The Impact of Agricultural Extension Services: The Case of Grape Production in Argentina

Abstract

In this paper we evaluate the impact of the provision of agricultural extension services to grape producers in Mendoza, Argentina, on yield and grape quality. Using fixed effects and matching techniques, we show that despite non-significant average treatment effects on yield, the program has large positive effects on productivity for producers who were in the bottom of the productivity distribution before launching of the program. There is also evidence of increased quality of their grapes, especially for large producers and those in the middle of the yield distribution ex-ante. However, large groups of producers did not see impact on yields or quality. Consistent with a previous qualitative evaluation of the program, these results point to the need to balance flexibility of the program with effective targeting mechanisms. Producers with different characteristics, such as land size, productivity or structure of production seek different objectives and have different needs, so that targeting these types of programs effectively to the different needs of producers would increase their effectiveness.

Full text

Page 1

The Impact of Agricultural Extension Services:
The Case of Grape Production in Argentina
Pedro Cerdán-Infantes, Alessandro Maffioli and Diego Ubfal
Inter-American Development Bank
Office of Evaluation and Oversight
Working Paper: OVE/WP-05/08
June, 2008

Page 2

Electronic version:
http://ove/oveIntranet/DefaultNoCache. aspx?Action=WUCPublications@ImpactEvaluations


Inter-American Development Bank
Washington, D. C.
Office of Evaluation and Oversight, OVE

The Impact of Agricultural Extension Services:
The Case of Grape Production in Argentina

Pedro Cerdán-Infantes, Alessandro Maffioli, and Diego Ubfal*





* Pedro Cerdan-Infantes, Office of Evaluation and Oversight, Inter-American
Development Bank, pedroce@iadb. org; Alessandro Maffioli, Office of Evaluation and
Oversight, Inter-American Development Bank, alessandrom@iadb. org; Diego Ubfal,
Department of Economics, University of California at Los Angeles, dubfal@ucla. edu

This paper is part of the project: “Ex-post Evaluation of the IDB’s Agricultural
Technology Uptake Projects”, by the Office of Evaluation and Oversight (OVE) of the
Inter-American Development Bank (IDB). We are grateful for the collaboration of the
IDB’s representatives in Argentina, the Program Coordination Unit (PCU) of the
PROSAP program and the Instituto Nacional de Vitivinicultura (INV). We are also
grateful to Paul Winters and Jeffrey Brown for very useful discussions and comments
on previous versions of this paper. The findings and interpretations of the authors do
not necessarily represent the views of the Inter-American Development Bank. The
usual disclaimer applies. Correspondence to: Alessandro Maffioli, e-mail:
alessandrom@iadb. org, Office of Evaluation and Oversight, Inter-American
Development Bank, Stop B-760, 1300 New York Avenue, NW, Washington, D. C.
20577.

Page 3

 
ABSTRACT

In this paper we evaluate the impact of the provision of agricultural extension
services to grape producers in Mendoza, Argentina, on yield and grape quality.
Using fixed effects and matching techniques, we show that despite non-
significant average treatment effects on yield, the program has large positive
effects on productivity for producers who were in the bottom of the productivity
distribution before launching of the program. There is also evidence of increased
quality of their grapes, especially for large producers and those in the middle of
the yield distribution ex-ante. However, large groups of producers did not see
impact on yields or quality. Consistent with a previous qualitative evaluation of
the program, these results point to the need to balance flexibility of the program
with effective targeting mechanisms. Producers with different characteristics,
such as land size, productivity or structure of production seek different objectives
and have different needs, so that targeting these types of programs effectively to
the different needs of producers would increase their effectiveness.


JEL Codes: Q12, Q16, H43

KEYWORDS: Technology Adoption, Productivity, Agriculture Sector, Policy
Evaluation

Page 4

 
 
