Food Policy 41 (2013) 11–17

Contents lists available at SciVerse ScienceDirect

Food Policy
journal homepage: www.elsevier.com/locate/foodpol

Does agricultural trade affect productivity? Evidence from Chilean farms

David A. Fleming a,b,⇑, David G. Abler a
a
Department of Agricultural Economics, Sociology, and Education, The Pennsylvania State University, University Park, PA 16802, United States
b
Social and Economic Sciences Program, CSIRO Ecosystem Sciences, Black Mountain Laboratories, ACT 2601, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper analyzes the relationship between trade and agricultural productivity in Chile, a middle-
Received 14 January 2012 income country with a recent noticeable history of agricultural trade. This study uses an agricultural
Received in revised form 1 November 2012 commodity trade exposure index in a cross-sectional analysis of more than 70,000 farms to study the
Accepted 2 April 2013
relationship between the trade exposure of agricultural commodities and the yields reported by these
Available online 7 May 2013
farms in the 1997 Chilean agricultural census. In order to capture both import and export exposure we
subdivide farms in two groups, according to the Chilean case: farms producing only importables such
Keywords:
as grains (traditional crops), and farms producing both traditional crops and non-traditional agricultural
Agricultural trade
Agricultural productivity
commodities (products more related to export markets). We exclude from our analysis farms producing
Trade exposure only non-traditional products because the census only reports yields for traditional crops. We employ a
Switching regression model switching regression model to analyze the effects of trade exposure on traditional crop yields for both
groups of farms. Results show that the trade exposure index is positively related to farm yields for both
groups, but with a larger effect on farmers producing both traditional and non-traditional commodities.
These results are important because they suggest that spillovers from both importables and exportables
produce gains in the productivity of traditional crops.
Ó 2013 Elsevier Ltd. All rights reserved.

Introduction and Rossi (2003), Jonsson and Subramanian (2001), Pavcnik

The benefits and costs of globalization continue to be debated in
the economic literature. Broadly speaking, two different views can
be found in the literature: one holds that international trade is
associated with productivity growth and improved economic per-
formance; and the other is concerned about impacts on equity and
local development (Carter et al., 1996; Estrades and Terra, 2012).
This study contributes to this debate by analyzing whether, and
to what extent, international trade (imports and exports) in agri-
cultural commodities is related to agricultural productivity in Chile
in the 1990s, a decade in which this country significantly increased
its presence in international markets and underwent a transition
from traditional to modern agriculture in many regions.
Several researchers have studied the impacts of trade liberaliza-
tion on industrial and agricultural productivity in a wide range of
countries. Using a cross-country sample (and generally relying on
national productivity aggregates), Coe et al. (1997), Edwards
(1998), and Badinger (2007) find that countries with fewer trade
barriers experienced more rapid productivity growth. Using indi-
vidual countries for analysis (generally relying on survey data at
the firm level), Amiti and Konings (2007), Hay (2001), Ferreira
⇑ Corresponding author. Tel./fax: +61 02 6246 4302.
E-mail addresses: David.Fleming@csiro.au (D.A. Fleming), d-abler@psu.edu
(D.G. Abler).
(2002), and Tybout et al. (1991) also find a positive link between
openness and productivity. The main channels giving rise to a po-
sitive trade/productivity relationship are generally hypothesized to
be international spillovers as a result of trade (especially in R&D
and best practices in production and supply chain management),
gains in productivity due to product specialization as a result of
trade, learning by doing through exporting, and pressures to raise
productivity due to international competition.
However, some researchers argue that the response of produc-
tivity to trade liberalization is more ambiguous (Krishna and Mitra,
1998; Winters et al., 2004). On the import side, although firms may
be stimulated to improve their productivity due to international
competition, there may also be an exodus of assets (human and
physical capital) from less competitive local firms. Under these cir-
cumstances productivity gains would only emerge if the irrevers-
ibility of investment in capital does not impede the exit of less
productive plants (Pavcnik, 2002). On the export side, although
firms are exposed to new markets through trade, local R&D and
innovation might be reduced as firms utilize R&D carried out
elsewhere.
In the literature on international trade in manufacturing, a con-
sensus has emerged that only highly productive firms participate
in the export market because of the presence of large fixed costs
of exporting (Bernard et al., 2003; Melitz, 2003). Agriculture is
different because trade in agricultural commodities is primarily

