The decline in maize prices, biodiversity, and subsistence farming in Mexico.

Seals, Alan ; Zietz, Joachim

I. Introduction

In terms of caloric intake, maize is the number one crop in the world according to the statistics of the Food and Agriculture Organization (FAOSTAT). The issue of preserving the plant's genetic diversity is thus of significant policy importance. Even though several scientific studies have been conducted by now, the latest one being the European Commission report by Messean et al. (2006), a resolution of the issue of transgenic (1) contamination of Mexico's native maize varieties is not likely to occur in the near future. (2)

As the debate on genetically modified (GM) maize is ongoing, not only in Mexico following the 1998 moratorium on growing GM maize, (3) but also elsewhere, such as in Europe, this paper will add some fundamentally economic arguments to the debate as it pertains to Mexico. (4) In particular, we examine how the current biodiversity of maize in Mexico may be endangered as subsistence farmers, who maintain and propagate the biodiversity, are faced with declining market prices for their produce as a consequence of the large and rapidly rising maize imports from the U.S. These imports not only worsen the terms of trade of subsistence farmers, but, as much of the imported maize is of the GM variety (such as Bt corn (5)), they also raise the risk of lower yields as indigenous varieties of maize may lose their resilience to environmental stress through contamination with GM maize.

The paper is organized as follows. The subsequent section will provide some institutional background on the connection between biodiversity and maize farming in Mexico. This is followed by a section that examines empirically the impact on the behavior of Mexican maize farmers of arguably the most important economic event that has affected them since the mid 1990s: the very large increase in maize imports from the U.S. This is done for two reasons. First, there is not much point in arguing about the loss in biodiversity through the impact of GM maize if one cannot predict that enough subsistence farmers with an interest in indigenous maize varieties will be left a decade from now to take on the job of preserving the biodiversity of maize. Second, by observing farmers' reactions to a major change in their economic environment, it may be possible to distill what drives farmers' behavior. That, in turn, will help predict how farmers may react to the lower yields that may arise from a contamination of their indigenous maize varieties with GM maize.

The section following the empirical analysis discusses to what extent the observed empirical regularities are consistent with a model of rational behavior of farmers. The model provides, among other things, an explanation of the puzzling fact that output of maize has reacted very little to the sharp decrease in the price of maize since NAFTA was enacted in 1994 (Ackerman et al. 2003; Nadal 2000 and 2002). Based on this model, some tentative policy recommendations can be formulated on what set of economic policies and incentives may support the objective of preserving the current biodiversity of maize in Mexico.

II. Institutional Background

Since the beginning of the Green Revolution in the 1940's, modernization of agricultural practices in the developing world has attracted the attention of policy makers. Increasing the scale of farm production through technological innovation has regularly been promoted as a substitute for low-output indigenous agriculture. Subsistence farming is often viewed by governments as an indication of economic inefficiency, and its eradication is perceived as a harbinger for a modern economy. (6) However, such views ignore that subsistence farmers, throughout the world, promote and protect the genetic diversity of native crop species and thus provide a significant public service to all of humanity. Due to their diversity, traditional varieties generally outperform modern varieties in the adverse conditions that the indigenous farmers face. The rich diversity of domestic varieties (7) not only meets local consumption requirements, which may be very specific, (8) but it also minimizes the agronomic risks posed by drought, climatic change, soil degradation, and insect infestation (Perales et al. 2003).

The genetic diversity that subsistence farmers propagate is also valuable to modernized agricultural nations, such as the United States. Capital-intensive farming in the industrialized world has created an increasing demand for genetically modified seeds that are resistant to pests or certain chemical applications. Industrial agriculturalists, due to the restrictions of mechanical farm production, can not promote genetic diversity and are not yet required to fully internalize the environmental degradation attributable to commercial fertilizers and pesticides. Thus, mechanized agriculture necessarily renders high levels of crop diversity economically infeasible. Potential pitfalls that attend low levels of crop diversity become evident when severe crop damage occurs due to disease or pest infestation, as happened in the United States in 1970 when approximately 25 percent of the U.S. maize crop was destroyed by the southern leaf blight (Boyce 1996; Nadal 2000). (9) Due to the ecological pressure of pests and disease, the average commercial life of a modified seed is only about seven years (Boyce 1996). Commercial plant breeders must continually use the genetic material from different varieties of a crop to obtain the desired pest and disease resistant qualities. Off-farm (10) conservation methods, such as germ plasm banks, preserve the native varieties only at a specific moment in time and can not capture the evolutionary changes of the crop. Thus, off-farm conservation is only a complement, not a substitute to the on-farm conservation performed by the farmers.