1
INTRODUCTION
Assisting farmers to become more productive has long been common practice in
development assistance, as well as the objective of ministries of agriculture and
sub-national governments around the world. Agriculture is still the main
economic activity for the majority of the world’s population and it is generally
correlated with low socioeconomic indicators, such as education, health and
poverty. In fact, as noted the World Development Report 20081, three out of poor
people in developing countries live in rural areas, and most of them depend on
agriculture for their livelihoods. In addition, agricultural products, whether in
bulk or processed form, are common exports in developing countries, and many
people depend on their production as their main economic activity. However,
while the objective of improving farmers’ productivity and agricultural workers
livelihood is common to numerous types of agriculture or rural development
programs, the approaches vary significantly by type of program. The most
common interventions include infrastructure development, market access, the
provision of fertilizer or other inputs, or provision of extension services to
producers, among others.
In this paper, we focus on extension services, and particularly a program in
Mendoza, Argentina that provides extension services to grape producers.
Extension services generally aim at transferring specific knowledge to producers,
such as the transfer of technology, the improvement of management practices or
the transfer of knowledge and capacities2. The provision of these services take a
wide range of forms (Purcell and Anderson, 1997), including availability of
occasional assistance by specialists when demanded by producers, formal
trainings on specific topics for groups of producers, or specialists working
directly with farmers. In addition, there are variations in the type of financing and
service provision, sometimes including co-payments or fees, and provision of the
services by public officials or private companies.
This study focuses on the PROSAP (Programa de Servicios Agricolas
Provinciales), which seek “to increase the value of agricultural exports by
improving quality and increasing production volumes3 by financing extension
services on key topics in grape producing. The services were free, publicly
financed and provided by consultants hired by the program. The services were
offered to all grape producers4, who had the option to receive the services. Since
                                                      
1 Agriculture for Development, World Development Report 2008, World Bank.
2 Similar definitions are found for example in Birkhaeuser et al. (1991), Evenson (2001), and
Owens et al. (2001).
3 From IDB AR-0061 Loan Document, page 10.
4 There were very few restrictions on eligibility, as we discuss in Section 3.

Page 5

 
 
2
the stated objectives of the program are to improve quality and increase
production, we focus our evaluation of its impact on these two outcomes, yield
(our measure of productivity) and sugar content of the grapes (a measure of
quality).
The results indicate that the program was effective at increasing yields for those
with low yields before participation, and increased the average quality of grapes
produced by beneficiaries. Our approach uses a combination of fixed effects and
different matching methods to isolate program effects, and provide evidence that
the program did have a positive and significant impact on productivity and
quality, though the effects differ widely by pre-program characteristics of the
producers. There is, however, no evidence of significant differences in
introduction of new varieties or switch in the production structure of
beneficiaries.
The paper presents first some theoretical justifications for the provision and/or
financing of extension services and existing evidence on the impact of such
programs in Section 2. Section 3 describes the extension services program
(PROSAP) in Mendoza, the results of a previous qualitative evaluation and the
characteristics of the beneficiaries at baseline. Section 4 describes the data, and
Section 5 presents the identification strategy including the different matching
methods used in the evaluation. Finally, Section 6 presents the results and
Section 7 concludes and provides some policy recommendations arising from the
analysis.

AGRICULTURAL EXTENSION SERVICES, THEORY AND EVIDENCE
The provision of agricultural extension services has been justified in the literature
on both equity and efficiency grounds. In the presence of market failures, for
example externalities, limited access to credit or non-competitive market
structures, producers will not face the correct incentives to produce certain
varieties, use new production techniques or adopt new technologies, resulting in
production levels that are not socially optimal5. In addition, if less advantaged
farmers are more exposed to these failures because of their limited resources
(lack of market power in olygopsonic markets, limited access to credit, low
capacity to pay for extension services6), the justification for solving these market
                                                      
5 See, for example, Hanson and Just (2001) and Feder et al. (2003).
6 See the World Development Report 2006, to see how inequality might create poverty through
limitations in market access.

Page 6

 
 
3
failures through public intervention gains relevance under both equity and
efficiency arguments.
Agricultural extension services seek to solve some of these inefficiencies,
through training and provision of technology for beneficiaries. The question is
why if producers are inefficient and extension services are effective, a private
market for extension services is not rampant in rural areas. The literature
identifies three main reasons for the absence of such developed markets. First,
services provided by these programs are non-rival, such that the use by one
farmer does not preclude other farmers from using it. In addition, they are non-
excludable, given the difficulty to limit the transfer of knowledge or technology
if effective. Finally, if services are non-rival and non-excludable, the
appropriability of the benefits of such programs are low, since a successful
technology would be immediately adopted (copied) by competitors, eliminating
the profits for the first adopter. Thus, the incentives to demand these services for
a positive price are low, limiting the development of these markets7.
Under these circumstances, public intervention would provide large positive
externalities if it directly targets these inefficiencies. However, even if public
intervention is necessary, the question still remains whether the government
should provide the services itself or instead set the institutional structure to create
a market and finance private providers for the services. In addition, despite the
existence of appropriability issues and externalities, the services provided are
expected to increase the value of production, and as such these services have a
monetary value that could be charged to the beneficiaries. However, the
incentives to set or accept market prices are complex, due to asymmetric
knowledge of the value of the service between government (principal) and
private provider (agent), the heterogeneity of beneficiaries and the uncertainty of
agricultural activities, and, as a consequence, the difficulty to create outcome-
based payments that create the right set of incentives between the government
and the provider. Together with the high cost of monitoring private providers
(remoteness & sparseness of services), the availability of information for
effective targeting by the public sector, might justify public financing and
provision of services in some settings.
Finally, if governments are generally concerned with equity considerations, even
in the absence of coordination failures or other market imperfections, public
intervention might be appropriate on equity grounds if targeted to low-income
farmers. In addition, and since poor farmers tend to be, among other things, less
educated, which combined with their lower capacity to pay for extension services
                                                      
