: a country that is a net exporter of an agricultural
   5
     See Barnum and Squire (1979) and Singh, Squire, and Strauss (1986) for a theoretical discussion of the
principle of separability. Benjamin (1992) tests for separability in Indonesia and cannot reject it.
   6
     To further simplify the exposition, we also assume separability between consumption and leisure in
utility.



                                                    9
good will benefit, on average, from price increases, whether they come from global market
fluctuations or from international policy reforms such as the international liberalization of
agricultural trade policies or subsidies; but net importers will probably be hurt by those
changes. In the context of the recent food crisis, for instance, this implies that countries
that are net importers of food were in principle hurt, while net-exporters were the winners.


3.2        Households, Net-Consumers, Net-Producers and Inequality

The net-consumer/net-producer result was introduced by Deaton (1989a), who launched a
literature to study the distributional effects of price changes. Deaton (1989a) used data from
the Thailand Socioeconomic Survey of 1981-82 to explore the distributional consequences of
an export tax on rice across Thai households. He found that an increase in the price of
rice (resulting from the elimination of the export tax) would benefit the average household.
The average poor, as well as the average rich, would benefit little. The benefits for the poor
are small because they tend to both consume and produce lots of rice, thus selling little.
The average benefit to the rich is also small because while sellers are often large, there are
few of them among the rich. The gains would be much higher in the middle of the income
distribution, indicating that the middle class would gain the most from higher rice prices.7
       The ideas introduced in Deaton’s work have been, and still are, extensively utilized
in the literature. Early examples include: Deaton (1989b), who reviews applications for
Cote d’Ivoire, Indonesia, and Morocco; Budd (1993), who investigates food prices and rural
welfare in Cote d’Ivoire; Benjamin and Deaton (1993), who study cocoa and coffee in Cote
d’Ivoire, too; Barret and Dorosh (1996), who look at rice prices in Madagascar; and Sahn and
Sarris (1991), who examine structural adjustments in several Sub-Saharan African countries.
Deaton (1997) provides an account of the early use of these techniques in distributional
analysis of pricing policies.
       Recently, there has been a rebirth of this type of work motivated by the food-price
“crisis” associated with skyrocketing prices in global markets. Ivanic and Martin (2008)
   7
       Notice, however, that this analysis does not take into account the fiscal implications of eliminating the
tax.



                                                       10
cover ten case studies using household survey data from Bolivia (2005), Cambodia (2003),
Vietnam (1998 and 2004), Malawi (2004), Nicaragua (2001), Pakistan (1999), Peru (2003),
and Zambia (1998). The authors find an overall poverty-increasing impact of higher food
prices because, in the sample, most poor households are net consumers of food (both in
urban and rural areas).
   Wodon et al. (2008) apply this method to a dozen West and Central African countries,
Burkina Faso (2003), Ghana (2005-2006), Democratic Republic of Congo (2005), Gabon
(2005), Guinea (2002-2003), Liberia (2007), Mali (2006), Niger (2005), Nigeria (2003-2004),
Senegal (2006), Sierra Leona (2003), and Togo (2006). The authors focus on rice, wheat,
maize, other cereals, milk, sugar and vegetable oils and find that a 50 percent increase in
prices for selected food items would cause an average increase in the share of the population
in poverty of between 2.5 and 4.4 percentage points. The average impact would be between
3.7 and 5.2 percentage points in urban areas, and between 2.2 and 4.1 points in rural areas.
They also find significant heterogeneity across countries, as expected.
   While the net-consumer/net-producer result is intuitive, it rests on two critical
assumptions, namely that the first-order approximation is valid and that labor markets are
unaffected by the price change (Deaton, 1997). The first-order approximation is analytically
useful, but it is valid conceptually only for small price changes. We say “conceptually,”
because in principle budget and income shares could be used to measure the distributional
effects of both small and large price changes, but the approximation error will be larger
when price changes are large. In these cases, it is crucial to incorporate supply and demand
responses, and we discuss these below. The role of labor markets is discussed next.


3.3    Labor Markets

Price changes also affect wages. The mechanisms are simple. Consider, for instance, a small
open economy that takes commodity prices as given in international markets. Changes
in relative product prices cause some sectors to expand and some others to contract. As
a result, labor demand for different types of labor in different sectors can change, thus



                                             11
affecting equilibrium wages.8 In the literature, these responses are captured by “wage-price”
elasticities, which measure the proportional response of wages to price changes. These
elasticities may vary across members of a household when they are endowed with different
skills (unskilled, semi-skilled or skilled labor) or if they work in different sectors (industry
premia). Clearly, if countries differ in technologies, endowments, or labor regulations, the
responses of equilibrium wages to prices can be heterogeneous across countries as well.
       The response of wages can generate first-order effects on household welfare. To account for
these responses, the standard net-consumer/net-producer proposition needs to be modified.
To see this, consider the extreme case where a farm-household consumes a product but does
not produce it at all. Instead, the household earns income from selling labor in neighboring
farms. By omitting wages, this household would be mistakenly classified as a net consumer
and could thus be presumably hurt by a price increase. But if wages respond positively to
prices, the final welfare effect may not necessarily entail a loss.
       This type of wage responses were studied in Ravallion (1990), who explored the conditions
under which net-consumers of food products in Bangladesh lose or gain in the face of
increased food prices when rural wages adjust.               Ravallion estimates low elasticities of
agricultural wages to food prices and concludes that responses are unlikely to be strong
enough to offset the short-run adverse distributional effects of higher food prices. The
long-run estimates appear to be more favorable to the poor, but it would nevertheless take
around four years for any gains to materialize. Boyce and Ravallion (1991) take another look
at this issue using newer data for Bangladesh. They set up a dynamic econometric model
of agricultural wages and rice prices. They find that increases in rice prices relative to the
prices of manufactured goods have adverse effects on the real wages (measured per quantity
of rice) in both the short and long runs.
       In a study of informal export barriers and poverty in Moldova, Porto (2005) amends the
first order approximation used by Deaton (1989a) to allow wages to adjust. In this setup,
exporters produce agro-manufactures (i.e., wines or apple juice) for international markets.
They purchase primary products from the farmers (grapes, apples) and manufacture exports
   8
   Notice that we are referring here to prices of tradable goods. For non-traded goods, the issues are more
complex since they are endogenous.