The incentive structure, which motivates the production process of the subsistence farmer, is markedly dissimilar to that of the conventional cash-crop farmer. This fact is clearly evident when one considers that U.S. producers do not face the same environmental and financial constraints as Mexican subsistence farmers, who are generally relegated to isolated lands marginally unfit for industrial agriculture, with no access to credit. A farmer who employs large amounts of physical capital expects to make a profit, while the expectation of the peasant farmer is to sell the surplus crop (if any), after own-consumption needs and seed requirements are met. Ashraf et al. (2005) contend that the agricultural provisions of the North American Free Trade Agreement (NAFTA) have had no discernible effect on the Mexican subsistence farmer. The initial fear that NAFTA would destroy the indigenous farmers of Mexico by forcing them to compete with the heavily subsidized farmers of the United States appears unfounded, as Mexican subsistence farmers have shown no significant agricultural diversification away from maize during a period in which the average price of maize in Mexico fell by 50 percent. Ashraf et al. (2005) also show that 75 percent of all the farmers surveyed report growing maize as their principal means of subsistence, while only 12-22 percent reported maize as the primary cash-crop. Of the poorest farmers surveyed from 1991-2000, 89-92 percent reported that maize was their primary crop for subsistence and 56-57 percent reported they did not produce maize to sell in the market. A survey of peasant farmers in the Guanajuato region of Mexico by Smale et al. (2001) reveals that farmers unanimously recognize maize as a critical component of their livelihood and grow maize even when it is unprofitable to do so.

Mexican subsistence farmers use labor-intensive methods to cultivate several varieties of maize, (11) with different planting and harvest times, to hedge against environmental risk. (12) Accordingly, indigenous farmers, with smaller plots of land, have a comparative advantage in labor-intensive farming over their larger and less diverse counterparts. Seed varieties favored by modern agriculture require large amounts of chemical inputs and are bred for low-stress environmental conditions not suitable for the small-scale farmers in Mexico (Soleri and Cleveland 2001). Most indigenous farmers raise their crops on peripheral lands that are primarily rain-fed, as opposed to the heavily irrigated farmland of industrial agriculturists. However, the cultivation of different varieties of maize is not only implemented to mitigate the environmental constraints of production, where irrigation and fertilizers are not readily available. Smale et al. (2001) find the determining factor in the allocation of maize varieties is the differential in consumption preferences for specific varieties. Subsistence farmers have also been found to cultivate crop varieties for the purpose of ensuring that the seeds from these crops remain available in their community. Perales et al. (2005), in a study of maize diversity between neighboring towns in the Chiapas highlands, find that maize varieties are cultivated "distinctly" according to ethnolinguistic groups. The authors show that farmers continue to use local maize varieties even when a superior and otherwise acceptable substitute is available from neighboring farmers. Knowledge of genetic resources (13) is generally well-defined among indigenous communities, due to the significance of securing reliable food supplies (Bellon 2001). Yet, diffusion of genetic knowledge between different ethnolinguistic groups is often costly due to language and ethnic barriers (Perales et al. 2005). Reluctance, on the part of indigenous farmers, to substitute away from their local maize varieties is cited as one possible explanation for the persistence of native varieties.

III. Empirical Regularities

1. Data and Methodology

The empirical results make use of data published by the Food and Agriculture Organization (FAO). The FAO data set is rather limited and extends from 1991 to 2004 for most variables. There are no separate data on commercial and subsistence farmers available from FAO. The data used are defined in Table 1.

The estimates are based on the structural time series approach, which is also known as unobserved component modeling, as advocated by Harvey (1989, 1997) and as implemented, among others, by Koopman et al. (2000). (14) Univariate structural time series models can be expressed as

[y.sub.t] = [[mu].sub.t] + [[summation].sub.i] [[summation].sub.j] [[alpha].sub.ij][x.sub.i, t-j] + [[epsilon].sub.t] = 1, ..., T,

where [[mu].sub.t], is a time-dependent intercept term, which is modeled as a stochastic process, and where the [x.sub.i] are observed regressors as in ordinary least squares regression. The stochastic term [[mu].sub.t], captures unobserved influences driving the dependent variable. It is assumed to follow a random walk with time dependent drift ([[beta].sub.t]). The drift parameter itself may follow a random walk,

[[mu].sub.t] = [[mu].sub.t-1] + [[mu].sub.t-1] + [[eta].sub.t] [eta] ~ NID(0, [[sigma].sup.2.sub.[eta]])

[[beta]].sub.t] = [[mu].sub.t-1] + [[xi].sub.t] [xi] ~ NID(0, [[sigma].sup.2.sub.[xi]])