7 Hanson and Just (2001).

Page 7

 
 
4
would result in large productivity differences based on socioeconomic status, the
efficiency and equity arguments would both hold for support of low income
farmers with the provision of extension services.
Evidence
Despite the theoretical considerations for public intervention through the
financing and/or provision of extension services, the existing evidence of their
effectiveness is scarce and inconclusive, partly due to the few rigorous impact
evaluations undertaken until now. In addition, these evaluations fail to address
questions on the effectiveness of new modalities of extension programs around
the world (e. g. , privatization, fee-for-service, decentralization, etc. ). As Alex
and Rivera (2005) point out after reviewing 44 case studies, reforms in extension
services include decentralization, privatization, demand-driven approaches, new
approaches for public services and national strategies. In addition, Hanson and
Just (2001) delimit five categories of extension under which those reforms can be
understood, including: i) public extension with public funding and traditional
delivery, ii) paid public extension with public provision, with a fee-for-service
funding, iii) partially public-funded private extension, delivered by private firms
and financed in part by public budgets and in part by user fees, iv) policy-
supported private extension, provided by firms and financed by users, but with
government subsidies or taxes for specific production techniques, and v) private
extension, provided by private firms that charges fees for their services. In this
section we review the evolution of the reforms and analyze their logic on the
basis of the arguments developed in the previous sections. Unfortunately, the
evidence on the comparable effectiveness of these five models is non-existent.
The main body of research on the effect of extension services on producers is
based on production functions using the extension variable as one of the inputs,
showing large positive rates of return to extension services (Birkhaeuser et al.,
1991). However, in the absence of random assignment to treatment and control
groups, this methodology is likely to provide biased estimates of causal effects,
due to endogeneity of program participation and the presence of unobservable
characteristics that might determine participation and be correlated with outcome
variable. In addition, as pointed out in Dinar et al. (2007), this approach assumes
that farms operate at technically efficient levels, an assumption that is disputed
by evidence of inefficiencies in production that serve as a justification for the
provision of public extension services. This led to a more sophisticated method,
based on stochastic frontier models, which relax the efficiency assumption by
defining a “best practice” frontier production function and measuring the distance

Page 8

 
 
5
from this frontier for each producer8. However, stochastic frontier models suffer
from the same identification problems if treatment is not random and includes
unobservable characteristics.
In demand driven programs, as is the case of PROSAP, when farmers initiate the
contact with extension agents on the basis of their own personal characteristics
and necessities, the treatment coefficient in production function estimations will
include unobservable characteristics that determine both participation and the
outcome of interest. The sign and importance of the bias will depend on these
unobservable characteristics. For example, if farmers who demand extension
services were more motivated, we would expect the production function
regression to overestimate the returns to extension services. Other examples of
selection bias might come from program implementation, such as if extension
agents are more prone to contact more educated productive farmers, or if the
programs are devoted to regions where there is more responsiveness to extension
services.
The results of the few pseudo-experimental evaluations generally show positive
results, though not uniform. Duflo & Kremer (2003) in the only randomized
experiment to date show that farmers do not use the optimal amount of fertilizer
due to risk aversion (hyperbolic discounters). Godtland et al. (2004) estimate the
effect of a farmer field school program and a traditional extension program on
farmers’ knowledge of integrated pest management practices, using both a
regression with controls and matching techniques, and show significant positive
effect of both programs, but of higher value for the farmer field school program.
With regards to the farmer field school approach, Feder et al. (2003) use a
modified differences-in-differences model and find no impact on yields or on the
reduction of pesticide use for Indonesia. Their modified model accounts for the
fact that the program was introduced at different times across villages, taking into
account that the diffusion of innovation would reduce the impact of extension
with the years. On the other hand, Praneetvatakul and Waibel (2006) claim that
two time-point observations are not enough to estimate the impact of this kind of
programs. They use a panel of four years that comprises eight rice-growing
seasons in Thailand and find a positive impact on knowledge and pest
management practices both on the short and long run. Nevertheless, in a
companion paper they do not find any impact of the program on rice production
yield.
                                                      
