The welfare impacts of rising quinoa prices: evidence from Peru

Introduction
Conceptual Framework
Data
Empirical Framework
Estimation Results
Summary and Concluding Remarks
The Welfare Impacts of Rising Quinoa Prices:
Evidence from Peru
Marc F. Bellemare, Johanna Fajardo-Gonzalez, and Seth R.
Gitter
International Center for Tropical Agriculture – Cali, Colombia
May 4, 2015
Bellemare, Fajardo-Gonzalez, and Gitter The Welfare Impacts of Rising Quinoa Prices

Introduction
Data
Empirical Framework
Estimation Results
Introduction
Quinoa (Chenopodium quinoa) is an Andean grain that is relatively
high in protein as well as in essential amino-acids. It was
domesticated over 3,000 years ago in the Andes and has been
grown in Bolivia, Ecuador, and Peru since then.
As a consequence of quinoa being seen as some kind of
“superfood” in the US, the UK, and other rich countries, there has
been a sharp increase in the demand for quinoa since 2007.
Peruvian exports of quinoa to the United States, for example,
totaled $80 million in 2013–up from $5 million in 2008
(AgroVision, 2014).
Similarly, the price of quinoa has tripled since 2007, and it shows
no sign of falling back down.

Introduction
Data
Empirical Framework
Estimation Results
Introduction

Introduction
Data
Empirical Framework
Estimation Results
Introduction
Given the foregoing, one might reasonably want to know what this
means for those who have traditionally relied on quinoa for their
subsistence, viz. Andean peasants.
And indeed, a few years ago, journalists made all kinds of
contradictory statements about the welfare impacts of rising
quinoa prices.

Introduction
Data
Empirical Framework
Estimation Results
Introduction
Johanna Blythman, The Guardian (Manchester), January 16, 2013:
[T]here is an unpalatable truth to face for those of us
with a bag of quinoa in the larder. The appetite of
countries such as ours for this grain has pushed up prices
to such an extent that poorer people in Peru and Bolivia,
for whom it was once a nourishing staple food, can no
longer a¤ord to eat it. Imported junk food is cheaper. In
Lima, quinoa now costs more than chicken. Outside the
cities, and fueled by overseas demand, the pressure is on
to turn land that once produced a portfolio of diverse
crops into quinoa monoculture.

Introduction
Data
Empirical Framework
Estimation Results
Introduction
Doug Saunders, The Globe and Mail (Toronto), January 19, 2013:
The people of the Altiplano are indeed among the
poorest in the Americas. But their economy is almost
entirely agrarian. They are sellers –farmers or farm
workers seeking the highest price and wage. The quinoa
price rise is the greatest thing that has happened to
them. And it is a deliberate strategy: Quinoa had all but
died out as a staple in Bolivia, replaced by beans and
potatoes, until farmers began planting it in the 1980s
with exports to North America in mind.

Introduction
Data
Empirical Framework
Estimation Results
Introduction
In my post, I highlighted three research questions one needs to
answer before knowing what the welfare impacts of rising quinoa
prices are:
1. Are most households in the Altiplano net buyers or net sellers
of quinoa, or are they autarkic?
2. Do net seller households produce under contract, as part of a
quinoa value chain, or do they sell to processors on the spot
market?
3. Is it possible to store quinoa for a relatively long period?
In this paper, we ask a question related to the …rst question above,
and we ask:
What e¤ects, if any, did rising quinoa prices have on
the welfare of quinoa-cultivating households?

Introduction
Data
Empirical Framework
Estimation Results
Introduction
To answer this question, we use ten rounds of the Peruvian
Encuesta Nacional de Hogares (ENAHO), a nationally
representative, large-scale household survey conducted annually.
Because the ENAHO is not longitudinal (i.e., it is a repeated
cross-section in that it surveys a new sample of households every
year), we construct a pseudo-panel (Antman and McKenzie, 2007)
by using geographical units (districts, provinces, and departments)
instead of households as our units of observation, and by looking
at geographical unit-level averages instead of household-speci…c
measures of welfare and quinoa cultivation.
We combine this with a di¤erence-in-di¤erences (DiD) approach to
assess the causal impact of quinoa cultivation on household welfare.

