Policies for reduced deforestation and their impact on agricultural production

Agricultural Rent

Agricultural rent can be defined as: r^a = p^ay^a – wl^a – qk^a − v^ad. Agricultural production per hectare (yield) is given (y^a). The produce is sold in a central market at a given price (p^a). The labor (l^a) and capital (k^a) required per hectare are fixed, with input prices being the wage (w) and annual costs of capital (q). The fixed wage assumption implies that labor can move freely in and out of agriculture. Transport costs are the product of costs per kilometer (v^a) and distance from the center (d). The rent declines with distance, and the agricultural frontier is where agricultural expansion is not profitable anymore: r^a = 0. Thus the frontier is defined at d = (p^ay^a – wl^a – qk^a)/v^a.

This model, illustrated in Fig. 1, yields several key insights into the immediate causes of deforestation. Temporarily ignoring the forest rent, deforestation will take place up to the distance A. Higher output prices and technologies that increase yield or reduce cost make expansion more attractive; i.e., they move the agricultural rent curve to the right. Lower costs of capital in the form of better access to credit and lower interest rates pull in the same direction. Higher wages, reflecting the costs of hiring labor or the best alternative use of family labor, work in the opposite direction. Reduced access cost (v^a), for example, new or better roads, also provides a stimulus for deforestation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Agricultural and forest rents and forest rent capture.

This simple framework served as the basis for a number of empirical investigations. A survey of >140 economic models of deforestation finds a broad consensus on three immediate causes of deforestation: higher agricultural prices, more and better roads, and low wages and shortage of off-farm employment opportunities (3, 6).

The basic model can be extended in several directions, for example, to allow farmers to be capital and/or labor constrained, to allow some markets to be missing or imperfect, to include uncertain tenure, to permit market feedback, to include the temporal dimension, and to account for multiple production systems and their interactions (7, 8).

Forest Rent

Forest rent is more complex, reflecting the different nature of products and services generated by standing forests. We distinguish between three main types: first, private forest products, such as timber and a large number of nontimber forest products (NTFP); second, local public goods, such as water catchment and pollination services; and third, global public goods, such as carbon sequestration and storage and biodiversity maintenance. We refer to the first type as extractive forest rent and the latter two as protective forest rent. Total forest rent is given by r^f = (p^ty^t – wl^t – qk^t – v^td) + p^ly^l + p^gy^g.

The extractive rent increases due to higher timber and NTFP prices (p^t); technological progress (y^t, l^t, k^t); and lower labor (w), capital (q), and transport (v^t) costs. Higher values of local (p^l) and global (p^g) forest public goods increase the overall forest rent further and should lead to less forest being put under agricultural use. However, such an outcome depends critically on that rent being captured by the actual land users, as returned to below.

Agricultural Policies

Increase and Capture of Forest Rent

Increasing forest rent over time is the second route to protect forest, “the forest scarcity path” of the FT (15). Higher demand for and limited supply of forest products stimulate forest cover stabilization and regrowth. This extractive forest rent can be influenced in similar ways to the agricultural rent, e.g., through tax policies and marketing arrangements that affect prices of timber and other forest products or through promotion of new technologies.

Large tracts of tropical forests are, however, characterized by weak, unclear, and contested property rights, making them de facto open access (38). Land users then have no incentives to include any forest rent in their decisions. If private property rights to the forests are established, we move from point A to point B in Fig. 1. Higher forest extractive rent then implies more forest will remain as forest. Factoring in degradation, the effects are more complicated. According to the standard Faustmann model, higher timber prices will shorten the rotation period and thereby reduce the average forest carbon stock (39).

Whereas the forest scarcity path historically has been linked to higher extractive forest rent, in the future it could be driven by increases in the protective forest rent. Because of its public goods nature, an increase in the protective rent has no impact on deforestation unless land users can capture some share if it. There are two principal ways of “internalizing the externalities”: (i) moving decisions to a higher scale and (ii) creating a market for the public goods.

In the popular debate assigning individual property rights to forest is commonly put forward as a solution to excessive deforestation. This reform in itself will not solve the problem of local and global public goods (externalities), but clear and secure property rights—either at the individual or at the community level—are a necessary step toward establishing systems for payments of environmental services (PES). It will also encourage more sustainable management of forests compared with an open access regime, with positive effects on degradation and carbon fluxes.