8 See Aigner et al. (1977) for the seminal paper on the literature. Coelli et al. (1998), Kumbhakar
and Lovell (2003) can also be consulted for a theoretical discussion of its options. For empirical
applications see, among others, Bravo-Ureta and Evenson (1994) and O’Neill et al. (1999).

Page 9

 
 
6
Several works using panel data have been carried out since then, but the number
is limited in comparison when compared to other areas of development
economics. With fixed effect or difference-in-difference models, they control for
farms’ and farmers’ time-invariant unobservable characteristics. Owens et al.
(2003) and Romani (2003) estimate the impact of traditional extension services
using panels of farmers for Zimbabwe and Ivory Coast respectively9. Both
studies find a positive impact of extension services on productivity and yields.
However, they note that this impact is neither present for all the years nor for all
the crops studied.
Another body of literature has used instrumental variables to identify program
effects, finding generally positive though small and heterogeneous program
effects. However, the instruments and assumptions are at least questionable. As
Romani (2003) and Feder et al. (2003) point out, finding a variable correlated
with the participation in extension programs but not with the studied outcome is
not an easy task since by program design the criteria used to select farmers for
extension services are usually correlated with the outcome. They argue that the
only option is to rely on strong distributional assumptions such as the joint
normal distribution of errors10. Akobundu et al. (2004), using as instruments the
distance from extension office, whether an individual was rejected a loan, total
farm debt, and the previous visit of an extension agent (not of the program) find a
positive impact on farm income only for individuals with a high number of visits.
The heterogeneity of results is also significant: farmers with better education,
more skills and wealth are more likely to adopt certain kind of innovations that
are dependent on knowledge. Wealth and size of the farm can have similar
effects on extension thorough the adoption rate. This is in accordance to the
results of Godtland et al. (2004) who find that the farmer field school approach is
effective for wealthier farmers.
In summary, the evidence on the impact of extension services on farmers
productivity and technological adoption is generally positive, though it generally
suffers from biases resulting from endogenous placement and omitted variable
bias. The results of pseudo-experimental studies highlight the heterogeneity of
program impacts based on farmer’s characteristics, such as education, experience
and wealth. Alston et al. (2000) review an important number of evaluations, with
the majority dedicated to agricultural research, and find a median rate of return of
58%. Evenson (2001) describes rates of return between 5 and 50% for
developing countries, but also notes that impacts vary widely. These results point
                                                      
9 Owens et al. (2003) also control for time effects, use clustered standard errors at the village level
and count with a measure of farmers’ skills.
10 With several periods of data a dynamic panel data estimator could be used, and further
differences or the moments of the distribution could act as appropriate instruments.

Page 10

 
 
7
to the specificity of impacts depending on the specific design of the program,
especially as it relates to the methodology for selecting beneficiaries.

THE PROSAP PROGRAM
The program object of this evaluation consists of the provision of free services11
to grape producers in the Mendoza province in Argentina. The provincial
program was part of a larger national program approved in 1995, designed to
support agricultural services, PROSAP (Programa de Servicios Agricolas
Provinciales), seeking “to increase the value of agricultural exports by improving
quality and increasing production volumes12. ” This broad objective was to be
achieved through a combination of interventions, including, for example, better
administration of water resources, providing basic agricultural infrastructure or
monitoring and ensuring both animal and plant health. Among them was the
program evaluated in this paper, the provision of “extension services” to
encourage the adoption of better technologies and methods of production,
providing integrated agricultural information services and strengthening the
programming capacity of the beneficiary provinces.
Extension Services Program Description
The program targeted one the factors which, according to the IDB’s loan
document, was contributing the inefficiency of grape production in Mendoza: the
lack of adequate technical services for grape producers13. The services were
publicly financed but provided by private agricultural specialists, contracted by
the program. There was no fee for service, though up-to-date payments on water
for irrigation fees were required for participation. The program provided farmers
with technical advise on production processes, especially on the use of variable
inputs (including water), with the stated objectives of increasing the efficiency of
production and improving the quality of the grapes produced14. The services were
promoted by inspectors of the irrigation system and consisted of periodic
meetings on different topics related to grape production. The menu of topics from
which groups of participants could choose included 8 topics, among others, the
                                                      