0306-9192/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodpol.2013.04.004
12
1,200
1,000
Thousands hectares
800
600
400
200
Fig. 1. Land use change in rural Chile, 1965–2007. Source: Portilla (2000) and Agricultural Census 2007 (http://www.censoagropecuario.cl).
carried out by specialized exporters or multinational enterprises
such as Dole or Chiquita that generate export volume through con-
tracts with large numbers of farms. Exporters sometimes enter into
‘‘resource provision’’ contracts with small farms in which they pro-
vide growers with specialized equipment, credit, technical assis-
tance, and/or insurance (Reardon et al., 2009). Farms also often
form cooperatives for input supply and domestic and international
marketing. In this way, agricultural trade differs from manufactur-
ing trade as firm (farm) heterogeneity does not necessarily exclude
producers from the export market.
In agriculture, Gopinath and Kennedy (2000) find a positive link
between trade and productivity for US states. In developing coun-
tries or regions, estimation of the agricultural trade/productivity
relationship is more complicated (especially at the farm level) be-
cause of difficulties in obtaining accurate measures of agricultural
inputs and outputs (Martin and Mitra, 2001). Chile is an interesting
exception, where data are consistently collected by local authori-
ties, the national agricultural censuses being a clear example of
this effort.1
In order to use international trade as explanatory variable in
economic models, researchers have employed different approaches
and methods to measure the importance of trade to an economy
(Harrison, 1994). One widely used and tested trade variable is
the trade dependency ratio, which is defined as the share of im-
ports plus exports in the total GDP of a region (Harrison, 1994;
Frankel and Romer, 1999; Jonsson and Subramanian, 2001). We
use a similar methodology to calculate the exposure of producers
to agricultural trade at a disaggregated level, on a per commodity
basis. Thus, this study constructs a product-specific trade exposure
index that measures the share of imports plus exports in the total
production of an agricultural commodity in a particular year.
Farm-level analyses are then carried out using the trade exposure
index in a cross-sectional study of more than 70,000 farms located
in the middle part of Chile.
Chilean agriculture in an open economy
Chile was one of the first countries in Latin America to shift
from import substitution to an open-economy model following
the import substitution policies that were popular in the region
during the 1960s and 1970s. This change led to several structural
adjustments in macroeconomic policies and institutions, and one
of the priorities given by authorities was to create an export-ori-
ented strategy supported by a market-friendly regulatory system.
This transformation included a strengthening of property rights
that helped to improve access to land ownership, a reduction in
1 2
Agricultural censuses are carried out every 10 years in Chile and collect data for
practically every single farm in the country.
D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17
1965
1976
1997
2007
public services and expenditures, the privatization of input and
product markets, a gradual elimination of price controls,2 and the
liberalization of trade (non-tariff barriers were eliminated and tariffs
on most imports were rapidly reduced) (Foster and Valdes, 2006).
However, it was not until 1984, with the reversal of Chile’s currency
appreciation policy, that agricultural commodities became more
competitive on global markets.3 In response the agricultural sector
started to receive major private investment and generate more in-
come. For rural areas this new system led to dramatic changes in
agriculture, land use and property rights. Fig. 1 shows the trends
in land use during recent decades by type of product.
As can be observed in Fig. 1, the area devoted to fruits has
shown significant growth since 1976, which clearly demonstrates
the export boom produced by trade liberalization (the same phe-
nomenon explains the boom in forest plantations). Thus, fruits be-
came a new export-oriented product in Chilean agriculture.
Following Barham et al. (1992) we refer to fruits and other prod-
ucts that were traditionally cultivated for local consumption but
that started being exported in recent decades as ‘‘non-traditional’’
crops. Fig. 1 indicates that the area devoted to cereals and grains
has fallen considerably over time, a trend explained by growth in
imports. For these products we use the term ‘‘traditional’’ crops,
since they were not part of the export boom in Chile, but rather
commodities with positive net imports.
Methods
In order to examine the effect of international trade on agricul-
tural productivity, a detailed farm-level evaluation is undertaken
through a cross-sectional analysis of crop yields on Chilean farm-
s.Crop yields are not a perfect proxy for productivity but they are
a commonly available measure and they do reflect productivity.
Consider a single-output production function of the form
Y = AF(X, L), where Y is output, A is total factor productivity, L is
land, and X is a vector of other inputs. If there are constant returns
to scale, then we can rewrite the production function as y = Af(x),
where y = Y/L is yield, x = X/L is other inputs per unit of land, and
f(x) = F(X/L, 1). If the farm’s choices with respect to other inputs
per unit of land depend on fixed or quasi-fixed farm-specific char-
acteristics (e.g. infrastructure, human capital, soils), denoted by k,
the output price (p), the level of total factor productivity, and input
prices (w), we can write:
y ¼ Ahðk; p ; wÞ; ð1Þ
where p = pA is the effective output price and h(k, p⁄, -
⁄
w) = f(x(k, p⁄, w)). An increase in total factor productivity affects in-
Except for wheat, oilseeds and milk.
3
However, price bands remained for wheat and oilseeds, and were added for sugar.
 D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17 13

Table 1
Product–level trade exposure indices 1990–92 and 1996–98, selected commodities. Source: Prepared by the authors using data from FAO (2008).

