You are on page 1of 22

“Agricultural Input Market Segmentation in Argentina:

How do Argentine farmers buy their expendable inputs?


The Case of the Seed Industry”

Feeney, Roberto (robertofeeney@gmail.com)


Associate Professor, Ph.D. in Business Administration
Austral University Argentina

Berardi, Valeria (valeriaberardi@hotmail.com)


Professor, Master in Applied Statistics,
Austral University Argentina

Steiger, Carlos (carlos.steiger@hotmail.com),


Director of the Center of Food and Agribusiness, Ph.D. in Business Administration,
Austral University Argentina

Center for Food and Agribusiness, Austral University Argentina


1950 Paraguay Street, City of Rosario, Province of Santa Fe, Argentina
Telephone: 54-341-522-3000
May 23rd, 2011

Symposium Paper Submission


IFAMA Conference, Frankfurt, June 2011

1
Abstract and Key Words

Abstract: In this paper we analyze the buying behavior of farmers for expendable
inputs. In particular, we will study the case of the seed industry in Argentina. We
segment them using cluster analysis, identifying 4 distinctive segments of farmers for
seed purchasing: performance, price, balance, and convenience. We also use a
multinomial logit model to predict Argentine farmers’ segment membership based on
key demographic, educational, commercial, and informational features.
Key Words: Expendable agricultural inputs, Seed markets, Cluster analysis,
Multinomial Logit Regression Model.

2
Executive Summary
Argentine agribusiness sectors have been moving forward very fast in the past fifteen
years, with agricultural production growing from less than 50 million tons in the mid
90’ to almost 100 million tons now-a-days, and with agro-industrial exports incomes
that have more than doubled. This has pushed the agricultural input demand
enormously, as for example the fertilizer market jumped from 1.2 million tons a year to
more than 3.5 million tons. The total expenditure of the agricultural sector is more than
17 billion dollars, with a market for expendable inputs such as seed, fertilizers, and
agrochemicals of more than 3.5 billion dollars. The seed market alone is worth around 1
billion dollars.
In this context in this work we study how Argentine farmers buy their seed inputs, and
how to segment these producers according to their purchasing behavior for seeds.
Previous works have explained this for the total expendable agricultural markets in
other countries such, as the US. However, up to now there has been no work done
intending to explain how input markets function in Argentina, and how farmers buy
their seed. Additionally, we envisioned to predict farmers’ segment membership
according to their observable characteristics, which would be useful for marketing
purposes in agricultural input companies.
In order to answer the research question we used data from a survey called the “The
Need of Argentine Farmers”, done in the second half of the year 2009 by the Center for
Food and Agribusiness of the Austral University in Argentina. The universe under study
were the farmers in the main agricultural area of Argentina (“Humid Pampa”) which
produce 750 or more annual tons in soybeans. The total universe was formed by 7,400
producers, which produce 70% of the total soybean in the main crop area of Argentina.
The sample was formed by 502 farmers producing soybeans, corn and wheat
responsible of farms with owned or rented land.
The main results of this work are that Argentine farmers can be segmented in four
clusters according to their seed buying behavior: performance, price, balance and
convenience segments, with different profiles. While farmers in the performance and
balance segments would be business purchasers, the ones in the price segment are cost-
oriented. The convenience farmers, on the other hand, are those who prioritize location
and convenience in their purchases, young and very educated small farmers who want to
grow fast. We also built a regression model provided with strong predictive power to
predict cluster membership according to the observable behavior of farmers as
customers of the input firms. Here we were able to establish six significant relationships
which would explain the odds of a farmer becoming a member a segment as the values
of his observable behavior increases, or the lesser likelihood to become a member to
another segment due to the negative values of his observable conduct.
Overall, this work intends to provide some answers of how farmers purchase their seed
in Argentina, how they can be segmented, and how to predict their input purchasing
membership that better fits according to their observable behavior. It has academic
value as very little has been written about these markets for Argentina, as well as
empirical value as it can be useful for marketing managers working in agricultural input
markets.

