Is Eco Certification A Win-Win For Developing Country

World Development Vol. xx, pp.
xxx–xxx, 2016
0305-750X/Ó 2016 Elsevier Ltd. All rights reserved.
www.elsevier.com/locate/worlddev
http://dx.doi.org/10.1016/j.worlddev.2016.01.004
Is Eco-Certification a Win–Win for Developing Country

Agriculture? Organic Coffee Certification in Colombia
MARCELA IBANEZ a and ALLEN BLACKMAN b,*
a
Courant Research Centre: Poverty, Equity and Growth, Goettingen, Germany
b
Resources for the Future, Washington, DC, USA
Summary. — According to advocates, eco-certification is a win–win solution to the problem of environmental degradation caused by
developing country agriculture, improving both the environmental and the economic performance of farmers. However, these notional
benefits can be undercut by the tendency of relatively wealthy farmers already meeting eco-certification standards to disproportionately
participate. Using original farm-level survey data along with matching and matched difference-in-differences models, we analyze the
producer-level effects of organic coffee certification in southeastern Colombia. We find that certification is associated with changes in
farm practices linked to improved environmental outcomes. It significantly reduces sewage disposal in the fields and increases the adop-
tion of organic fertilizer. However, we are not able to discern economic benefits.
Ó 2016 Elsevier Ltd. All rights reserved.
Key words — organic certification, coffee, Colombia, difference-in-differences matching
1. INTRODUCTION translates into benefits for producers who receive price premi-
ums for certified commodities and/or have improved access to
Although agricultural production is an important source of output and credit markets. Those private economic benefits, in
income and employment for developing countries, it also is turn, motivate producers to improve their environmental per-
responsible for serious environmental damage, including aqui- formance. In addition to these private economic incentives,
fer depletion, land degradation, water pollution, soil erosion, eco-certification may improve environmental and economic
deforestation, biodiversity loss, and the spread of invasive spe- performance by helping to disseminate technical information
cies (Foley et al., 2011; Laurance, Sayer, & Cassman, 2014; about best management practices and molding private and
Sebastian & Schulz, 2015; Tilman et al., 2001). Addressing public sector actors’ environmental preferences and standards
these problems using conventional command-and-control reg- (Bartley, 2007a, 2007b).
ulation is challenging for a number of reasons: regulatory However, in practice, both the environmental and the
institutions are often weak, political will for stringent enforce- economic benefits of eco-certification may be limited at the
ment is limited, and producers tend to be small, numerous, producer level. In the case of environmental benefits,
and geographically dispersed (Chomitz, 2007; Sterner & self-selection is a major barrier. Producers already meeting
Coria, 2011; Wehrmeyer & Mulugetta, 1999). environmental certification criteria tend to disproportionately
Organic, Rainforest Alliance, UTZ, and other eco- obtain certification (Barbosa de Lima et al., 2009; Blackman &
certification initiatives award labels to farmers conditional Rivera, 2011). Such producers have relatively strong incentives
on their meeting specific environmental and social perfor- to participate: the costs are low because they do not have to
mance criteria. These initiatives represent a nonstate, change production practices to meet certification standards,
market-based approach to addressing environmental problems and the benefits, including price premiums and improved mar-
in developing countries that has the potential to sidestep the ket access, can be significant. However, if the bulk of certified
constraints on conventional command-and-control regulation producers are already meeting certification criteria, then on
noted above (Auld, 2010; Blackman, 2010; Cashore, Auld, & average, certification will have only limited additional effects
Newsom, 2004; Jordan, Wurzel, & Zito, 2003). Moreover,
unlike conventional environmental regulation, which generally
imposes economic costs on producers, in principle eco- * This study was funded by the Latin American and Caribbean
certification can generate economic benefits, such as price Environmental Economic Program (LACEEP); the Adlerbert Research
premiums and improved market access (Bolwig, Gibbon, & Foundation; the Swedish Research Council, Formas, through the Human
Jones, 2009; Dragusanu, Giovannucci, & Nunn, 2014; Cooperation to Manage Natural Resources (COMMONS) program; and
Giovannucci & Ponte, 2005). This feature makes it a particu- The Swedish Foundation for Strategic Environmental Research,
larly attractive environmental management strategy in the MISTRA, through its Environment and Trade in a World of
agricultural sector, which typically employs the poorest of Interdependence (ENTWINED) program. We are grateful to the
the poor. Hence, eco-certification has the potential to be a Federación Nacional de Cafeteros de Colombia and Cenicafe for
win–win solution to the problem of environmental degrada- assistance with our survey. The paper has benefited from discussions
tion caused by developing country agriculture. with Gerhard Riener and comments from participants in LACEEP
According to proponents, eco-certification can generate both workshops, the AgroFood Conference in Göttingen, and the
environmental and economic benefits at the producer level ENTWINED Policy Seminar on Emerging Trends in Sustainable Trade
(Milder et al., 2014; Romero et al., 2013). Eco-certification in Stockholm. Special thanks to Tomas David Vargas for assistance with
enables consumer to differentiate among commodities based field research and to Joyce Bond and Sally Atwater for careful editing.
on their environmental attributes. This improved information Final revision accepted: January 6, 2016.
1
Please cite this article in press as: Ibanez, M., & Blackman, A. Is Eco-Certification a Win–Win for Developing Country Agriculture?
Organic Coffee Certification in Colombia, World Development (2016), http://dx.doi.org/10.1016/j.worlddev.2016.01.004
2 WORLD DEVELOPMENT
on production practices. Hence, it is not clear ex ante Takahashi and Todo (2013) and Rueda and Lambin (2013)
whether—after controlling for self-selection effects—eco-certifi focus on Rainforest Alliance, not organic, certification.
cation actually has significant environmental benefits. Several less rigorous studies analyze environmental effects
Just as the environmental effects of coffee eco-certification by comparing farm practices or environmental outcomes for
are uncertain, so too are the economic effects. In principle, certified farms before and after certification or comparing out-
the price premiums, improved market access, and technical comes for certified farms and unmatched uncertified farms.
change associated with eco-certification can boost growers’ Most find certification has little or no effect. Quispe Guanca
profits. But eco-certification also generally requires changes (2007) compares environmental management practices before
in production practices that raise some costs. For example, and after five types of eco-certification in Costa Rica and finds
organic cultivation is typically more labor intensive than con- that aside from herbicides, certification did not lead to reduc-
ventional farming (Lygbaeck, Muschler, & Sinclair, 2001; Van tions in the use of agrochemicals. Philpott, Bichier, Rice, and
der Vossen, 2005). In addition, eco-certification entails fixed Greenberg (2007) find no differences in ecological indicators
transaction costs associated with red tape and variable trans- for (unmatched) organic, Fair Trade, and uncertified farms
action costs associated with monitoring and reporting. in Chiapas, Mexico. Finally, Martı́nez-Sánchez (2008) finds
Finally, just as it dilutes environmental benefits, self- that compared with unmatched conventional farms in Costa
selection also can dilute economic benefits. The reason is that Rica, organic farms do not have significantly different shade
relatively wealthy producers may be more likely to obtain eco- levels, bird diversity, or bird abundance.
certification because they can more easily cover the fixed trans- Empirical evidence on the producer-level economic effects of
action costs (Barham, Callenes, Gitter, Lewis, & Weber, 2011; coffee eco-certification is more plentiful than that on environ-
Mendez et al., 2010). Given all these factors, ex ante, the net mental effects and includes several studies that purport to con-
effect of eco-certification on producers’ economic status is trol for self-selection. Overall, this evidence also is quite
uncertain. mixed. Of the studies that control for self-selection, Bolwig
As discussed below, existing evidence on the producer-level et al. (2009) and Arnould, Plastina, and Ball (2009) reach
environmental effects of coffee eco-certification is limited, and the most optimistic conclusions. Relying on a Heckman
that on the economic effects is quite mixed. The present paper model, Bolwig et al. (2009) find that the net revenues of Ugan-
investigates both the environmental and the economic effects dan certified organic growers are 75% higher than those of
of organic certification in the southeastern part of Colombia, uncertified growers. 1 Using a combination of matching and
one of the world’s leading coffee producers. We rely on an regression, Arnould et al. (2009) find that Fair Trade certifica-
original panel data set and use matching and matched tion is positively correlated with volume of coffee sold and
difference-in-differences (DID) estimators to control for selec- price obtained (but less consistently correlated with various
tion effects (Rosenbaum & Rubin, 1983; Smith & Todd, 2005). indicators of educational and health status).
We do not directly observe environmental outcomes such as Other studies that control for selection effects reach more
water and soil quality, and we therefore use as proxies farm measured conclusions. For example, Chiputwa, Spielman,
practices linked to these outcomes (e.g., Blackman & and Qaim (2015) use propensity score matching to analyze
Naranjo, 2012; Quispe Guanca, 2007). We find that organic the effects of three coffee certification schemes in Uganda—
certification in southeastern Colombia fosters the adoption Fair Trade, organic, and UTZ—and find that only Fair Trade
of cleaner farm practices: it significantly reduces the disposal raises household living standards. Using fixed effects panel
of sewage in field and spurs the adoption of organic fertilizer. regression models, Barham and Weber (2012) find that the
However, we are not able to discern an effect on producers’ net returns to Fair Trade/organic and Rainforest Alliance cer-
