Robust agricultural TFP growth is also important for poverty reduction

Analysis of the production frontier reveals a dynamic inverse relationship between farm size and frontier productivity, where technical change has increased the frontier for larger farms at a faster rate than for smaller farms, weakening the inverse relationship along the frontier of productivity. Despite these changes at the frontier, the farm size – average TFP relationship has remained constant due to technical inefficiencies growing faster for larger rather than smaller farms. In essence, many of the larger farms were not able to keep up with technical change at the frontier, suggesting that successfully reducing technical inefficiency for this group could mediate, if not reverse, the farm size – productivity relationship. To the extent that the inverse relationship between farm size and TFP has flattened along the frontier for Mexican family farms, it suggests that size may fade as one of the key determinants of productivity differences as agricultural sectors modernize. Policies that help family farms keep abreast of improvements in agricultural technology, such as farmer education, will be needed to reduce growing technical inefficiency. These findings support the claim that family farms have struggled in the wake of NAFTA era market liberalization, and we echo the calls of Pérez et al. that investment in rural infrastructure and assistance for smallholder transition into niche markets would support productivity growth for family farms. By growing the food supply more rapidly than demand, falling prices benefit poor consumers wherever they may live.

And for the small farms that continue to exist, gutter berries either because they are competitive or because they have few other opportunities, TFP growth helps to boost income. Where farms are too small, as in many parts of Mexico, increased productivity may still be insufficient to lift households out of poverty. Households in regions with access to non-agricultural employment may persist, and some will escape poverty, but migration is likely to continue. An important extension of this work would assess the potential impact of productivity growth on rural economic development and poverty alleviation. An important limitation of analysis conducted here is the absence of non-family commercial farms in the Mexican sample. Future research should extend this analysis to a nationally representative sample of farms, such as the 2007 Mexican Agricultural Census, which would include family and non-family agricultural operations. Extending the analysis to the entire range of farm sizes and farm types would allow for a more complete analysis of the farm size – productivity relationship. Together with a theoretical analysis of a dynamic farm size – TFP relationship, such extensions would inform policy efforts to increase agricultural productivity.The panel of family farms is drawn from the Mexican Family Life Survey , a nationally representative sample of Mexican households asking detailed information on households’ assets, incomes, consumption patterns, and well-being with survey waves in 2002, 2005-2006, and 2009-2012. For those households engaging in agriculture, MxFLS gathers plot level information on plot size and the most important crops produced and household information on asset ownership, on farm labor, and expenditures on other inputs. These farming households are the focus of the analysis, and the nature of the survey make it a rich source of data for analyzing the relationships between agricultural productivity and poverty.

Drawing from the MxFLS sample of family farms identified in chapter 2 as having complete data on their agricultural operations, this study uses farms engaging in agricultural production in both the 2002 and 2009 survey waves.3 Due to concern over measurement error in the income measure, the top and bottom 5% of the income percapita distribution were dropped from the sample, resulting in a final sample of 224 farms. The sample contains a broad range of family farms: while the median and mean farm sizes in 2002 were 2.45 ha and 113.0 ha, respectively, the largest farms are over 1,000 ha. Median farm size fell to 2.1 ha and mean farm size grew to 179.3 ha by 2009,suggesting that while small farms are getting smaller, large farms are growing in size. Figure 3.1 below displays kernel density estimates of the distribution of farm size for 2002 and 2009. Construction of the measure of agricultural output closely follows that of chapter 2, with the exception that the current analysis does not construct an output quantity index. Rather, the current analysis uses the real value of output, valued in 2002 Mexican pesos. Although an output quantity index is often preferable for productivity analysis, the use of an output variable measured in value terms provides a direct link between land productivity and household income, facilitating the current analysis. Average land productivity fell over time, with the mean logged land productivity falling from 7.71 in 2002 to 7.56 in 2009. Figure 3.2 shows the kernel density estimates of the land productivity distributions for 2002 and 2009, revealing that most of this decline occurred in the bottom half of the land productivity distribution. Given the inverse relationship found between farm size and productivity, this increase in average farm size is likely contributing to the observed decline in land productivity. Factors of production other than land include measures of physical capital, draft animals, purchased inputs, family labor, and non-family labor.