Introduction
Data
Empirical Framework
Estimation Results
Introduction
Our results suggest that rising quinoa prices have had positive
e¤ects on the welfare of quinoa-producing households through two
channels:
I First, the average level of consumption (our proxy for income,
and thus for welfare) of quinoa-producing households
increased at a higher rate than that of households that did not
produce quinoa in 2012 and 2013, after the sharp increase in
quinoa prices.
I Second, the variance of consumption of quinoa-producing
households decreased at a higher rate than that of households
that did not produce quinoa in 2012 and 2013.
Thus, it looks as though quinoa production has both …rst- and
second-order e¤ects on household welfare.

Introduction
Data
Empirical Framework
Estimation Results
Outline
1. Introduction
2. Conceptual Framework
3. Data
4. Empirical Framework
I Estimation Strategy
I Identi…cation Strategy
5. Results
6. Summary and Concluding Remarks

Introduction
Data
Empirical Framework
Estimation Results
In order to investigate the welfare e¤ects of a change in the price
of good i on a household’s welfare, Deaton (1989) de…ned the
concept of net bene…t ratio as follows:
NBRi =
pi qi
m
, (1)
where pi denotes the price of good i, qi denotes the household’s
net purchases of good i (net sales if qi < 0), and m denotes the
household’s income.
Obviously, the sign of NBRi depends directly on the sign of qi .
With qi > 0, NBRi < 0 and the household loses out from a price
increase. With qi < 0, NBRi > 0 and the household bene…ts from
a price increase.

Introduction
Data
Empirical Framework
Estimation Results
The idea of the NBR has found much support in the empirical
literature (Budd 1993; Barrett and Dorosh 1996; Lasco et al.
2008).
One also needs to realize that the presence of transactions costs
will also drive whether some households will be net buyers (q > 0)
or net sellers (q < 0). See for example de Janvry et al., 1991;
Goetz, 1992; Key et al., 2001; Bellemare and Barrett, 2006; etc.
So as the price of quinoa goes up, theory suggests those
households who are net sellers will bene…t, and those who are net
buyers will lose out.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Net Buyers and Net Sellers
Data
We use the Peruvian ENAHO Condiciones de Vida y Pobreza for
the period 2004-2013.
We have ten years worth of nationally representative data on
227,400 households across 1,401 districts in 194 provinces in 25
departments.
The data cover 8,216 quinoa-producing households household-year
observations, or about 3.5% of the sample.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Welfare Measure
For our welfare measure, we use total household consumption,
which the Peruvian INEI imputed based on a two-week recall of
consumption at the household level.
If you are not an economist, treating household consumption as
“welfare” might seem strange. Ideally, one would want to use a
measure of income to capture welfare— for economists, welfare is
increasing in income— but as Deaton (1997) points out, collecting
data on income is both tedious and di¢ cult. Consumption is a
good proxy measure for income, and so it is our proxy measure for
welfare here.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Welfare Measure
In order to avoid obvious endogeneity issues, we subtract the
consumption of quinoa from our consumption measure (we include
it as part of robustness checks later).
Our consumption measure also includes the consumption of goods
produced by the household— for quinoa growers, this represents
about 40% of their total consumption.
All data is expressed in 2004 PEN using in‡ation data obtained
from the Central Reserve Bank of Peru.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Welfare Measure
Quinoa producers have total consumption that is about one third
that of households that do not grow quinoa. In other words, they
start out much poorer than households that do not produce quinoa.
The NBR for quinoa is much larger (in absolute value) for quinoa
sellers than it is for quinoa buyers. In other words, the budget share
of quinoa for sellers exceeds the budget share of quinoa for buyers.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Welfare Measure
But we are also interested in second-order e¤ects. In other words,
we are interested in knowing whether the variability of
consumption is di¤erent for those households that produce quinoa.
In an expected utility theory sense, assuming people are risk averse
(which is not unlikely for households in a country such as Peru), it
would be interesting to know if the production of quinoa is
associated with a more stable household consumption.
So we use the variance of total household consumption as an
additional measure of welfare.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Welfare Measure
0.2.4.6
Density
4 6 8 10 12 14
Ln(Real Annual Expenditure)
Growers of quinoa Non-growers of quinoa
Data source: ENAHO
Figure: Distribution of Welfare by Quinoa Production Status.