Commodity 1990–92 1996–98 Commodity 1990–92 1996–98

Traditional crops Non-traditional crops
Artichokes 0.041 0.028 Apples 0.485 0.514
Asparagus 0.223 0.241 Apricots 0.151 0.110
Barley 0.041 0.485 Avocados 0.353 0.357
Beans, dry 0.699 0.419 Cherries 0.309 0.256
Chillies and peppers 0.022 0.022 Grapes 0.377 0.326
Lentils 0.301 2.189 Kiwi fruit 0.730 0.908
Lettuce and chicory 0.008 0.017 Lemons and limes 0.030 0.084
Maize 0.248 0.661 Oranges 0.006 0.004
Oats 0.034 0.043 Melons and cantaloupes 0.009 0.002
Onions, dry 0.190 0.124 Peaches and Nectarines 0.355 0.338
Peas, dry 0.199 2.116 Plums and sloes 0.414 0.414
Peas, green 0.004 0.000 Strawberries 0.003 0.003
Wheat 0.161 0.346 Walnuts 0.745 0.311

Note: The figures represent 3-year averages of the trade exposure indices.

put choices because production becomes more profitable for any gi-
ven levels of output and input prices, and hence the producer has an
incentive to increase output.For reasons discussed above we
hypothesize that a farm’s productivity depends on international
trade, and it may also depend on fixed or quasi-fixed farm specific
characteristics such as human capital. Letting T denote some mea-
sure of the exposure of a farm to international trade, we can write:
y ¼ AðT; kÞhðk; p  AðT; kÞ; wÞ:
Eq. (2) shows two routes through which trade affects productiv-
ity and hence yields: first, yields increase for any given levels of
non-land inputs per unit of land; and second, increases in produc-
tivity affect the optimal levels of non-land inputs per unit of land. If
the supplies of non-land inputs to the farm are more elastic than
the supply of land, and if none of the non-land inputs are inferior
inputs, then it can be shown that the optimal levels of the non-land
inputs per unit of land will tend to increase as total factor produc-
tivity increases.
This study focuses on the role of trade on productivity and
therefore yields. In particular, for our empirical approach, we mea-
sure a farm’s exposure to international trade through an agricul-
tural commodity-specific trade exposure index as in Fleming
et al. (2010), which is defined as the sum of exports plus imports
divided by the production of that commodity:
tural census. This census was chosen because in the 1990s Chile,
with the arrival of a democratic regime, began opening its com-
mercial borders, leading to significant growth in agricultural trade
(see Fig. 1). Data on how much of a specific farm’s production is
sold on international markets versus domestic markets are not
available. However, we can say something about the importance
of international trade to an individual farm based on the mix of
commodities produced by that farm. Farms producing commodi-
ð2Þ ties for which imports or exports are significant nationally are
more exposed to trade than farms producing commodities for
which imports and exports are both small.
The farm-level analysis is geographically focused on the middle
part of Chile, from the Valparaiso Region to the Biobio Region (area
that encompasses five regions). This is the zone that covers most
non-traditional crop production,4 has goods soils for agricultural
production, and ideal weather conditions for production of both tra-
ditional and non-traditional crops. These five regions encompass
more than 90,000 farms.
We define the trade exposure index (FTj) for farm j as a crop
area-weighted average of the national-level trade exposure indices
across 45 commodities:
!, !
X
45 X
45
FTj ¼ Lij T i Lij ; ð4Þ
i¼1 i¼1

T i ¼ ðExportsi þ Importsi Þ=Productioni ;
where the subscript i refers to the commodity. This index captures
the degree to which a country produces commodities that are lar-
gely internationally traded as opposed to commodities for which
international trade is small. The lower bound for T is zero in the case
of commodities that do not cross the border, but there is no upper
bound because domestic production could be zero. Thus, a crop that
is not produced in the country and is imported to meet domestic
demand would have an infinite value for T. Such a case will not oc-
cur in this study because we are looking at the effects of trade on
crop yields, and barring a total crop failure, yields are only observed
for crops with non-zero production.
The trade exposure index is calculated in this study for 25 tra-
ditional crops and 20 non-traditional crops, which together consti-
tute Chile’s principal agricultural products. Data to construct these
indices were obtained from the FAO’s FAOSTAT database. In order
to mitigate the effects of supply and demand shocks in any given
year, we calculate the indices using three-year averages. Table 1
presents a list of selected values for 1990–1992 and 1996–1998,
where the later period is used for the analyses described below.
The analysis here examines the relationship between interna-
tional trade and yields using data from the 1997 Chilean agricul-
ð3Þ where Lij is the area cultivated with crop i by farm j. Note that even
if a particular farm is producing solely for the domestic market, it
still has high exposure to the economic forces of global markets
and trade if its value of FTj is high.
We consider only farms that produce at least one traditional
crop, because the agricultural census reports yields only for tradi-
tional crops. We also exclude observations that correspond to com-
panies or associations of farmers, focusing the analysis upon
individual producers. Farms that reported yields equal to zero in
one of their reported crops were also eliminated from the final
sample.5 The final dataset consists of 73,333 farms.
Table 2 reports descriptive statistics for the farm-level trade
exposure index for the entire sample, broken down by crops pro-
duced (only traditional versus both traditional and non-tradi-
tional), and for the five Chilean regions under study (ordered by
4
Although northern regions have significant grape production, most of these
grapes are used in the Pisco industry (a liquor only produced in these northern
regions), and Pisco is a product that is neither importantly exported nor imported.
Southern regions of Chile have some apples and berries, but do not have any major
significant presence of other fruits.
5
As the agricultural census does not indicate if the zero yields were caused by total
crop failure or unwillingness to report data on the part of the farmer, the option of
excluding all these observations was chosen.
14 D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17

Table 2
Farm-level trade exposure index, descriptive statistics. Source: Prepared by the authors using data from INE (1997) and FAO (2008).

