Comparing Results

Jump to a section:

Why is there a need to compare the impacts of biofuels, according to predictions from different studies?

From Zhang et al. 2009:

The world has seen rapid growth in biofuel production in recent years. Correspondingly, the use of major feedstock crops for biofuel production has increased dramatically. Beyond contributing to the world‘s supply of energy, the recent surge in biofuel production has the potential of inducing a wide spectrum of consequences—both negative and positive.

The rapid expansion of biofuel production has generated considerable interest within the body of empirical economic literature that has sought to understand the impact of biofuel growth on the global food economy. While the consensus within the literature is that biofuel emergence is likely to have some effect on future world agricultural market, there is a considerable range in the estimated size of the impact. Some studies claim strong linkages (e.g., Qiu et al., 2009). Others suggest that the linkages between biofuels and commodity prices are relatively weak (e.g., Banse et al., 2008). Studies that project the impact of future biofuel production on agricultural prices provide important guidelines for setting long-term agricultural, food security, and energy policies, as well as development agenda. Despite the importance of this topic to policy makers, there has been no study that has tried to reconcile the differences among various outlook studies.
This paper undertakes an in-depth review of some key outlook studies which quantify the impacts of biofuels on agricultural commodities, and which are based on either general-equilibrium (GE)* or partial-equilibrium (PE)** modeling approaches. To reach this goal we have two specific objectives. First, we will describe the range of projections from a group of papers that ar focused on forecasting prices and production of the three key biofuel feedstock crops globally as well as in different parts of the world. Second, we seek to explain the differences in the projections by examining the differences in underlying assumptions and model structures.

How are the comparisons made?

From Zhang et al. 2009:

We focus on the prices and production of three biofuel feedstock crops, maize, sugar cane and oilseeds. To make the results of the studies comparable, we make necessary adjustments and organize some of the studies in ways that make the inter-model comparisons as straightforward as possible. First, we organize the studies by the modeling approach taken by the authors. In particular, we need to examine and compare the results of PE studies and GE studies separately. Second, since we are interested in isolating the effects of biofuels on agricultural prices and production, we consider a “reference scenario” that is closest to a no-biofuels case for each of the modeling exercises. Third, we look at the results of a particular time period, i.e. 2015, that is shared by almost all of the PE and GE studies. Finally, we considered differences in coverage and consistency.

What is the range of projection results among PE studies?

From Zhang et al. 2009:

According to the PE studies, the emergence of biofuels (relative to the reference scenario—no biofuels) will have a positive impact on both prices and production in 2015. The results from all of the models show that the world prices of all commodities in our study—maize, sugar cane and vegetable oil—rise. The price of maize in the US also is shown to rise. The PE models also demonstrate the special, demand-side nature of the emergence of biofuels on agriculture. Although prices are shown to rise, they do so even though production also rises. All of the models demonstrate that the expected demand for agricultural commodities that is associated with the emergence of biofuels is strong enough to lead to a rise of world maize, sugar cane and vegetable oil production as well as maize production in the US.

While all of the PE studies project upward trends in prices and production, there are differences in the magnitudes of the estimated impacts among studies. Above all, the WEMAC model consistently projects the highest effects on the prices and production of maize in 2015. In contrast, the estimated effects of the OECD, IFPRI, and FAPRI models are much lower and are largely consistent with one another. For example, the estimates of the rise of world/US maize prices by 2015 range from 14.6% (OECD model) to 16.1% (IFPRI model) to 16.2% (FAPRI model). Maize production forecasts are equally close for the three models, ranging from 2.9% (OECD model) to 5.8% (FAPRI model). However, there are differences among the models in the projections of sugar cane prices.

What is the range of projection results among GE studies?

From Zhang et al. 2009:

As in the case of the PE models, the predictions from the GE models demonstrate that both prices and production of key feedstock commodities in key countries rise with the emergence of biofuels. Specifically, the prices of US maize, the prices of Brazil sugar cane and the prices of EU oilseeds all rise. At the same time the strong demand-side effect of the emergence of biofuels is clear: at the same time that prices rise, the production of US maize, Brazilian sugar cane, and EU oilseeds also rise according to the predictions of all of the models.