                                                    12
using, among other inputs, labor. Farmers are endowed with land and labor and allocate
time (net of leisure) to the “farm” labor market, the home (land) plot to produce primary
inputs (grapes) for sale, or to work in nearby farms. In this framework, Porto derives
additional welfare impacts given by the direct impact on wages. Also, a given farm can
either sell or purchase labor in the “farm-labor” market, and, consequently the estimation
of the first-order welfare effects needs to include the additional impacts that arise due to
changes in wage earnings or paid wages. Porto (2005) finds that increases in the prices of
agro-manufactured exports such as wines (a major export sector in Moldova) have sizeable
poverty-reducing impacts. Wages respond positively to export prices and this causes first
order gains that dominate both the consumption losses due to higher consumer prices and
the profit losses due to higher wages paid to hired labor.
   Porto (2006) studies a case where households consume products, sell labor in
the market, but do not produce farm products.             His model applies well to urban,
middle-income economies, where farm activities are not relevant (and are not captured
by the available household surveys either). Porto explores the distributional consequences
of MERCOSUR—a regional trade agreement between Argentina, Brazil, Paraguay and
Uruguay—in Argentina and finds welfare gains for the average poor and middle-income
households (and negligible effects for the wealthiest households). This is because, on top of
gains from price reductions due to tariff cuts, there are changes in wages that favor unskilled
workers over skilled workers, and unskilled workers are concentrated at the bottom of the
income distribution.
   Three recent studies are also worth mentioning. Ferreira, Fruttero, Leite and Lucchetti
(2013) examine the consequences of food price inflation in Brazil, a large food producer
with a predominantly wage-earning agricultural labor force. They find large and negative
consumption effects, which are also markedly regressive, as expected. However, they also
uncover a positive and progressive income effect, particularly in rural areas. Overall, thus,
the Brazilian middle-income household suffered larger proportional losses than the very poor
or the richer households. Jacoby (2013) reaches a similar conclusion in his study of food prices
in India. Specifically, once the wage gains are accounted for, he finds that rural households


                                              13
across the income spectrum actually benefit from higher agricultural commodity prices.
Finally, Nicita, Olarreaga, and Porto (2013) examine the presence of a pro-poor bias in
                                                                                      ote
trade protection of six Sub-Saharan African (SSA) countries, Burkina Faso, Cameroon, Cˆ
d’Ivoire, Ethiopia, Gambia, and Madagascar. Their model includes consumption effects,
agricultural income effects, and labor income effects. They find that trade policies tend to
be biased in favor of poor households because protection increases the agricultural prices of
goods that are sold by African households, and this effect dominates the impacts on both
expenditures and wages.


3.4    Spillovers

Broadly speaking, spillovers are the impacts of a change in market i on the activity in
market j . There are two types of spillovers. Production linkages occur when the expansion
of a sector affects upstream activities (backward linkages) or downstream activities (forward
linkages). Expenditure linkages occur when the increase in income due to the expansion
of a sector raises the demand for outputs and thus the derived demand for inputs in other
sectors. Given our focus on commodities, we emphasize here spillovers in agriculture. The
large literature on the topic uncovers two main regularities. First, the linkages are strong
in rural areas because households earn a large share of their income from and spend a large
share of their total expenditures in goods produced locally. Second, both production and
expenditures spillovers are important (Mellor and Johnston, 1984; Delgado, 1996; Harriss,
1987). Evidence on spillovers in agriculture in Latin America can be found in the regional
study by De Ferranti et al. (2005). They report that while the share of the primary sector
in GDP is 12 percent, there are massive spillovers to the rest of the economy. In fact, the
multiplier factor from the primary sector to national GDP is 0.22 so that, for instance, a
one percent growth of the rural natural resource sector is associated with an increase of 0.22
percent increase in national GDP (much higher than the 0.12 percent increase to be expected
from the sector’s share of GDP). The multiplier for the income of the poorest households is
even higher at 0.28.
   Note that spillovers can be interpreted as general equilibrium effects, but they are not


                                             14
necessarily the same. If there is some sort of labor immobility in general equilibrium, for
instance, then a change in the price of good i may have no effect on the wages paid in
sector j . In such a scenario, there are no spillovers even in a general equilibrium setting.
This distinction is useful to organize our discussion. Of the papers reviewed above, Ravallion
(1990) assumes no spillovers or general equilibrium effects from rice prices to non-agricultural
wages. The author estimates a partial equilibrium model. An alternative interpretation is
that labor markets are segmented. In contrast, Porto (2006) and Nicita (2009) take a general
equilibrium approach and assume that a change in a given price pi affects economy-wide
wages. This assumes that labor markets are integrated and workers are freely mobile across
sectors; an increase in wages in one sector spills over to other sectors.
   Porto (2008) describes a variant of the spillover mechanism. Markets can be segmented so
that wages can differ across sectors. However, sectors are related via forward and backward
linkages, so that an expansion of one sector may have implications for other sectors. In
such a setting, changes in corn prices in Mexico can affect the corn sector directly as well
as various rural services associated with corn production. For instance, higher corn prices
may require more labor in corn planting, weeding, and harvesting (thus causing wages in the
corn sector to increase) and may also require farm services (road graveling, farm-gate and
fence repairing, etc.). Porto finds that allowing for those spillovers can make a big difference
is assessing the welfare impacts of higher corn prices. In particular, the average Mexican
low-income household tends to be a net-consumer of corn and consequently is made worse off
by an increase in corn prices. However, once spillovers from the corn sector to agricultural
wages are allowed to take place, even low-income net consumers stand to gain from higher
corn prices.
   Clearly, the type of assumptions used to model labor markets is crucial to determine
the impacts of a so-called commodity-price crisis. While in theory assumptions may help
illustrate a result, in practice it is important to work with realistic assumptions. In the end,
the existence or lack of spillovers, and the nature of those spillovers, will be specific to the
case under study, and it is a priori difficult to make generalizations.
   Different types of spillovers arise when other product markets, rather than labor markets,


                                              15
are affected. These spillovers are likely to take place in non-traded goods. As we have shown
above, changes in commodity prices affect factor prices, including wages. If the wages earned
in non-traded sectors are affected, then the cost of producing these goods will change, and
this in turn will affect the equilibrium prices of these goods. As a result, there are additional
welfare impacts on the consumption side. Notice that these are first order impacts. Porto
(2006) provides an example of how to estimate these impacts for the case of the countries
that are members of the MERCOSUR customs union. He uses time series of prices to
recover the elasticity of the price of non-traded goods with respect to the prices of traded
products; import tariff cuts by MERCOSUR appear to cause the prices of non-traded goods
to decline and households to benefit. It is noteworthy that these spillovers are similar to
those highlighted by, for instance, Burstein, Neves, and Rebelo (2003) for the computation
of the real exchange rate and the adjustment of prices to stabilization programs.