income or net returns. This last result implies that maintaining tification depend critically on how certification affects yields.
and expanding the current level of organic certification in Ruben and Fort (2012) use propensity score matching to ana-
Colombia may be challenging. lyze the economic effects of Fair Trade certification in Peru
Empirical evidence on the environmental effects of coffee and conclude it has only a modest direct effect on income
eco-certification is limited (Blackman & Rivera, 2011; IFAD, and production. Using the same methods, Jena, Chichaibelu,
2003; Parrot, Olesen, & Høgh-Jensen, 2007). To our knowl- Stellmacher, and Grote (2012) find that Fair Trade certifica-
edge, only three quantitative studies consider the producer- tion had limited effects on small-scale coffee producers’ liveli-
level environmental effects of coffee and attempt to control hoods in Ethiopia. Using rudimentary matching, Lygbaeck
for self-selection bias. Blackman and Naranjo (2012) find that et al. (2001) find that although the price premium from
organic certification in Costa Rica reduces the use of chemical organic coffee partly compensates for lower yields, once the
pesticides, herbicides, and fertilizers and increases the use of cost of certification is included, organic production generates
organic fertilizer—all farm practices linked to environmental lower net revenues than conventional production. Finally,
outcomes like soil and water quality. They attribute this find- two separate studies of coffee growers in Nicaragua that do
ing partly to the fact that in Costa Rica, the vast majority of not control for selection bias—Valkila (2009) and Beuchelt
coffee growers rely heavily on agrochemicals and therefore and Zeller (2011)—conclude that in some circumstances, Fair
do not meet organic certification standards. As a result, Trade organic certification can perpetuate or exacerbate low
opportunities for the type of self-selection described above productivity and poverty among marginal growers.
are limited. Using remote sensing data, Takahashi and Todo Our paper aims to make several contributions to the litera-
(2013) conclude that Rainforest Alliance certification reduces ture on eco-certification. First, to our knowledge, it is only the
deforestation in Ethiopia. Finally, relying on both satellite second study to examine both the environmental and the eco-
and survey data, Rueda and Lambin (2013) find that Rain- nomic effects of an eco-certification initiative (the other being
forest Alliance certification in northern Colombia increases Rueda & Lambin, 2013). As discussed above, studies focused
adoption of environmentally friendly farm practices but does on each effect separately reach varying conclusions. An impor-
not have a significant effect on deforestation. Like the present tant unanswered question is whether positive environmental
study, Blackman and Naranjo (2012), Takahashi and Todo effects are associated with positive economic ones (Milder
(2013), and Rueda and Lambin (2013) use propensity score et al., 2014). If eco-certification is in fact a win–win proposi-
matching to control for self-selection bias. Unlike our study, tion, as some proponents contend, then it is more likely to
IS ECO-CERTIFICATION A WIN–WIN FOR DEVELOPING COUNTRY AGRICULTURE? ORGANIC COFFEE 3
be an effective nonregulatory strategy for mitigating the conservation and pollution prevention practices. Certified pro-
adverse environmental effects of agriculture in developing ducers are monitored at least once a year to ensure they con-
countries. Second, our study is one of few that control for tinue to meet organic standards. They also need to establish an
self-selection bias when analyzing the environmental effects internal control system that verifies compliance with the
of coffee eco-certification (Blackman & Rivera, 2011). And norms.
finally, it is among the first such studies to analyze coffee Colombia’s experience with organic coffee certification
eco-certification in Colombia, one of the world’s leading coffee began in the late 1980s. It was promoted by multiple organiza-
producers. tions, including Instituto Mayor Campesino, Asociación de
The remainder of the paper is organized as follows. The next Caficultores Orgánicos de Colombia (ACOC), and Corpo-
section provides background on organic coffee cultivation in ración Suna Hisca. Since then, the production of certified
Colombia. Section 3 discusses our methodology. Sections 4 organic coffee has increased (Esguerra, 2001). In 2010, Colom-
and 5 present our data and the results. Section 6 discusses bian growers produced 8,056 tons of certified organic coffee
some of the limitations of our analysis, and the final section on 8,773 certified organic hectares (Ministerio de
sums up and discusses policy implications. Agricultura, 2012). 2
Our study focuses on organic growers in Cauca, a depart-
ment (state) in the southeastern part of the country. Cauca,
2. COFFEE CULTIVATION IN COLOMBIA one of Colombia’s leading centers of organic coffee produc-
tion, is home to 16% of Colombia’s coffee farms and 8% of
Coffee is the most important agricultural product in Colom- its coffee acreage (FNC, 2008a). We focus on five municipali-
bia, generating 12.4% of the total agricultural revenue and ties in Cauca with particularly high rates of organic certifica-
employing about 553,000 growers, 95% of whom cultivate tion: Inzá, Cajibı́o, Tambo, Timbı́o, and La Sierra. In all,
farms smaller than 5 hectares (FNC, 2011; Giovannucci, 331 growers harvesting 587 hectares were certified organic in
2002). these municipalities in 2008 (FNC, 2008b). In addition, 162
Traditionally, coffee was grown in the shade of natural or growers harvesting 211 hectares were in transition to certifica-
planted forests, with few agrochemicals. Over the past tion (FNC, 2008b). The vast majority of these growers were
30 years, however, new coffee varieties have been developed certified in the decade prior to our survey. Very few were cer-
that can be cultivated in full sun. Almost half of Colombian tified prior to 1998. On average, farmers in our sample had
coffee is ‘‘sun-grown,” and in some regions the fraction only 2.5 years of experience in certified organic production.
exceeds three-quarters (FNC, 2008a). In Colombia, yields Three organizations certified the organic growers in our study
per hectare from sun-grown coffee are about four times those area: Bio-Latina, Institute for Marketecology (IMO) Control,
from shade-grown coffee (FNC, 2008a). and Organic Crop Improvement Association. Each of these
But there is a disadvantage: sun-grown coffee has serious organizations holds multiple accreditations for multiple mar-
adverse environmental effects. The reduction in tree cover is kets. 3 The operation of the three organizations is similar in
associated with a loss in biodiversity. For example, terms of certification criteria and procedures. The price pre-
Greenberg, Bichier, Cruz, and Reitsma (1997) found that mium for organic coffee that growers in our sample received
full-sun coffee plantations support 90% fewer bird species than depended on the grower association and ranged from 5.2%
shade-grown coffee. In addition, sun-grown coffee is associ- to 7.7% of the price of similar uncertified coffee.
ated with higher rates of soil erosion, is more susceptible to
weeds and pests, and is typically associated with increased
use of chemical fertilizers and pesticides that can pollute sur- 3. METHODOLOGY
face and groundwater and cause worker health problems
(Ataroff & Monasterio, 1997; Babbar & Zak, 1995; We evaluate the effect of organic certification on the adop-
Bermúdez, 1980). For example, in Costa Rica, sun-grown cof- tion of cleaner farm practices and economic outcomes. Let
fee uses 30 kilograms of nitrogen per hectare per year com- y citþs be an outcome variable for grower i at time t + s (for
pared with little or no use in shade-grown plantations s > 0) conditional on a certification indicator variable C which
(UNDP, 2010). Shade-grown plantations benefit from leaf lit- takes a value equal to one when the producer was certified and
ter and other organic residues that help improve soil fertility. a value equal to zero otherwise. The causal effect of certifica-
Also, some tree species harbor nitrogen-fixing bacteria that tion for grower i at time period t + s is defined as
increase soil fertility (Beer, 1988).
Concerns about the environmental harms of coffee cultiva- y 1itþs y 0itþs ð1Þ
tion, among other things, spurred organic certification. The average effect of certification on all certified growers is
Organic agriculture certification requires producers to adhere 1
to five broad production principles (IFOAM, 2010; Van der E y itþs y 0itþs jC ¼ 1 ¼ E y 1itþs jC ¼ 1 E y 0itþs jC ¼ 1 ð2Þ
Vossen, 2005):
use of composted organic matter instead of chemical fer- The general challenge of causal inference is that the quantity
tilizers to maintain soil quality; y 0itþs is unobservable for certified growers. Hence, we must con-
use of natural methods for controlling disease, pests, and struct a contrafactual against which the impacts of the inter-
weeds instead of synthetic insecticides and herbicides; vention can be evaluated. The average outcome for certified
use of soil conservation practices, including contour growers had they not been certified. We rely on matching tech-
planting, terracing, planting cover crops, mulching, and niques to construct this counterfactual. That is, we pair each
planting shade trees; certified grower with an uncertified grower based on a vector
minimal use of fossil fuels in the production process; and of Z characteristics observed before certification.
minimal pollution during postharvest handling. It may not be possible to find uncertified growers with iden-
Organic certification requires growers to complete a transi- tical values of all elements of Z, however. Rosenbaum and
tion period of two to three years, during which they must dis- Rubin (1983) demonstrate that it is necessary to match agents
continue use of chemical inputs and adopt various only on the basis of their propensity score, the probability of
4 WORLD DEVELOPMENT
treatment (here certification) as predicted by a probit regres- variables. DID matching estimators control for time-invariant
sion. This prediction can be interpreted as a weighted average unobserved confounding factors and therefore are more
of the characteristics in Z, where the weights reflect the impor- robust than simple matching estimators (Smith & Todd,
tance of each characteristic in explaining treatment. This 2005). The DID matching estimator is defined as
approach collapses the thorny problem of exact matching on
all observable characteristics to the much simpler problem of E Dy 1itþs Dy 0itþs jC ¼ 1; P ðZ i Þ