Both [[mu].sub.t] and [[beta].sub.t] are driven by white-noise disturbances, [[eta].sub.t] and [[zeta].sub.t]. These disturbances are assumed to be independent of each other and of [e.sub.t]. (15) The general trend model can be tested down to a simpler form, such as a model with no drift parameter, for which [[mu].sub.t] would be written as

[[mu].sub.t] = [[mu].sub.t-1] + [[eta]].sub.t-1] [eta] ~ NID(0, [[sigma].sup.2.sub.[eta]]),

or, for example, a model with deterministic trend, which arises when the disturbances [[eta].sub.t] and [[zeta].sub.t] have zero variance. OLS is a limiting case of the structural time series model. It arises when [[beta].sub.t] and the variance of the disturbance terms [[eta].sub.t] are both zero.

The advantage of the structural time series model over OLS is that it can capture movements in the data that are not represented by the observed independent variables. This can play a significant role in applications such as the present one where the data set is rather limited in the sense that potentially relevant variables are missing because they are not measured or are not known theoretically. In the absence of allowing for unobserved components in these cases, the left-out variables will typically show up in OLS estimates as spurious trends, unexplained lags on variables, or residual statistics that suggest misspecification. It should be obvious that the inclusion of unobserved stochastic components is a second-best approach, like all black-box methods. (16) Ideally, one would want to replace unobserved components with observed variables. Oftentimes, the movement of the unobserved components over time will provide some hints as to what variables may be driving them. Hence, unobserved component modeling may help in the process of identifying the data generating process. In fact, if all relevant variables are being employed in a particular application of structural time series modeling, no unobserved components should be statistically significant any longer and the model collapses to OLS.

2. Estimation Results

A key element in understanding the behavior of Mexican maize farmers is the relationship between maize imports from the U.S. and the producer price of maize in Mexico. Anecdotal evidence (Lambrecht 2005; Campbell and Hendricks 2006) suggests that farmers find it difficult to survive when the output price of maize drops. Most commentators take it for granted that the massive influx of U.S. maize into Mexico following the implementation of NAFTA in 1994 is responsible for the decrease in the maize price. A recent study by the World Bank (Fiess and Lederman 2004), however, appears to suggest that U.S. imports do not play much of a role for the price of maize.

Since there is little statistical evidence of a stochastic trend, the structural time series model that explains the maize price as a function of maize imports and maize yield collapses to OLS. A negative sign is expected for the explanatory variables imports and yield. The estimated equation in log-linear format for the time period 1991 to 2003 is given as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where p-values are provided in parenthesis underneath the estimated coefficients. P-values are also given for a test of first-order autocorrelation (Auto), the Ljung-Box test of autocorrelation up to lag order four (LB), the Jarque-Bera normality test (JB), and a test for heteroskedasticity (Het). None of the p-values suggest any statistical problem at conventional levels of statistical significance. The estimates suggest that a 10 percent rise in imports has lowered the maize price by 1.7 percent over the sample period. Since imports tripled over the period from the pre-NAFTA average for the years 1991 to 1993 to the year 2004, this elasticity estimate suggests that imports are responsible for about a fifty percent drop in the price of maize.

Based on previous research (Fiess and Lederman 2004) and anecdotal evidence (Lambrecht 2005; Campbell and Hendricks 2006), the acreage cultivated of maize has reacted little to the dramatic change in the price of maize since the implementation of NAFTA. This observation is consistent with regressions on the FAO data. Similar to the price equation, no unobserved component appears significant for the regression of acreage on the price of maize ([price.sub.-1]) and the consumer price index ([cpi.sub.1]), both lagged by one year, (17)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

Although there is no statistical problem evident with the estimated equation, it clearly does not explain acreage. Neither the price of maize nor the consumer price appears to influence acreage.

It is often suggested that maize farmers may be forced to leave the agricultural sector and migrate to the cities as economic conditions worsen on the farm (Lambrecht 2005). A worsening of conditions could be associated with lower output prices, rising inflation, or lower yields associated with a contamination of the maize crop with GM maize. The migration data used in this study are derived from FAO data on total population growth and agricultural population figures (Table 1). Migration is explained as a function of the acreage and yield of maize. As more acreage is planted, one would expect more work opportunity for agricultural workers. This should reduce migration. Similarly, as yields go up, everything else constant, subsistence farmers are better off. Again, this should reduce off-farm migration. Over the time period 1991-2004, the structural time series model contains a smooth trend, which is brought about by the variance of [eta] being zero in combination with the variance of [zeta] being positive. The estimated coefficients of the fixed regressors and some statistical adequacy tests are given as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

Starting the regression sample one year later in 1992 raises the parameter values of both area and yield considerably. At the same time, the unobserved trend becomes statistically insignificant. An OLS regression over the period 1992-2004 yields

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where none of the statistical adequacy tests suggests a statistical problem.