3.5    Price Pass-Through

So far, the review has focused on a generic analysis of price changes. Often, these price
changes, especially if they refer to commodities, are generated by external shocks. It is thus
important to review the literature on the pass-through of international prices to the domestic
economy and, furthermore, to the household. There are various literatures that are relevant
here: international economics, international trade, and economic development. While this
collection of theories and empirical analyses play different roles in our framework, we can
learn from all three of them.
   Standard models of international trade and international economics assume competitive
markets (and homogeneous goods) and friction-less trade. In this scenario, markets are
integrated and the law of one price holds: domestic prices are equal to the international
prices converted to the local currency. Any difference between these prices is due to transport
and distribution costs as well as to trade policy. In this family of models, a proportional
change in the exchange rate, in the international price, or in the tariff is fully transmitted
to domestic prices. There is, however, strong evidence against this prediction. The papers
reviewed by Rogoff (1996) consistently reject the law of one price for a variety of products


                                              16
and countries.
       The second type of impact arises when distribution costs change. Distribution activities
are mostly non-tradable services, which require factors of production such as labor and land.
If factor prices react to changes in international prices or exchange rates, then distribution
costs will themselves change, and this increase can feed into a larger-than-proportional
increase in the domestic price of imported goods.9 There is evidence that this mechanism can
be relevant in practice. In Argentina, for instance, Burstein, Neves and Rebelo (2003) show
that the distribution margin in wholesale and retail is around 60 percent. Using calibration
methods to compute the impact of a stabilization program (i.e., they study a decline in the
exchange rate), the authors find that the increase in the relative prices of non-tradable goods
can be 42 percent higher in a model with distribution costs than in a model without such
costs.10
       Another strand of literature is the work on exchange-rate pass-through (ERPT), which
is defined as the percentage change in import prices resulting from a 1 percent change in the
exchange rate. ERPT models are often associated with imperfect competition in product
markets (but they are consistent with perfect competition as well). Goldberg and Knetter
(1997) identify two conditions for full pass-through: i) constant mark-ups and ii) constant
marginal costs. These conditions are unlikely to hold in a given market and, as a result, the
literature has found strong evidence of imperfect pass-through. In their review, Goldberg
and Knetter (1997) conclude that we should expect a pass-through rate of around 60 percent.
Because this literature includes measures of costs in the derivation of the pass-through rate,
the results imply that around 40 percent of the change in the exchange rate is actually
offset by changes in markups. In a comprehensive analysis of pass-through rates, Campa
and Goldberg (2005) find large differences across developed countries. The estimated rate
for the U.S. is 42 percent, it is 98 percent in France, 80 percent in Germany, and 46 percent
in the United Kingdom.
   9
    Note that this is a type of spillover as those discussed above.
  10
    With perfect pass-through, Burstein et al. (2003) assume that the prices of tradable goods decrease in
the same proportion as the decrease in the exchange rate. When non-traded prices adjust, the calibration
results predict an increase in the relative price of non-tradable goods of 15.4 percent (without distribution
costs) and of 25.7 percent (with distribution costs).



                                                     17
      An interesting paper that attempts to incorporate the estimation of tariff-rate
pass-through into the trade and poverty literature is Nicita (2009). Nicita argues that while
imperfect price transmission has been established in the literature (as our review suggests),
most of the work on the impacts of trade on poverty, and more generally on the distributional
impacts of pricing policies, implicitly assume full pass-through. He attempts to fill this gap
by estimating tariff-rate pass-through for Mexico. He finds lower pass-through rates for
both manufacturing (27 percent) and agriculture (33 percent) and supports these findings
by arguing that pass-through is likely to be smaller in a developing country like Mexico due
to lack of domestic transport infrastructure.
      Another important strand of the literature is the work on spatial market integration and
the efficiency of arbitrage. In the presence of transport costs, prices across regions differ in
the size of the transport costs but should move together due to arbitrage. In practice, three
scenarios can be observed: efficiency, characterized by price differentials that can be perfectly
accounted for by transport costs; segmented efficiency, characterized by price differentials
that are smaller than transport costs (so that markets are competitive but segmented and
high transfer costs make trade and arbitrage not profitable); and inefficiency, characterized
by price differentials that are larger than transfer costs and thus imply unexploited arbitrage
profits (due to market power, for instance).
      There are various procedures to test for spatial market efficiency.11 The state-of-the-art
method is the parity bound model of Sexton, Kling and Karman (1991), Baulch (1997), and
Park et al. (2002), among others. The method relies on maximum likelihood methods to
estimate a switching regression model for the three states, efficiency, segmented efficiency
and inefficiency. As expected, the evidence is mixed. Baulch (1997) finds that rice markets
in the Philippines are spatially efficient and that differences in prices levels are due to
transaction costs. Instead, Cirera and Arndt (2006) report inefficiencies in maize markets
in Mozambique. Notably, these inefficiencies are somewhat reduced by market liberalization
(but not by road rehabilitation) but tend to persist over time.
 11
      See, for instance, Fackler and Goodwin (2002).




                                                       18
4     Household             Adjustment               in      Consumption                and
      Production
The discussion has followed the traditional first-order approach to assessing the distributional
effects of commodity price changes.        A major concern with this approach is that it
does not allow households to respond to the changes in the economic environment. In
fact, even when wage adjustments are incorporated, the standard approach takes the
net-consumer/net-producer position of the household as “exogenous.” This is just a datum
in the analysis that we can assess using survey data, as we did when we described the
dependency of households in Latin America and Africa in section 2. The net-position of the
household is, however, endogenous when households respond.
    It is often argued that the first-order approximation can be accurate for small price
changes. The “endogeneity” of a household choice is in these cases, a second order concern
in the sense of being unimportant. This is true, and there are numerous examples where the
framework works very well. The evaluation of the Doha Development Agenda is one instance,
see Hertel and Winters (2006) or Hoekman and Olarreaga (2007). For larger price changes,
as those that can be expected from a persistent food-price crisis (or a collapse of global
commodity prices), the approximation error grows larger. This section thus explores how
large this error can actually become and discusses how the approximation can be improved
by measuring household responses in consumption and production.
    A consumer will always lose from an increase in prices, but the losses can only be
ameliorated by reducing purchases of the more expensive goods. A producer will instead
always lose from a price decline. In this case, adjustments can mitigate the losses by shifting
resources to more productive activities. Similarly, a farm that hires labor will lose from
higher wages, even if labor is substituted with other factors of production. Finally, a worker
will lose from a reduction in wages, even if he decreases labor supply to enjoy more leisure.
    For a farm-household, or more generally in a model where the household participates in
multiple markets, adjustments can play a bigger role. In fact, Porto (2008) shows that it is
possible for a household to change sides of the market (that is, to become a net producer


                                              19
after a sufficiently high price increase) so that even an ex-ante net consumer can benefit
from a price increase. In what follows, we illustrate various sources of adjustment and
discuss attempts to quantify them.