3
Introduction
Agricultural input markets for products such as seeds, crop protection, and fertilizer are
important markets in Argentina. Although it varies greatly from year to year,
expendable input purchases by Argentine producers can amount for several billion
dollars a year: These markets for seeds, crop protection and fertilizer inputs were worth
around 3.5 billion dollars for the year 2010. For the seed case, the total informal and
formal seed market for major crops in Argentina is about one billion dollars a year: It
would be around 400 million dollars for soybean seeds, 400 million dollars for corn
seeds, and 200 million dollars for wheat and other seeds. (Vilela et al., 2009; FIS, 2010,
CREA 2010)
The market for seeds in Argentina is divided in two: The self-pollinated/self-fertilizing
crop markets, such as soybeans and wheat; and the hybrid crops markets, such as corn
and sunflower. In the first type of markets there are serious intellectual property issues
as 70% of the seeds are sold through informal markets, by which there are no easy ways
for seed companies to make profits. Here in these markets there are present few
international companies and there are more local ones. In the hybrid markets, on the
other hand, there are no such intellectual property issues, firms can capture value, and
the main players are international firms. Incomes would be split half and half between
these two types of markets: Roughly, 500 million dollars each, although the hybrid
markets would be more profitable. (Vilela et al., 2009; USDA-FAS, 2011)
However, Argentine farmers are not a homogeneous group, nor buy seeds in the same
way: They differ in terms of key dimensions such as farm size, educational background,
age, location, land tenure, attitudes, risk management practices, technology adoption,
and so forth. Grouping farmers by more homogeneous classes, in terms of segmenting
farmers and defining their profiles are important issues for agricultural input companies
in order to define their marketing strategies. In this paper we will try to define how
argentine farmers buy their seeds for crops, segmenting these farmers in different
classes with different purchasing profiles.
Marketing segmentation helps firms define particular marketing mix strategies that
enable them to target customers with specific profiles and needs in each segment. This
results from the fact that rarely customers in a particular market have exactly the same
needs and expectations. By segmenting their customers companies can get closer to
each customer by developing an appropriate marketing mix (Kotler, 1997).
Previous work has been done to segment farmers who buy agricultural inputs, especially
for the US (Hooper, 1994; Bernhardt, Allen, Helmers, 1996; Gloy and Akridge, 1999;
Foley, 2003; Alexander, Wilson and Foley, 2005; Reimer, Downey, Akridge, 2009).
However, there is not much work done for Argentine agricultural input markets, in spite
of the importance of these markets, as we have explained above.
There are some differences, though, with what previously has been studied in the US
and what we intend to do: While they have focused on larger farmers, we will also
target mid-size producers as well. Also, we will deal with famers that produce soybeans,
corn and wheat; while previous author have studied farmers in a wide range of different
crops.
Accordingly, the problem we want to study is how Argentine farmers buy their seed
inputs, and how to segment these producers in order to understand better their

4
purchasing behavior for seeds. We will restrict the study to Argentine farmers in the
geographic area of the ‘humid pampas’ (which is equivalent to the US Corn Belt) that
produce more than 750 tons of soybeans a year.
In this way, the main goal of this paper is to identify distinctive market segments for
argentine farmers purchasing seeds. The idea is to segment farmers into buying
characteristics according to their purchasing behavior, and to be able to predict farmers’
segment membership. This will allow us to answer the problem of how Argentine
producers purchase their agricultural inputs.

Data
The data we use to segment the farmers’ input markets is based on the survey on “The
Need of Argentine Farmers”, done in the second half of the year 2009 by the Center for
Food and Agribusiness of the Austral University in Argentina, with the partnership of
the University of Purdue in the US 1, and the help of the Rosario Stock Exchange of
Argentina. This survey was done between August 17th and September 17th 2009,
through personal interviews in the farms, surveying 502 producers.
The universe under study were the farmers in the main agricultural area of Argentina
(“Humid Pampa”) which produce 750 or more annual tons in soybeans. This covers the
provinces of Santa Fe, Córdoba y Buenos Aires. It includes the counties in which the
sowing area represents more than 10% of the total production area. The total universe
was formed by 7,400 producers, which produce 70% of the total soybean in the main
crop area of Argentina. The sample was formed by 502 farmers (producing soybeans,
corn and wheat) responsible of farms with owned or rented land, with a degree of
statistical confidence of 95%.
Surveyed farmers were heads of farms (owned or leased properties) located in selected
departments (counties) of the Provinces of Buenos Aires, Santa Fe and Córdoba with a
production greater than 750 tons of soybeans per annum (year 2008); 70% of their
income came from soybean and the rest (30%) from other crops.
Accordingly these farmers located in above defined departments (counties) in the main
cropland of Argentine that produced more than 750 tons of soybeans per annum were
divided into medium size farmers (250-500 hectares), commercial size (601-1840
hectares) and large (more than 1840 hectares). In annex 1 we show the segmentation of
the sample`s population.
The survey was done interviewing farm operators responsible for the farmers they
manage, by a team of professional people, with personal interviews for each of the 502
farmers. The questionnaire had 37 questions, and took around 60 minutes to answer.
Only one was an open question, 29 were closed questions, and seven were semi-
structured questions. The sample data we gathered in this survey was cross-checked
with last National Agricultural Census available in Argentina, from 2002, by which the
results of this survey are statistically significant for producers of more than 750 tons a
year for soybeans, in the in the main agricultural area of Argentina (“Humid Pampa”).