11 According to a previous evaluation, the project originally planned for co-payments from
beneficiaries in some departments that would increase in time till reaching 100% by the 4th year.
However, this co-payment was never materialized in practice, and the services were considered
free.
12 From IDB (1995), page 10.
13 IDB (1995), Chapter 1, starting page 6.
14 From IDB’s loan document.

Page 11

 
 
8
use of fertilizers, irrigation, pruning, use of different machinery or fitosanitary
plans.
Selection of Beneficiaries
A previous qualitative evaluation of the program, commissioned by PROSAP
and conducted in 200615, noted that the selection of beneficiaries had no clear
targeting mechanism and both the objectives of beneficiaries upon joining and
their perception of the impact of the program after participation varied widely
depending on the characteristics of the beneficiaries. Program promotion was
done for all producers, who were instructed to create groups of 12-14 producers
who would receive the services together. The result of the lack of clear targeting
mechanisms was a large heterogeneity among these groups, both in terms of the
characteristics of the producers and their objectives upon entering the program.
The lack of clear targeting rules makes the program difficult to evaluate for two
reasons. On the one hand, the objectives of the participants varied widely. The
main needs for producers outlined in the qualitative evaluation, which were
gathered from conversations with different beneficiaries, included increasing the
productivity of their vineyards, improving the quality of their grapes, introducing
new varieties, and improving market access and increasing efficiency of
production by reducing costs (better water management being a key objective).
The evaluation pointed out that producers within groups were largely
homogenous, which allowed them to broadly categorized the groups in
“established producers seeking access to markets and industrialization”,
“experienced producers with limited resources seeking to improve production
methods” and “less experienced producers seeking basic advice on production
methods”16. However, the beneficiaries note that the menu of options often did
not serve their needs, especially regarding access to markets.
Since the program was not targeted to specific types of producers and offered a
wide arrange of topics for services, we cannot identify the specific goals of each
beneficiary upon entering the program. Since according to the qualitative
evaluation the objectives of these groups varied widely, to the extent that the
impact of the program was small or not significant on objectives not sought by
participants the estimated average treatment effects are likely to understate the
true program impacts by including producers in the beneficiary group that did not
seek that particular objective. Though, as we will see, our methodology considers
these issues by dividing producers into different categories according to their pre-
                                                      
15 Evaluación del Programa de Servicios Agrícolas Provinciales, Martinez and Posada (2006).
16 Note that we do not have individual information for producers categorized in these groups, so we
cannot use individual information on these categories in our empirical analysis.

Page 12

 
 
9
program observable characteristics, we cannot explicitly control for the
objectives of the participants due to lack of information.
On the other hand, since groups were formed voluntarily, we would expect the
characteristics of beneficiaries and non-beneficiaries to be different. Since these
include both observable and unobservable characteristics that are likely to be
correlated with the outcomes of interest, simple OLS estimations are likely to
produce biased estimates of the program impact. The bias from these
characteristics is unclear, and it will be our goal to eliminate most of this bias
using fixed effects and matching techniques, as we discuss in the methodology
section.
DATA
Since the program did not collect information on outcome variables, this
evaluation relies on secondary data sources, which were difficult to gather17. We
gained access to restricted administrative datasets that allowed us to evaluate the
program18. The main source of information is the National Institute of
Vitiniviculture (INV for its name in Spanish), which compiles yearly information
on production and basic input use for all grape producers in the region of
Mendoza. We have access to annual data for every producer (8,873 in total) in
the region starting in 2002, before the inception of the program, and ending in
2006. These data include production for each variety of grape19, total land size
and productive land, and some data on input use (type of irrigation system, for
example). Though most of the grapes from producers in INV’s dataset have as
destination the production of wine, about 25 percent of the production goes to
table grapes for direct consumption. However, producing more than one variety
of different quality levels is usual practice, and only 10 percent of producers have
their entire production in table grapes. Thus, most of the producers have grapes
for wine production. Data for beneficiaries was provided by the coordination unit
of the program, and included a unique producer identifier number which allowed
the identification of beneficiaries in the INV dataset. The result is a rich panel
dataset of all producers that allows us to identify beneficiaries.
                                                      