4.1    Consumption Responses

We begin with consumption because research in this area has been prolific. To incorporate
consumption responses, research needs to estimate a system of demand elasticities (own-
and cross-price elasticities) to identify the pattern of substitution in consumption. The
estimation of such a demand system is difficult, however.
   There is a long tradition in economics in estimating systems of demand equations. The
type of estimation method depends critically on the available data. In the early days,
when only national account data were available, estimation relied on time series of various
consumption aggregates and price indices. The search for a utility-consistent empirical model
of demand converged to the Almost Ideal Demand System (AIDS) model of Deaton and
Muellbauer (1980), which is now the leading framework to study demand elasticities.
   The mass-production of household surveys in more recent years—a process in which
the World Bank has been instrumental— has broadened the options for demand analysis.
However, while expenditure surveys provide detailed accounts of what people consume, they
collect little information on prices. In some cases, community questionnaires collect price
data at the village level, and this variation can be exploited to identify demand elasticities.
For those cases in which the surveys collect information on quantities purchased as well
as on expenditures, demand elasticities can be recovered from unit values—the ratio of
expenditures over quantity. Deaton (1987, 1988, 1990) developed a model of demand for
quantity and quality that allows for the extraction of the price signal embedded in the
unit values at the level of a market (or cluster of households residing near each other and
thus shopping for consumption goods in the same markets) and these price signals identify
the demand elasticities. Because the estimating equations are not necessarily derived from a
utility framework (that is, the equations do not form a full-blown AIDS model), the estimated
parameters cannot be used to recover the parameters of the utility function. However, they


                                              20
do reveal how household consumption responds to price changes.
       Deaton’s unit values procedure has been used extensively to look at issues of tax reforms
(see, for instance, Deaton (1989b), Deaton and Grimard (1992), and Deaton (1997) for
case studies of Thailand, Pakistan, India, and Cote d’Ivoire).12              A nice application of
Deaton’s methods to welfare analysis is the research conducted by Friedman and Levinsohn
(2002), who studied consumption responses during the Indonesian financial crisis of 1997.
They compared the welfare effects resulting from a first order approximation (that is,
without substitution effects) with the ones resulting from a second order approximation
(thus including substitution effects). However, they did not take into account wages or,
more generally, income responses. Two major findings are worth emphasizing in the context
of the discussion about the magnitude of consumption responses to market signals. On the
one hand, the losses estimated as first order approximations for the urban poor were high,
roughly about 50 percent of the pre-crisis level of per capita expenditures. In rural areas, the
losses of the poor were equivalent to nearly 40 percent of their pre-crisis real consumption.
On the other hand, allowing for substitution effects in consumption made a big difference.
The estimated losses were cut by half for the urban poor and by more than 40 percent.
These results imply that first order approximations can be quite large when price shocks are
large, as those observed during the Indonesian crisis (when prices increased by 100 percent
or more). In cases of economic crisis, when large price changes may be the norm rather
than the exception, the results of Friedman and Levinsohn (2002) warn us about the perils
of ignoring second order terms in welfare analysis. In light of the stylized facts presented in
section 3, these adjustments need to be taken into account in the context of the recent surge
of commodity prices.


4.2       Farm Supply Responses

We turn now to farm supply responses. As throughout all of our analysis, we are interested
in responses at the household level and especially for the poorest households.13                  Since
  12
     In short, the difference is that compensated demand elasticities are needed for welfare analysis and
uncompensated elasticities are needed to analyze tax reforms.
  13
     We do not cover aggregate supply responses in our review.


                                                  21
these are households typically involved in agriculture, much of our review is concerned with
agricultural farm responses.
   Conceptually, the issues are those that we discussed within consumption responses. Faced
with changed relative prices, the question is whether there is any adjustment in production
and what factors prevent or facilitate that adjustment. It is perhaps not really surprising to
learn that supply responses are somewhat feeble in developing countries. The literature is
too large for a comprehensive review, but some of most fascinating recent examples include
Cadot, Dutoit and de Melo (2009) on vanilla-market reforms in Madagascar and McMillan,
Welch, and Rodrik (2003) on cashews in Mozambique.
   The evidence on supply responses is not always so bleak, however, as two strands of
literature illustrate. First, a common finding in empirical work is that sizeable responses
can be expected when price changes are accompanied by appropriate complementary factors
such as access to inputs, information and credit (McKay, Morrissey, and Vaillant, 1997),
productive assets like animals and tools (Deininger and Olinto, 2000), productive capital,
                              opez, Nash, and Stanton, 1995; Heltberg and Tarp, 2002).
education, and land quality (L´
Second, there is a growing literature showing responses to marketing reforms under successful
contract farming arrangements. Two examples will make the point: Brambilla and Porto
(2011) estimate large adjustments in cotton production after changes in the outgrower
contract environment in Zambia; Maertens and Swinnen (2009) estimate similarly large
responses in the vegetable sector in Senegal.
   Another way to explore these issues is to look at the performance of subsistence farmers
relative to commercial farmers. Subsistence farmers produce for home-consumption, whereas
commercial farmers earn income by selling their produce in markets. The evidence indicates
that commercial farming is associated with higher consumption and incomes than subsistence
farming. Cadot, Dutoit and Olarreaga (2006) estimate an average loss in income of those in
subsistence of 43 percent relative to commercial farmers that produce the same crop. Even
with much smaller margins, what is puzzling is why subsistence farmers do not transform
themselves into commercial farmers by selling their harvest in markets? This question is
essentially about supply responses to market signals.


                                                22
    Cadot, Dutoit and Olarreaga (2010) argue that there are barriers to exit from subsistence
agriculture, but the issue highlights explanations of why supply responses to market signals
might be sluggish. There are two major barriers. One is risk: cash crops are riskier than
food crops so that part of the difference in return is actually a risk premium (Fafchamps,
1992; Rosenzweig and Binswanger, 1993). The other barrier is related to missing markets:
the differential rate of return is due in part to the shadow price of a good with missing
markets such as for labor or food crops (de Janvry, Fafchamps, and Sadoulet, 1991). Notice
that missing markets can arise due to variable transaction and transportation costs, fixed
transaction costs (searching for partners, enforcing contracts with distant buyers, establishing
quality), or sunk costs of transacting. One example is Key, de Janvry and Sadoulet (2000),
who illustrate how supply responses are affected by transaction costs. Another interesting
example is Cadot, Dutoit and Olarreaga (2006), who show that sunk costs can be particularly
large, ranging from 124 to 153 percent of the market value of the average annual farm output
in Madagascar.
    We end this section with a caveat. Positive and large supply responses to price changes
are always an indication that economic incentives work. However, in the context of our
analytical framework of section 3, these supply responses comprise second order effects. In
consequence, they are likely to generate only small welfare impacts because of the opportunity
cost of the resources used in the production of different goods. While this argument does
not invalidate the search for supply responses in the data, it strongly argues for a careful
interpretation of the findings.



