Booming markets for Moroccan argan oil appear to benefit some rural households while threatening the endemic argan forest

In this section, we first synthesize results from a recent microanalysis of the local impacts of the argan oil boom on how rural households interact with argan markets, what benefits arise from their market participation, and how their use of the forest has evolved (29). Building on this microanalysis, we present the results of mesoanalyses that test the win–win claims at a more aggregated landscape level.

The argan oil case offers an excellent opportunity to investigate the dynamic linkages between local benefits and conservation gains. Consider the interplay between locals’ response to argan fruit price appreciation, argan markets, and forest impacts. In 1999, high-value argan oil extractors insisted on purchasing only whole fruit in local markets as a guarantee that the kernel inside had not passed through a goat's gut (40). Since then, rapid appreciation of argan fruit prices has prompted locals to collect fruit by hand and prevent their goats from ingesting whole fruit. Thus, there are far fewer goat kernels in local markets. Knowing this fact, high-value extractors are increasingly willing to purchase kernels in spot markets, which have become nearly as active as those markets for argan fruit. Moreover, this change has increased the volume of kernels sold as well as doubled the price premium of kernels over fruit between 2005 and 2009.^†† In short, initial market changes induced a behavioral response that resolved a market for lemons problem (41) and enabled the emergence of high-value kernel markets. Because the laborious task of extracting the kernel has yet to be mechanized, this induced kernel market may be the most promising path from high-value argan markets to rural poverty alleviation.

Next, consider the dynamic pressures on tenure institutions in the argan region. Booming argan markets have, not surprisingly, led to greater privatization pressure (11, 42). In the argan forest, unique tenurial arrangements (SI Discussion and Fig. S2) give important nuance to implications of these privatization pressures. Households that previously only had seasonal usufruct rights to specific tracts of the forests are now enclosing these tracts to deny others access all year long. These barriers, which are technically allowed if temporary, are becoming increasingly permanent. Because private usufruct rights encroach on collective grazing rights during the nonharvest season, the forest may well benefit from better management—albeit at the cost of excluding poor households with limited private productive resources from these forest commons. The poverty of these households may persist despite booming argan markets.

From a conservation perspective, a key dynamic involves the apparent mismatch between locals’ conservation incentives and the long-run sustainability of the argan forest. Although locals are now less likely to let their goats browse in the tree canopy, this forest-friendly change is motivated more by immediate concerns about the fruit harvest than by any longer-term concern for tree or forest productivity. Goats still regularly climb and browse trees outside the fruit harvest season, when most trees are treated as an open access resource, and therefore, they continue to tax the forest. Ever-growing goat herds, thus, remain a concern. This mismatch between private incentives and sustainability of the forest not only bodes poorly for the argan forest in general but interacts with drought cycles and more aggressive fruit harvesting in potentially negative ways. Although the argan tree is well-adapted to arid conditions and can survive in a dormant state for several years during extreme drought, it is quite vulnerable during lesser drought conditions that do not trigger this dormant response. Locals and their goats generally exploit the argan tree more heavily during such episodes of moderate drought, because other options dry up before the argan tree goes dormant—a dependence that can only increase as households shift their livelihoods more heavily to argan fruit. As fruit production falls during these periods, harvesting tactics are likely becoming even more aggressive. Precisely, this scenario played out in 2008—a year of low fruit production, high fruit prices, conflicts among collectors, and frequent and aggressive fruit harvesting. Although the argan tree is robust to many pressures, recovery from these episodes of intense climatic and exploitation pressures can be painfully slow. Because it serves as foundation species for over 1,200 species (7), this cyclical and intensifying pressure on the argan tree threatens the broader biodiversity of the Acacia–Argania ecoregion.

We surveyed a representative sample of rural households in Essouira Province in 1999 and again in 2007 (Fig. S1). Analysis of the 1999 data, collected before the argan boom, suggested that differentiation in argan oil markets would prevent locals from tapping high-value markets directly (40) and that conservation gains might be disappointing, because the slow growth of the tree implies low returns to local conservation investments (43). We compare our 1999 and 2007 data to evaluate the welfare effects of induced changes in households’ argan activities. Specifically, we assess how households’ access to argan fruit in 1999—an indicator of ex ante potential to benefit from the argan boom—affected three types of household welfare outcomes: consumption (spending at market), assets (livestock holdings), and children's education (complete details in ref. 29).

Our broader mesolevel analysis exploits differences in forest coverage across communes and pixels to assess the impact of the argan boom on two response variables: changes in educational outcomes and forest canopy density. As the treatment, we use booming argan markets after 1999. Initial argan forest coverage (as collected in a 1994 Moroccan forest inventory) indicates the dosage of the argan boom treatment in a given location.