Average Std. dev. Minimum Maximum

Entire sample (n = 73,333) 0.234 0.207 0 2.189
By crops produced:
Both traditional and nontraditional (n = 5,656) 0.220 0.159 0 1.613
Only traditional (n = 67,677) 0.235 0.210 0 2.189
By Region:
Valparaiso region (n = 2,203) 0.158 0.220 0.0001 2.189
Metropolitan region (n = 3,377) 0.184 0.190 0.0001 2.017
O’Higgins region (n = 14,266) 0.364 0.232 0 2.189
Maule region (n = 22,502) 0.220 0.190 0 2.189
Biobio region (n = 30,985) 0.195 0.180 0 2.189

Table 3
Definitions and summary statistics for farm-level variables. Source: Prepared by the authors using data from INE (1997).

Variable Definition Mean Std. dev.

Yield Average normalized yields (0–100 scale) 28.679 18.089
Nontraditional crops Dummy for production of nontraditional crops (1 if yes, 0 if no) 0.077 0.267
Land Land utilized for agricultural production (hectares) 11.022 34.764
Irrigation Proportion of farmland covered by irrigation 0.549 0.610
Hired Manager Dummy for presence of a hired manager (1 if yes, 0 if no) 0.060 0.237
Adults Number of adults in the farm household 1.945 2.706
Modern machinery Dummy for use of modern machinery (1 if yes, 0 if no) 0.694 0.461
Formal title Dummy for whether farm has official land ownership records (1 if yes, 0 if no) 0.683 0.465
Storage capacity The sum of the capacity of wells, warehouses, and silos (m2) 45.796 375.483
Infrastructure Total surface with roads, and other infrastructure on the farm (hectares) 0.519 1.775
Gender Dummy for farmer’s gender (1 if male, 0 if female) 0.838 0.369
Age Farmer’s age 55.616 14.045
Oxen Dummy for presence of oxen on farm (1 if yes, 0 if no) 0.195 0.397
Forests Dummy for presence of forests on farm (1 if yes, 0 if no) 0.336 0.472
Live on farm Dummy for whether farm household lives on the farm (1 if yes, 0 if no) 0.714 0.452
Mechanical irrigation Dummy for use of mechanical irrigation (1 if yes, 0 if no) 0.010 0.101
Basic education Dummy for whether farmer obtained basic education (maximum of 8 years) (1 if yes, 0 if no) 0.665 0.472
High school education Dummy for whether farmer obtained high school education (maximum of 12 years) (1 if yes, 0 if no) 0.128 0.334
Technical education Dummy for whether farmer obtained technical education obtained (maximum of 14 years) (1 if yes, 0 if no) 0.026 0.161
Superior education Dummy for whether farmer obtained superior education (maximum of 17 years) (1 if yes, 0 if no) 0.052 0.221
No education Dummy for whether farmer has no education (1 if yes, 0 if no) (reference category) 0.129 0.348

their geographical location from north to south).The agricultural
census does not contain data on every variable that might be
included in the vector k in Eq. (2), and has no data on prices
(p, w).6 However, data are available on a number of quasi-fixed farm
inputs and other farm characteristics. Applying a logarithmic trans-
formation to Eq. (2) and then a first-order Taylor series approxima-
tion yields the following model that can be estimated:
ln yj ¼ a0 þ a1 FT j þ c0 kj þ uj ; ð5Þ
where yj is the average yield on farm j and uj is an error term. The
variables in kj in Eq. (5) include the log of the irrigated area divided
by total farm area, a dummy variable for the presence of a hired per-
son as manager of the farm, the log of the number of adults in the
farm household per hectare (a measure of farm household labor
supply), a dummy variable for the use of modern agricultural
machinery, a dummy variable for farmers with possession of formal
land titles, the log of fixed capital per hectare, the log of infrastruc-
ture (i.e. roads) per hectare on the farm, a dummy variable for the
gender of the producer, the log of the producer’s age, the square
of the log of the producer’s age, and dummy variables for the pro-
ducer’s level of education.7 kj also includes dummy variables for
6
The absence of data on prices is a concern only to the extent that there is spatial
variability in output and input prices.
7
A small positive constant (one) was added to irrigated area, labour, fixed capital,
and infrastructure before taking logarithms to deal with cases where the values of one
or more of those inputs were zero.
the farm’s agro-ecological zone and the government region in which
it is located. Descriptions of the variables and main statistics are pre-
sented in Table 3.
Farm yield in Eq. (5) is an average across crops and is calculated
as an average of the normalized yields of every traditional crop
that the farm produced. The normalization scales the yield of each
crop on a farm j to lie between 0 and 100 by dividing the farm’s
crop yield by the second-highest yield of that commodity reported
by a farm located in the same region as farm j, and then multiply-
ing by 100. An average is then taken across crops. The average is
used in order to help isolate the effects of international trade on
productivity from its effects on a farm’s optimal mix of crops.
The optimal output mix will depend on relative output prices,
which may change as a result of trade liberalization, and on rela-
tive changes in total factor productivity across crops. The second-
highest yield in a region is used as the divisor rather than the high-
est yield in order to mitigate against outliers and data errors.8 Scal-
ing in this way is somewhat analogous to a distance function where
a region’s second-highest yield represents that region’s productivity
frontier.
Approximately 92% of the farms in our sample produced only
traditional crops, while the other 8% produced both traditional
and non-traditional crops. Both types of farms, as producers of tra-
ditional crops, faced pressures to increase productivity due to im-
8
The highest and second–highest yields of a particular commodity, reported
among farms of the same region, received the value of 1.
 D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17 15