Protected Areas (PAs)

Forest protected areas (PAs) within International Union for Conservation of Nature (IUCN) categories I–VI make up 13.5% of the world's forests (54), the share being significantly higher (20.8%) for rainforests (tropical lowland evergreen broadleaf forests). There is a broad consensus in the literature that (i) the degree of protection is <100%, but (ii) rates of deforestation within PAs are lower than outside them, also after controlling for “passive protection” (PAs are often located in remote areas with lower deforestation pressure) (55, 56).

A study of PAs in Costa Rica found substantial passive protection: Without controlling for observable covariates, PAs reduce deforestation by 65%; the degree of protection drops to 10% after controlling for differences in location and other characteristics (57). A methodologically similar study from Sumatra finds the difference between deforestation rates in PAs and wider areas during the 1990s to be 58.6%; this difference falls to 24% after propensity score matching (58). None of the studies finds any significant leakage (deforestation activities shift from inside to outside PAs), although the methods required to estimate leakage are complex and go beyond simple comparison of the (adjusted) deforestation rates inside and outside PAs.

Various types of PAs have also significantly reduced deforestation in the Amazon. Indigenous lands occupy one-fifth of the Brazilian Amazon, and Nepstad and coauthors (59) find the inhibitory effect for the period 1997–2000 to be 8.2 (the deforestation ratio between 10-km-wide strips of land outside and inside the PA border). These and other results led a World Bank forest policy review (60) to suggest that “protected areas may be more effective than is commonly thought” (page 126).

There is less consensus on other aspects related to PAs, e.g., the livelihood benefits and to what extent an inclusive or an exclusive approach of local communities is more effective when it comes to conservation effectiveness (61). This lack of consensus also holds for the integrated community development programs (ICDPs), which can be seen as a mix of a traditional “park and fence” approach and an attempt to provide alternative income opportunities to reduce agricultural rent and nonsustainable forest extraction. One study (62) concludes that “it is not that the principle of linking protected area management with local social and economic development is flawed, [but] the expectations and implementation that have been problematic” (page 514). The alternative livelihoods created were often small compared with the income from deforestation and forest degradation, and the benefits were not made conditional on forest conservation (as they are in a PES system).

Successful PAs can be expected to have similar effects on agricultural yield as policies to increase and capture forest rent. However, one can hypothesize that a PA approach will lead to higher loss of agricultural production per hectare forest saved, because there is less assurance that the least productive land is saved from agriculture.

The Global Food Equation

The global food equation (GFE) is a simple identity that links population and food consumption per capita with agricultural yield and land area:

Put simply, without an increase in yield, agricultural area must expand to feed a growing population and meet higher per capita food consumption. GFE has been used to draw conclusions about the need for higher yield to spare forests (63). Waggoner and Ausubel (64) refer to it as “the popular image of farming's encroachment on forests” (page 241). This line of reasoning also underlies the Borlaug hypothesis (17), which suggests that the Green Revolution has had a positive effect on forest cover.

Using the GFE logic, Balmford and coauthors (65) predict that the agricultural land area in developing countries will increase by 2–49% between 2000 and 2050, depending on assumptions of population growth (23% being the medium variant scenario). This scenario assumes an extrapolation of current yield trends (with a mean of 1.13% per year). A more optimistic scenario with an annual yield increase of 1.53% virtually eliminates the agricultural area increase.

The GFE provides no direct link between agricultural and forest areas, nor does it account for two facts: Much agricultural production is not food and countries trade. Moving to the national level and further decomposition gives a national deforestation equation (NDE):

or

Agricultural yield is just one of many factors affecting deforestation, and changes in yield have indirect effects on these factors. First, countries increasingly trade in agricultural products. The trade intensity (trade/GDP ratio) has increased from 60 to >100% since the early 1970s (66). Developing countries as a group have over the same period moved from being net agricultural exporters to net importers. Higher yield boosts the competitiveness of domestic agriculture and raises self-sufficiency.

Second, a lower share of agricultural output being for foodstuffs (Δ food share) can boost deforestation, as illustrated by the boom in biofuel crops. Oil palm expanded by 1.9 and 3.0 million ha in Malaysia and Indonesia, respectively, during the period 1990–2005 (67). Most of the smallholder crops on forests cleared in Indonesia, following the economic crisis in the late 1990s, were not food crops (e.g., rubber) or not staples (e.g., cocoa, pepper, and coffee) (68). Higher yield can reduce the food share, as food demand is typically more price inelastic than demand for nonfood commodities.