5     Quantifying The Mechanisms: The Case of Mexico
In this section, we use data from Mexico to quantify the different channels uncovered by our
review of the literature. The main objective is to illustrate how the mechanisms work. Our
results are reported in Table 6.
    We begin with the net consumer-net producer proposition. In the first panel of Table
6, we compute the first order welfare impacts of a (hypothetical) increase of 20 percent in


                                              23
corn prices in rural Mexico (Porto, 2008). Net producers would enjoy gains equivalent to
1.78 percent of their initial (i.e., pre price increase) average expenditure. Net consumers
would instead suffer a loss of 1.93 percent. For the whole sample, the average impact of
an increase in corn prices is negative, but small (equivalent to only 0.39 percent of average
national income).
   To incorporate labor market responses, recall that the impacts depend on how wages
respond to price changes, and this in turn depends on whether we think that labor markets
are integrated or segmented and on whether there are spillovers and backward or forward
linkages (Ravallion, 1990; Porto, 2005; Porto, 2006). Continuing with the case of corn in
Mexico, we give two examples in Table 6. In the second panel, we assume that labor markets
are “segmented” and thus we only allow wages in the agricultural sector to respond to corn
prices. We use a wage-price elasticity of 0.40 (as estimated in Porto, 2008). The income
gains of net producers jump to 7.22 percent and their net gain is now 6.23 percent—nearly
3.5 times higher than before. For net consumers, the income gains are more modest, of
around 1.99 percent, and these gains are not enough to offset the consumption losses. In the
end, even with wage responses, net consumers would be made worse off by the rise in corn
prices. The national average effect would, however, be positive: higher corn prices would
bring welfare increases of 2.17 percent, on average.
   In another example, in the third panel of Table 6, we allow for spillovers from corn
prices to the wages of self-employed individuals (in rural areas). The idea is that increases
in agricultural prices may raise the derived demand for labor in services, odd-jobs and,
more generally, in local rural labor markets (not exclusively in agriculture). Using the same
wage-price elasticity as before (0.40), we estimate the following welfare impacts: i) the income
gain of net producers would be equivalent to 10.22 percent of their initial income, and the
net gain would be 9.78 percent; ii) net consumers would also gain, with an income gain of
3.44 percent and a net gain of 0.44 percent; iii) the average national gain would be equivalent
to 4.32 percent on initial income.
   As argued above, the net position of the household can be endogenous and it can thus
change in response to a price shock. Consider, for example, an increase in the price of


                                              24
corn and let households be able to adjust. If consumption responses are sufficiently strong
(because households can substitute easily to other food products) and/or if supply responses
are large enough, then a net-consumer can in principle become a net-producer and thus
benefit from the increase in corn prices. The standard argument against this caveat is
that second-order adjustments of the type described here are often small and thus unlikely
to change the implications of the first-order calculations. An example is shown in the
fourth panel of Table 6, where we estimate consumption substitutions and supply responses.
Allowing for consumption and production second order effects does not, in fact, affect the
results. The gains for net producers are slightly larger and the losses for net consumers are
slightly smaller, but the welfare impacts are not affected much.



6     Conclusions
This paper provided an overview of the impacts of price changes on household welfare. We
began by exploring commodity dependence in the data and by looking at expenditure and
income dependence in the household surveys from Latin American and Africa. Our analysis
suggests a strong dependence on commodity prices, a dependence that is biased towards the
poor, especially on the consumption side.
    The paper then tackled the conceptual issues related to commodity dependence. It did
so by laying out a theoretical framework, which described how changes in commodity prices
affects consumption, production, wages, and household adjustments. Each of these aspects
of the analysis of commodity dependence is illustrated with evidence from existing literature.
    Our review tells the story of heterogeneous responses to commodity price changes. Most
shocks to commodity prices are international in nature and the pass-through to domestic
prices is imperfect (that is, changes in international prices are not translated one to one
to domestic prices). The pass-through is also heterogeneous across countries and depends
on infrastructure, institutions and market structure (the nature of imperfect competition in
domestic markets).
    Faced with a price shock, households are affected on both the consumption and income


                                             25
sides. On the consumption side, higher prices cause welfare losses. In Latin America and
particularly in Africa, higher food prices can cause large welfare losses, especially for the
poor. On the income side, rising commodity prices can bring gains or losses, depending
on which goods are produced and also on how wages respond to those price changes.
The evidence shows that households tend to earn a significant share of their total income
from commodities so that the consumption losses can be ameliorated, sometimes to a large
extent. This is most likely to occur when wages and spillovers from commodities to other
sectors are taken into account, in particular if wages respond positively to increases in labor
demand in primary sectors and if the spillovers to non-primary sectors are positive (via either
production or expenditure linkages). In the end, the overall welfare impacts depend on the
net consumer-net producer position of the household, after taking into account wage income.
Net consumers will be hurt by price increases, and net producers will benefit. As expected,
these impacts are heterogeneous across countries. Some countries will gain and some others
will lose. Also, in a country that gains from higher prices, there will be households that gain
and households that lose out. Conversely, there will be winners and losers within countries
that lose as an aggregate. If anything, the evidence seems to suggest a larger probability of
welfare losses for poor households.
   Household adjustments can work as a cushion to protect against price changes. Given
a price increase in commodities, consumption responses can ameliorate the consumption
losses and production responses can boost the income gains. Typical adjustments under the
separability assumption appear not to be enough to overturn the initial impacts derived from
a static view of commodity dependence. Non-separability, which is likely to be a real feature
in many developing countries, can work in both directions. Sometimes, market failures
can shut down markets completely so that price changes can become irrelevant for many
households. While this can work as a protective device, it can also hinder the realization of
any efficiency gains and of expanded market opportunities. Sufficiently large price changes
in the presence of market failures can generate discontinuities in household responses that
can potentially multiply the typically small responses found in the empirical literature – in
the real world, the price is not always right.


                                                 26
   We end with some caveats. First, due to space limitations, our review has focused on
welfare impacts and we have neglected issues of insurance, coping mechanisms, and social
protection policies. Second, the review centered the discussion on one-shot price shocks. In
particular, we have not covered the literature on the effects of commodity price volatility.
In addition, even in the case of one-shot price shocks, we should note that there might be
hysteresis in the effects. In other words, a price shock may have long term implications,
even if the price change reverts at a later date (Baird, Friedman and Schady, 2011). These
themes require careful consideration in future assessments of the developmental implications
of price signals.



References
Baird S., J. Friedman and N. Schady (2011). “Aggregate Income Shocks and Infant Mortality
in the Developing World,” Review of Economics and Statistics, 93 (3), pp. 847–856.

Barnum, H. and L. Squire (1979). “A model of an Agricultural Household. Theory and
Evidence,” World Bank Ocassional Papers No 27.

Barrett, C. and P. A. Dorosh (1996). “Farmers’ Welfare and Changing Food Prices:
Non- Parametric Evidence from Rice in Madagascar,” American Journal of Agricultural
Economics, vol. 78, pp. 656–69.