port competition. But those producing non-traditional crops also Table 4

had the opportunity to take advantage of the channels for raising Log of yield OLS estimates, entire sample.

productivity that exporting provides. Even though these channels Variable Coefficient Std. error
only apply to non-traditional crops, the yields that we measure Constant 1.195 **
0.354
(for traditional crops) would be expected to increase also because Trade Exposure Index (FT) 0.205*** 0.012
profit-maximizing farms will produce both traditional and non- ln(Irrigation) 0.465*** 0.012
traditional crops only if they are equally profitable at the margin. Hired Manager 0.070*** 0.011
ln(Adults/Land) –0.122*** 0.004
Achieving equal profitability at the margin for traditional and Modern Machinery 0.290*** 0.006
non-traditional crops could occur through the adoption of new ln(Storage Capacity/Land) 0.017*** 0.002
techniques for traditional crop production, switching land that is ln(Infrastructure/Land) –0.126*** 0.017
only marginally profitable for traditional crops into non-traditional Formal Title 0.041*** 0.006
Gender 0.072*** 0.007
crops (thereby increasing the average yield on the remaining tradi-
ln(Age) 0.930*** 0.185
tional cropland), or both. This implies that the effects of trade on [ln(Age)]2 –0.134*** 0.024
farm yields may differ between these two types of producers, Basic education 0.032*** 0.007
and that estimating a separate model for each group may be appro- High school education 0.100*** 0.010
priate. Because producers are unlikely to fall at random into the Technical education 0.107*** 0.017
Superior education 0.153*** 0.013
two groups, an endogenous switching regression model is
suggested. R2 0.29
Adjusted R2 0.29
As part of the endogenous switching regression model, let the
adoption of non-traditional crops be a dichotomous choice, where Notes: The model also includes dummy variables for agro-ecological zones and
farm jdecides to grow these crops (Bj = 1) if they perceive a net po- regions whose estimated coefficients are not reported here.
⁄
Denote significance at the 1% level.
sitive benefit (Bj ) and does not grow them otherwise (Bj = 0). This **
Denote significance at the 5%.
decision-making criterion can be represented by a probit model: ***
Denote significance at the 10%.

Bj ¼ g zj þ ej ; ð6Þ

where ej is the error term and the vector zj includes the same vari-
ables that are in kj, except for FTj (which could not explain the pres-
ence of non-traditional crops). zj also includes a dummy variable for
the farm’s residency status (whether or not the farm family lives on
the farm), a dummy for whether or not there is forestland on the
farm, a dummy for the presence of mechanical irrigation, and a
dummy for the presence of oxen on the farm. Farmers living on
the farm are expected to be more likely to have non-traditional
crops, since they require care during the winter season (more easily
provided if the farmer lives on-site). The presence of forestland on
the farm suggests a diverse pool of agricultural products, which
may be associated with a greater likelihood of non-traditional
crops. The presence of mechanical irrigation suggests a more mod-
ern and dynamic farm, which may also be associated with non-tra-
ditional crops. In contrast, the presence of oxen suggests more
traditional methods of production and a lower likelihood of non-
traditional crop production.
The second stage of the switching regression model involves
estimating separate equations for yields on the two groups of
farms, with an inverse Mills ratio variable (derived from the probit
results) included in each equation in order to control for sample
selectivity.
Results and discussion
Results of OLS estimation of the farm-level model for the entire
sample of farms are shown in Table 4.9 The farm-level trade expo-
sure index is positive and statistically significant. When expressed in
terms of elasticities, the results indicate that a 1% increase in the
farm-level trade exposure index is associated with an increase of
about 0.05% in crop yields.10 This effect is not large in magnitude
but it is statistically significant. Other things equal, an increase in
9
We also ran the farm-level models in this paper with the dummy variables for
agro-ecological zones and government regions replaced by dummy variables for
comunas. Comunas (districts) are minor administrative divisions in Chile. The data
reported 177 comunas with traditional crops within the area under study. The results
were generally similar to the results reported here.
10
An alternative model using the log of the tradability index rather than the
tradability index itself was also tried, with similar results to those shown here.
the trade exposure index of 0 to a value of 1 would be associated
with about a 20% increase in crop yields.
Results for the other variables in the farm-level model are gen-
erally plausible. Since the dependent variable is the log of yields,
elasticities are equal to the coefficient estimates except for age
where the log of age and its square enter into the model. The re-
sults indicate that the age of the farmer has dome-shaped relation-
ship with crop yields, consistent with other studies on age or work
experience and productivity. The peak of the dome occurs at about
33 years of age, which is younger than that found in many studies
of age and earnings but consistent with a country undergoing sig-
nificant change in its agricultural sector (especially in the 1990s)
where young farmers may have a comparative advantage over old-
er farmers in knowledge of new technologies and innovative agri-
cultural management techniques. At the sample mean of about
56 years, the elasticity of yields with respect to age is –0.14. The re-
sults also indicate that higher levels of education are associated
with greater productivity as one would expect. Farm management
offers opportunities for applying education through the use of new
technologies and management techniques, especially in a country
transitioning from traditional to modern agriculture like Chile (Lo-
pez and Valdes, 2000).
The negative estimated coefficient for the number of adults in
the farm household per hectare may appear surprising. However,
if there is a thin market for hired labor, then a larger number of
adults in the farm household would imply a lower (shadow) price
for labor, favouring the use of labor-intensive production tech-
niques over techniques intensive in other inputs, including land.
The result would be less land-intensive production methods and
therefore lower yields. The negative estimated coefficient for the
infrastructure per hectare variable, which measures the presence
of roads (and other similar infrastructure) within or immediately
adjacent to the farm, may also appear surprising. However, as most
rural roads in Chile are dirt roads that produce high levels of dust
during the dry spring/summer season, the negative sign is not nec-
essarily counterintuitive. Dust can reduce agricultural yields be-
cause it reduces plant respiration and facilitates the presence of
pests and plant diseases.
The first-step results from the switching regression model are
reported in Table 5. Marginal effects are included; they indicate
16 D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17