The regressions explaining off-farm migration for Mexico for the 1990s and early 2000s suggest that increases in both acreage and yield have a retarding effect on migration. Given that maize acreage has changed little since the early 1990s, while yields have been rising somewhat, the results indicate that off-farm migration would have been higher in the absence of these two trends. They also reveal that a drop in yields that may be brought about by GM maize contaminating the traditional maize varieties may have significant consequences for off-farm migration.

IV. A Model to Explain the Observed Behavior

The purpose of this section is to check whether the empirical regularities described in the last section are consistent with common assumptions of maximizing behavior on the part of farmers. This is done by postulating a simple utility maximization problem for a maize farmer and checking whether the empirical findings can be encompassed by this model. An analysis of this type is useful for two reasons. First, there has been some suggestion (Fiess and Lederman 2004) that Mexican maize farmers have somehow behaved irrationally in response to the large decrease in the maize price by increasing production. Second, without an understanding of the core driving forces behind farmers' behavior, it is difficult to formulate economic policy prescriptions about preserving biodiversity.

Hymer and Resnick (1969) develop a theoretical model to explain the positive production response of subsistence farmers who are faced with price volatility. Barnum and Squire (1980) extend Hymer and Resnick's work to incorporate a number of different scenarios where farmers can choose among heterogeneous crops, the acreage they cultivate, and between farming and non-agricultural employment. However, neither Hymer and Resnick (1969) nor Barnum and Squire (1980) distinguish between tradable and non-tradable agricultural output. We extend Barnum and Squire's model to include the farmer's choice between consumption of market goods and the own-consumption of agricultural goods.

The farmer's decision problem is to maximize a utility function,

u = [[theta].sup.[alpha]][(m - [bar.m]).sup.[beta]] [l.sup.[delta],

where utility depends on consuming (a) a given fixed amount of maize that is taken from own production ([theta]), (b) household products that are purchased from outside the farm (m and [bar.m]), and (c) leisure (l). The parameters a, [beta], and d identify weights. The preference for own consumption ([theta]) is discussed in a previous section, but it should again be stressed that [theta] is the farmer's preference for a specific maize variety which is endemic to the farmer's region or particular ethnicity. A key component of the farmer's utility function is its dependence on a certain minimum number of household products which need to be purchased off the farm ([bar.m]). Following the Stone-Geary utility function, household products purchased off-farm (m) raise utility only to the extent that their quantity exceeds this minimum requirement.

Utility is maximized subject to a time constraint and a budget constraint. According to the time constraint, total available time, which is set to unity for simplicity, has to be divided between leisure (l), and time spent working on the farm (n), 1 = n + l. Maximization of the utility function is also subject to the budget constraint

p(y - [theta]) = m,

where the left-hand side is the revenue from selling maize in the open market and where the right-hand side contains all expenditures on off-farm goods and services. Revenue from selling maize is the product of the price of maize relative to that of off-farm products (p) (18) and the quantity of production that is not destined for own consumption (y - [theta]). Production is assumed to be given by the function

y = z[(1 - l).sup.[phi]],

where z is a productivity parameter, perhaps representing the idiosyncratic genetic characteristics of the farmer's indigenous maize. There are two production factors: land or acreage planted and labor (n = 1 - l). Only labor is treated as a decision variable in production. The parameter [phi] is the corresponding weight of labor in the production function. Land is assumed constant and normalized to unity for simplicity. It is assumed that the farmer does not enter the credit market. Hence, all off-farm purchases have to be paid for from the market sale of maize.

The specification of the utility function with respect to market goods suggests consumption of market goods m exceeds the minimum [bar.m], otherwise utility could be zero or negative. A Kuhn-Tucker formulation replaces strict equality measures and the farmer's optimization problem is therefore given by

max [[theta].sup.[alpha]][(m - [bar.m]).sup.[beta]] [l.sub.[delta]] s.t.

p[z[(1 - l).sup.[phi]] - [theta]] [greater than or equal to] m, m [greater than or equal to] [bar.m], l [greater than or equal to] 0,

where variable n has been substituted out by the time constraint. The variables for off-farm purchases of household items (m) and leisure (l) are the farmer's decision variables. The Lagrangean for the farmer's optimization problem can be written as,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The Kuhn-Tucker conditions for a maximum are given by,

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The sign of the bordered Hessian for [lambda] > 0, [psi] = [xi] = 0 is positive which is necessary to establish a maximum. The key comparative static result [partial derivative]l/[partial derivative]p, is unambiguously positive (See Mathematical Appendix), which indicates that, as the relative price of maize decreases, leisure declines and, hence, farm labor increases, and with it output. As a result, the income effect of a price change dominates.