Baulch, B. (1997). “Transfer Costs, Spatial Arbitrage, and Testing for Food Market
Integration.” American Journal of Agricultural Economics, 79(2): 477–487.

Benjamin, D. (1992). “Household Composition, Labor Markets, and Labor Demand: Testing
for Separation in Agricultural Household Models,” Econometrica, vol. 60, No.2, pp. 287–322.

Benjamin, D. and A. Deaton (1993). “Household Welfare and the Pricing of Cocoa and Coffee
in Cte d’Ivoire: Lessons from the Living Standards Surveys,” The World Bank Economic
Review, 7, pp. 293–318.




                                            27
Boyce, J. and M. Ravallion (1991). “A Dynamic Econometric Model of Agricultural Wage
Determination in Bangladesh,” Oxford Bulletin of Economics and Statistics, vol. 53(4), pp.
361–76.

Brambilla, I. and G. Porto (2011). “Market Structure, Outgrower Contracts and Farm
Output. Evidence from Cotton Reforms in Zambia,” Oxford Economic Papers 63(4), pp.
740–766.

Budd, J. W. (1993). “Changing Food Prices and Rural Welfare:             a Nonparametric
Examination of the Cte d’Ivoire,” Economic Development and Cultural Change, 41, pp.
587–603.

Burstein, A., J. Neves, and S. Rebelo (2003). “Distribution Costs and Real Exchange Rate
Dynamics During Exchange-rate-based Stabilizations,” Journal of Monetary Economics, 50,
pp. 1189–1214.

Cadot, O., L. Dutoit and M. Olarreaga (2006). “The Cost of Moving Out of Subsistence,”
mimeo University of Lausanne.

Cadot, O. L. Dutoit and M. Olarreaga (2010). “Barriers to Exit from Subsistence
Agriculture,” in G. Porto and B. Hoekman (Eds.) Trade Adjustment Costs in Developing
Countries:      Impacts, Determinants and Policy Responses, CEPR and World Bank,
Washington DC.

Cadot, O. , L. Dutoit, and J. de Melo (2009). “The Elimination of Madagascars Vanilla
Marketing Board, 10 Years On,” Journal of African Economics, Vol. 18, Issue 3, pp. 388–430.

Campa, J.M. and L. S. Goldberg (2005). “Exchange Rate Pass-through into Import Prices,”
The Review of Economics and Statistics, 87(4), pp. 679–690.

Cirera, X. and C. Arndt (2008). “Measuring the Impact of Road Rehabilitation on Spatial
Market Efficiency in Maize Markets in Mozambique,” Agricultural Economics, Vol. 39 Issue
1, pp. 17–28.



                                            28
Dang, H., P. Lanjouw, J. Luoto and D. McKenzie (2011). “Using Repeated Cross-sections
to Explore Movements In and Out of Poverty,” Policy Research Working Paper Series 5550,
The World Bank.

De Ferranti, D., G. Perry, D. Lederman, A. Valdes, and W. Foster (2005). Beyond the City:
The Rural Contribution to Development in Latin America and the Caribbean, World Bank,
Washington DC.

De Janvry, Alain, M. Fafchamps, and E. Sadoulet (1991). “Peasant Household Behaviour
with Missing Markets: Some Paradoxes Explained,” Economic Journal 101, pp. 1400–1417.

Deaton, A. (1987). “Estimation of Own- and Cross-Price Elasticities From Household Survey
Data,” Journal of Econometrics, vol. 36, pp. 7–30.

Deaton, A. (1988). “Quality, Quantity, and Spatial Variation of Price,” American Economic
Review, vol. 78 No 3, pp. 418–430.

Deaton, A. (1989a). “Rice Prices and income Distribution in Thailand: A Non-Parametric
Analysis,” Economic Journal, vol. 99 No 395, pp. 1–37.

Deaton, A. (1989b). “Household Survey Data and Pricing Policies in Developing Countries,”
The World Bank Economic Review, 3, pp. 183–210.

Deaton, A. (1990). “Price Elasticities from Survey Data,” Journal of Econometrics, vol. 44,
pp. 281–309.

Deaton, A. (1997). The Analysis of Household Surveys. A Microeconometric Approach to
Development Policy, John Hopkins University Press for the World Bank.

Deaton, A. and J. Muellbauer (1980). “An Almost Ideal Demand System,” American
Economic Review, vol 70, pp. 312–336.

Deaton, A., and F. Grimard (1992). “Demand Analysis and Tax Reform in Pakistan,” Living
Standards Measurement Study, Working Paper No 85, The World Bank.



                                            29
Deininger, K. and P. Olinto (2000). “Why Liberalization Alone Has Not Improved
Agricultural Productivity in Zambia: The Role of Asset Ownership and Working Capital
Constraints.” World Bank Working Paper No 2302.

Delgado, C. (1996). “Agricultural Transformation: The Key to Broad-Based Growth and
Poverty Alleviation in Africa,” in B. Ndulu and N. van de Walle (eds.), Agenda for Africa’s
Economic Renewal, New Brunswick, NJ: Transaction Publishers.

Dixit, A. and V. Norman (1980). Theory of International Trade. A Dual, General Equilibrium
Approach, Cambridge: Cambridge Economic Handbooks, 1980.

Fackler, P.L., and B.K. Goodwin (2002). “Spatial Price Analysis,” in B.L Gardner and G.C.
Rausser (eds.) Handbook of Agricultural Economics, Amsterdam: Elsevier Science.

Fafchamps, M. (1992). “Cash Crop Production, Food Price Volatility, and Rural Market
Integration in the Third World,” American Journal of Agricultural Economics 74(1) pp.
90–99.

Ferreira, F., A. Fruttero, P. Leite, and L. Lucchetti (2013). “Rising Food Prices and
Household Welfare: Evidence from Brazil in 2008,” Journal of Agricultural Economics 64(1)
pp. 151-176.

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, pp. 397–423.

Goldberg, P. and M. Knetter (1997). “Goods Prices and Exchange Rates: What Have We
Learned?,” Journal of Economic Literature, vol. XXXV, pp. 1243–1272.

Goldberg, P. and N. Pavcnik (2004). “Trade, Inequality, and Poverty: What Do We Know?
Evidence from Recent Trade Liberalization Episodes in Developing Countries,” Brookings
Trade Forum 2004, pp. 223–269.

Goldberg, P. and N. Pavcnik (2007). “Distributional Effects of Globalization in Developing
Countries,” Journal of Economic Literature March, 45(1), pp. 39–82.

                                            30
Harrison, A., J. McLaren and M. McMillan (2011). “Recent Perspectives on Trade and
Inequality,” Policy Research Working Paper 5754, The World Bank.

Harriss, B. (1987). “Regional Growth Linkages from Agriculture,” Journal of Development
Studies, 23:2, pp. 275–89.

Heltberg, R. and F. Tarp (2002). “Agriculture Supply Response and Poverty in
Mozambique,” Food Policy 27:2, pp. 103–24.