1
The University of Purdue did a similar work “Serving Producers in Volatile Times” (2008), on which the
Argentine survey is based.

5
Methodology
Following Gloy and Akridge (1999) and Alexander, Wilson and Foley (2005), we will
use cluster analysis to segment the seed input markets. The goal of cluster analysis is to
divide a data set into different groups or clusters, based on observed attributes, so that
the attributes in a group/cluster are as similar as possible to each other and as dissimilar
as possible to the observations in other groups/clusters.

According to Aldenderfer (1984) there are five basic steps that characterize all cluster
analysis studies:
a. Selection of a sample to be clustered;
b. Definition of a set of variables on which to measure the entities in the sample;
c. Computation of the similarities among the entities;
d. Use of a cluster analysis method to create groups of similar entities;
e. Validation of the resulting cluster solution.

In a cluster-based segmentation we first have to select the sampled data, which in our
case are the argentine farmers in the Humid Pampas producing more than 750 tons of
soybeans a year, as we explained in the previous section. Then, identify the key
variables that ought to characterize the purchasing behavior of argentine farmers for
seeds. In this case, our variables are price, performance, convenience and location,
personal factors, customer services, and support services for seed inputs. As in
Alexander et al. (2005), the key question used in the segmentation analysis asked
farmers to weigh the influence of six purchasing factors that they may use in order to
purchase their agricultural inputs. The influence of these factors had to sum up 100%.
Next, the data on these variables is processed in order to place respondents with similar
answers in the same segment/group or cluster. The idea is that through cluster analysis
we can group observations in a way that there will be a high level of natural association
between group members than those that are not.
What follows is to define the cluster analysis method to be used. The two main cluster
analysis methods to create groups of similar entities are the hierarchical and non-
hierarchical (or partitioning clustering) clustering methods. The hierarchical method
joins observations until the researcher decides to stop, while non-hierarchical methods
require the researcher to define previously the number of clusters. (Everitt et al. (2001))

Finally, to validate the clusters we used different criteria: the pseudo F statistical value,
the cubic clustering criterion (CCC) and the R2 test. The Pseudo F-value, is used to
compare variability obtained with K and K+1 groups or clusters, evaluating the relative
reduction of variability as we add new clusters. The higher the F value, the higher the
variability reduction that is obtained as we add one additional cluster. The cubic
clustering criterion (CCC), establishes a comparative measure of the deviation of the
segments regarding the expected distribution if the observations would have been
obtained from a uniform distribution. A value above two would suggest that the
structure of the clusters would be good; value of zero to two would suggest not a very
clear structure of a cluster. The negative values of the CCC criteria would be attributed
to the presence of out-layers. Finally we have the R2 test, as the proportion of variance
explained by the observations belonging to the conglomerate, the higher its value the
better the conglomerate.

6
Hierarchical clustering involves creating clusters that are hierarchically nested within
clusters at earlier iterations, in that each cluster can be included as a member of a larger,
more comprehensive cluster at a higher level of similarity. The most familiar expression
of the results of hierarchical clustering methods is the tree diagram or ‘dendrogram’,
which shows graphically the hierarchical structure entailed by the similarity matrix and
clustered by the linkage rule. Among agglomerative hierarchical methods, we have the
Ward Method. This procedure is designed to optimize the minimum variance within
clusters, and it works by joining those groups or clusters that result in the minimum
increase in the variance. (Aldenderfer, 1984)