The theoretical specification is consistent with the fact that few subsistence farming households are completely autarchic. Subsistence farmers need to purchase market goods that they cannot produce at home (e.g., pharmaceuticals and professional medical care). It is easy to imagine that some of these market goods are also used as complementary goods to leisure, such as a television. Leisure is not worth as much without these complementary goods. As the output price of maize drops, fewer market goods would be available without a concurrent increase in agricultural work effort and additional market sales resulting from this increased work effort. Thus, the key contribution of the theoretical derivations is to show that when a farmer must consume a minimum amount of market goods and also has preference for own-consumption, the output response to a price decrease is positive.

The evidence provided by Ackerman et al. (2003), the empirical results of the last section, and the work of Fiess and Lederman (2004) suggest that the Mexican farming sector in total has not reacted to the price decrease in maize with a reduction in output. The fact output has not fallen in response to the sharp drop in the price of maize is fully consistent with the theoretical model. There is little irrational about this behavior when one considers the constraints maize farmers are likely to face.

It is interesting to postulate what may happen if the price of agricultural output were to fall to a point where farmers could no longer purchase a minimum quantity of market goods. Harris and Todaro (1970) argue that higher expected earnings in the non-agricultural sector will induce rural farmers to migrate to urban areas if those farmers are maximizers of expected utility. Our simple utility maximization model does not explicitly incorporate a stopping rule for agricultural production that is linked to deterioration in the terms of trade of subsistence farmers, although such an extension would be possible in principle. Barnum and Squire (1980) provide an example of a similar model which incorporates the basic Harris and Todaro (1970) predictions when time spent in nonagricultural employment is included as a choice variable. (19) Even without an explicit rule for farm out-migration, the model suggests an intuitively appealing explanation for out-migration. Given the lack of capital available to subsistence farmers and their positive output response to a price decrease, there must be some minimum threshold level of leisure and of utility that induces a farmer to leave the farm and to search for off-farm employment. One may speculate that the farm would be purchased and used by a more efficient farmer, which ensures that, across all farms, acreage does not fall but yields rise in the long run.

Given that the comparative static results are sufficiently consistent with the empirical evidence, it is interesting to hypothesize how subsistence maize farmers would react to the contamination of their fields with GM maize. Since GM maize is primarily used for feeding livestock, it is reasonable to assume that farmers would have trouble selling their crop in the market for domestic maize, and as a result would have to accept a lower market price. Thus, the reaction of the subsistence farmer to genetic contamination may be analogous to that of a decrease in the price of maize.

Given the difficulty of identifying infected maize, it would be almost impossible to stop the process of contamination. Most likely, contaminated maize would be reused as seed even if an infection is obvious if for no other reason than lack of funds on the part of subsistence farmers to root out the contamination and start with clean seed for several seasons. How a contamination is ultimately affecting the indigenous gene pool of maize and the properties of maize is an open question. However, it appears fairly certain that the total output of maize will be declining, at least for a short time, as farmers are unfamiliar with the agronomic properties of the new contaminated seed stock. In addition, the GM maize varieties are not intended for reuse as seed and GM maize is more dependent on fertilizer and pesticide, which subsistence farmers are not using to any significant degree. (20) In addition, the new hybrid maize varieties may be less resistant to severe weather, in particular drought, because GM maize is intended for irrigated fields. All this suggests an increase in the risk of catastrophic crop loss for subsistence farmers.

When seen in conjunction with the empirical analysis of the last section, the predictions of the theoretical model suggest at least two conclusions that are of relevance for the preservation of biodiversity in Mexico. First, further sharp increases in imports of maize from the U.S. will likely cause many subsistence farmers to leave their land and migrate to the cities of Mexico or the U.S. This is independent of whether there is any contamination of the indigenous varieties of maize with GM maize. The fact that, so far, maize output has reacted positively to the surge in imports from the U.S. and the subsequent large decrease in the price of maize should not be taken as a sign that Mexican maize farmers are not under stress. On the contrary, it is a sure sign that farmers do react to the price decrease and that they react rationally.

Their response entails more work effort, fewer purchases of off-farm products for household use, and, as a consequence, lower levels of utility. This will make off-farm migration ever more likely over time. However, if subsistence farmers leave the countryside in large numbers, the current levels of biodiversity can not be maintained: with no subsistence farmers, there is no biodiversity. Again, this is completely independent of the issue of contamination of the gene pool by GM maize.

