Maize yields in developing countries

Since 1960, the differences in the trajectories of maize yields that different countries have followed respond to socioeconomic and geographic factors. In China, the average yield in the 1980s exceeded the three tons per hectare (Table 1), which was associated to the changes in agrarian and technological policies. In this decade, the system of communes in effect since 1950 was replaced by the system of family responsibility, with which the peasant in family farms produced and sold his harvest to the state through contracts (^{McMillan et al., 1989}; ^{Fan, 1991}; ^{Lin, 1992}; ^{Salvador, 2008}). Additionally, research institutes were allowed to self-finance through the sale of applied technology (^{Pray et al., 1997}), with which agricultural research was fostered and highly productive agronomic systems were established thanks to irrigation (^{Wang, 2000}). This allowed for China to become global leader in agrarian biotechnology by the end of the 1990s, being one of the countries with highest investment in rural research and development in the world (^{Huang and Rozelle, 2009}).

In Argentina, since the decade of the 1990s a structural change of the agricultural sector took place, due to the technological advancement, the emergence of new forms of organization between producers and suppliers, and to the changes in economic and fiscal policy, with which it was managed to obtain maize yields of more than four tons per hectare (Table 1). The “sowing pools” arose as new forms of organization that attained agreements for producers to be in charge of performing farming tasks, while the suppliers granted technical resources, financing and transport, with the aim of generating scale economies and high yields, allowing for earnings to be distributed at the end of the harvest (^{World Bank, 2006}; ^{Sturzenegger and Salazni, 2007}; ^{Lence, 2010}).

The increase in maize yields in Argentina was due to diverse factors such as investment in research and development by transnational agrichemical companies, as well as the increase in use of fertilizers, the incorporation of supplementary irrigation, and the renovation of machinery and equipment, culminating with the adoption of transgenic seeds since the 1998/1999 cycle (^{Gear, 2006}; ^{Gutman and Lavarello, 2007}). This was possible, to a large extent, because of the free trade policy that reduced the imports tariffs on inputs and fixed assets, which, in turn, allowed a greater flexibility in the commercialization of the crops (^{Lence, 2010}).

In Brazil, the significant changes in yields began since the 1990s, decade when the trade policy allowed greater access to international markets, during which strong investments were made on infrastructure and agricultural research and a strengthening of the private sector in the agriculture and livestock sector (^{Rosegrant et al., 2006}). Also in the 1990s, improvements took place in the yields of different crops derived from the research and development that EMBRAPA (Empresa Brasileira de Pesquisa Agropecuaria) fostered, institute in charge of agricultural research, both in local and state institutions. This drive allowed for research centers to have a national presence and to develop new varieties of seeds of various crops, among them, maize (^{Magalhaes and Diao, 2009}).

In 2008/2009, approximately ten years after the Argentinian case, the sowing of transgenic maize was approved, which led to an increase in the yields’ trend, as had happened in Argentina (^{Parentoni et al., 2013}).

In sum, the increases in maize yields in the main developing countries are linked to: 1) trade policies, with which there was access to external markets, both to export and to import inputs; 2) research and development applied to agriculture, as well as the adoption of transgenic seeds; and 3) strengthening of commercial chains, coordinated by the state in the case of China and promoted privately in the case of Argentina and Brazil.

Maize yields in Mexico

Maize is a crop native to Mexico. The first vestiges of the plant, called teocintle, were found in Tehuacán, Puebla, and have an age of between 4500 and 7000 years (^{McClung de Tapia, 1997}). The creation of maize varieties with higher yields, more resistant and apt to be sown in different climates was a process that took Pre-Hispanic peoples thousands of years. This process involved seed selection, sowing techniques, and its cultivation in different climates (^{Benz, 1997}). Presently, there are more than forty maize races and hundreds of varieties.

In Mexico, in the 1970s and beginning of the 1980s, the state was the promoter of agricultural and livestock activity, so that institutions were created in favor of agricultural producers that had government infrastructure which carried out stockpiling, storage, commercialization and distribution of grains; likewise, there were large subsidies both to producers and to consumers of maize and its byproducts. Facing climate fluctuations and the scarce transfer of technology, guarantee prices arose to foster production, which increased in a relevant way the public expenditure, which is why at the end of the 1980s and in the decade of the 1990s, economic and constitutional reforms were implemented that liberalized various sectors of the economy, among them the agriculture and livestock sector, with which the government ceased to play a prominent role in the commercialization of agricultural and livestock products, the guarantee prices were eliminated, and the subsidies were dissociated (^{Appendini, 2005}).

In Mexico, the trade policies that liberalized the maize market with the entry of NAFTA in 1994 do not seem to have generated an important increase in the average of maize yields at the national level (Figure 1).