matching a single summary variable. ¼ E Dy 1itþs jC ¼ 1; P ðZ i Þ E Dy 0itþs jC ¼ 1; P ðZ i Þ ð5Þ
Using this method, we first estimate the probability of being
certified organic using a probit model. That is, we estimate where D is the before–after change in outcome. Hence, the
DID matching estimator considers the difference between the
ProbðC i ¼ 1Þ ¼ F ðZ i Þ ð3Þ mean before–after change in outcomes for the sample of certi-
We use the estimated parameters to generate a propensity fied growers and a matched sample of uncertified growers
score for each grower in our sample. Various algorithms are selected using propensity scores. Intuitively, this estimator
available to pair certified growers with uncertified growers indicates whether growers who adopted the organic certificate
on the basis of their propensity scores (Caliendo & improve farm practices more than uncertified growers. 4
Kopeining, 2008; Morgan & Harding, 2006). To ensure Calculating standard errors for ATT estimated using
robustness, we report results from five: (i) nearest neighbor propensity score matching is not straightforward because
1-to-1 matching, wherein each certified grower is matched to these errors should, in principle, account for the fact that
the uncertified grower with the closest propensity score; (ii) propensity scores are estimated and for the imputation of
nearest neighbor 1-to-4 matching, wherein each certified the common support (Heckman, Ichimura, & Todd, 1997,
grower is matched to the four uncertified growers with the 1998). Therefore, following Dehejia and Wahba (2002) and
closest propensity scores and the counterfactual outcome is others, we bootstrap standard errors (using 1,000 replica-
the average across these four; (iii) nearest neighbor 1-to-8 tions).
matching; (iv) nearest neighbor 1-to-16 matching; and (v) ker-
nel matching, wherein a weighted average of all uncertified
growers is used to construct the counterfactual outcome. 4. DATA AND VARIABLES
Propensity score matching depends on two identifying
assumptions. The first assumption, ‘‘ignorability” or ‘‘condi- (a) Survey and sample
tional independence,” is that conditional only on agents’
observed characteristics, the treatment (certification) decision The data used for our analysis come from an original survey
is ignorable for purposes of measuring outcomes. That is, we of 382 coffee growers in five municipalities in the department
are able to observe and control for all potentially confounding of Cauca, Colombia. As noted above, we selected this depart-
variables that simultaneously affect the treatment decision and ment and these municipalities because they are home to a rel-
the outcome variables. This first assumption is untestable. The atively large concentration of organic growers. We randomly
second assumption, ‘‘common support” or ‘‘overlap,” is that selected our survey sample from lists of coffee growers. The
the distribution of observed characteristics for nonparticipants survey was conducted between March and June 2008. The sur-
is similar to that for participants, such that agents with similar vey questionnaire was administered on-site by trained enumer-
characteristics have a positive probability of being participants ators in face-to-face sessions that typically lasted 40 min.
and nonparticipants. For all five models, we enforce a com- In conducting our survey, we worked with three main
mon support and allow matching with replacement. grower organizations: Empresa Cooperativa del Sur del Cauca
The average treatment effect for the treated (ATT) can be (COSURCA), Juan Tama Association, and Federación
estimated using cross-sectional data as Nacional de Cafeteros. Each of these organizations has their
own organic certification program and works with a different
E y 1itþs y 0itþs jC ¼ 1; P ðZ i Þ ¼ E y 1itþs jC ¼ 1; P ðZ i Þ certifier. Cosurca works with Institute for Marketecology
(IMO), Juan Tama Association with by Bio-Latina, and Fed-
E y 0itþs jC ¼ 1; P ðZ i Þ ð4Þ
eración de Cafeteros with Organic Crop Improvement Associ-
where P ðZ i Þ is the estimated propensity score. That is, ATT is ation. As noted above, these programs use similar certification
simply the difference between the mean outcomes for the sam- criteria and procedures.
ple of certified growers and a matched sample of uncertified The survey solicited information on both grower character-
growers selected using propensity scores. A disadvantage of istics (e.g., age, education) and farm characteristics (e.g., eco-
this approach, however, is that a violation of the conditional certification, hectares cultivated, types of inputs used) for two
independent assumption—that is, the assumption that we have years: 2007 and 1997. 5 All but five of the organic growers in
controlled for the effects of unobservable confounding fac- our survey sample were certified organic after 1997. Hence,
tors—can bias the results. For example, say the grower’s envi- for all but these five growers, 1997 grower and farm character-
ronmental consciousness is unobservable and highly istics predate and therefore are exogenous to the decision to
correlated both with environmental outcomes and with selec- obtain organic certification. 6 As noted below, we dropped
tion into certification. In that case, ATT estimates may be these five growers from our regression sample.
biased upward. In the analysis of environmental effects, which relies on
To control for such unobservables, in addition to simple recall data from 1997 as well as more recent data from 2007,
matching we also use a difference-in-differences (DID) match- we solicited and used data on farm practices linked to environ-
ing estimator where our data permit. This estimator, which mental outcomes, rather than on the exact amount of agro-
exploits information on changes in outcomes over time as well chemicals and other inputs, because the former are easier to
across growers, takes advantage of the fact that we have two- remember and therefore less susceptible to recall bias
period panel data for some (but not all) outcome variables. (Beegle, Carletto, & Himelein, 2012). Among the production
Specifically, as discussed below, we have panel data for farm practices for which we have data, six are monitored by organic
practice outcome variables, but not for our economic outcome certifiers. Of these farm practice outcome measures, four are
‘‘negative” practices that must be discontinued for organic cer- outcome variables and grower and farm characteristics. We
tification: present sample means for the entire sample and certified and
use of chemical fertilizers; uncertified subsamples.
use of chemical insecticides; Our 13 outcome variables correspond to those in the bul-
use of chemical herbicides; and leted list in the previous subsection. The seven farm practice
disposal of sewage in open fields. outcome variables are chemical fertilizer, insecticide, herbicide,
sewage, organic fertilizer, shade, and pulp. These variables are
Two ‘‘positive” practices must be adopted: indicator variables that take the value of one for the use of
use of organic fertilizer; and that practice and takes the value of zero otherwise. The six
use of shade cover for coffee trees. economic outcome variables are labor cost, input cost, total
In addition to those six practices monitored by organic cer- cost, yield, income, and net revenue per hectare per year. Labor
tifiers, we included another positive practice that is not: cost is estimated by summing the cost of hired and family
use of coffee pulp to fertilize. labor. Family labor is valued at the market wage. Input cost
is estimated using input prices and quantities reported by the
We included this practice because there is some concern that producer. Total cost is the sum of labor cost and input cost.
the requirement to use organic fertilizer forces growers to buy It considers only the variable cost and not the fixed cost of
additional organic matter. This seventh practice flags growers planting of a new hectare of coffee. Income comprises the
that use on-farm materials. income from selling coffee at the market price, plus additional
Our survey data include detailed information on six mea- income from price premiums or bonuses. Net return is the dif-
sures of the economic outcomes of coffee cultivation: ference between income and total production cost.
labor costs; To match certified and uncertified growers, we used propen-
input costs; sity scores generated by regressing an organic certification
total costs; dummy onto a rich set of grower and farm characteristics.
yields; The grower characteristics are female, a dichotomous dummy
income; and equal to one for female growers; age, the age of the grower in
net revenue. 2007; education, the highest grade completed in 2007; family
size, the size of the family in 2007; and cooperative, a dichoto-
Starting with the 382 randomly selected growers whom we mous dummy equal to one if the grower was a member of a
surveyed, we constructed our regression sample as follows. coffee committee affiliated with the Federación Nacional de
First, we eliminated 44 growers who obtained a certification Cafeteros de Colombia in 1997. 7
other than, or in addition to, organic, specifically, 33 growers The farm characteristics, all of which correspond to the year
who were Rainforest Alliance certified and 11 who were Fair 1997, are own farm, a dichotomous dummy equal to one if the
Trade certified. We dropped these growers so that we could grower owns (versus rents or leases) the farm 8; farm size, the
disentangle the environmental and economic effects of organic total area of the farm in hectares; area coffee, the size of all
certification from those of other types of certification. How- coffee lots in 1997, no. lots, the number of geographically dis-
ever, as a robustness check, we estimate ATTs using a larger tinct lots on the farm; organic matter, an estimate of the num-
sample that also includes growers with those certificates. Sec- ber of kilograms of manure produced by all animals on the
ond, we dropped eight growers who were certified after 1997 farm 9; capital index, a count of the number of common capital
but gave up their certifications prior to 2007. Third, we items owned (depulper, mill, silo, fumigator, motor, other);
dropped five growers who were certified or transitioning to Borbon, Caturra, Colombia, and Castillo, the proportions of
certification prior to 1998 to control for the endogeneity prob- the farm’s coffee trees that are the most common varieties in
lem noted above. We dropped these growers because they the study area; pasilla, federacion, especial, and calidad, the
would be counted as uncertified in our matching sample but proportions of coffee sold in the four quality grades of Colom-
may have had outcomes or characteristics that were affected bian coffee; transport with vehicle, a dichotomous dummy
by having been certified. Fourth, we dropped 14 growers equal to one if the grower transports the coffee by vehicle (ver-
who reported an economic outcome (labor cost, input cost, sus by animal or on foot); proportion on farm and proportion
total cost, yield, income, or net return) that was an outlier, off farm, the proportions of household members who work
specifically, one more than three times the interquartile range on and off the farm; and Inzá, Cajibı́o, Tambo, Timbı́o, and
above the third quartile or below the first quartile. We did that Sierra, dichotomous dummy variables indicating in which of
to avoid findings that could be driven by the upper or lower the five study municipalities the farm is located.
tail of the distribution. Finally, we dropped 87 growers who
provided incomplete responses, including 69 who did not pro- (c) Descriptive statistics
vide complete responses for 1997 (often because their farms
were not yet producing coffee) and 18 who did not provide There was significant variation across growers in precertifi-
complete responses about economic outcomes. cation (1997) use of the seven farm practices. For the four neg-
Hence, our regression sample comprises 224 growers, all of ative practices, rates range from 4% for herbicide to 51% for
whom had been producing coffee since at least 1997, none of chemical fertilizer. For the positive practices, rates range from
whom were certified as organic in 1997, and none of whom 21% for organic fertilizer to 91% for shade cover. In 1997, for
had ever obtained an eco-certification other than organic. Of two of the seven farm practices, growers who went on to
these 224 growers, 56 were certified organic in 2007, and 168 obtain organic certification in subsequent years performed
had never been eco-certified at any time. ‘‘better” than those who were never certified: they were signif-
icantly less likely to use chemical fertilizer and more likely to
(b) Variables use organic fertilizer. Hence, it is reasonable to expect that a
disproportionate share of growers who in 1997 were de facto
Table 1 lists, defines, and presents descriptive statistics for organic—that is, those already meeting many organic
the variables used in our matching analysis, including both standards—self-selected into organic certification, and as a
6 WORLD DEVELOPMENT
Table 1. Variables, definitions, means, and difference-in-means tests