Second, the analysis has suggested that a contamination of the indigenous varieties of maize with GM maize may have similar consequences as a further reduction in the relative price of maize. However, this conclusion is based on the as yet unproven assumption that any maize variety that is an unplanned hybrid of the indigenous varieties and GM maize will be more susceptible to environmental stress, such as droughts and pest infestation, than the current indigenous varieties and, as a consequence, average yields of maize farmers decline.

V. Summary and Conclusions

The purpose of this paper is to model the economic behavior of Mexican maize farmers in order to predict what would be needed from an economic perspective to ensure continued biodiversity.

To that end, the paper attempts to establish empirically the connection between the large imports of maize from the U.S., the price of maize, acreage planted, and off-farm migration. The results suggest that U.S. imports have depressed the price of maize. Acreage, however, has reacted little. Finally, both declining acreage and maize yields are key driving forces of off-farm migration.

The paper develops a simple theoretical model to examine whether the empirical results are consistent with rational behavior on the part of farmers and to suggest policy actions to maintain biodiversity. The comparative static properties of the theoretical model are consistent with the key empirical facts. In particular, it is shown that an increase in production is fully consistent with a declining relative price of maize. But as maize farmers work more and can afford ever fewer off-farm products, their utility levels decline, which will eventually induce them to leave the farm in search of employment in the urban areas of Mexico or the U.S. It is suggested that the contamination of the indigenous maize varieties with GM maize may be interpreted as an alternative unfavorable movement in the terms of trade that subsistence farmers face. As a consequence, they may in the long run react to such a contamination in a manner that is similar to that of a reduction in the relative price of maize: they choose to migrate off-farm as utility levels fall below certain threshold levels.

Off-farm migration, however, has significant consequences. First, as many indigenous farmers stop production, the maize gene pool will contract, possibly by a very sizable amount. Although it is difficult to foresee all the consequences of such a result, it does not appear to bode well for the future security of the world's food supply since Mexico is home to the world's only self-sustaining genetic repository for maize. Second, as farmers leave their land, possibly in large numbers, Mexico's cities are likely to experience significant stress when the now landless farmers arrive and are looking for employment. Based on past experience, it appears unlikely that a large number of former subsistence farmers will find employment. An increase in illegal immigration to the United States is a likely consequence.

In the light of these results, the key policy issue appears to be how to stop a sufficient number of subsistence farmers from leaving their land. That is the prerequisite of keeping biodiversity, even in the absence of GM maize contamination. Given political reality, maize will continue to be imported from the U.S. Some effort may be worthwhile to contain the growth rate of imports. If that is not politically feasible and the relative price of maize continues to decline, cash subsidies may be an option to keep farmers on the land. These subsidies would be the price to be paid for maintaining biodiversity. They would constitute a transfer scheme that internalizes the positive external effects that are derived from biodiversity. The subsidies would also be the price to pay to keep Mexican farm workers from illegally immigrating to the U.S. Since Mexico, the U.S., and the world at large reap the benefits of continued Mexican biodiversity, it appears sensible to pay for the subsidies from an international fund rather than from the budget of a single country.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Both the first and second terms are positive, which gives the result [partial derivative]l/[partial derivative]a > 0.

References

Ackerman, F., Wise, T.A., Gallagher, K.P., Ney, L. and R. Flores. 2003. Free Trade, Corn, and the Environment: Environmental Impacts of US--Mexico Corn Trade Under NAFTA. North American Commission for Environmental Cooperation. G-DAE Working Paper No. 03-06.

Ashraf, N., M. McMillan, and A. P. Zwan. 2005. My Policy or Yours: Have OECD Agricultural Policies Affected Incomes in Developing Countries? NBER Working Paper Series. Working Paper 11289.

Bellon, M.R., 2001. Demand and Supply of Crop Infraspecific Diversity on Farms: Towards a Policy Framework for On-Farm Conservation. CIMMYT Economics Working Paper 01-01. Mexico, D.E: CIMMYT.

Bellon, M.R., Michelle Adato, Javier Becceril, and Dubravka Mindek. 2005. Poor Farmers' Perceived Benefits from Different Types of Maize Germplasm: The Case of Creolization in Lowland Tropical Mexico. World Development 34(1): 113-129.

Boyce, J. 1996. Ecological Distribution, Agricultural Trade Liberalization, and In Situ Genetic Diversity. Journal of Income Distribution 6(2): 265-286.

Barnum, H.N. and L. Squire. 1980. Predicting Agricultural Output Response. Oxford Economic Papers 32(2): 284-295.

