Land Accumulation Dynamics in Developing Country Agriculture

Abstract

Understanding land accumulation dynamics is relevant for policy-makers interested in the economic effects of land inequality in developing country agriculture. We thus explore and simultaneously test the leading theories of micro-level land accumulation dynamics using unique panel data from Paraguay. The results suggest that farm growth varies systematically with farm size – a formal rejection of stochastic growth theories (that is, Gibrat’s Law) – and that titled land area may have considerable influence on land accumulation. Furthermore, our estimates indicate that a dualistic agrarian structure is the likely product of the unfettered operation of land markets.

Acknowledgement

The authors would like to thank Laura Schechter for providing the data for the project, Kevin Curran for assisting with data cleaning, and the Inter-American Development Bank (IDB) for financial support. Further, for their helpful comments we would like to thank participants of the IDB’s 2012 workshop Evaluating the Impact of Tenure Security Interventions in Developing Countries, as well as those of the 2013 Annual World Bank Conference on Land and Poverty. The views expressed here are those of the authors and should not be attributed to the IDB or its member countries.

Notes

1. It is important here to clarify our use of the term ‘accumulation’. We define land accumulation as ‘the acquisition or gradual gathering of land use rights or land access for purposes of agricultural production’. Notably, this includes expansion of the farm unit through legal (for example, ownership, rental or sharecropping arrangements) or extralegal (for example, squatting) means. As opposed to the term farm growth, land accumulation appears less ambiguous, as farm growth can occur along multiple dimensions (for example, growth in the quantity or value of output, capital accumulation, employment increases, and so forth). We do, however, use the terms land accumulation and farm growth synonymously throughout.

2. It is important to note that such positive effects are by no means a universal finding. Several studies have found that land-related investment (Brasselle, Gaspart, & Platteau, 2002; Gavian & Fafchamps, 1996; Migot-Adholla, Hazell, Blarel, & Place, 1991) and land market activity (Barnes & Griffith-Charles, 2007; Deininger & Jin, 2005; Gould, 2006) are not appreciably affected by land formalisation.

3. The reforms, for example, included measures to remove government from all forms of direct land redistribution, to end prohibitions on land rental and sale, and to activate private credit markets.

4. See Champernowne (1953), Reed (2001, 2003), or Reed and Jorgensen (2004) for other notable stochastic growth theories.

5. In the wake of Inégalités Économiques, a wealth of empirical literature has emerged seeking to test Gibrat’s Law, much of which has focused on the agricultural sector and farm size growth. Such empirical tests typically consist of estimating some variant of the following:

(9)

where represents the size of farm i at time t, and are parameters to be estimated, and is the error term. Rejection of the null hypothesis entails a rejection of Gibrat’s Law (Weiss, 1999).

6. More specifically, labour market imperfections entail that hired labour requires supervision and agents who seek off-farm employment face a distinct probability of unemployment. Credit market imperfections entail that the quantity of working capital available to a given agent depends on that agent’s land endowment.

7. For the sake of brevity, all other arguments in are suppressed.

8. In other words, ‘the model identifies an agrarian structure composed of mid-sized farms, and poverty refuge minifundias as a likely outcome of the unfettered operation of the land market’ (Carter & Mesbah, 1993, p. 1097).

9. The life cycle was said to begin with the marriage of the nuclear couple, then proceed through child-bearing and rearing, the entrance of the children into the family work force, and finally end with the exit of the children from the household to form families of their own.

10. See Sumner and Leiby (1987) for an alternative, albeit similar, theoretical model.

11. Agriculture-specific human capital is assumed to be primarily determined by learning-by-doing, though formal education may also play an important role. General human capital is assumed to be determined by formal education, employment history and inherent ability.

12. Rodgers also suggested that agricultural producers with relatively high may also grow faster due to the fact that they are able to spread fixed technology adoption costs over a greater quantity of output.

13. The data presented above pertain to the year 2008.

14. The latifundia-minifundia system is a dualistic agrarian structure composed of latifundias, or large hacienda-type estates or landholdings, and minifundias, which are small subsistence-oriented farms.

15. These regions were selected primarily because they are precisely those regions where much of the country’s agricultural production and land scarcity problems are concentrated.

16. See Section 4 for a detailed discussion of the relationship of each variable to the hypotheses put forth in Section 2.

17. It is worth nothing that in Table 3 the means of land operated and titled area are of a noticeably lower magnitude for 2007. This is primarily due to the addition of new units in 2007, though to some extent the differential persists even after omitting these observations.

18. See Table 2 for variable definitions and abbreviated names. Note also that the predetermined nature of the additional explanatory variables is a testable assumption and we undertake tests for over-identifying restrictions to substantiate this claim (discussed below). We should further note here that an anonymous reviewer suggested adding land owned to the empirical model. Due to its statistical insignificance, however, we retain the more parsimonious specification outlined above.

19. In situations where scaling is necessary to permit logarithmic transformation of zero-valued explanatory variables, we employ the dummy variable procedure outlined in Battese (1997) as alternative approaches (for example, adding an arbitrarily small constant) can bias parameter estimates. The method is simple: recode all zero values of explanatory variables to 1 and include in the regression a corresponding dummy variable that takes on the value of 1 if that observation was recoded and 0 otherwise. Note that dummy variables for land operated and experience are not necessary as these variables never take on a value of 0.

20. See Figure 1 in Carter and Mesbah (1993) for a depiction of this relationship.

21. Nevertheless, below we present a simpler specification for comparison purposes.

22. Data limitations preclude distinguishing between agricultural-specific and general human capital.

23. The results of the tests are available upon request.

24. Full OLS and within-groups results are available upon request.

25. The non-linear effects associated with these theories are further explored below. It is worth mentioning, however, that we attempted to include higher-order polynomials in the model, but their insignificance gave way to the more parsimonious specification presented.

26. An alternative approach would be to plot predicted farm growth against initial land operated while holding constant all other covariates at their sample means. Given the aforementioned correlation, however, it is not clear that this would result in economically meaningful predicted values.

27. An anonymous reviewer suggested testing the statistical significance of the predicted positive growth of farms in this first regime. Accordingly, using the CR model, we tested the null hypothesis that predicted growth is negative for a regime-representative farm. The p-value from the one-tailed hypothesis test was 0.06, and we thus reject the null hypothesis at the 10 per cent level. Detailed results and computer code are available upon request.

28. While the estimates indeed imply continual growth among the latter regime, the number of observations in this regime is relatively small and more information may reveal a new equilibrium at the high end of the farm size spectrum.