17 Note that the Loan Document listed no specific outcome indicators for the project other than
“production” and non-measurable indicators such as “producers show knowledge of the appropriate
techniques of production” (IDB (1995), page 9, Annex III-1). The Project Performance Monitoring
Report largely reports on the completion of activities and program outputs (people trained), but not
on the proposed outcome variables (IDB (2006) PPMR).
18 These data were managed under strict confidentiality by an institutional agreement between the
INV and the IDB, under the guidance of OVE.
19 There is a total of 107 varieties in the dataset, but 14 varieties concentrate 90 percent of
production. We categorize in highest quality, high quality and low quality following INV’s
classifications.

Page 13

 
 
10
There are, however, some limitations to the data provided. First, there is no
demographic information about the producer, such as age, level of education or
experience in grape production. Though we can control for all time-invariant
characteristics20 by including producer fixed-effects, this lack of information on
the producer does not allow us to explore differential effects of the program by
demographic characteristics, which, as we saw in the literature review, have
been shown to be determinant in other programs (for example, more educated vs.
less educated producers). In addition, the dataset does not have information on
specific fertilizer or water use, two of the main inputs for production, which
makes it impossible to observe short-term changes in production practices that
might lead to lower costs, a potential objective of the program. Lastly, we do not
have information on the specific menu of topics received by each beneficiary.
Finally, we lack information on prices received by the producer, but instead we
obtain yearly average prices for varieties that are traded in the Mendoza
centralized market, which are higher quality varieties. Varieties not traded are
those of less quality. Thus, as a proxy, we make the assumption that lower
quality varieties have a lower price than the higher quality varieties with the
lowest price, and thus assign the same price to all low quality grapes, the
minimum price of higher quality varieties. Since we lack price information for
about 50 percent of the production, we are forced to make this conservative
assumption in order to evaluate the impact of the program in monetary value. In
practice, we are probably overestimating the price for these grapes and thus
assigning a higher value than low quality producers are likely to get. However,
since not all varieties are traded in that market and we cannot differentiate prices
received by different producers, we use these prices to monetarize the production
as a means to homogenize the measure of production (since better quality grapes
with higher prices might have different yields, money is a better measure than
weight to compare different producers) than to consider program impacts on the
value of production driven by quality.
Baseline Characteristics
The program benefited 372 producers in the 2003-2006 period, though we only
have complete information for 311 of them. As we can see in Table 1, the
program take-up was slow at the launching of the program in 2003, but the
number of beneficiaries increased notably in 2005 and 2006. According to the
previous evaluation of the program, this increase was a result of a greater
                                                      
20 In addition to obviously time-invariant characteristics such as gender or location, fixed-effects
control for characteristics such as age and experience, to the extent that they vary uniformly with
time for all producers. In addition, assuming that most producers do not increase their stock
education, fixed-effects also controls for education.

Page 14

 
 
11
emphasis on program promotion. Geographically, the program targeted
beneficiaries in three departments within Mendoza, Junin, Rivadivia and San
Martin.
Table 1. Number of Beneficiaries and Non-Beneficiaries by Department, 2002-2004
20022003
Department
Junin
La Paz
Lavalle
Rivadavia
San Martin
Santa Rosa
BenefNo BenefBenefNo BenefBenefNo BenefBenef No BenefBenefNo Benef
0
0
0
0
0
0
1695
54
1256
1678
3190
697
3
0
0
0
1620 401583
52
1170
1569
3028
668
1101547
54
1206
1520
3010
676
1141560
54
1231
1542
3014
690
510
2
0
2
0
21152
1594
3034
673
27
51
103
108
112
14135
0011
Total
Note: “Beneficiaries” are those that participated in the program at any point since inception, so the number of
beneficiaries should be thought of as cumulative. Source: Own calculations using INV dataset.
08570 388124 1208070 3248013 3708091
200620042005