Campbell, M. and T. Hendricks. 2006. Mexico's Corn Farmers See Their Livelihoods Wither Away: Cheap U.S. Produce Pushes Down Prices Under Free-Trade Pact. San Francisco Chronicle, Monday, July 31.

Commission for Environmental Cooperation. 2004. Maize and Biodiversity: The Effects of Transgenic Maize in Mexico: Key Findings and Recommendations. Montreal, Canada, August 11; http://www.cec.org/files/PDF//Maize-andBiodiversity_en.pdf.

Diamond, J. 1997. Guns, Germs, and Steel. New York, NY: W.W. Norton and Company, Inc. FAOSTAT, http://www.fao.org/es/ess/chartroom/factoid04.asp.

Fiess, N. and D. Lederman. 2004. Mexican Corn: The Effects of NAFTA. Trade Note 18, International Trade Department, Washington, D.C.: World Bank, September 24.

Gilbreth, C. and G. Otero. 2001. Democratization in Mexico: The Zapatista Uprising and Civil Society. Latin American Perspectives 28(4): 7-29.

Harris, J.R. and M.R Todaro. 1970. Migration, Unemployment and Development: A TwoSector Analysis. American Economic Review 60(1): 126-142.

Harvey, A. C. 1989. Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge University Press: Cambridge and New York.

Harvey, A. C. 1997. Trends, Cycles and Autoregressions. Economic Journal 107: 192-201.

Hodgson, J. 2002. Maize Uncertainties Create Political Fallout. Nature Biotechnology 20: 106-107.

Hymer, S. and Resnick, S. 1969. A Model of An Agrarian Economy with Nonagricultural Activities. American Economic Review 59(4): 493-506.

Koopman S.J., Harvey, A.C., Doornik, J.A. and N. Shephard. 2000. Stamp: Structural Time Series Analyser, Modeller and Predictor. London: Timberlake Consultants Press.

Lambrecht, B. 2005. Low Prices Force Mexicans from Fields. St. Louis Post Dispatch, Sunday, Oct. 30.

McAffee, K. 2003. Corn Culture and Dangerous DNA: Real and Imagined Consequences of Maize Transgene Flow in Oaxaca. Journal of Latin American Geography 2(l): 18-42.

Messean, A., Angevin, F., Gomez-Barbero, M., Menrad, K., and E. Rodriguez-Cerezo. 2006. New Case Studies on the Coexistence of GM and Non-GM Crops in European Agriculture. European Commission, Joint Research Centre (DG JRC), Institute for Prospective Technological Studies, Technical Report EUR 22102 EN.

Nadal, A. 2000. Mexican Corn: Genetic Variability and Trade Liberalization. Programa sobre ciencia, tecnologia y desarrollo. Documento de trabajo no. 1-06.

Nadal, A. 2002. Corn in NAFTA: Eight Years After. Research Report for the North American Commission for Environmental Cooperation.

Perales, H. R., S.B. Brush, and C.O. Qualset. 2003. Landraces of Maize in Central Mexico: An Altitudinal Transect. Economic Botany 57(1): 7-20.

Perales, H.R., B.F. Benz, and S.B. Brush. 2005. Maize Diversity and Ethnolinguistic Diversity in Chiapas, Mexico. Proceedings of the National Academy of Sciences 102(3): 949-954.

Quist, D. and I.H. Chapela. 2001. Transgenic DNA Introgressed into Traditional Maize Landraces in Oaxaca, Mexico. Nature 414(6863): 541-543.

Smale, M., Bellon, M.R., and J.A.A. Gomez. 2001. Maize Diversity, Variety Attributes, and Farmers' Choices in Southeastern Guanajuato, Mexico. Economic Development and Cultural Change 50: 201-225.

Soleri, D. and D.A. Cleveland. 2001. Farmers' Genetic Perceptions Regarding Their Crop Populations: An Example with Maize in the Central Valleys of Oaxaca, Mexico. Economic Botany 55(1): 106-128.

Vershuren, C. 2006. Against the Grain. Business Europe, Wall Street Journal, Jan. 4.

Notes

(1.) Transgenic denotes contamination of native plant varieties with genetically modified varieties.

(2.) Qist and Chapela (2001) allege that GM maize has polluted the native varieties in the Oaxaca region of southern Mexico. This article set off a firestorm of debate (Hodgson, 2002) and has come under intense scrutiny from the scientific community. The primary concern is that GM varieties could displace native varieties and possibly cause introgressive hybridization with the wild relatives of maize, such as teosinte, which would forever alter the gene pool.