Source: Authors’ elaboration based on data from FAO’s Statistics Division (FAOSTAT).

Figure 1 Mexico: Maize yields from 1961 to 2012. Figures in tons per hectare. 

However, there is great heterogeneity in the yields at the regional level, where the highest yields were concentrated in the Northeast and Bajío regions of the country, situation that contrasts with the Center and South regions, where in some states the average yields did not exceed the three tons per hectare in the period between 2000 and 2010 (Figure 2).

Figure 2 Average maize yields in Mexico (2000-2010). Figures in tons per hectare. 

According to data from the Agriculture and Livestock Census 2007, published by the National Institute of Statistics and Geography (Instituto Nacional de Estadística y Geografía, INEGI), there are more than two million producers devoted to maize cultivation. It is important to mention that it was the last census that INEGI performed up until this study, while the previous Agriculture and Livestock Census was in 1991.

According to data from the Agrifood and Fishing Information Service (Servicio de Información Agroalimentaria y Pesquera, SIAP)⁷ and the Statistics Division from the Food and Agriculture Organization of the United Nations (FAO)⁸, currently, the most common maize varieties sown in Mexico are white and yellow. In 2010, 23.3 million tons of maize were produced in Mexico, of which approximately 91.0 % corresponded to the white maize variety and the rest to the yellow. This allowed Mexico to be the fourth maize producer, after the United States, China and Brazil, globally, with a participation of 2.8 % in that year.

The differences at the regional level can be explained in part by the new instruments for promotion of agricultural and livestock production and trade that were created starting in 1994, such as PROCAMPO, and the programs for support to commercialization operated by Supports and Services to Agriculture and Livestock Commercialization (Apoyos y Servicios a la Comercialización Agropecuaria, ASERCA), which is currently a decentralized administrative organization of the Ministry of Agriculture, Livestock Production, Rural Development, Fishing and Food (Secretaría de Agricultura, Ganadería, Desarrollo Rural, Pesca y Alimentación, SAGARPA). These new policies allowed for capitalized producers to be organized according to the openness and promotion of ejido privatization, controlling the market and trading their harvests directly, thanks to their competitive abilities, complementary supports, and backing by ASERCA; the large-scale producers located in the Northwest and Bajío regions of the country were especially favored (^{Aguilar, 2004}). In particular, an important growth took place in maize yields of the irrigation zones where commercial producers were concentrated, which differed from the yields in the rainfed zones that did not exceed the two tons per hectare during the first decade of the 21^st century, and which concentrated the small-scale or subsistence producers (^{Yunez and Serrano, 2010}).

Literature review

Various previous studies have focused on understanding the determinants of maize yields. ^{Evenson and Kislev (1973}) found a strong and persistent relation between agricultural research and maize yields, with a sample of 49 producing countries. ^{Gowon et al. (1978}) analyzed the response of yields in California, Colorado and Utah, states of the American Union, in face of factors such as irrigation under a specific type of soil and whose results differed, depending on the phenological stages of the crop. Other more recent studies (^{Kaylen et al., 1992}; ^{Schlenker and Roberts, 2006}; Roberts et al., 2012) incorporated agronomic knowledge to climate variables such as temperature and precipitation to relate them econometrically with the yields and, in the case of the last two studies, to project the possible impact of climate change on maize yields in the United States.

In their turn, ^{Kaufmann and Snell (1997}) estimated a model for maize yields in the municipalities of the United States, incorporating socioeconomic determinants such as market conditions, production scales, technology adoption and public policies, as well as climate variables. ^{Ruttan (2002}) performed a detailed analysis of the technical, institutional and resource restrictions for the growth of agricultural productivity in the last 50 years of the 20^th century, where he concluded that most of the developing countries should invest more in agricultural research and development, technical training and infrastructure to increase the yields in this century. ^{Kamruzzaman et al. (2000}) examined the effect of credit on the yields and on the technical efficiency of rice production in the district of Comilla, Bangladesh, and found that the producers who have credit have a higher yield potential than those who do not, concluding that credit has a positive impact to increase the technical efficiency of producers.

The effects of climate change have also been studied through regressions of the yields at the level of municipality in climate variables. ^{Olivera (2013)} finds that climate change will have negative effects on maize yields in Mexico, especially in rainfed zones. Likewise, ^{Galindo (2009)} indicates that the impacts on the yields derived from climate change will be heterogeneous by regions and each productive cycle will have different response sensitivities to temperature and precipitation in Mexico.

Data

Derived from the literature review, it was determined that the factors that could impact a sustained increase in maize yields in Mexico would be those related to the production (infrastructure, biotechnology, use of agrichemicals, and use of credit), geography, climate, agriculture and livestock policies, and socioeconomic factors.