In the commune-level education analysis, we use primary and secondary school enrollment data from Essaouira Province disaggregated by sex and collected annually from 2000 to 2005, a period of rapid appreciation in argan prices that roughly corresponds to our household-level data. Although we can measure the transition from primary to secondary school for each child in our surveyed households, we can only measure total enrollment by sex with the commune data, because these data do not distinguish between primary and secondary students. We regress changes in enrollment as a percentage of commune population on the percent of the commune covered by argan forest in 1994, two commune-level infrastructure indexes, interaction terms between forest cover and infrastructure indexes, and a time trend. We construct the infrastructure indexes using factor analysis (SI Materials and Methods) and include them as controls, because rural enrollment decisions are heavily shaped by accessibility to schools and geographic isolation (e.g., accessibility by different forms of transportation, availability of electricity, etc.). Furthermore, the impact of argan benefits on these enrollment decisions may vary according to local infrastructure: increased household income may partly compensate for weak infrastructure and therefore, could have a bigger marginal impact on enrollment in places with poor infrastructure. To test whether the enrollment impact of the argan boom is mediated by infrastructure, we include the interaction of argan cover and these indexes in our specification. We estimate this model separately by sex for 1- and 4-y enrollment differences. The lack of similar panel data prevents us from conducting this kind of commune-level analysis for other provinces in the argan region or using measures of household consumption, poverty, or livestock as dependent variables.

We use our NDVI analysis to evaluate anthropogenic changes in forest and canopy density caused primarily by local exploitation of the forest, especially grazing. Based on the validation exercise described in SI Materials and Methods, we are confident that NDVI is a robust measure of canopy coverage (Table S1). We compare changes in the argan forest to changes over the same time period in nearby nonargan forest using an empirical model that is relevant to argan trees, nonargan trees, and shrubs alike—all of which function as foundation species in very similar ecosystems. Our approach relies on NDVI differencing, which has been shown to measure landscape changes in regions more arid and more sparsely vegetated than the argan forest (44, 45). In contrast to earlier work that assessed changes throughout Morocco using NDVI (46), we focus on the argan forest region, incorporate rainfall data, and isolate the NDVI trend for the forest canopy by focusing on dry season trends, during which there is virtually no vegetation besides tree canopies. The data that we use consist of 10-d composite NDVI data covering 29 y (1981–2009), which have a resolution of 8 × 8 km and were derived from advanced very high-resolution radiometer available from the US Geological Survey (47). The data were based on composite images generated by the National Oceanic and Atmospheric Administration. As in other settings (37), we expect the NDVI to be strongly correlated with species richness and biomass in the argan forest region, because in the dry season, it reflects forest and canopy density of foundation species in these ecosystems.

In the arid argan forest region, rainfall is critically important to vegetative productivity. On average, over 90% of precipitation in the region falls during the wet season between November 1 and May 31. Although the forest floor can produce various grasses and rain-fed farmers can grow a meager barley crop during these months, the floor of the forest—whether argan or nonargan—is largely void of vegetation during the dry season. We leverage the fact that trees and shrubs are the only vegetation in the forest during the dry season to isolate the NDVI trend for the forest canopy. Specifically, we estimate an autoregressive switching regression model for each pixel that takes the following form (Eqs. 1 and 2):

where Rain_it is contemporaneous rainfall measured at a provincial weather station for the 10-d period t and season i, CumRain is cumulative rainfall since the beginning of the last wet season, Wet and Dry are wet and dry season dummies, respectively, and t is a trend variable. In this specification, the switching regression on trend enables us to focus specifically on the dry season trend (), while controlling for factors that affect NDVI in both seasons as well as season-specific factors such as lagged cumulative rainfall, which directly affects the forest canopy. During the dry season, vegetation captured by the NDVI is contained almost entirely in the forest canopy, and therefore, specifically captures changes in the forest canopy. We use a stepwise procedure to determine the order of autoregression. Given the distinct seasonality of rainfall in the region, this procedure yields a high order of autoregression (K = 39). The complete time series includes over 800 observations for each of 7,765 pixels. We use the onset of the boom in 1999 to split this data into preboom and boom phases and estimate the dry season trend over these two sets of years for each pixel separately (Fig. 1).

To formally test how the boom has affected the argan forest, we use the two estimated trend coefficients for each pixel to construct a differential trend variable , which captures the change in the estimated canopy trend for pixel z associated with the onset of the argan boom in 1999.^‡‡ Next, we compare for each pixel in the argan forest to one in similar and nearby nonargan forest drawn from pixels adjacent to the argan forest. Because the pixels in each of these pairs are subject to comparable climatic and other pressures^§§ but only the argan pixel is affected by booming argan markets,^¶¶ we can pool all these pairwise comparisons together to estimate an average treatment effect of the argan boom on the argan forest (Table 2).

Before matching argan and nonargan pixels, we exclude pixels that include irrigated land and urban fringes. We use nearest-neighbor matching based on geographic proximity, mean dry season NDVI, and the coefficient of variation of NDVI. This matching process pairs pixels in the argan forest with nearby pixels with other tree and shrub species that otherwise have similar vegetative density. We use these matched pairs to estimate the average treatment effect of the argan boom on argan forest density using the nonargan pixels with other species as controls. The average treatment effect on the treated indicates the impact of the argan boom on the argan forest relative to its impact on nonargan forest. Because this test is a relative test, it captures genuine improvements in argan forest, displaced degradation in nonargan forest, or both, which makes finding a positive (negative) impact of booming argan markets on argan forest easier (harder). For example, if locals on the edge of the argan forest move their livestock to the nonargan forest in the wake of rapid argan price appreciation to increase their fruit collection, this increased pressure would decrease the nonargan forest canopy trend relative to the neighboring argan forest. As one of the three robustness tests in Table 2, we estimate treatment effects excluding pixels with sparse tree or shrub cover (i.e., those pixels in the lowest quantile of mean dry season NDVI). Notice: Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official agency policy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.