Introduction
Data
Empirical Framework
Estimation Results
Data Source
Welfare Measure
Most quinoa-producing households produce quinoa for their own
consumption. In 2013, only 17% of them sold any quinoa, up from
roughly 8% for the period 2004-2010.
Less than 0.5% of quinoa producers reported having purchased
quinoa over the last two weeks.
Roughly 30% of all Peruvian households consumed quinoa over the
last two weeks. This …gure was the same in 2004 and 2013, with
some variation in between.

Introduction
Data
Empirical Framework
Estimation Results
Estimation Strategy
Identi…cation Strategy
Estimation Strategy
As I mentioned above, the ENAHO is a repeated cross-section, i.e.,
it does not follow households over time, and so it is not possible to
use standard panel techniques (e.g., household …xed e¤ects) to
identify the potential causal relationship ‡owing from quinoa
production to welfare.
What we can do, however, is to treat geographical units (i.e.,
districts, provinces, and departments) as our unit of observation.
Because households are randomly selected in each community in
each round, this is akin to matching households across rounds
along both their observable and unobservable characteristics.
It is this random sampling (as well as the use of controls for the
time period) on which our identi…cation strategy hinges— more on
that in a minute.

Introduction
Data
Empirical Framework
Estimation Results
Estimation Strategy
Estimation Strategy
So to be clear: We treat a geographical unit as our unit of
observation. For each unit, we take the average of household
consumption, and the proportion (i.e., average) of households who
grow quinoa and regress the former on the latter.
We do this for three di¤erent geographical units (i.e., district,
province, and department, with varying sample sizes) so as to
ensure that our results are robust.

Introduction
Data
Empirical Framework
Estimation Results
Estimation Strategy
Estimation Strategy
The equation we estimate is thus
ln cgt = α0 + β0Dgt +
2013
∑
t=2005
αt Tt
+
2013
∑
t=2005
βt Dgt xTt +
G
∑
g =2
γg dg + gt , (2)
where cgt is average household consumption in region g in year t,
Dgt is the proportion of households who produce quinoa in
geographical unit g in year t, T are …xed e¤ects for each year t,
dg are …xed e¤ects for each geographical unit g, and gt is an error
term with mean zero

Introduction
Data
Empirical Framework
Estimation Results
Estimation Strategy
Might this not only capture a correlation between household
consumption and quinoa cultivation instead of the causal e¤ect of
quinoa cultivation on household consumption?
We argue that our combined use of (i) pseudo-panel techniques,
and (ii) a di¤erence-in-di¤erences design yield the causal e¤ect of
quinoa production on household welfare.

Introduction
Data
Empirical Framework
Estimation Results
Estimation Strategy
This is because
1. The fact that households are randomly selected each year in
each community means they are matched along both
observables and unobservables, and the inclusion of time …xed
e¤ect corrects for changes over time in their characteristics.
2. Given that, the di¤erence-in-di¤erences estimator (with
clustered standard errors) should yield an estimate of the
e¤ect of quinoa production on household welfare. that is
plausibly causal.

Introduction
Data
Empirical Framework
Estimation Results
Estimation Results: Household Consumption
Variables District Province Department
Year 2005*Quinoa -0.155* -0.093 -0.091
(0.084) (0.127) (0.141)
Year 2006*Quinoa -0.004 0.092 0.191
(0.085) (0.106) (0.223)
Year 2007*Quinoa 0.005 0.098 0.207
(0.093) (0.163) (0.308)
Year 2008*Quinoa 0.06 0.033 0.107
(0.098) (0.195) (0.413)
Year 2009*Quinoa 0.079 0.129 0.228
(0.084) (0.161) (0.378)
Year 2010*Quinoa 0.061 0.082 0.369
(0.09) (0.157) (0.233)
Year 2011*Quinoa 0.022 0.029 0.278
(0.09) (0.144) (0.292)
Year 2012*Quinoa 0.161* 0.257* 0.474
(0.085) (0.146) (0.336)
Year 2013*Quinoa 0.377*** 0.562*** 1.050**
(0.092) (0.164) (0.502)
N 9613 1919 250
R2 0.217 0.426 0.778
Number of districts 1401
Number of provinces 194
Number of departments 25
Signi…cant at: *** p<0.01, ** p<0.05, * p<0.1
Robust standard errors in parentheses
clustered at regional level