Non-hierarchical or partitioning clustering, on the other hand, are methods that divide a
data set into a number of clusters by trying to minimize some defined error function.
Partitioning methods do not depend on previously found clusters. Partitioning clustering
methods work directly upon the raw data, therefore offer the opportunity of handling
distinctly larger data sets than hierarchical methods. As they make more than one pass
through the data and can compensate for a poor initial partition of the data, thereby
avoiding one of the major drawbacks of hierarchical agglomerative methods.
Partitioning clustering methods, however, suffer from some drawbacks, as they posit
explicit assumptions about the shape of the clusters; calls for an initial guess at the
number of clusters that will eventually be found; and are influenced by the choice of
initial seeds, the presence of outliers, and by the order in which the seeds are observed
and analyzed. (Aldenderfer, 1984)

As previous authors have done (Gloy and Akridge, 1999), we will first use a Ward
hierarchical clustering method to identify the number of cluster and to get the starting
points (seed values) for a second non-hierarchical algorithm procedure, which is the k-
means technique. This second algorithm rearranges the results optimally given the
previous results about the cluster means. The next steps would be the segmentation
validation through tests of significance for group differences, and finally we have the
interpretation of the results.
As Alexander et al. (2005) we will use a ‘multinomial logit model’ to predict segment
membership for seed purchases by Argentine farmers, based on observable factors such
as demography, behavior, and business management attitudes. For our case, the
multinomial logit is a probability model that explains the odds ratio of belonging to a
certain cluster if an observable behavior or characteristic of a farmer is present. This is,
if the logit value is positive it means that if the regressor/s values (observable behavior)
increase the odd that the regressand equals one increases too (become a member of the
segment). (Gujarati, 2003)

Results
The data in order to make the clustering analysis was processed, according to how mid-
size, commercial and large farmers purchased their seed inputs in Argentina. As in
Alexander et al. (2005) the key question in our questionnaire was the one that asked
respondents to weight the influence of six factors farmers may use to choose a seed
provider. These six selected factors were price, performance, convenience and location,
personal factors, customer services, and support services.

7
The Ward Hierarchical procedure was performed to define the number of clusters that
there would be for seed inputs. However, as there are no completely satisfactory
methods for determining the number of population clusters for any type of cluster
analysis, we validate the segmentation through the three methods explained above: the
pseudo F statistical value, the cubic clustering criterion (CCC) and the R2 test.
(Everitt2001)

The results we obtained using the above mentioned criteria are that the Argentine
farmers are segmented in four clusters according to their seed buying behavior;
performance, price, balance and convenience segments. As it has been mentioned in
Alexander et al. (2005) and Gloy and Akridge (1999), we will also show that the results
meet the validation criteria by which the members of the segments differ in non-
clustering variables such as demography, behavior, and business management attitudes.
In the next sections we will present the result of our cluster analysis. In the first section
we present the segmentation of farmers in different clusters, as we shall see in Table 1.
In the next part, we introduce the description of farmers in each of these segments by
non-clustering variables, as described in Tables 2 to 9. Finally, in the last section, we
present the logit multinomial regression used to predict segment membership, presented
in Table 10.
Characteristics of the Segments
The largest cluster is Performance, followed by Price and Balance segments, as we can
see in Table 1 on next page. The smallest cluster is Convenience, with only 5% of the
population. Performance segment has a 37% share. Members of this cluster search for
high quality products and services. On average members of this segment placed 77%
weight on performance, and only 11% on the second most important factor, which is
price. These Performance farmers are quite young (45.24 years versus 46.42 years for
the entire population), and well educated, as 50% of the members have a college degree.

8
In tables 2 and 3 (next page) we can see the demographics, education, farm size, income
level, and future growth expectation for each group. The Performance segment is the
one with the largest farms (18% farm more than 1840 hectares), with relatively high
incomes (41% earn more than half a million dollars), but with low expectation of future
growth (31% growth average). As we saw, they are relatively young and educated.

The second largest cluster is the Price segment, with a share of 29%. Price as a
purchasing factor has a weight of 48%, which would mean that these farmers are cost-
oriented: they buy their seed at a lowest price. In spite of this, performance is the second
factor with a weight of 34%, and these two factors account for 82% of the total weight
of purchasing factors. The members of this segment are the oldest, with an average of
47.59 years, and less than half of them have a college degree. This group also has
relatively high level of incomes (44% earn more than half a million dollars), 12% farm
more than 1840 hectares, and have a fairly high growth expectancy in the future with
41% growth.