Hertel, T. and A.L. Winters (Eds.) (2006). Poverty Impacts of the Doha Development
Agenda, Palgrave-McMillan for the World Bank, Washington DC.

Hoekman, B. and M. Olarreaga (Eds.) (2007). Global Trade and Poor Nations: Poverty
Impacts and Policy Implications of Liberalization, Brookings Institution Press.

Ivanic, M. and W. Martin (2008). “Implications of Higher Global Food Prices for Poverty
in Low-Income Countries,” World Bank Policy Research Working Paper Series No 4594,
Washington DC.

Jacoby, H. (2013). “Food Prices, Wages, and Welfare in Rural India,” World Bank Policy
Research Working Paper No 6412.

Key, N., E. Sadoulet and A. de Janvry (2000). “Transactions Costs and Agricultural
Household Supply Response,” American Journal of Agricultural Economics, 82, pp. 245–259.

Lederman, D. (2013). “International Trade and Inclusive Growth: A Primer,” Indian Growth
and Development Review, Vol. 6 No. 1, pp. 88–112.

 opez, R., J. Nash and J. Stanton (1995). “Adjustment and Poverty in Mexican Agriculture:
L´
How Farmers’ Wealth Affects Supply Response,” World Bank Policy Research Working Paper
No 1494.

Maertens, M. and J. Swinnen (2009). “Trade, Standards, and Poverty: Evidence from
Senegal,” World Development Vol. 37, No. 1, pp. 161–178.



                                            31
McKay, A., O. Morrissey and C. Vaillant (1997). “Trade Liberalization and Agricultural
Supply Response: Issues and Some Lessons,” European Journal of Development Research,
9:2, pp. 129–47.

McMillan, M., K. Welch, and D. Rodrik (2003). “When Economic Reform Goes Wrong:
Cashews in Mozambique,” Brookings Trade Forum, pp. 97–151.

Mellor, J. and B. Johnston (1984). “The World Food Equation: Interrelations among
Development, Employment and Food Consumption,” Journal of Economic Literature, 22:2,
pp. 524–31.

Nicita, A. (2009). “The Price Effect of Trade Liberalization: Measuring the Impacts on
Household welfare,” Journal of Development Economics, vol. 89(1), pp. 19–27.

Nicita, A., M. Olarrega, and G. Porto (2013). “Pro-Poor Trade Policy in Sub-Saharan
Africa,” mimeo Universidad de La Plata and University of Geneva.

Park, A. H. Jin, S. Rozelle, and J. Huang (2002). “Market Emergence and Transition
Arbitrage, Transaction Costs and Autarky in Chinas Grain Markets,” American Journal
of Agricultural Economics 84(1), pp. 67–82.

Porto, G. (2005). “Informal Export Barriers and Poverty,” Journal of International
Economics, 66, pp. 447–470.

Porto, G. (2006). “Using Survey Data to Assess the Distributional Effects of Trade Policy,”
Journal of International Economics 70, pp. 140–160.

Porto, G. (2008). “Estimating Household Responses to Trade Reforms. Net Consumers and
Net Producers in Rural Mexico,” Policy Research Working Paper No 3695, World Bank,
Washington

Ravallion, M. (1990). “Rural Welfare Effects of Food Price Changes Under Induced Wage
Responses: Theory and Evidence for Bangladesh,” Oxford Economic Papers, vol. 42, pp.
574–585.


                                              32
Rogoff, K. (1996). “The Purchasing Power Par- ity Puzzle?,” Journal of Economic Literature,
34(2), pp. 647–68.

Rosenzweig, M. and H. Binswanger (1993). “Wealth, Weather Risk and the Profitability of
Agricultural Investment,” Economic Journal 103, pp. 56–78.

Sahn, D. and A. Sarris (1991) “Structural Adjustment and the Welfare of Rural Smallholders:
A Comparative Analysis from Sub-Saharan Africa,” World Bank Econonomic Review, vol.
5, pp. 259–89.

Sexton, R., C. Kling and H. Carman (1991). “Market Integration, Efficiency of Arbitrage and
Imperfect Competition: Methodology and an Application to US Celery,” American Journal
of Agricultural Economics 7: 568–580.

Singh, I., L. Squire, and J. Strauss, eds. (1986). Agricultural Household Models: Extensions,
Applications and Policy. Baltimore, Johns Hopkins Press for the World Bank.

Winters, A., N. McCulloch, and A. McKay (2004). “Trade Liberalization and Poverty: The
Evidence So Far,” Journal of Economic Literature, XLII, pp. 72–115.

Wodon, Q., C. Tsimpo, P. Backiny-Yetna, G. Joseph, F. Adoho, and H. Coulombe (2008).
“Potential Impact of Higher Food Prices on Poverty: Summary Estimates for a Dozen West
and Central African Countries,” World Bank Policy Research Working Paper Series No 4745,
Washington DC.




                                             33
                      Table 1
            Commodity Expenditure Shares
                  Latin America


                 Quintiles p/c expenditures            Total
                (1)  (2)     (3)   (4)   (5)
Argentina       61.8   48.9   42.3    37.5   29.0      43.9
Bolivia         82.9   78.7   73.0    67.6   54.4      71.3
Colombia        80.1   69.1   61.7    55.6   41.6      61.6
El Salvador     65.2   59.6   54.6    49.4   39.0      53.5
Mexico          40.6   36.2   32.4    28.4   20.8      31.7
Nicaragua**     63.5   60.9   57.3    54.4   41.6      55.5
Panama**        60.0   49.0   43.9    37.4   28.3      43.7
Peru            60.1   50.6   42.6    35.8   26.1      43.0
 Note: Own elaboration based on expenditure surveys.



                      Table 2
            Commodity Expenditures Shares
                       Africa


                  Quintiles p/c expenditures           Total
                 (1)  (2)     (3)   (4)   (5)
Burundi         56.3   60.7    63.1   61.9   53.2       59.1
Benin           65.5   63.8    62.9   59.8   56.5       61.7
Burkina Faso    77.1   72.5    67.4   60.9   48.0       65.2
 ote d’Ivoire
Cˆ              57.2   56.6    55.8   49.3   39.4       51.6
Cameroon        71.7   69.7    67.1   62.3   53.3       64.8
Ethiopia        84.0   79.7    77.0   70.5   55.5       73.3
Ghana           71.2   68.5    67.1   66.6   65.4       67.8
Gambia          67.6   71.3    71.0   70.9   68.6       69.9
Guinea Bissau   74.5   76.6    76.5   77.3   71.5       75.3
Kenya           85.0   82.7    80.7   76.6   67.8       78.5
Madagascar      89.6   90.5    90.7   90.6   88.1       89.9
Mali            79.5   74.4    73.1   67.5   68.8       72.7
Malawi          81.2   79.9    78.4   74.9   69.2       76.7
Nigeria         92.4   82.6    81.6   80.5   74.7       82.3
Rwanda          78.8   69.1    60.8   49.7   27.7       57.2
Senegal         63.7   63.2    62.7   59.8   54.7       60.8
Tanzania        90.0   88.4    85.8   82.9   74.9       84.4
Uganda          74.7   75.5    71.9   67.1   53.0       68.4
South Africa    63.6   55.7    46.7   32.1   15.6       42.7
 Note: Own elaboration based on expenditure surveys.