Table 5
Probit results for switching regression model and marginal effects.

Variable Probit estimates Marginal effects

Coefficient Std. error Coefficient Std. error
Constant –7.679*** 1.270
ln(Land) 0.156*** 0.007 0.017*** 0.001
ln(Irrigation) 0.753*** 0.040 0.081*** 0.004
Hired Manager 0.181*** 0.028 0.022*** 0.004
ln(Adults/Land) 0.050** 0.023 0.005** 0.002
ln(Storage Capacity/Land) 0.041*** 0.004 0.004*** 0.000
ln(Infrastructure/Land) 0.053** 0.019 0.006*** 0.002
Mechanical Irrigation 1.257*** 0.049 0.303*** 0.003
Oxen 0.033*** 0.026 0.004 0.018
Live on Farm –0.124*** 0.036 0.013*** 0.004
Forests 0.057** 0.018 0.006*** 0.002
Formal Title 0.206*** 0.020 0.021*** 0.002
Gender 0.042* 0.022 0.004** 0.002
ln(Age) 2.362*** 0.650 0.255*** 0.070
[ln(Age)]2 –0.290*** 0.083 0.031*** 0.009
Basic Education 0.078** 0.025 0.008*** 0.003
High School Education 0.285*** 0.032 0.036*** 0.005
Technical Education 0.412*** 0.046 0.060*** 0.009
Superior Education 0.518*** 0.038 0.080*** 0.008
Pseudo R2 0.15

Notes: The model also includes dummy variables for agro-ecological zones and regions whose estimated coefficients are not reported here. The marginal effect for a dummy
variable is the discrete change from 0 to 1.
*
Denote significance at the 1% level.
**
Denote significance at the 5% level.
***
Denote significance at the 10% level.

Table 6
Log of yield estimates with farms classified by types of crops produced.

OLS estimates Second stage of switching regression model

Farms with traditional and Farms with only traditional Farms with traditional and Farms with only traditional
nontraditional crops crops nontraditional crops crops
Variable Coefficient Std. error Coefficient Std. error Coefficient Std. error Coefficient Std. error
Constant 2.725* 1.514 1.157*** 0.370 4.304*** 1.526 3.385 0.382
Trade Exposure Index (FT) 0.410*** 0.060 0.194*** 0.013 0.438*** 0.060 0.239*** 0.013
ln(Irrigation) 0.515*** 0.054 0.455*** 0.012 0.419*** 0.056 0.283*** 0.014
Hired Manager 0.074*** 0.027 0.061*** 0.012 0.028 0.027 –0.003 0.012
ln(Adults/Land) –0.173 0.019 –0.118*** 0.004 –0.140*** 0.020 –0.087*** 0.004
Modern Machinery 0.244*** 0.024 0.291*** 0.006 0.227*** 0.024 0.265*** 0.006
ln(Storage Capacity/Land) 0.017** 0.007 0.016*** 0.002 0.005 0.007 0.000 0.002
ln(Infrastructure/Land) –0.105 0.068 –0.129*** 0.018 –0.078 0.068 –0.092*** 0.018
Formal Title 0.072*** 0.024 0.036*** 0.006 0.023 0.025 –0.035*** 0.006
Gender 0.053** 0.026 0.072*** 0.007 0.031 0.026 0.045** 0.007
ln(Age) 0.173 0.775 0.951*** 0.191 –0.388 0.776 0.255 0.070
[ln(Age)]2 –0.035 0.099 –0.135*** 0.025 0.033 0.099 –0.031* 0.009
Basic Education 0.016 0.032 0.032*** 0.008 0.007 0.032 0.014* 0.008
High School Education 0.072* 0.037 0.100*** 0.010 0.017 0.038 0.022** 0.011
Technical Education 0.080 0.050 0.102*** 0.018 –0.000 0.051 –0.012 0.018
Superior Education 0.106** 0.040 0.147*** 0.015 0.002 0.043 0.003 0.016
Inverse Mills Ratio –0.203*** 0.030 –0.285*** 0.013
R2 0.31 0.29 0.31 0.30
Adjusted R2 0.30 0.29 0.31 0.30