Variable Definition 1997 2007
Mean uncert07 Mean cert07 Mean all Mean uncert07 Mean cert07 Mean all
(n = 168) (n = 56) (n = 224) (n = 168) (n = 56) (n = 224)
TREATMENT
Certified Certified organic(0/1) 0.00 0.00 0.00 1.00 0.25
OUTCOMES
Negative farm practices
Chemical fertilizer Applies chemical fertilizer (0/1) 0.51 0.23*** 0.44 0.45 0.09*** 0.36
Insecticide Applies insecticide (0/1) 0.20 0.13 0.19 0.16 0.02** 0.12
Herbicide Applies herbicide (0/1) 0.04 0.04 0.04 0.13 0.07 0.11
Sewage Disposes of sewage in field (0/1) 0.47 0.48 0.48 0.32 0.14* 0.27
Positive farm practices
Organic fertilizer Applies organic fertilizer (0/1) 0.21 0.39*** 0.26 0.42 0.88*** 0.54
Shade Uses shade (0/1) 0.91 0.91 0.91 0.96 0.96 0.96
Pulp Uses pulp to fertilize (0/1) 0.55 0.46 0.53 0.80 0.96** 0.84
Economic effects
Labor cost Total cost wages (000 COP) 819.53 684.91 785.87
Input cost Total cost inputs (000 COP) 304.53 97.73*** 252.83
Total cost Total production cost (000 COP) 1,124.06 782.64* 1,038.70
Yield Number of arrobas per hectare 37.48 24.52** 34.24
Income Total income from coffee (000 COP) 1792.63 1,415.82 1,698.42
Net revenue Net revenue from coffee (000 COP) 668.57 633.18 659.72
CONTROLS
Grower
Female Female (0/1) 0.29 0.21 0.27
Age Age (years) 50.77 51.88 51.05
Education Highest grade completed 3.68 4.05 3.77
Family size Total persons in family 4.88 5.27 4.97
Cooperative Member coffee cooperative (0/1) 0.33 0.21 0.30
Farm
Farm owner Own (vs. rent/lease) farm (0/1) 0.92 0.93 0.92
Farm size Farm size (ha) 2.05 2.38 2.13
Area coffee Area cultivated in coffee (ha) 1.27 1.53 1.33
No. lots No. lots farm comprises 1.51 1.84* 1.59
Organic matter Organic material available (000 kg) 2.61 3.53 2.84
Capital index No. capital items owned (1–6)a 1.00 1.14 1.04
Borbon Prop. coffee trees Borbon variety 0.03 0.09* 0.04
Caturra Prop. coffee trees Caturra variety 0.56 0.47 0.53
Colombia Prop. coffee trees Colombia variety 0.09 0.12 0.10
Castillo Prop. coffee trees Castillo variety 0.01 0.01 0.01
Pasilla Prop. coffee sales Pasilla grade 0.11 0.58* 0.23
Federacion Prop. coffee sales federacion grade 0.90 0.56 0.82
Especial Prop. coffee sales especial grade 0.01 0.01 0.01
Calidad Prop. coffee sales calidad grade 0.05 0.27*** 0.10
Transport wt vehicle Transp. to market w/ vehicle (0/1) 0.69 0.64 0.68
Proportion on farm prop. household works on farm 0.58 0.56 0.58
Proportion off farm Prop. household works off farm 0.20 0.19 0.20
Inzá Municipality Inzá (0/1) 0.31 0.55*** 0.37
Cajibı́o Municipality Cajibı́o (0/1) 0.20 0.05* 0.16
Tambo Municipality Tambo (0/1) 0.16 0.14 0.16
Timbı́o Municipality Timbı́o (0/1) 0.18 0.02** 0.14
Sierra Municipality La Sierra (0/1) 0.16 0.23 0.17
Cert07 refers to growers who were certified as organic in 1997–2007. Uncert07 refers to growers who were not certified as organic in the period. Prop.
stands for proportion.
a
Capital items are depulper, mill, silo, fumigator, motor, and other.
result certification had only limited additional influence on the itive, albeit small, implying that a handful of the 52 certified
use of cleaner technologies for the average grower in our sam- growers in our sample used chemical inputs or disposed of
ple. Our empirical analysis aims at determining whether that their sewage in their fields in 2007. 10 Similarly, the means of
was the case. the positive farm practice outcome variables for certified farm-
Note that all of the means of the negative farm practice out- ers all are less than one, implying that some of the certified
come variables for growers who were certified by 2007 are pos- growers in our sample had not adopted the environmental
management practices we consider. There are two reasons why Table 2. Probit regression results (dependent variable = organic
certified organic farmers may not meet all certification criteria. certification)
First, certification rules permit the occasional use of chemical Variable Marginal effect S.E.
inputs and other deviations in certain cases. 11 Second, grow-
Grower
ers may fail to conform with certification standards. Certifying
Female 0.008 0.063
bodies directly inspect 20% of certified farms; within the Age 0.001 0.002
grower cooperatives, the peer control system inspects 100% Education 0.015 0.013
of certified growers. Both certifying bodies and cooperatives Family size 0.020 0.013
recognize that this system does not guarantee perfect confor- Cooperative 0.068 0.059
mance. For example, a report from a certifying body working
in our study area reported a 7% nonconformance rate among Farm
the one-fifth of the sample inspected and a cooperative Area coffee 97 0.000 0.026
reported a 5% rate (Biolatina, 2006). Farm owner 0.093 0.094
For both certified growers (i.e., those certified by 2007) and Farm size 0.003 0.014
uncertified growers, mean use of most negative farm practices No. lots 0.077*** 0.027
fell during 1997–2007, and mean use of all positive farm prac- Organic matter 0.000 0.000
tices rose. The statistical analysis aims to establish whether the Capital index 0.007 0.026
effect was significantly larger for certified producers once we Borbon 0.308** 0.131
control for potential self-selection. Caturra 0.092 0.070
Among our seven farm practice outcome variables, in five Colombia 0.179* 0.104
cases we observe significant differences in 2007 means for Castillo 0.019 0.348
Pasilla 0.078 0.064
certified and uncertified growers. Means for three of our
Federacion 0.008 0.021
six economic variables are significantly lower for certified
Especial 0.083 0.270
growers than for uncertified growers. Specifically, mean
Calidad 0.336*** 0.080
input costs for certified growers are about 40% that for
Transport with vehicle 0.021 0.056
uncertified growers, and total production costs per hectare
Prop. on farm 0.020 0.057
are about 75%. Surprisingly, even though organic technolo- Prop. off farm 0.003 0.064
gies are generally thought to be relatively labor intensive, Inzá 0.037 0.084
labor costs for certified and uncertified growers are not sta- Cajibı́o 0.263** 0.116
tistically different. Mean yield per hectare for certified grow- Timbı́o 0.351*** 0.132
ers is about 70% that for uncertified growers, a significant Sierra 0.032 0.087
difference. Mean income for certified growers is not signifi-
cantly different from that for uncertified growers. Finally, N 224
mean net revenues for certified and uncertified growers are Pseudo R2 0.268
not statistically different. LR chi2(28) 67.530
As for our control variables, our sample of growers is com- * ** ***
, , significant at 10%, 5%, and 1% level, respectively.
posed of small-scale growers with relatively low socioeco-
nomic status. The average 1997 farm size in our sample was
two hectares, and the average 2007 coffee net revenue per hec- Table 3 presents the balance test for the five matching estima-
tare was 668,000 Colombian pesos (US$267). The average tors used. All the estimators balanced all 26 covariates
grower had 3.7 years of school in 1997. included in the model; that is, they generated a statistically
There are significant differences in the precertification (1997) insignificant difference in means for certified and matched
means of six control variables for certified and uncertified uncertified farms for all 26 covariates. Following
growers. Compared with uncertified growers, certified growers Rosenbaum and Rubin (1983), we also used median standard-
were more likely to have had farms comprising multiple lots, ized bias to measure matching quality. 12 The highest median
to have grown the Borbon variety of coffee, to have produced standardized bias is 12.8% for the nearest neighbor 1–1 estima-
high-quality coffee, and to have lived in certain municipalities tor, and the lowest is 3.8% for the nearest neighbor 1–16 esti-
and not others. These preexisting characteristics are likely to mator. Although a clear threshold for acceptable median
affect our outcomes. Hence, it is important to control for them standardized bias does not exist, according to Caliendo and
in estimating our treatment effects. Kopeining (2008), a statistic below 3–5% is generally viewed
as sufficient. These reasonable balance statistics are likely
due to the fact that even though our probit selection model
5. RESULTS has 26 covariates, our sample includes more than three uncer-
tified growers for each certified grower. As a result, we were
(a) Propensity scores and balance tests able to find fairly close matches for each certified farm.
Table 2 presents the marginal effects from the probit regres- (b) Average treatment effect on the treated
sion (of organic certification on grower and farm characteris-
tics) used to generate propensity scores. The results indicate We first discuss farm practice outcomes and then examine
that, all other things equal, certified growers are more likely economic outcomes. Table 4 presents results for the four neg-
to have more individual coffee plots, to have grown the ative farm practices. Results are from five standard matching
Borbon and Colombia coffee varieties, and to have sold estimators (Eqn. 4) and from the DID estimator (Eqn. 5).
calidad-grade coffee, and they are less likely to be located in The standard matching estimators indicate that certification
Cajibı́o or Timbı́o than in Tambo. has positive additional effects, reducing the use of three of
We used propensity scores from our Probit regression to the four negative practices: chemical fertilizer, chemical insec-
match certified organic producers with uncertified producers. ticides, and sewage disposal in fields. For each ATT, at least
8 WORLD DEVELOPMENT
Table 3. Matching quality: covariates achieving balance (N) and median standardized bias (SB) after matching, for five propensity score matching
methodsa,b,c
Method N SB R2 Treated Controls
(i) Nearest neighbor 1–1 26 12.8 31.4 56 168
(ii) Nearest neighbor 1–4 26 5.9 5.0 56 168
(iii) Nearest neighbor 1–8 26 5.5 4.8 56 168
(iv) Nearest neighbor 1–16 26 3.8 3.6 56 168
(v) Kernel 26 5.5 4.1 56 168
a
The model includes 26 covariates.
b