The third segment is Balance, with a share of 28%. These farmers value all factors fairly
equal, but give special importance to performance and price (22% and 23%) and then to
services/information and personal factors (18% each). This group has the lowest
number of college graduates (39%), and is the second oldest segment (47.05 years).
Balance farmers are relatively large farmers (14% farm more than 1840 hectares), have
high income levels as 46% of them earn more than half a million dollars.

9
The last group is the Convenience segment, is the smallest as we already mentioned,
with 5% share. They place most importance to convenience and location, with a 60%
weight. The rest of the factors have a lower weight: 13% for price, 8%
service/information, 7% support services, 5% personal factors, and only 6% to
performance. They have the highest level of education as 56% have a college degree,
and are the youngest group with an average of 44.92 years.

The Convenience segment is not only the smallest group, but also they are the ones who
farm the smallest amount of land, as only 4% farm more than 1840 hectares. However,
for 60% of these farmers their income is between 200 thousand and half a million
dollars, and they have the largest expectation of future growth: 52% growth in the
amount of land increase they expect to farm for the next years.

This gives us a general profile of farmers in each cluster. In the next section we will
analyze the commercial attitudes of farmers in these groups.

Commercial Attitudes of Farmers in each Segment

Here we analyze how farmers consider the concepts ‘Brand Similarity’, ‘Brand
Loyalty’, ‘I Purchase at the Lowest Price’, and ‘Loyalty with the Local Dealer’, in terms
if they consider themselves loyal to brands, to local dealers, or if they buy at the lowest
price, or if they consider brands similar. Their answers are on table 4, using a five point
Likert scale, in which a 1 would mean “I strongly disagree”, and a 5 “I strongly agree”.
An answer around 3 would convey some neutral standing regarding the question.

10
Farmers in the Performance segment appear to consider brands dissimilar (2.07/5), tend
to relatively loyal to brands (3.72/5), they reject buying seeds at the lowest price
(1.62/5). These results for the Performance segment would be quite different than the
ones for the other groups, in which farmers tend to consider brands a little bit more
similar than the Performance group, less loyal to brands, and a little bit more neutral
regarding to purchasing at the lowest price. Regarding loyalty to the local dealer, all
segments express a weak loyal, with an average of 3.76/5).

Another aspect that we can analyze in table 4, in the next page, is the statistical
differences between clusters in two of the four observations. Both ‘I purchase at the
lowest price level’ and ‘loyalty with the local dealer’ have low probabilities of no
association, which means that the hypothesis that segments are similar is rejected in
both cases. Each cluster has a significant difference in behavior regarding these two
observations. In both cases the Performance segment is the one that most rejects the
hypothesis of purchasing at the lowest price, and is the one who is most loyal to dealers.
The less loyal to local dealers would be the Price segment, while the ones who are less
likely to reject the hypothesis of buying at a lowest price are the Price and the
Convenience segments.

Information Sources Useful for Managing/Purchasing Decisions: Personal and


Communication Media

Tables 5 presented in the next page shows the useful information sources that farmers
use to make managing and purchasing decisions. We present both more personal
oriented sources and the communication media information sources. We make the
observation that in both cases farmers in different segments have significant statistical
differences for all the observations related with information sources they use to make
managing and purchasing decisions.

11
Regarding the importance of information sources we can see in table 5 that the
information of local dealers is valued by all segments, a little bit less by the Price
segment, and more than average by farmers belonging to the Convenience group.

Now going to the communication media information sources, we observe that both
Performance and Convenience segments value all the different sources (all the values
above three), although values are higher for the Convenience cluster than Performance.
The Price segment does not have any relevant information source. The Balance group
values the agricultural section of newspapers, and very weakly the meetings with
suppliers.

As a summary, local dealers, emails, and agronomic websites would be the main
sources of useful information for farmers in order to make managing and purchasing
decision. The former source used by all farmers, the two latter ones more by
Performance and the Convenience groups. To a lesser extent, the agricultural section of
newspapers is relative good information sources for the Balance and Convenience
groups.

Use of Consultants by Farmers

Table 6 presents the information regarding the usage of consultants by farmers, in terms
of percentage of farmers who use consultants in different fields. In the case of
accounts/tax consultants there is relative high usage for all segments, with almost 90%
of all farmers using their services. Then we have the usage of independent crop
consultants, which goes from 44% for the case of the Convenience segment to 70% of
the Price Segment. For the Balance and Performance groups in average a little bit more
than 60% of the farmers use independent crop consultants.