Notes: The model also includes dummy variables for agro-ecological zones and regions whose estimated coefficients are not reported here.
*
Denote significance at the 1% level.
**
Denote significance at the 5% level.
***
Denote significance at the 10% level.

the effect of a one unit change in an exogenous variable on the
probability that the farm will have non-traditional crops. The re-
sults indicate that farms with higher levels of fixed or quasi-fixed
inputs, formal title to their land, and higher levels of education
are more likely to adopt non-traditional crops. The farmer’s age
has a dome-shaped relationship with the probability of adopting
non-traditional crops, with the peak of the dome occurring at
about 59 years.
The second-step results of the endogenous switching regression
model, that is the separate production functions for farms sepa-
rated according to whether they produce non-traditional crops,
are shown in Table 6. For comparison purposes, OLS results for
these two groups of farms are also shown in Table 6. The last
row of the switching regression results shows that the estimated
coefficients on the inverse Mills ratio variable are statistically sig-
nificant, implying that self-selection occurs.
The results indicate that the farm-level trade exposure index
has a positive and statistically significant impact on yields for both
groups of farms, whether using the switching regression model re-
sults or the OLS results. In both cases, though, the estimated coef-
 D.A. Fleming, D.G. Abler / Food Policy 41 (2013) 11–17
ficient on the trade exposure index variable is larger for farms pro-
ducing both traditional and non-traditional crops than for farms
producing only traditional crops, and this difference is statistically
significant in both cases. Interestingly, the estimated coefficients of
the switching model on the trade exposure index for both groups
of farms are also larger than the estimated coefficient for all farms
from the OLS model in Table 4. These findings indicate that it is
important to take into account the determinants of crop produc-
tion choices when estimating the impact of trade on crop produc-
tivity. When expressed in terms of elasticities, the results from the
switching regression model indicate that a 1% increase in the farm-
level trade exposure index is associated with an increase of about
0.1% in crop yields for farms producing both traditional and non-
traditional crops, and an increase of about 0.06% for farms produc-
ing only traditional crops. Other things equal, an increase in the
trade exposure index of 0 to a value of 1 would be associated with
about a 44% increase in crop yields for the former group of farms
and about a 24% increase in yields for the latter group.
These results confirm that both groups of farms responded to
import competition by increasing productivity. However, our re-
sults show that farms in the group producing non-traditional crops
also took advantage of the channels for raising productivity due to
exposure to production of exportables. Based on the literature,
these channels include international technology and knowledge
spillovers as a result of trade, gains in productivity due to product
specialization as a result of trade, and learning by doing through
exporting. Although we do not have evidence of exactly which
channels are responsible for the yield gains on these farms, our re-
sults indicate that productivity for traditional crops is higher on
farms exposed to export markets, signaling the importance of spill-
overs from trade to farmers.
Summary and conclusions
This study analyzes the relationship between agricultural trade
and agricultural productivity in Chile, a middle-income country
that has seen significant growth in agricultural trade in recent dec-
ades. We used an agricultural commodity trade exposure index in a
cross-sectional analysis of Chilean farms analyzing the relationship
between trade exposure and yields of traditional crops. We also in-
cluded a switching regression model designed to explain whether a
farm (1) produces only traditional crops, or (2) produces both tra-
ditional and non-traditional crops. Traditional crops in Chile are
largely import-competing, while non-traditional crops are largely
exportables.
The switching regression results show that international trade
has a larger effect on crop yields for farms producing both tradi-
tional and non-traditional crops than for farms producing only tra-
ditional crops. The implication is that exporting creates channels
for productivity growth above and beyond those created by
importing even in traditional crops. Based on the literature, these
channels include international technology and knowledge spill-
overs as a result of trade, gains in productivity due to product spe-
cialization as a result of trade, and learning by doing through
exporting. More research is needed on the relative importance of
each of these channels in agricultural trade, although our results
suggest the importance of spillovers for the entire farm (as gains