As we noted in the previous section, arising from the previous evaluation and the
discussion on the selection of beneficiaries we expect that i) beneficiaries differ
significantly from non-beneficiaries, and ii) there is a large heterogeneity in the
characteristics of beneficiaries.
The differences between beneficiaries and non-beneficiaries are evident in Table
2, which presents summary statistics of key variables in 2002 for producers that
would later benefit from the program and those that never participated.
Beneficiaries have less land (12. 4 ha vs 17. 7 ha of non-beneficiaries), but have
levels of production that are not significantly different from non-beneficiaries. As
a result, they are about 30 percent more productive before participation than
producers that never received the services, as measured by yield (total
production/land size).
There are, however, no significant differences in the broad types of grapes
produced. The percentages of production in the highest quality varieties (“fina”)
and the high quality varieties (“especial”) are the same. In addition, grape
quality, as measured by sugar content, was not significantly different between
participants and non-participants. On input use, we see some differences in the
percentage of land covered by the different types of irrigation systems, with
beneficiaries having a slightly larger extension of their land covered by irrigation
systems (31 vs 26 percent covered by reservoir, for example). There are no
significant differences, ex-ante, on the use of drop irrigation, considered to be the
most sophisticated system and rarely used by grape producers in Mendoza in
2002.

Page 15

 
 
12
Table 2. Baseline Characteristics of Beneficiaries and Non-Beneficiaries
PROSAP
N Mean StdN MeanStdDif-Means
t-stat
Outcome variables
Total Land Size (ha)
Productive Land (ha)
Production (kg)
Yield (kg/ha)
Yield ($/ha)
% production in highest quality varieties
% production in high quality varieties
% production in wine varieties
% production in table wine varieties
Sugar content
344
344
344 100,019.7 (156229.1)
34413,033.7
344 10,293.5
300 0.2
3000.3
3000.6
3000.4
300222.0
12.4
8.1
(26.52)
(11.70)
7229
7229
7229
7207
7207
5839
5839
5839
5839
5839
17.7
10.0
(47.64)
(17.24)
-5.31
-1.88
-2291.60
2632.33
2029.46
0.03
0.00
0.00
0.00
-0.43
3.46
2.84
0.26
-2.51
-2.63
-1.54
-0.15
0.22
-0.05
0.73
102,311.3 (226948.6)
10,401.3
8,264.0
0.2
0.3
0.6
0.4
222.5
(19063.8)
(14029.8)
(0.35)
(0.36)
(0.37)
(0.37)
(9.80)
(16768.84)
(12995.6)
(0.31)
(0.36)
(0.38)
(0.38)
(10.88)
Input variables
% of land with reservoir irrigation
% of land with aspersion irrigation
% of land with superior irrigation
% of land with irrigation turno
% of land with drop irrigation
% of plantation with anti-hail
Number of Reservoir
Reservoir Capacity
Harverst machine, own
Harvest machine, rents
Harvest machine, doesn't use
Number of tractors

344
344
344
344
344
344
344
344
344
344
344
344
0.31
0.00
0.79
0.65
0.00
0.01
0.07
6,119.0
0.00
0.00
0.99
0.71
(0.39)
(0.00)
(0.26)
(0.36)
(0.05)
(0.09)
(0.25)
(107836.4)
(0.00)
(0.05)
(0.08)
(0.60)
7229
7229
7229
7229
7229
7229
7229
7229
7229
7229
7229
7229
0.26
0.00
0.76
0.62
0.01
0.01
0.09
37,992.7
0.00
0.00
0.97
0.66
(0.38)
(0.01)
(0.30)
(0.38)
(0.07)
(0.07)
(0.35)
(567360.4) -31873.74
(0.03)
(0.04)
(0.16)
(0.79)
0.05
0.00
0.03
0.03
1.05
0.00
-0.03
-2.39
2.26
-1.95
-1.39
0.00
-0.95
1.84
3.60
0.00
0.00
0.00
-1.50
2.24
-0.33
-4.79
0.05
Full Sample, Non-Beneficiaries

In order to explore the heterogeneity of producers who become beneficiaries of
the program, we present in Figure 1 the density function for key variables
(production, area, yield and sugar content) at baseline, for producers who became
beneficiaries in later years and those who did not. As is evident from the figures,
the variability in the characteristics of beneficiaries is at least as large as that of
non-beneficiaries in all 4 variables, as shown by distributions of the density
functions. In fact, beneficiaries seem to have a wider distribution in yield, both in
quantity and monetary value, and sugar content (a proxy of grape quality). It is
important to note that yield and quality, the two stated objectives of the program
at inception, vary significantly among participants. Among beneficiaries we find
producers both at the top of the yield distribution and at the bottom. Similarly, we
find producers at the highest levels of sugar content, pointing to potentially
different objectives of producers upon joining the program.