For a given covariate, the standardized bias (SB) is the difference of means in the certified and matched uncertified subsamples as a percentage of the
square root of the average sample variance in both groups. We report the median SB for all covariates.
c
Median SB before matching is 18.3.
Table 4. Negative farm practices: average treatment effect on treated (ATT) estimates, by outcome variable and matching method; critical value of
Rosenbaum’s C
Propensity score matching DID matched
Mean Certified ATT S.E.a
C*b
Coef. S.E.a
Chemical fertilizer
(i) Nearest neighbor 1–1 0.085 0.404*** 0.125 >3.00
(ii) Nearest neighbor 1–4 0.085 0.378*** 0.097 >3.00
(iii) Nearest neighbor 1–8 0.085 0.388*** 0.086 >3.00
(iv) Nearest neighbor 1–16 0.085 0.396*** 0.078 >3.00
(v) Kernel 0.085 0.388*** 0.084 >3.00 0.074 0.087
Insecticide
(i) Nearest neighbor 1–1 0.021 0.170** 0.075 >3.00
(ii) Nearest neighbor 1–4 0.021 0.122** 0.060 >3.00
(iii) Nearest neighbor 1–8 0.021 0.130** 0.052 >3.00
(v) Kernel 0.021 0.128*** 0.047 >3.00 0.056 0.061
Herbicide
(i) Nearest neighbor 1–1 0.085 0.043 0.066 –
(ii) Nearest neighbor 1–4 0.085 0.005 0.064 –
(iii) Nearest neighbor 1–8 0.085 0.003 0.057 –
(iv) Nearest neighbor 1–16 0.085 0.025 0.056 –
(v) Kernel 0.085 0.013 0.052 – 0.047 0.054
Sewage
(ii) Nearest neighbor 1–4 0.170 0.181* 0.096 >3.00
(iii) Nearest neighbor 1–8 0.170 0.194** 0.085 >3.00
(v) Kernel 0.170 0.178** 0.084 >3.00 0.179* 0.098
* ** ***
a
Standard Errors computed using bootstrap with 1,000 repetitions.
b
Critical value of odds of differential assignment to organic certification due to unobserved factors (i.e., value above which ATT is no longer significant).
four of the five standard matching estimators (nearest neigh- pulp fertilizer. For organic fertilizer, all five standard matching
bor 1–1 and so forth) are negative and significant. The results ATTs are positive and statistically significant. The magnitude
suggest that the positive effects of certification are not only due of the effect is substantial, ranging from 45 to 53 percent-
to self-selection of more environmentally friendly producers. age points. The DID ATT is 27 percentage points. For the
The adoption of the certification is associated with a reduction use of pulp fertilizer, all five standard matching ATTs are pos-
in the use of chemical fertilizer by 39–40 percentage points, itive and statistically significant suggesting that certification
chemical insecticides by 12–17 percentage points, and sewage boosts the practice by 21–32 percentage points. The DID esti-
disposal in fields by 18–22 percentage points. Qualitative mator indicates a 25 percentage point increase. For shade,
results from the DID matching estimators comport with those none of the simple matching ATTs or the DID estimator is
from the standard matching estimators only in the case of sew- statistically significant. This is not surprising, given that 95%
age disposal in the field. The DID suggests that certification of the matching sample used shade in 1997.
reduces sewage disposal in fields by 18 percentage points. Table 6 presents results for the economic effects of certifica-
Given the discrepancy between our matching and matched tion. As noted above, for these outcome variables, we are not
DID results, our results for chemical fertilizer and insecticide able to use a DID estimator because we do not have 1997 data.
must be interpreted with caution. The results suggest that organic certification decreases the
Table 5 presents results for the three positive practices. They input costs and yields but does not have a significant effect
suggest that certification increases use of organic fertilizer and on labor costs, total costs, income or net return. For input
Table 5. Positive farm practices: Average treatment effect on treated (ATT) estimates, by outcome variable and matching method; critical value of
Rosenbaum’s C
Propensity score matching DID matched
Mean certified ATT S.E.a
C*b
Coef. S.E.a
Organic fertilizer
(v) Kernel 0.872 0.451*** 0.091 >3.00 0.266*** 0.095
Shade
(v) Kernel 0.957 0.051 0.064 – 0.004 0.051
Pulp fertilizer
(v) Kernel 0.979 0.225*** 0.071 2.66 0.245*** 0.089
* ** ***
a
Standard errors computed using bootstrap with 1,000 repetitions.
b
costs, ATT is both negative and significant for four of the five tors; and so forth. The probability value on the Wilcoxon sign-
matching estimators. The magnitude is substantial, ranging rank or Mantel and Haenszel statistic is a test of the null
from 194,000 to 249,000 pesos per hectare. 13 For yield, hypothesis of a zero ATT given unobserved confounding vari-
ATT is both negative and significant for three of the five esti- ables that have an effect given by C. So, for example, a prob-
mators. Certified producers produce 31–36% less than uncerti- ability value of 0.01 and a C of 1.2 indicate that ATT would
fied producers. The insignificance of ATTs for income and net still be significant at the 1% level even if matched pairs differed
return probably reflects the fact that lower yields associated in their odds of certification by a factor of 1.2 because of
with certification are offset by lower input costs and poten- unobserved confounding factors.
tially by higher price premiums. Given these results, it is clear We calculate C*, the critical value of C at which ATT is no
that because of the prices and other conditions that existed in longer significant at the 10% level, in each case—that is, for
Cauca during our study period, economic incentives for certi- each combination of outcome variables and matching estima-
fication were limited. tor—where ATT is significant (Tables 4–6, column 4). Among
the negative farm practices, C* is greater than 3 in the case of
(c) Sensitivity analysis all 15 estimators for chemical fertilizer use, insecticide, and
sewage disposal in fields. Among the positive practices, C* is
Might endogeneity drive our results? As noted above, the greater than 3 in the case of all five estimators for organic fer-
effectiveness of our matching estimators in controlling for tilizer and is no lower than 2.66 in the case of pulp fertilizer.
selection bias depends on the untestable identifying assump- For economic effects, C* is larger than 3 for all nine statisti-
tion that we are able to observe confounding variables that cally significant estimators (five for input costs, one for total
simultaneously affect growers’ decisions to obtain organic cer- cost, and three for yield). Hence, our sensitivity tests suggest
tification and our outcome variables. That is, we essentially that unobserved confounders would need to be quite strong
assume endogeneity is not a problem. We calculate Rosen- to undermine our statistically significant matching ATTs. In
baum bounds to check the sensitivity of our results to the fail- other words, endogeneity is unlikely to drive these results.
ure of this assumption (Aakvik, 2001; Rosenbaum, 2002). 14
Rosenbaum bounds indicate how strongly unobserved con- (d) Robustness check
founding factors would need to influence growers’ decisions
about obtaining organic certification to undermine the match- As previously discussed, to disentangle the effects of organic
ing result. To be more specific, the Rosenbaum procedure gen- certification from those of other types of certification, we elim-
erates a probability value for Wilcoxon sign-rank statistic (in inated from our study sample 44 growers who obtained Rain-
the case of continuous variables) or a Mantel and Haenszel forest Alliance or Fair Trade certification in addition to
statistic (in the case of binary variables) for a series of values organic certification. However, one might argue that neither
of C, an index of the strength of the influence that unobserved Rainforest Alliance nor Fair Trade certification is likely to
confounding factors have on the selection process. C = 1 affect environmental outcomes because Rainforest Alliance
implies that confounding factors have no influence, such that environmental criteria are generally less stringent than organic
pairs of growers matched on observables do not differ in their and because Fair Trade criteria are mostly economic. There-
odds of obtaining organic certification; C = 2 implies that fore, as a robustness check, we estimated environmental ATTs
matched pairs could differ in their odds of certification by as using a larger sample (n = 251 versus n = 224) that includes
much as a factor of 2 because of unobserved confounding fac- the certified organic growers among farmers with Fair Trade
10 WORLD DEVELOPMENT
Table 6. Economic effects: average treatment effect on treated (ATT) estimates, by outcome variable and matching method; critical value of Rosenbaum’s C
Mean certified Propensity score matching
ATT S.E.a C*b
Labor cost (000 COP)
(v) Kernel 704.156 97.151 119.315 –
Input cost (000 COP)
(i) Nearest neighbor 1–1 102.757 199.180* 107.959 >3.00
(v) Kernel 102.757 249.309*** 82.148 >3.00
Total Cost (000 COP)
(v) Kernel 806.913 346.460* 184.805 >3.00
Yield
(ii) Nearest neighbor 1–4 24.301 13.432* 7.102 >3.00
(iii) Nearest neighbor 1–8 24.301 11.126* 6.060 >3.00
(v) Kernel 24.301 14.110** 7.142 >3.00
Income (000 COP)
(i) Nearest neighbor 1–1 1,400.205 352.142 405.015 –
(ii) Nearest neighbor 1–4 1,400.205 371.151 358.889 –
(iii) Nearest neighbor 1–8 1,400.205 273.220 317.806 –
(iv) Nearest neighbor 1–16 1,400.205 141.407 296.460 –
(v) Kernel 1,400.205 409.897 339.152 –
Net return (000 COP)
(v) Kernel 593.292 63.437 226.866
* ** ***
a
Standard Errors computed using bootstrap with 1,000 repetitions.
b
or Rainforest Alliance certifications. 15 The results, which are tions. We include this test to show that dropping or keeping
reported in the Appendix, are qualitatively identical and q very these economic outlier observations does not affect our results
similar to those detailed above (Tables A1 and A2). At least for farm practices. The results for both our cross-sectional and
three ATTs from cross-sectional matching models for chemi- DID models, which are presented in the Appendix (Tables A3