12
Pest control consultants are also used to a certain extent with between 20 to 34% of the
farmers using their services. Another type of consultant used by farmers is the
management consultant, with 14% to 19% of farmers using their services. Financial
consultants are not largely used, but the in the Balance cluster 16% of all farmers
demand these services. Environment consultants are not extensively used. The only type
of consultant in which there would be significant differences in usage between farmers
is the case of independent crop consultants.

Salespeople

Salespersons are a critical assess for input suppliers in order to sell their products. In
tables 8 we evaluate how farmers in different segments value the different
characteristics that salespeople may have, such as honesty, technical competence,
represents my interests, he is a friend, knows well my business. In table 9, on the other
hand, intends to answer which are the activities that salespeople do that farmers value
the most.

The most valued characteristic of a salesperson for farmers in most clusters is that he
may have a high level of technical competence. Farmers in three of the four clusters
valued this ranked this feature as the most important characteristic for a seed
salesperson: almost half of the farmers in each of these clusters rated it as the most
important feature.

The Convenience group was the one who valued this characteristic the highest (52% of
farmers), while the farmers in the Balance segment ranked this feature as the second
more important (32%). For the performance and Price segments 49% and 44% of
farmers rated this feature as the most relevant for a salesperson.

13
The Balance segment rated honestly as the most important characteristic of a
salesperson (35%), while for the rest of the clusters this was the second most important
feature, with an average of 30% rating. The third most important characteristic rated by
farmers was that the salesperson knows their operations well, with the Balance segment
rating it the highest (18%), and the rest with an average value of 10%. The Convenience
and Price groups were not very convinced about the value of this feature choosing
instead ‘represents my interest’ which was the fourth most valued feature for the rest of
the clusters (Performance and Balance). ‘Friendship’ as a salespersons’ characteristic
was rated as the least important feature, with a rating of between 4 to 0 %. It is
important to remark that in the ranking of these characteristics there are significant
statistical differences of behavior between the different clusters.

14
Regarding the activities that a salesperson performs that are more valued by farmers,
there are specially three that farmers tend to value the most: ‘Brings me the best price’,
‘Provides relevant/timely information’, ‘Provides good follow up service’. In all these
cases farmers tend to agree that they value these activities relatively strongly (4/5 points
in the Likert scale), and for ‘Provides good follow up service’ and ‘Brings me the best
price’ there are significant statistical differences between clusters. Farmers in the Price
group would value the most these activities, while the Convenience segment would
value them the least.

The other activities such as ‘Provides access to supplier resources’, ‘Is a consultant for
my business’, ‘Brings me innovative ideas’, and ‘He is a consultant for my business’ are
also well rated for all clusters, while salespersons telephone calls was not valued highly
although there are significant differences between clusters regarding this activity.

Predicting Segment Membership

Table 9 on next page reports the results of a logit regression model, which shows the
odds ratio of belonging to a certain cluster if an observable behavior or characteristic of
a farmer is present. If the logit value is positive, it means that if the observable behavior
increases the odd the farmer becomes a member of a certain cluster increases too. In
case it is negative, the higher the value of the observable behavior would decrease the
odd that a certain farmer would belong to that cluster.

This is potentially useful for marketing managers because as we will see, observing key
behaviors of a client they would be able to predict to which cluster that farmer belongs,
and in this way, know what that person values most in his purchases.

The model χ2statistic (89.37 with 30 degrees of freedom) is significant at a level of 1%


level of probability. Likewise, the predicted share for each cluster is consistent with the
actual share in each one of the segments. In all the groups, expect for the Balance
segment which has only two significant observable variables, there are at least three to
six significant observable characteristics that supply significant statistical predictive
power for each one of the cluster membership.
In this logit regression we find six significant relationships that provide predictive value
to the segments’ membership, explaining why a positive value of that observable
characteristic would make a farmer belong to a certain cluster, and the negative value
why another farmer would belong to another.
In first place, the usage of independent crop consultant observation explains in an
8.25% the likelihood of a farmer belonging to the Price Segment, and in a 3.44% it
would be less likely for a farmer to becoming a member of the Convenience Segment.
This results from the fact that Convenience Farmers are technological self-sufficient and
educated people, as we saw in Tables 2 and 6.
The frequent consultation of the manufacturer salesperson as a useful information
source makes it a 5.79% more likely for a farmer to be a member of the Performance
Segment, and in a 3.18 % less likely to belong to the Price Segment.