Third, forest, cropland, and pasture are not the only land uses; large areas of fallow, savannah, bush, and other land categories are available for agricultural expansion (Δ ag/forest ratio is not stable). Waggoner and Ausubel (64) find changes in cropland and forest area to be uncorrelated in the period 1900–1995, although this might partly be due to poor data for many countries. Their average “encroachment factor” (share of agricultural expansion into forests) is assumed to be 1/3, but is also highly variable across crops and countries. Fifty-five to 60% of the recent oil palm expansion in Indonesia and Malaysia was at the expense of forests (67).

Other potential impacts of higher yield include a price effect on food consumption per capita (inelastic food demand suggests this effect will be small) and a Malthusian effect (higher population growth due to increased food consumption).

The GFE, NDE, and similar identities are useful in providing a consistent accounting framework, but are also potentially dangerous to use as predictive models and for policy analysis if they do not factor in how a yield change impacts the other factors through behavioral and market changes. This change by moving from a mechanical simulation to empirical analysis using a regression model is illustrated by the results of Ewers and coauthors (69), using country-level data for the period 1980–2000. If “perfect land-sparing” yield change were occurring, the land-yield elasticity should be −1; i.e., all other factors in NDE remain constant. The authors find a much lower elasticity: −0.152 (t = −1.78) for developing and −0.089 (t = −0.57) for developed countries due to effects such as those discussed. The impact on forests (not included in their analysis) would be even smaller as long as the encroachment factor is below unity.

Discussion and Conclusion

The starting point of the von Thünen model is the plot level, and deforestation is framed as a contest between agricultural and forest rents. The GFE starts at the other end of the scale and asks how much land we need to feed the global population. The von Thünen model is at one extreme where demand is perfectly elastic (prices fixed), whereas the GFE assumes demand to be perfectly inelastic (quantities fixed). They present two contrasting views on the forest impact of higher agricultural yield, but they converge when modified to include behavioral and market effects. Whereas overall food demand may not respond much to price changes, this does not necessarily hold for particular crops or for nonfood agricultural products where substitutes are available.

The demand elasticity and thereby the forests impact of higher yield also depend critically on the scale of analysis. Angelsen and Kaimowitz (17) conclude that “situations that are win–lose [agricultural production and forest conservation] at the local level may be win–win at the global level” (page 400).

An illustration of the limited trade-off between production and conservation at higher scales is given by comparing recent agricultural production and area increases in developing countries. Crop and livestock production grew by 3.3–3.4% per year during the period 1985–2004 (66). Gross annual deforestation (1990–2005) for agricultural uses represents ∼0.3% of the total agricultural area (66, 70, 71). Because productivity of cleared forest land can be expected to be well below average productivity (production is less intensive, and most productive land is already cleared), these numbers suggest that only a small share (<<10%) of the agricultural output increase has come from deforestation.

REDD is currently being promoted as a low-cost climate mitigation option. The report that underlies the Stern review (50) and the Eliasch report (72) finds the opportunity costs (foregone agricultural rent and logging revenue) of completely eliminating deforestation in eight countries (accounting for 6.2 million ha of annual deforestation, about half the global number) to be approximately USD 6.5 billion per year (73). Due to increasing marginal costs, spreading a 50% reduction across all deforesting countries is significantly cheaper. Other studies such as Kindermann and coauthors (74) have cost estimates in the range of USD 17–28 billion for a 50% global reduction. These numbers include REDD rents to developing countries, which are not true economic costs but transfers and typically inflate cost numbers by a factor of ≥3 (51). Yet, the relatively low opportunity costs of avoided deforestation, particularly for the initial reductions, suggest the conflict between production and conservation is modest.

At the national level, higher volumes of agricultural trade have delinked domestic and local consumption from production and deforestation. Moreover, high rates of deforestation for several decades have made forested areas recede, frequently into relatively inaccessible areas. The issues of forest conservation and agricultural production are therefore becoming increasingly spatially delinked.

In summary, at global and national levels policy makers are only to a limited degree presented with a trade-off between conserving the forests and feeding the hungry. Potential conflicts between production and conserving forests do, however, exist at the forest margins. Stimulating agriculture in forest-rich areas through, for example, better technologies, improved roads, and more secure tenure to “reduce the need for new agricultural land” is a highly risky conservation strategy. Agricultural policies that target low-forest areas, or crops and production systems that are unsuitable at the agricultural frontier, are more likely to reduce pressure on forests. Such policies are complementary to, and will increase the effectiveness of, efforts that more directly target forest conservation: protected areas and institutional arrangements and payment mechanisms that enable land users to capture a higher share of the local and global benefits provided by tropical forests.