Introduction
Data
Empirical Framework
Estimation Results
How should you interpret those …ndings? Here, the marginal
e¤ects are complicated by the semi-logarithmic speci…cation of the
equation of interest. In order to recover a marginal e¤ect, the
estimates need to be transformed as per the formula given by
Kennedy (1980).
So for Year 2013*Quinoa, for example, the marginal e¤ect is 46
percentage points. That is, for a community whose proportion of
quinoa growers increases by 10%, households would see an increase
in their consumption that is 4.6% faster on average.

Introduction
Data
Empirical Framework
Estimation Results
We conduct a number of robustness checks:
I We use only quinoa-producing regions.
I We control for baseline (i.e., 2004) quinoa production
I We include quinoa purchases as part of overall household
consumption
I We did the analysis at the household level (with all the
endogeneity problems that this entails)
By and large, our …ndings are preserved: Signi…cant positive
welfare impacts of quinoa production, but only in most recent years
(i.e., 2012 and 2013).

Introduction
Data
Empirical Framework
Estimation Results
Estimation Results: Variance of Household Consumption
Variables District Province Department
Year 2005*Quinoa 0.021 0.081 0.314
(0.075) (0.122) (0.19)
Year 2006*Quinoa 0.015 0.014 0.138
(0.065) (0.13) (0.217)
Year 2007*Quinoa -0.159** -0.143 -0.198
(0.072) (0.115) (0.197)
Year 2008*Quinoa -0.125 -0.151 -0.083
(0.08) (0.124) (0.25)
Year 2009*Quinoa -0.064 -0.061 -0.12
(0.084) (0.125) (0.162)
Year 2010*Quinoa -0.072 0.083 -0.099
(0.082) (0.142) (0.156)
Year 2011*Quinoa -0.106 -0.033 0.032
(0.082) (0.115) (0.254)
Year 2012*Quinoa -0.183** -0.264** -0.363**
(0.079) (0.121) (0.155)
Year 2013*Quinoa -0.079 -0.198 -0.356**
(0.082) (0.121) (0.149)
Intercept 0.368*** 0.460*** 0.563***
-0.008 -0.013 -0.017
N 9613 1919 250
R2 0.008 0.032 0.239
Signi…cant at: *** p<0.01, ** p<0.05, * p<0.1
Robust standard errors in parentheses
clustered at regional level

Introduction
Data
Empirical Framework
Estimation Results
Estimation Results: Variance of Household Consumption
Our results indicate that over time, within district variance of
consumption has been increasing.
It was only in 2012, however, that the production of quinoa
consistently led to a decrease in the variance of consumption.
The same might be true for 2013, although the result here is much
less robust.

Introduction
Data
Empirical Framework
Estimation Results
Given our empirical results, it looks as though the dramatic
increase in quinoa prices translated into improved welfare outcomes
for quinoa-producing households in Peru, but only in 2012 and
2013, after the quinoa price spike.
Before the price spike, quinoa-producing households had
consumption growth rates similar to the rest of the country.

Introduction
Data
Empirical Framework
Estimation Results
Our next steps are to conduct additional robustness checks to
make sure that our …ndings robust to various speci…cations of our
equation of interest.
Two of us are currently doing …eldwork in Peru, surveying a sample
of 150 households every four months, in an e¤ort to study seasonal
patterns of production and consumption, and truly get at the
welfare e¤ects of rising quinoa prices.

The welfare impacts of rising quinoa prices: evidence from Peru

Recommended

Recommended

More Related Content

What's hot

What's hot (13)

Viewers also liked

Viewers also liked (16)

Similar to The welfare impacts of rising quinoa prices: evidence from Peru

Similar to The welfare impacts of rising quinoa prices: evidence from Peru (20)

More from CIAT

More from CIAT (20)

Recently uploaded

Recently uploaded (20)

The welfare impacts of rising quinoa prices: evidence from Peru