15
Table 9 Results of a Multinomila Logit Model Predicting Segment Membership Marginal Effects (with standard errors in parantheses)
Producer Segment
Performance Price Balance Convenience
Variable dy/dx Std. Err. dy/dx Std. Err. dy/dx Std. Err. dy/dx Std. Err.
Sales -.0651826 .0118682 .0475028 .0058116
(0.0317)** (0.02937) (0.03005) (0.01084)
Crop Consultant -.0140938 .0824578 -.0339639 -.0344001
(0.04774) (0.04322)* (0.04459) (0.01977)*
Manufacturer salesperson .0579527 -.0318182 -.021036 -.0050985
(0.01932)*** (0.01743)* (0.01756) (0.00625)
Local Dealer -.00019955 -.0314128 .0266879 .0067203
(0.01989) (0.01812)* (0.01877) (0.00736)
EMAILS .0664933 -.0152157 -.0450566 -.0066609
(0.02226)*** (0.02052) (0.02034)** (0.00786)
Ag websites -.0198009 .0044311 -.004261 .0196308
(0.02186) (0.02049) (0.02038) (0.00804)**
Brand similarity -.0453926 .0164886 .0184951 .010409
(0.01835)** (0.0165) (0.01665) (0.00586)*
Lowest Price -.0577939 .0605335 -.010912 .0081725
(0.02374)** (0.0196)*** (0.0211) (0.00694)
Local Dealer Loyalty .0401864 -.023215 -.0199755 .0030042
(0.0174)** (0.01572) (0.0159) (0.0063)
Salesperson offers best prices -.0190974 .0769323 -.0500795 -.0077554
(0.02595) (0.02678)*** (0.02296)** (0.00787)
Predicted share 37.65% 29.13% 29.39% 3.82%
Real Share 37.45% 29.28% 28.28% 4.98%
χ2=88.37*** (30 d.f.) Prob> χ2 > 0.001; Single, double, and triple asterisks (*) denote statistical significance at the 0.10, 0.05, and 0.01 level respectively.

The Email usage as a useful source of information from communication media makes it
6.65% more likely for a farmer to become a member of the Performance Segment, and a
4.5% less likely to become a member of the Balance Segment. Performance farmers are
open to all sources of information, while this would not so much the case for the price-
oriented farmers. This congruent with the information we analyzed in Tables 5 and 6.
Likewise, if a farmer considers brands similar he would be 1.04% more likely to belong
to the Convenience Segment and 4.54% less likely to become to the Performance
segment. The commercial attitudes that are shown in Table 4 explain that Performance
farmer do not consider brands similar, while farmers in the Convenient group would
tend to be almost neutral regarding brand similarity.
If a farmer purchases seeds at the lowest prices, he would be 6.05% more likely to
belong to the price segment, and 5.52% less likely of becoming a member of the
Performance Segment. Also this is show in Table 4.
Finally, if a farmer considers the most important of salesperson activity as ‘bringing the
best price’, he is 7.69% more likely to belong to the Price Segment, and 5% less likely
to become a member of the Balance Segment. Again, this is congruent with the
information provided in Table 8.
Overall, the logit model has strong predictive power, which is shown by the six pairs of
significant relationships we explained above.
Conclusions
The main goals of this paper was to identify distinctive market segments for argentine
farmers purchasing seeds, segmenting farmers into buying characteristics according to
their purchasing behavior and in this way provide some answers of how Argentine
producers purchase their agricultural inputs. Argentine farmers were segmented in four
16
clusters according to their seed buying behavior: performance, price, balance and
convenience segments. While farmers in the performance and balance segments would
be business purchasers, the ones in the price segment are cost-oriented. The
convenience farmers, on the other hand, are those who prioritize location and
convenience in their purchases.
The data from tables 2 to 8 fitted well with the different segments we defined in this
work, as well with the regression model established to predict segment membership. It
was clearly established that farmers in the Performance segment value the information
coming from the manufacturer salesperson and emails, they do not consider brands
similar, they are loyal to brands and would not buy seed at the lowest price. The price-
oriented farmers belonging to this segment would be the ones which would purchase
seeds at the lowest price and value the activities of salesperson regarding to bring the
lowest price; also would tend to hire independent crop consultants.
This would represent quite clearly the profiles of Performance and Price Segments of
farmers. For the case of the Balance buyers the results are also congruent with the
information provided in tables 2 to 8.
As in Alexander et Al. (2005) the convenience buyer is the smallest segment. Farmers
in this segment in Argentina tend to have less loyalty to brands than other segments,
demand a very high level of technical competence to salespersons, and use relatively
few independent crop consultants. They do not seem to be significantly more loyal to
local dealers than other groups, although they value the information dealers provide
them.
The Convenient segment farmers in Argentina result to be young and well-educated
farmers, intending to grow fast. They are a self-sufficient group, in terms of using few
independent consultants, regarding brands more or less similar and not being very loyal
to brands. As we can see in annex 2, Convenience farmers are mostly mid-size and
commercial operations; while large producers are mainly in the Performance group. So
from the data we have, Convenience farmers in Argentina have small operations but are
progressive willing to grow fast, more than relational farmers as presented in other
works. They are convenience buyers in that location and convenience are important for
them to grow their business and save time, not in order to make less effort.
In this way we were able to define a profile for each segment, which we summarize in
the Table 10 that we present in the next two pages. This work provides two main
contributions which are, in first place, identifying four different segments for the seed
markets in Argentina. Secondly, defining a very special segment of farmers called
‘convenience’ with singular characteristics, different from the way they have been
defined before for other markets.