Therefore, the following data were obtained, relative to the use of chemical fertilizers, improved seed, insecticides, tractor, natural fertilizer, controlled burning, credit, as well as the scale and schooling, aggregated at the municipal level for the cultivation of maize grain, from the Agriculture, Livestock and Forest Census 2007, provided by INEGI, which were requested from that organization to carry out this study. The irrigation variable was obtained from the data of the irrigation surface from 2005 of the Agrifood Information System for Consult (Sistema de Información Agroalimentaria de Consulta, SIACON) of the SAGARPA.

The socioeconomic variables of total population, availability of public services and population recipient of public health services, were obtained from the Population and Housing Census 2005 elaborated by INEGI. These were included to get a better coverage of possible variables that could influence the maize yields, through their indirect incidence on the labor market. For example, ^{Barbier and Burgess (1996}) find a positive relation between the population and the agricultural surface sown in Mexico, which impacts directly the agricultural yields. ^{López-Feldman (2012}) estimates a positive relation between the marginalization index of the municipalities in Mexico, and the expansion of the agricultural frontier. In turn, ^{Deininger and Minten (1999}) find a negative relation between poverty and agricultural land use in Mexico.

Likewise, the price of maize was included, which represents the average of the price between 2006 and 2007 per municipality. The variable Amount of PROCAMPO per producer at the municipal level was obtained from the Municipal System of Databases (Sistema Municipal de Bases de Datos, SIMBAD) of the INEGI for the years of the Census and represents the average amount of transferences performed to producers by PROCAMPO in each municipality.

Geographic data were also obtained, such as distance to the closest city, density of roads, distance to the coast, length of the road, and slope of the land from the geospatial data of Use of Land and Vegetation from INEGI. These variables have an impact on the surface sown and on the conversion of land use (^{Barbier and Burgess, 1996}).

Similarly, the characteristics of the land per municipality were incorporated, using the geospatial soil data elaborated by the National Institute of Forest, Agriculture and Livestock Research (Instituto Nacional de Investigaciones Forestales Agrícolas y Pecuarias, INIFAP) and the National Commission for the Knowledge and Use of Biodiversity (Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, CONABIO) from 1995. These characteristics were classified per type of soil, physical composition and texture. There are 21 types of soil, six categories for the physical composition and three for texture.

Finally, in the case of the climate variables, the data of accumulated precipitation and average growing degree days (GDD), of the years 2003 to 2007 per municipality present during the growth stage of maize, were included. The data concerning the precipitation and temperature at the level of meteorological station were provided by the National Meteorological Service and processed by the authors to obtain data at the municipal level in agricultural zones.

The growing degree days are used to capture the exposure to heat during the cultivation in the growth stage, which is why with the data of temperature the calculation was carried out in the following way (^{Roberts et al., 2012}):

min{(Th+Tl) /2 ,Tm}-Tb

where Th is the maximum temperature, Tl the minimum temperature, and Tb the base temperature (10 °C) and Tm the upper limit (30 °C).

Tables 2a and 2b from the Annex show the descriptive statistics of the main variables. Table 2a present averages per municipality for all the variables used in this study and Table 2b presents the averages at the national level. It is done this way because there are important variations in terms of the degree of use of some inputs in maize production, whether if measured as proportion of the national surface or as average of the municipal surface. As can be observed, the penetration of improved seed is 12 % in average in the municipalities (Table 2a), but measured in terms of the total surface sown with maize in the country (Table 2b) represents 18 %, figure close to that reported by other studies⁹.

Empirical model

The econometric estimation of the yields as function of the variables specified in the previous subsection is detailed in this section. Three specifications of the model are performed: 1) controlled for technological and credit variables; 2) geographic, climate and access to infrastructure data are added; and 3) socioeconomic variables are included. The model is based on ^{Kaufmann and Snell (1997}), and it is expressed in the following way:

where Rend indicates yields (tons of maize per hectare), c represents the cycle (Spring-Summer or Fall-Winter), and m the municipality; Tech represents the set of technology variables like use of improved seed, irrigation, fertilizers, insecticides, controlled burning, natural fertilizer, use of tractor, scale, credit and surface sown with white maize; Geog denotes the set of controls for geographic characteristics such as distance to cities, average slope, distance to the coast and type of soil, as well as precipitation and growing degree days (GDD). Finally, the vector Socio contains socioeconomic variables such as schooling, total population, availability of public services, population that has access to public health services, average amount of PROCAMPO assigned per producer and mean rural maize price.

Each specification is estimated for three cases: using the units that sow in the spring-summer cycle (PV), fall-winter (OI), and in both cycles. In the three cycles it is controlled for dummies of federal state µ _e and, additionally, in the third cycle it is controlled for sowing cycle µ _c .