cal fertilizers, insecticides, disposal of sewage in the field, and A4), are qualitatively and quantitatively very similar to
organic fertilizer, and pulp fertilizer are statistically significant, those for our main modes. We expect at least some results
and ATTs from DID models for disposal of sewage in fields, for economic effects to be different when economic outliers
organic fertilizer, and pulp fertilizer are statistically significant. are included, and indeed, we find that is the case. Overall qual-
The economic effects of organic certification are also similar to itative results are quite similar—certification affects input costs
the ones obtained with a smaller sample. At least three ATTs and yields—but point estimates and significance levels differ.
from cross-sectional matching models for input cost, total
cost, and yield are statistically significant.
As explained earlier in Section 4, we dropped from our 6. DISCUSSION
regression sample observations where economic outcomes
(labor cost, input cost, total cost, yield, income, or net return) In considering the implications of our results, it is important
were more than three times the interquartile range above the to keep in mind that our analysis has several limitations. First,
third quartile or below the first quartile. We did that to avoid we do not have data on environmental outcomes such as water
findings that could be driven by the upper or lower tail of the quality, soil quality, and biodiversity. Therefore, we use as
distribution. As a robustness check, we relax this restriction proxies data on farm practices linked to such outcomes. This
and estimate the model with a larger sample of 236 observa- linkage is stronger in some cases than in others. Reductions
in the use of insecticides, herbicides, and disposal of sewage in environmental outcomes for the most recent completed har-
the fields are likely to improve water and soil quality vest cycle (2007) and for the cycle 10 years earlier (1997).
(Robinson & Mansingh, 1999; Soares, Leão, Vianna Neto, Growers may misremember information from 10 years earlier,
& Oliveira, 2012) and increased shade is likely to enhance bio- which could bias results, particularly if the information con-
diversity and reduce soil erosion and nitrogen leaching cerns environmental and economic outcomes. To minimize
(Ataroff & Monasterio, 1997; Greenberg et al., 1997). How- this risk, we use as environmental outcomes simple 0/1 indica-
ever, the environmental effect of replacing chemical fertilizers tors of whether growers used important farm practices such as
with organic ones depends on the types and amounts of fertil- organic fertilizers, shade, and disposal of sewage in fields. Such
izers used. Like chemical fertilizers, organic fertilizers can ‘‘salient” information is relatively easy to remember and less
degrade water and soil quality. For example, poultry manure susceptible to recall bias than information about quantities
is associated with water pollution and can increase the concen- of farm inputs (Beegle et al., 2012). It is worth emphasizing
tration of heavy metals in soil (Gao et al., 2006; Pell, 1997). A that we do not ask for or use 1997 data on economic outcomes
back-of-the envelope calculation indicates that given the types (labor, yields, income, etc.). Unfortunately, we do not have
and amounts of chemical and organic fertilizers used by the contemporaneous reported 1997 data needed to determine
growers in our sample, switching from chemical fertilizers to the reliability of our recall data.
organic ones is likely to reduce nitrogen, phosphorus, and Fourth, we estimate average treatment effects for growers
potassium used per hectare, but may increase calcium oxide who were certified within the last two years of the study period
use. 16 and for those who were certified up to 10 years earlier. In prin-
Adding to the uncertainty about the net environmental ciple, effects of certification may be different for these two sub-
effects of the farm practices associated with organic certifica- groups. Unfortunately, however, our sample size (56 certified
tion is their effect on forest cover. In principle, these practices growers) is not large enough to enable us to estimate subgroup
could exacerbate forest degradation and/or clearing in several effects. For similar reasons we cannot disentangle the effect of
ways. First, if organic fertilizers are less effective than chemical certification by the different certifying bodies.
ones in boosting yields, growers may compensate by expanding Finally, we do not control for the spillover effects that certi-
the area planted in coffee at the expense of nearby forests. Sec- fication may have had on uncertified growers’ outcomes. In
ond, if yields from organic farms are relatively low and rates of particular, the adoption of certain practices by certified grow-
certification are sufficiently high, then certification could in the- ers may spur adoption by uncertified ones. Such effects can
ory reduce the local supply of coffee and raise coffee prices. occur when uncertified growers learn about the private benefits
That, in turn, could create an economic incentive for forest of adoption (e.g., lower input costs) from certified growers in
clearing and degradation to plant more coffee. Alternatively, their social networks (Angelucci & Di Maro, 2015; Weber,
price premiums for certified organic coffee could in principle 2012). Such effects would tend to bias our ATTs downward
create an economic incentive for conversion of forest to coffee by reducing average differences between treated and control
production (Rappole, King, & Vega Rivera, 2003; Tejeda- groups.
Cruz, Silva-Rivera, Barton, & Sutherland, 2010; Zwane, 2007).
A second limitation of our study is that it relies on self-
reported data on environmental outcomes, specifically, farm 7. CONCLUSION
practices associated with organic certification. Certified survey
respondents may have a tendency to strategically report prac- We have used rich original survey data on 224 coffee farms
tices—both for 2007 and for 1997—that meet certification cri- in southeastern Colombia to identify the effects of organic cof-
teria, whether or not they actually used these practices. That fee certification on farm practices and economic indicators.
phenomenon would have different effects on our cross- We have used propensity score matching and DID estimators
sectional matching and DID results. It would overstate differ- to control for self-selection bias.
ences in average 2007 outcomes between certified and uncerti- Our findings strongly suggest that certification has signifi-
fied growers, thereby biasing our cross-sectional matching cant additional effects on the adoption of farm practices that
ATTs upward. However, it would understate before–after are associated with positive environmental outcomes. It
(1997–2007) differences in outcomes for certified growers, reduces the use of one of the negative practices for which we
thereby biasing our DID ATTs downward. Unfortunately, have data (disposal of sewage in fields) and increases the use
we do not have third-party data that could help us determine of two of the three positive practices for which we have data
whether self-reporting bias is an issue in our sample. (use of organic fertilizer and pulp fertilizer). In each case,
A third limitation of our study is that it relies on recall data; our conventional matching estimators and DID matching esti-
our survey solicited information on grower characteristics and mator generate a positive and significant ATT that is robust to
Table 7. Mean (standard deviation) estimated application of nitrogen, phosphorus, potassium, and calcium oxide per hectare per year, and difference-in-means
tests
Chemical Mean All (n = 224) Mean Cert07 (n = 56) Mean Uncert07 (n = 168) t-Test
***
Nitrogen (kg) 22.170 0.334 29.449
(50.960) (2.325) (57.034)
**
Phosphorus (kg) 10.376 1.925 13.193
(24.320) (6.462) (27.278)
***
Potassium (kg) 15.372 1.094 20.131
(34.472) (3.405) (38.624)
***
Calcium oxide (kg) 114.604 211.044 82.457
(214.585) (259.098) (187.669)
* ** ***
, , significant at 10%, 5%, and 1% level, respectively. Standard deviation in parentheses.
possible hidden bias. For two additional negative practices— that these are the hallmarks of certification programs that tend
the use of chemical fertilizer and insecticide—our results also to generate significant producer-level benefits (Darnall &
suggest that certification reduces use. However, these results Sides, 2008; deLeon & Rivera, 2009; Rivera, de Leon, &
are weaker. In each case, our conventional matching estima- Koerber, 2006). Hence, our study indicates that commodity
tors generate a positive ATT that is not sensitive to hidden certification schemes with these characteristics can have signif-
bias, but our DID matching estimator, which purports to con- icant effects on farm practice associated with environmental
trol for unobserved confounders, is not significant. Hence, quality, even in areas where self-selection threatens to dilute
overall, we find that organic certification affects the use of these effects.
farm practices in a way that would be expected to improve Our results regarding organic certification’s economic effects
growers’ environmental performance. are much less encouraging. We are not able to discern robust
These findings are consistent with those of Blackman and effects of certification on labor costs, income, or net returns.
Naranjo (2012), who find that organic certification spurs farm Only in the case of input costs and yields does certification
practices associated with higher environmental quality and have a significant (negative) effect. These findings reflect the
attribute this to the fact that in Costa Rica, growing coffee emerging conventional wisdom that organic cultivation has
is highly technified, so most farms must change their manage- mixed effects on net returns, typically limiting overall eco-