Physical capital is measured as the real value of tractors and other machinery owned, and draft animals is similarly measured as the real value of horses, donkeys, and mules owned by the household. Purchased inputs are the real value of aggregate reported expenditures on nine inputs, including fertilizer, manure, pesticides, seeds, tractor services, animal power, wage labor, water, and fuel. As with output, these inputs are measured in 2002 prices. Non-family labor is measured as the number of non-family workers working on the farm over the course of the year, and family labor is an index of the hours worked on the farm by household members during the year. Figures 3.3 and 3.4 show the distributions of family labor per ha and purchased inputs per ha in each period, as these are the most commonly used agricultural inputs in the sample, and were found to be the two most significant inputs in chapter 2. While the intensity with which family labor is used falls during the sample period, the share of farms using purchased inputs is increasing. As with changes in farm size, these changing input intensities are potentially contributing to the changing land productivity distribution.No complete measure of household income is included by MxFLS, but detailed questions on income sources and economic livelihoods throughout the individual and household level components of the survey are used to construct a measure of total household income. As introduced in section 3.3 above, income is comprised of agricultural income and non-agricultural income. Whereas agricultural income is measured as the value of agricultural output, whether brought to market or used on the farm, non-agricultural income includes both labor and non-labor components. Labor income includes that derived from wage labor and/or self-employment, strawberry gutter system whereas nonlabor income includes profits from businesses, government transfers, rental income, non-agricultural household production, remittances, and pensions, among other sources. See Appendix C.2 for a detailed discussion of the construction of total household income. Where possible, labor income is grouped into non-agricultural and agricultural related activities. Table 3.1 shows the share of households deriving income from each income source by year, alongside the average share of total household income derived by that source . The first group of income sources includes labor income derived from participation in labor markets. Agricultural households increasingly have adult household members participating in wage labor, in both agricultural and non-agricultural occupations, while the participation of household children in wage labor is on the decline. For those family farms with adult household members engaging in wage labor, such wages are important components of total household income – income from non-agricultural wage labor, for example, makes up half of the household incomes of participating households. The second group includes income from entrepreneurial activities, including agricultural production and non-agricultural businesses, self-employment in agricultural and non-agricultural industries, and home production. On average, agricultural income accounts for 26-29% of total household income for this sample of Mexican family farms – the economic livelihoods of these family farms are diverse, with the majority of household incomes being generated by activities other than agricultural production. Further, whereas fewer households are participating in agriculture related entrepreneurial activities, households are increasingly participating in non-agricultural activities. The non-agricultural sector of the rural economy appears to be increasingly prevalent in the livelihood strategies of Mexico’s family farms.

The third and final group include non-labor sources of income, where participation in Oportunidades and other government programs are increasingly prevalent. On average, these programs are important for family farms – households receiving payments through Oportunidades, for example, receive a fifth of their household income from the program. Overall, the livelihood strategies of this sample of family farms are consistent with previous findings from Mexico. In their study of ejido households in 1997, de Janvry and Sadoulet found that off-farm income generated half to three-fourths of ejido household incomes, that self-employment activities and non-agricultural wage labor were more common than agricultural labor for households participating in off-farm employment, highlighting the importance of the rural non-agricultural sector for the economic livelihoods of Mexican family farms. Real monthly income per-capita fell in the sample, from a mean of 1,147 to 1,016 Mexican pesos over the span of the panel. Despite this, income per-capita grew for the poorest quartile of the income distribution. These changes had heterogenous impacts on the income-based poverty measures used in this study. For reference, Table 3.2 displays the urban and rural overall and extreme poverty lines for 2002 and 2009, adopted from Mexico’s National Council for the Evaluation of Social Development Policy . Table 3.3 shows the poverty measures introduced in using the rural poverty and rural extreme poverty lines, for both survey years. The final two columns calculate the percentage change. As both mean and median incomes fell over this period, the poverty rate increased. While this change in overall poverty incidence was marginal, increasing by just 0.8%, the extreme poverty rate saw a notably large 6 percentage-point increase, equivalent to a 15% increase in the prevalence of extreme poverty for this sample of family farms. A more complex picture unfolds in light of poverty depth and poverty severity, measured by the poverty gap and the poverty gap squared, respectively. The overall poverty gap index experienced a 1.2% increase as farmers slipped into extreme poverty. However, the extreme poverty gap index fell by 1.0% and the squared poverty gap index declined by 2.2% when using the overall poverty line and by 1.7% when using the extreme poverty line. To summarize, although the incidence of overall poverty increased marginally and the overall poverty gap increased, poverty severity showed declined. Movement in these poverty measures is primarily driven by changes to the extremely poor – as poverty deepened for some, as reflected in the 15% increase in the incidence of extreme poverty. Rising incomes in the bottom quartile of the income distribution meant that the extreme poverty of the poorest of the poor was partially mediated, leading to a reduction in overall poverty severity.As a starting point for unpacking the direct poverty reduction potential of agricultural productivity, Table 3.4 displays the counterfactual poverty measures that would have prevailed if land productivity had not changed over time, alongside the observed poverty measures in 2009 and the percentage difference. In short, the counterfactual poverty measures answer the question, what would have been the change in poverty if farm income generating activities had evolved as observed, but agricultural productivity had remained at observed 2002 levels? Similarly, the percentage difference provides an estimate of the reduction in poverty that would have been achieved if base year land productivity had been maintained. Results from the poverty rate suggest that if land productivity had been held at 2002 levels, poverty incidence on family farms would have been more than 4% lower. This would amount to a 4.1-4.6% reduction in poverty given a 16% increase in land productivity, for a point elasticity of approximately -0.26 to – 0.29. Irz et al. , in comparison, find that poverty falls by 5-7% given a 10% increase in land productivity, for a point elasticity of -0.5 to -0.7. This is reasonable, given that the current approach assesses only the direct links between land productivity and poverty, whereas Irz et al. assess the linkages more comprehensively.


Posted

in

by