(3.) The Mexican moratorium was enacted largely due to strong political opposition from activist groups representing the country's indigenous farmers, not due to scientific evidence. The ban does not include other genetically modified crops and it does not include imports of GM maize for the purpose of consumption. See in this context Gilbreth and Otero (2001) for an overview of the armed uprising against the Mexican government in the wake of NAFTA.

(4.) A non-economic approach is taken by the recent report on maize and biodiversity in Mexico published by the Commission for Environmental Cooperation (2004), and the background studies that were commissioned for that report.

(5.) Bacillus thuringiensis is a soil bacteria that is toxic to certain pests, especially the European corn borer. Bt-toxin, genetically derived from the above mentioned bacteria and currently patented by Monsanto Co., creates crystalline formations on the stalks of maize which act as insecticide.

(6.) See Nadal (2002) for an account of the agriculture reform measures taken by the Mexican Government after signing the North American Free Trade Agreement (NAFTA) in 1994.

(7.) Boyce (1996) notes that the subsistence farmers of Mexico have also incorporated hybrid modified seeds for years, and artificially selected for desirable traits from these seed stocks. Most researchers agree that this assimilation of "improved" seeds into the gene pool is at a very low level. However, GM seeds pose different risks that are not yet well understood by either the farmers or commercial plant breeders (McAfee 2003).

(8.) Mexico's ethnolinguistic diversity with more than 200 language groups among the indigenous peoples, is believed to facilitate local attachments to specific maize varieties (Perales et al. 2005).

(9.) According to Boyce (1996), Bipolaris maydis, the fungus responsible for Southern Leaf Blight, was infective to plants with the genetic makeup shared by approximately 85 percent of the maize grown in the U.S. in 1970.

(10.) Ex situ: off site. Organizations such as the International Maize and Wheat Improvement Center (CIMMYT) are engaged in facilitating the genetic diversity of wheat and maize to aid developing countries in establishing food security and overall agricultural productivity. See Bellon (2001).

(11.) Although this paper only concerns the effects of GM maize, it should be noted that subsistence farmers in Mexico have shown some preferences for creolized varieties derived from cross-pollination between native varieties and modern hybridized varieties. However, Bellon et al. (2005) have shown that in areas with high genetic diversity such as Chiapas, farmers are relatively indifferent to the benefits of creolization.

(12.) American farmers often use several different varieties of maize with different plant and harvest dates, albeit on separate plots of land. This was pointed out to one of the authors in a conversation with Matthew Garner, a Tennessee farmer.

(13.) This is also one of the central themes of Diamond (1997).

(14.) In SAS, unobserved component modeling can be found in the ETS package under the name UCM.

(15.) After estimation of the model parameters, a Kalman filter is applied to determine the state vectors [[mu].sub.t] and [[beta].sub.t] for each time period.

(16.) For completeness, it should be mentioned that more unobserved components can be added to a structural time series model than just a stochastic trend. Other components may be a stochastic cycle or a stochastic seasonal or a stochastic autoregressive component.

(17.) The consumer price index is included because it has been suggested (Campbell and Hendricks 2006) that its increase has caused subsistence farmers to raise acreage.

(18.) p represents the terms of trade for the subsistence farmer.

(19.) See equations 8 through 16 in Barnum and Squire (1980) for further reference.

(20.) In fact, distributors of genetically modified maize varieties mandate that new seed is purchased for every new planting season. This raises intellectual property rights issues. Compare on that the controversial 2001 Monsanto Inc. vs. Percy Schmeiser verdict in Canadian Supreme Court. Schmeiser was convicted of patent right violation for saving and knowingly replanting the seeds from his canola field, after being infected with Roundup-Ready[R] Monsanto Co. canola.

Alan Seals, Department of Economics and Finance, Meinders School of Business, Oklahoma City University. Special thanks to James Culpepper for his helpful comments.

Joachim Zietz, Corresponding author: Department of Economics and Finance, P.O. Box 129, Middle Tennessee State University, Murfreesboro, TN 37132, U.S.A. Fax: 615-898-5596, email: jzietz@mtsu.edu, url: www.mtsu.edu/~jzietz.

TABLE 1.
Variable Definitions

Variable    Definition

price       Producer price of maize (US $/ton)
imports     Import quantity of maize (1,000 tons)
area        Area harvested of maize (1,000 Ha)
yield       Yield per hectare of maize (tons/Ha)
cpi         Consumer price index, derived from the
              cpi inflation rate
mig         Off-farm migration, calculated as (population
              growth rate at t times agricultural
              population at t - 1)--agricultural
              population at t

Notes: All data relate to Mexico and cover the time period
1991-2004, except price, which ends in 2003. The data
are taken from FAO, http://faostat.fao.org/default.aspx.