ment practices to obtain organic certification. In Cauca, nomic benefits (Lygbaeck et al., 2001; Van der Vossen,
Colombia, by contrast, growing coffee is less technified. As 2005). Although our results do not comport with recent stud-
noted above, except in the case of organic fertilizer use, the ies that suggest that organic certification can cause economic
lion’s share of the uncertified growers in our sample use prac- harm at the producer level—exacerbating or perpetuating pov-
tices consistent with organic certification. Hence, one would erty (Beuchelt & Zeller, 2011; Valkila, 2009) or diverting
expect that the environmental effects of certification would investment from coffee growing (Barham et al., 2011)—they
be diluted by self-selection. Contrary to this intuition, how- do suggest that the economic benefits of organic certification
ever, we find that organic certification encourages environ- are limited. That implies that promoting further organic certi-
mentally friendly farm practices at least to some extent. fication and even maintaining existing certifications may be
Thus our study suggests that the effectiveness of organic cer- challenging.
tification in spurring cleaner farm practices has less to do with As a caveat to these conclusions, we hasten to note that our
the preexisting characteristics of coffee growers than with the study is among the first to examine these issues using quanti-
design characteristics of the certification program. Although tative methods that aim to control for sample selection. More
many eco-certification programs feature fuzzy standards, evidence from other study sites is needed before we can draw
self-monitoring, and participation by cooperatives, we have general conclusions about whether and under what circum-
examined a program that has relatively well-defined standards stances eco-certification, or even just organic certification,
enforced at the individual farm level by both third-party mon- has environmental and economic benefits.
itors and a peer-monitoring system. Previous studies suggest
NOTES
1. A caveat, however, is that, as the authors make clear, certified farmers 7. A limitation of the survey is that it did not ask about family size for
were also ‘‘contract farmers” for a large agroindustrial company, an 1997.
arrangement that likely affected their economic outcomes.
8. Note that growers who own their farms may also rent lots outside
2. This represents a relatively small share of the total production of those farms.
coffee. In 2010, the annual production of coffee was 578,000 metric tons,
harvested from about 921 thousand hectares.
9. These estimates are based on Muñoz and Moreno’s (2001) simple
linear model, which relates the number and type of farm animals to the
3. Bio-Latina is accredited by the German Deutsches Akkredi-
tierungssystem Prüfwesen (DAP), the US Department of Agriculture quantity of manure.
National Organic Program (USDA NOP), Japan Agricultural Standards
(JAS), and the Conseil des Appellations Agroalimentaires du Québec 10. Certified growers in our regression sample reported using four types
(CAAQ). IMO Control is accredited by the Swiss Accreditation Service of chemical fertilizers: 17-6-18, DAP, Urea and Malation.
(SAS), USDA NOP, and JAS. The Organic Crop Improvement Associ-
ation is accredited by USDA NOP, the International Organic Accredi-
11. In general, inspectors enforce prohibitions against negative practices
tation Service, the European Union Equivalent Standards, JAS, and
(e.g., use of agrochemicals) more stringently than they require the positive
CAAQ.
ones (e.g., soil conservation).
4. We estimated the model using the Stata command diff, which uses
kernel matching to implement the DID matching estimators (Villa, 2011). 12. Standardized bias is the difference of the sample means in the certified
and uncertified subsamples as a percentage of the square root of the
5. Coffee prices in 1997 were the highest in the past two decades, reaching average of sample variances in both groups.
US$1.31 per pound. This price spike may have induced changes in farm
practices, although we expect that transformation of the productive 13. This represents a 65–70% reduction in input costs.
system is slow.
14. An example of an unobserved confounder might be environmental
6. All certified producers in our regression sample obtained their consciousness or managerial skill. Each could cause growers to select into
certificates during 1998–2007. Because of the limited number of observa- organic certification and—independent of certification—use fewer nega-
tions, we do not test for subgroup effects for early and late adopters. tive practices and more positive ones.
15. Only two of the producers in the sample had more than one formula, we estimated the amount of nitrogen, phosphorus, potassium,
certificate. One had Rainforest Alliance and organic certification, and the and calcium oxide used by each grower per hectare per year. For
other had Rainforest Alliance and Fair Trade certification. nonchemical fertilizers, we used estimates of chemical content in North
and Bell (1984) and Ramos Agüero and Alfonso (2014). Our results from
16. Our survey data include the type and amount of fertilizers growers this exercise, presented in the Appendix, indicate that certified organic
used per year. Based on the technical specifications of the formula for the producers applied significantly less nitrogen, phosphorus, and potassium
chemical fertilizers, and assuming that there was no adulteration of the but more calcium oxide (Table 7).
REFERENCES
Aakvik, A. (2001). Bounding a matching estimator: The case of a Caliendo, M., & Kopeining, S. (2008). Some practical guidance for the
Norwegian training program. Oxford Bulletin of Economics and implementation of propensity score matching. Journal of Economic
Statistics, 63, 115–143. Surveys, 32, 31–72.
Angelucci, M., & Di Maro, V. (2015). Programme evaluation and spillover Cashore, B., Auld, G., & Newsom, D. (2004). Governing through markets:
effects. Journal of Development Effectiveness, 1–22 (Epub ahead of Forest certification and the emergence of non-state authority. New
print). Haven, Connecticut: Yale University Press.
Arnould, E., Plastina, A., & Ball, D. (2009). Does fair trade deliver on its Chiputwa, B., Spielman, D. J., & Qaim, M. (2015). Food standards,
core value proposition? Effects on income, educational attainment, and certification, and poverty among coffee farmers in Uganda. World
health in three countries. Journal of Public Policy and Marketing, 28, Development, 66, 400–412.
186–201. Chomitz, K. (2007). At loggerheads? Agricultural expansion, poverty
Ataroff, M., & Monasterio, M. (1997). Soil erosion under different reduction, and environment in the tropical forests. Washington, DC:
management of coffee plantations in the Venezuelan Andes. Soil World Bank.
Technology, 11, 95–108. Darnall, N., & Sides, S. (2008). Assessing the performance of voluntary
Auld, G. (2010). Assessing certification as governance: Effects and broader environmental programs: Does certification matter?. Policy Studies
consequences for coffee. Journal of Environment and Development, 19, Journal, 36, 95–117.
215–241. Dehejia, R., & Wahba, S. (2002). Propensity score-matching methods for
Babbar, L., & Zak, D. (1995). Nitrogen loss from coffee agroecosystems in non experimental causal studies. Review of Economics and Statistics,
Costa Rica: Leaching and denitrification in the presence and absence 84, 151.
of shade trees. Journal of Environmental Quality, 24, 227–233. deLeon, P., & Rivera, J. (Eds.) (2009). Voluntary environmental programs:
Barbosa de Lima, A., Novaes Keppe, A. L., Maule, F. E., Sparovek, G., A policy perspective. Lanham, Maryland: Rowman & Littlefield.
Corréa Alves, M., & Maule, R. F. (2009). Does certification make a Dragusanu, R., Giovannucci, D., & Nunn, N. (2014). The economics of
difference? Impact assessment study on FSC/SAN certification in Brazil. Fair Trade. Journal of Economic Perspectives, 28, 217–236.
Piracicaba, Brazil: Imaflora. Esguerra, G. (2001). La Caficultura orgánica en Colombia. División de
Barham, B., Callenes, M., Gitter, S., Lewis, J., & Weber, J. (2011). Fair Estrategia y proyectos Especiales de Comercialización. Bogotá: Fed-
trade/organic coffee, rural livelihoods, and the ‘‘agrarian question”: eración Nacional de Cafeteros de Colombia (FNC).
Southern Mexican coffee families in transition. World Development, 39, Federación Nacional de Cafeteros de Colombia (FNC) (2008a). Infor-
134–145. mación económica cafetera. Estadı́sticas Históricas. Bogotá: FNC.
Barham, B., & Weber, J. (2012). The economic sustainability of certified Federación Nacional de Cafeteros de Colombia (FNC) (2008b). Sistema
coffee: Recent evidence from Mexico and Peru. World Development, 40, de Información Cafetera SICA. Bogotá: FNC.
1269–1279. Federación Nacional de Cafeteros de Colombia (FNC) (2011). Infor-
Bartley, T. (2007a). How foundations shape social movements: The mación económica cafetera. Estadı́sticas Históricas. Bogotá: FNC.
construction of an organizational field and the rise of forest certifica- Foley, J., Ramankutty, N., Brauman, K., Cassidy, E., Gerber, J.,
tion. Social Problems, 54, 229. Johnston, M., ... Zaks, D. (2011). Solutions for a cultivated planet.
Bartley, T. (2007b). How certification matters: Examining mechanism of Nature, 478, 337–342.
influence. Appendix C in Toward sustainability: The roles and limits of Gao, D., Chen, T. B., Liu, B., ZhengZheng, Y. M., Zheng, G., & Li, Y. X.
certification. Steering Committee of the State-of-Knowledge Assess- (2006). Releases of pollutants from poultry manure in China and
ment of Standards and Certification. Washington, DC: RESOLVE. recommended strategies for the pollution prevention. Geographical
Beegle, K., Carletto, C., & Himelein, K. (2012). Reliability of recall in Research, 2, 14.
agricultural data. Journal of Development Economics, 98, 34–41. Giovannucci, D. (2002). Colombia: Coffee sector study. Washington, DC:
Beer, J. (1988). Litter production and nutrient cycling in coffee (Coffea World Bank, Latin American and Caribbean Regional Office.
arabica) or cacao (Theobroma cacao) plantations with shade trees. Giovannucci, D., & Ponte, S. (2005). Standards as a new form of social
Agroforestry Systems, 7, 103–114. contract? Sustainability initiatives in the coffee industry. Food Policy,
Bermúdez, M. (1980). Erosión hı́drica y escorrentı́a superficial en el 30, 284–301.
sistema de café (Coffea arabica L.) por (Erythrina poeppigiana, Walpers Greenberg, R., Bichier, P., Cruz, A., & Reitsma, R. (1997). Bird
O. F. Cook) y laurel (Cordia ailiodora, R & P. Cham.) en Turrialba, populations in shade and sun coffee plantations in central Guatemala.
Costa Rica (Tesis de Maestrı́a inédita). Turrialba, Costa Rica: CATIE. Conservation Biology, 11, 448–459.
Beuchelt, T., & Zeller, M. (2011). Profits and poverty: Certification’s Heckman, J. J., Ichimura, H., & Todd, P. (1997). Matching as an
troubled link for Nicaragua’s organic and fairtrade coffee producers. econometric evaluation estimator: Evidence from evaluating a job
Ecological Economics, 70, 1316–1324. training programme. Review of Economic Studies, 64, 605–654.
Biolatina (2006). Informe de inspección programada. Area A. Inzá, Cauca: Heckman, J. J., Ichimura, H., & Todd, P. (1998). Matching as an
Asociación de Cabildos Indı́genas de Juan Tama, Mimeo. econometric evaluation estimator. Review of Economic Studies, 65,
Blackman, A. (2010). Alternative pollution control policies in developing 261–294.
countries. Review of Environmental Economics and Policy, 4, 234–253. International Federation of Organic Agriculture Movements (IFOAM).
Blackman, A., & Naranjo, M. (2012). Does eco-certification have (2010). Principles of organic agriculture. <http://www.ifoam.org/
environmental benefits? Organic coffee in Costa Rica. Ecological about_ifoam/principles/index.html> Accessed January 13, 2009.
Economics, 83, 58–66. International Fund for Agricultural Development (IFAD) (2003). The
Blackman, A., & Rivera, J. (2011). Producer-level benefits of sustainability adoption of organic agriculture amongst small farmers in Latin America
certification. Conservation Biology, 25, 1176–118585. and the Caribbean: Thematic evaluation Report no. 1337. Rome:
Bolwig, S., Gibbon, P., & Jones, S. (2009). The economics of smallholder IFAD.
organic contract farming in tropical Africa. World Development, 37, Jena, P. R., Chichaibelu, B. B., Stellmacher, T., & Grote, U. (2012). The
1094–1104. impact of coffee certification on small-scale producers’ livelihoods: A
case study from the Jimma Zone, Ethiopia. Agricultural Economics, 43, Rosenbaum, P. (2002). Observational studies (2nd ed.). New York:
429–440. Springer-Verlag.
Jordan, A., Wurzel, R., & Zito, A. (2003). ‘‘New” environmental policy Rosenbaum, P., & Rubin, D. (1983). The central role of the propensity
instruments: An evolution or a revolution in environmental policy?. score in observational studies for causal effects. Biometrika, 70, 41–55.
Environmental Politics, 12, 201–224. Ruben, R., & Fort, R. (2012). The impact of Fair Trade certification for
Laurance, W. F., Sayer, J., & Cassman, K. (2014). Agricultural expansion coffee farmers in Peru. World Development, 40, 570–582.
and its impacts on tropical nature. Trends in Ecology and Evolution, 29, Rueda, X., & Lambin, E. F. (2013). Responding to globalization: Impacts
107–116. of certification on Colombian small-scale coffee growers. Ecology and
Lygbaeck, A., Muschler, R., & Sinclair, F. (2001). Productivity and Society, 18, 21. http://dx.doi.org/10.5751/ES-05595-180321.
profitability of multistrata organic versus conventional coffee farms in Sebastian, S., & Schulz, R. (2015). Agricultural insecticides threaten
Costa Rica. Agroforestry Systems, 53, 205–213. surface waters at the global scale. Proceedings of the National Academy
Martı́nez-Sánchez, J. C. (2008). The role of organic production in of Sciences, 112, 5750–5755.
biodiversity conservation in shade coffee plantations (Ph.D. disserta- Smith, J., & Todd, P. (2005). Does matching overcome LaLonde’s critique
tion). Seattle: Department of Biology, University of Washington. of nonexperimental estimators?. Journal of Econometrics, 125,
Mendez, V. E., Bacon, C. M., Olson, M., Petchers, S., Herrador, D., 305–353.
Carranza, C., ... Mendoza, A. (2010). Effects of Fair Trade and organic Soares, A., Leão, M., Vianna Neto, M., & Oliveira, S. (2012). Risk
certifications on small-scale coffee farmer households in Central estimate of water contamination by pesticides used in coffee crops.
America and Mexico. Renewable Agriculture and Food Systems, 25, Revista Brasileira de Engenharia Agrı́cola e Ambiental, 16, 425–432.
236–251. Sterner, T., & Coria, J. (2011). Policy instruments for environmental and
Milder, J., Arbuthnot, M., Blackman, A., Brooks, S., Giovannucci, D., natural resource management. Washington, DC: Resources for the
Gross, L., ... Zrust, M. (2014). An agenda for assessing and improving Future Press.
conservation impacts of sustainability standards in tropical agriculture. Takahashi, R., & Todo, Y. (2013). The impact of a shade coffee
Conservation Biology, 29, 309–320. certification program on forest conservation: A case study from a
Ministerio de Agricultura. (2012). Consumo interno y exportación de wild coffee forest in Ethiopia. Journal of Environmental Management,
productos ecológicos (2008–2013). <https://www.minagricultura.gov. 130, 48–54.
co/tramites-servicios/Paginas/Sello-Ecologico.aspx> (Agricultura Eco- Tejeda-Cruz, C., Silva-Rivera, E., Barton, J., & Sutherland, W. (2010).
lógica). Why shade coffee does not guarantee biodiversity conservation.
Morgan, S., & Harding, D. (2006). Matching estimators of causal effects: Ecology and Society, 15, 13, [online] URL: <http://www.ecologyand-
Prospects and pitfalls in theory and practice. Sociological Methods and society.org/vol15/iss1/art13/>.
Research, 35, 3–60. Tilman, D., Fargione, J., Wolff, B., D’Antonio, C., Dobson, A., Howarth,
Muñoz, C., & Moreno, A. (2001). Potencial de tres comunidades R., ... Swackhamer, D. (2001). Forecasting agriculturally driven global
campesinas para caficultora orgánica. Cenicafé, 52, 289–302. environmental change. Science, 292, 281–284.
North, M. O., & Bell, D. D. (1984). Breeder management: Commercial United Nations Development Programme (UNDP) (2010). Coffee scoping
chicken production manual. Westport, Connecticut: AVI Publishing, paper Green Commodities Facility, Internal Working Document. New
240–21. York: UNDP.
Parrot, N., Olesen, J., & Høgh-Jensen, H. (2007). Certified and non- Valkila, J. (2009). Fair Trade organic coffee production in Nicaragua:
certified organic farming in the developing world. In N. Halberg, H. Sustainable development or a poverty trap?. Ecological Economics, 68,
Alrøe, M. Knudsen, & E. Kristensen (Eds.), Global development of 3018–3025.
organic agriculture: Challenges and promises (pp. 153–180). Walling- Van der Vossen, H. (2005). A critical analysis of the agronomic and
ford, Oxfordshire, UK: CAB International. economic sustainability of organic coffee production. Experimental
Pell, A. N. (1997). Manure and microbes: Public and animal health Agriculture, 41, 449–473.
problem?. Journal of Dairy Science, 80(10), 2673–2681. Villa, J. M. (2011). DIFF: Stata module to perform differences in
Philpott, S., Bichier, P., Rice, R., & Greenberg, R. (2007). Field testing differences estimation. In Statistical software components. USA: Boston
ecological and economic benefits of coffee certification programs. College Department of Economics.
Conservation Biology, 21, 975–985. Weber, J. (2012). Social learning and technology adoption: The case of
Quispe Guanca, J. L. (2007). Caracterización del impacto ambiental y coffee pruning in Peru. Agricultural Economics, 43, 73–84.
productivo de las diferentes normas de certificación de café en Costa Rica Wehrmeyer, W., & Mulugetta, Y. (Eds.) (1999). Growing pains: Environ-
(Master’s thesis). Centro Agronómico Tropical de Investigación y mental management in developing countries. Sheffield, UK: Greenleaf
Enseñanza (CATIE). Publishing.
Ramos Agüero, D., & Alfonso, E. T. (2014). Generalidades de los abonos Zwane, A. (2007). Does poverty constrain deforestation? Econometric
orgánicos: Importancia del Bocashi como alternativa nutricional para evidence from Peru. Journal of Development Economics, 84, 330–349.
suelos y plantas. Cultivos Tropicales, 35, 52–59.
Rappole, J. H., King, D. I., & Vega Rivera, J. H. (2003). Coffee and
conservation. Conservation Biology, 17, 334–336.
Rivera, J., de Leon, P., & Koerber, C. (2006). Is greener whiter yet? The
Sustainable Slopes Program after five years. Policy Studies Journal, 34,
195–224. APPENDIX A. SUPPLEMENTARY DATA
Robinson, D., & Mansingh, A. (1999). Insecticide contamination of
Jamaican environment: Transport of residues from coffee plantations Supplementary data associated with this article can be
in the Blue Mountains to coastal waters in eastern Jamaica. Environ-
found, in the online version, at http://dx.doi.org/10.1016/
mental Monitoring and Assessment, 54, 125–142.
Romero, C., Putz, F., Guariguata, M., Sills, E., Cerutti, P., & Lescuyer, j.worlddev.2016.01.004.
G. (2013). An overview of current knowledge about the impacts of forest
management certification: A proposed framework for its evaluation
Occasional paper 91. Bogor, Indonesia: Center for International
Forestry Research.
Available online at www.sciencedirect.com
ScienceDirect

Is Eco Certification A Win-Win For Developing Country

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Is Eco Certification A Win-Win For Developing Country

Uploaded by

Copyright:

Available Formats

World Development Vol. xx, pp.

Is Eco-Certiﬁcation a Win–Win for Developing Country

Key words — organic certification, coffee, Colombia, difference-in-differences matching

Table 1. Variables, deﬁnitions, means, and diﬀerence-in-means tests

Available online at www.sciencedirect.com

You might also like