from exporting affect productivity for importables).
Our study is not entirely exempt from endogeneity issues that
can arise in empirical studies analyzing the effects of trade. It can
be difficult to determine if countries (or firms or farms) export
more because of pre-existing high levels of productivity or if trade
openness is a factor spurring productivity growth. However, in our
17
case, the measure of trade openness for each crop is calculated at
the national level and is exogenous to any individual farm, which
is the spatial level at which our productivity measure is calculated.
Most studies in this area show that the trade dependency ratio
(from which we derive our farm trade exposure index) is a valid
and reliable indicator for assessing the impacts of trade, especially
when imports are also included (Yanikkaya, 2003). Another issue
to consider in interpreting our results is that they may be biased
due to missing variables (like soil quality) or problems with data
reliability. However, even allowing for the margins of uncertainty
that are inherent in any empirical work, and the particular realities
of Chile, our results are robust across estimations, suggesting that
farms do derive productivity benefits from international agricul-
tural trade.
References
Amiti, M., Konings, J., 2007. Trade liberalization, intermediate inputs, and
productivity: evidence from Indonesia. Am. Econ. Rev. 97, 1611–1638.
Badinger, H., 2007. Market size, trade, competition and productivity: evidence from
OECD manufacturing industries. Appl. Econ. 39, 2143–2157.
Barham, B., Clark, M., Katz, E., Shurman, R., 1992. Nontraditional agricultural
exports in Latin America. Latin Am. Res. Rev. 27, 43–82.
Bernard, A., Jensen, B., Schott, P., 2003. Falling Trade Costs, Heterogeneous Firms,
and Industry Dynamics. NBER Working Paper No. 9639.
Carter, M., Bradford, B., Mesbah, D., 1996. Agricultural export booms and the rural
poor in Chile, guatemala, and paraguay. Latin Am. Res. Rev. 31, 33–65.
Coe, D., Elhanan, H., Hoffmaister, A., 1997. North–south R&D spillovers. Econ. J. 107,
134–149.
Edwards, S., 1998. Openness, productivity and growth: what do we really know?
Econ. J. 108, 383–398.
Estrades, C., Terra, M.I., 2012. Commodity prices, trade and poverty in uruguay.
Food Policy 37, 58–66.
FAO. FAOSTAT Rome: Food and Agriculture Organization of the Unite Nations. 2008.
<http://faostat.fao.org>.
Ferreira, P., Rossi, J., 2003. New evidence from brazil on trade liberalization and
productivity growth. Int. Econ. Rev. 44, 1383–1405.
Fleming, D., Abler, D., Goetz, S., 2010. Agricultural trade and poverty in Chile: a
spatial analysis of product tradability. Agric. Econ. 41, 545–553.
Foster, W., Valdes, A., 2006. Chilean agriculture and major economic reforms:
growth, poverty and the environment. Region Dev. 23, 187–214.
Frankel, J., Romer, D., 1999. Does trade cause growth? Am. Econ. Rev. 89, 379–399.
Gopinath, M., Kennedy, P.L., 2000. Agricultural trade and productivity growth: a
state level analysis. Am. J. Agric. Econ. 82, 1213–1218.
Harrison, A., 1994. Openness and growth: a time-series, cross-country analysis for
developing countries. J. Dev. Econ. 48, 419–447.
Hay, D., 2001. The post-1990 Brazilian trade liberalization and the performance of
large manufacturing firms: productivity, market share and profits. Econ. J. 111,
515–529.
INE Agricultural Census. 1997 Santiago: Instituto Nacional de Estadísticas de Chile.
CD-ROM.
Jonsson, G., Subramanian, A., 2001. Dynamics gains from trade: evidence from
South Africa. IMF Staff Papers 48, 187–224.
Krishna, P., Mitra, D., 1998. Trade liberalization, market discipline and productivity
growht: new evidence from Inida. J Dev. Econ 56, 447–462.
Lopez, R., Valdes, A., 2000. Fighting rural poverty in Latin America: new evidence of
the effect of education, demographics, and access to land. Econ. Dev. Cult.
Change 49, 197–211.
Martin, W., Mitra, D., 2001. Productivity growth and convergence in agriculture and
manufacturing. Econ. Dev. Cult. Change 49, 403–422.
Melitz, M., 2003. The impact of trade on intra-industry reallocations and aggregate
industry productivity. Econometrica 7 (6), 1695–1725.
Pavcnik, N., 2002. Trade liberalization, exit, and productivity improvements:
evidence from Chilean plants. Rev. Econ. Stud. 60, 245–276.
Portilla, B., 2000. La Política Agrícola en Chile: Lecciones de tres Décadas. Serie
Desarrollo Productivo No. 68. (Santiago: División de Desarrollo Productivo y
Empresarial, CEPAL).
Reardon, T., Barrett, C.B., Berdegué, J.A., Swinnen, J.F.M., 2009. Agrifood industry
transformation and small farmers in developing countries. World Dev. 37 (11),
1717–1727.
Tybout, J., De Melo, J., Corbo, V., 1991. The effects of trade reforms on scale and
technical efficiency: new evidence from Chile. J. Int. Econ. 31, 231–250.
Winters, L., McCulloch, N., McKay, A., 2004. Trade liberalization and poverty: the
evidence so far. J. Econ. Lit. 42, 72–115.
Yanikkaya, H., 2003. Trade openness and economic growth: a cross-country
empirical investigation. J. Dev. Econ. 72, 57–89.