17
18
Annexes

Annex I

19
Annex II

20
References
Aldenderfer, Mark S.; Blashfield, Roger K, 1984.Cluster Analysis. Sage University
Papers Series. Quantitative Applications in the Social Sciences

Alexander, Corinne; Wilson, Christine; Foley, Daniel. 2005. “Agricultural Input Market
Segments: Who is Buying What?”, Journal of Agribusiness, Vol. 23, Issue 2
(fall 2005), pages 113-132
Bernhardt, K.J.; Allen, J.C.; Helmers, G.A. 1996. “Using Cluster Analysis to Classify
Farms for Conventional/Alternative Systems”. Review of Agricultural
Economics, Vol. 18, No. 4 (Oct., 1996), pages 599-611
CREA. 2010. De la tierra al país: Una Radiografía de la Inversión 2010/11 en la
Argentina, September 2010, CREA Groups

Everitt, Brian et al., Second Edition. 2001Handbook of Statistical Analyses Using SAS.
Chapman and Hall Publishers

Feeney, R; Berardi, V.; Bertossi, O.; Steiger, C.; Piazzardi. 2009. “The Needs of
Argentine Farmers”, Center for Food and Agribusiness, Austral University
Argentina
FIS/ASSINSEL. The International Seed Trade Federation/The International Association
of Plant Breeders for the protection of Plan Varieties. Accessed at URL
http://www.worldseed.org/isf/seed_statistics.html, May 13th, 2011.

Foley, D. H. 2003. “Segmenting the commercial producer marketplace for agricultural


inputs.” Unpublished Master’s thesis, Department of Agricultural Economics,
Purdue University, West Lafayette, IN.

Gloy, B. A., and J. T. Akridge. 1999. “Segmenting the Commercial Producer


Marketplace for Agricultural inputs.” International Food and Agribusiness
Management Review 2, 145S163.

Gujarati, D. 2003. Fourth Edition. Basic Econometrics. McGraw Hill Higher Education

Hooper, M. A.1994. “Segmentation of the Market for Agricultural Inputs: A Nested


Approach”. Unpublished M. S. thesis, Department of Agricultural Economics,
Purdue University
Kotler, P. 1997. Marketing and Management: Analysis, Planning, Implementation, and
Control, 9th edition. Upper Saddle River, NJ: Prentice-Hall.

Reimer, Aaron; Downey, Scott; Akridge, Jay. 2009. “Market Segmentation Practices of
Retail Crop Input Firms”, International Food and Agribusiness Management
Review, Vol. 12, Issue 1, 2009

Roucan-Kane, M.; Gray, A; Downey, S.; Alexander, C.; Boehlje, Gunderson, M. 2008.
“Serving Producers in Volatile Times”, Center for Food and Agribusiness
Purdue University

21
USDA FAS. 2010.“Argentina-Annual Biotechnology Report”, December 29th 2010

Vilela, F. et al. 2009. El Sistema de Agronegocios de la Soja en la Argentina, su Cadena


y Prospectiva al 2020.Ediciones Horizonte, ACSOJA- FAUBA

22

You might also like