                              34
                         Table 3
                 Commodity Income Shares
                   Latin America: 2004


                          Quintiles p/c   income       Total
                      (1)   (2)    (3)      (4)  (5)
Argentina             3.1   3.1    3.2      3.2  3.5    3.2
Bolivia              38.2 19.6 17.1        13.8 10.6   19.9
Brazil               22.3 14.8 10.1         7.8  5.5   12.1
Chile                17.7 15.6 13.4        10.8 9.0    13.3
Colombia             13.7 14.7 11.0         8.3  6.2   10.8
Costa Rica           26.0 23.9 19.1        13.1 8.9    18.2
Dominican Republic   22.0 16.1 12.7        12.4 7.4    14.1
Ecuador              44.1 33.9 24.5        17.5 11.5   26.3
El Salvador          19.6 17.1 10.9         7.8  5.4   12.2
Guatemala            25.2 23.0 20.6        14.3 11.1   18.8
Honduras             43.7 34.8 21.8        14.1 11.1   25.1
Mexico               24.1 14.0 11.0         7.1  6.4   12.5
Nicaragua            36.9 29.6 20.8        15.7 15.0   23.6
Panama               32.9 17.3 11.6         9.7  4.5   15.2
Paraguay             38.4 27.9 22.5        17.1 15.5   24.3
Peru                 37.6 27.1 17.3        11.2 9.5    20.5
Uruguay               9.6   6.9    4.7      3.8  3.5    5.7
Venezuela            23.1 19.3 13.5        11.8 8.2    15.2
   Note: Own elaboration based on household surveys.




                              35
                       Table 4
                Commodity Income Shares
                        Africa


                        Quintiles p/c income          Total
                 (1)      (2)    (3)    (4)  (5)
Burundi          54.0    57.9    60.8   59.6   51.7   56.8
Benin            60.6    58.3    56.9   54.1   52.2   56.4
Burkina Faso     73.9    69.4    64.3   57.3   43.7   61.7
 ote d’Ivoire
Cˆ               57.1    56.5    55.5   49.0   38.6   51.3
Cameroon         68.4    66.8    64.3   59.9   51.7   62.2
Ethiopia         82.4    78.0    75.2   68.3   52.6   71.3
Ghana            67.5    65.2    64.3   63.8   62.5   64.7
Gambia           67.3    70.8    70.0   69.4   66.4   68.8
Guinea Bissau    68.4    72.6    72.0   73.0   67.3   70.7
Kenya            83.0    80.9    78.8   74.6   65.6   76.6
Madagascar       85.7    87.5    88.0   87.8   85.0   86.8
Mali             76.8    71.0    69.4   63.1   63.7   68.8
Malawi           58.8    61.0    60.9   59.2   55.6   59.1
Nigeria          88.5    75.3    75.0   74.8   69.6   76.6
Rwanda           77.7    67.3    59.1   47.9   26.4   55.7
Senegal          61.6    61.0    60.2   57.1   51.8   58.3
Tanzania         90.0    88.4    85.8   82.9   74.9   84.4
Uganda           71.2    73.0    69.3   64.3   49.5   65.5
South Africa     58.8    51.6    43.9   31.0   15.4   40.1
 Note: Own elaboration based on household surveys.




                                36
                              Table 5
                     Commodity Income Shares
              Methodology Using Repeated Cross-Sections
                     Latin America : 2004-2008


                        year       Quintiles p/c   income       Total
                               (1)   (2)    (3)      (4)  (5)
Argentina               2004   3.1   3.1    3.2      3.2  3.5   3.2
                        2008   4.2   3.8    3.2      3.0  2.5   3.3
Brazil                  2004   22.3 14.8 10.1        7.8  5.5   12.1
                        2008   16.9 14.6 12.0        9.6  5.9   11.8
Chile                   2003   17.7 15.6 13.4       10.8 9.0    13.3
                        2009   16.0 16.0 14.0       12.6 10.9   13.9
Costa Rica              2004   26.0 23.9 19.1       13.1 8.9    18.2
                        2008   18.9 20.3 17.5       13.6 8.1    15.7
Dominican Republic      2004   22.0 16.1 12.7       12.4 7.4    14.1
                        2008   17.2 13.9 11.4        9.5  6.7   11.7
Ecuador                 2004   44.1 33.9 24.5       17.5 11.5   26.3
                        2008   32.3 28.6 24.4       19.8 12.3   23.5
El Salvador             2004   19.6 17.1 10.9        7.8  5.4   12.2
                        2008   28.3 22.4 17.9       14.0 8.0    18.1
Honduras                2004   43.7 34.8 21.8       14.1 11.1   25.1
                        2008   44.4 39.4 29.4       20.6 12.6   29.3
Mexico                  2004   24.1 14.0 11.0        7.1  6.4   12.5
                        2008   18.9 13.9 11.1        9.4  7.0   12.1
Panama                  2004   32.9 17.3 11.6        9.7  4.5   15.2
                        2009   20.7 14.6 10.9        9.4  4.2   12.0
Paraguay                2004   38.4 27.9 22.5       17.1 15.5   24.3
                        2008   33.0 26.9 22.9       19.1 14.7   23.3
Peru                    2004   37.6 27.1 17.3       11.2 9.5    20.5
                        2008   39.1 29.4 20.7       13.1 7.7    22.0
Uruguay                 2004   9.6   6.9    4.7      3.8  3.5   5.7
                        2008   14.9 14.5 12.8       11.3 8.6    12.4
         Note: Own elaboration based on household surveys.




                                    37
                        Table 6
             Welfare Effects of Price Changes
                 Examples from Mexico


                                Net          Net      All
                             Producers    Consumers

First Order Effects
  consumption effects            -0.99        -2.70    -1.99
  production effects             2.78          0.77     1.60
  net effects                     1.78        -1.93    -0.39
Segmented Labor Markets
  income effects                   7.22       1.99     4.16
  net effects                      6.23       -0.71    2.17
Labor Market Spillovers
  income effects                10.77          3.14    6.31
  net effects                    9.78          0.44    4.32
Second Order Effects
  consumption effects            -0.93        -2.53    -1.87
  production effects             3.05          0.84     1.76
  net effects                     2.12        -1.69    -0.11
  Note: Own calculations based on Porto (2008).




                             38