However, as is also found in the case of the PE models, the price and production outcomes vary considerably across different models. In fact, the ranges of production and price forecasts are even wider for the GE models. For example, differences in estimated price impacts vary from 4.7% to 70% for maize; 1.5% to 109.1% for sugar cane; and 5.5% to 66.7% for oilseeds. Differences in production projections are actually larger, ranging from increases of 4.0% to 70% for maize; 1.5% to 109.1% for sugar cane; and 5.5% to 66.7% for oilseeds.

Although large differences characterize the predictions, a close examination reveals distinct patterns among the findings of the GE models. Specifically, the LEITAP model consistently predicts the lowest prices and production values. In contrast, the GF model projects the highest levels of price and production impacts. The other three models are in between, but also are (relatively) clearly ranked.

How do we explain these differences?

From Zhang et al. 2009:

We focus on several key factors that we believe may be driving a significant share of the differences among the studies. Specifically, for the PE models, we review scenario design, key assumptions, and model structure. For the GE models, we focus on key model assumptions and parameters. Notably, the set of factors that we use to explain differences among the models differs for the PE models and the GE models. We do not include, for example, model structure for the GE models because, unlike the PE models that differ in the model representation of biofuels, most of the GE models share the standard GTAP structure (except for FARM II).

What were the conclusions?

From Zhang et al. 2009:

Overall, all outlook studies reviewed indicate that biofuel growth will lead to higher prices and production levels for the three primary feedstock crops of 1st generation biofuels by 2015. In other words, all modeling efforts believe that—to a greater or lesser extent—biofuel development is likely to remain an important driving force in world agricultural markets over the medium term. Since most of these basic results are driven by comparing the baseline of ?no-biofuels‘ with an alternative scenario which allows the policy-driven emergence of biofuels, the impacts of ambitious policy objectives in major biofuels-producing countries are shown to be significant.

Despite the fact that all models predict positive impacts on prices and production, there are large differences among the studies. Our findings point to a number of key assumptions and structural differences in the modeling approach that seem to jointly drive the variation we observe, across these studies. First, the scenario design appears to be an important factor for the PE models and has likely contributed to relatively high price impact of biofuels in the WEMAC projections. Second, the presence or absence of biofuel trade, and the structural way in which agriculture and energy market linkages are modeled, are likely to account for some of the differences we see between the OECD and IFPRI model projections for sugar and vegetable oil. Third, relaxing restrictions on total agricultural land supply may be the driving force for LEITAP‘s relatively low estimate of price impacts, relative to the other GE-based projections. Fourth, accounting for the possible contribution of DDG by-products to animal feed is the distinguishing difference between the GTAP-based Purdue I and Purdue II models. Fifth, the high degree of substitutability between petroleum and biofuels (especially when combined with assumptions on future crude oil prices) is the distinguishing feature of the GF study and contributes towards its relatively high predictions on price and production impacts. The assumption—whether true or false—relates closely to what is envisioned in terms of future technology adoption within the transportation sector. Policy and economic factors will weigh in heavily on determining which types of the flexible-fuel vehicles will become widely available and also the types of fuel sold at filling stations. For example, if policies are slow to encourage the adoption of flexible-fuel vehicles or impose “blending walls” such as those which exist in some regions of the US, then the degree to which biofuels are substitutable with gasoline and diesel may be limited.9 Furthermore, if biofuels are sold at or below their energy values with respect to gasoline and diesel, there will be enough consumer demand to encourage the building of E-85 or other alternative fuel pumps. Because these differences in assumptions make a significant difference, policy makers should take into account the underlying assumption-based and structural differences among models when using model-generated outcomes to evaluate economic and environmental impacts, and to guide decision-making.

What are the implications for future research?

From Zhang et al. 2009:

Based on our findings, we have identified a number of urgent knowledge gaps and uncertainties that need to be addressed by future research. First, there is a need to learn more about key model parameters such as the elasticity of substitution between oil-based fuels and biofuels because of their importance in driving GE model results. Knowledge on these parameters is extremely limited, so far, especially with regard to how they might evolve over time, given that their values are often set based on calibration to a relatively short series of historical data or by expert judgment. Second, better predictions of future crude oil prices are needed, ideally, in the same way that IPCC harmonizes the assumptions underlying quantitative assessments of future climate change impacts, and examines the model-based sensitivity and other key sources of uncertainty that are embodied in the wide range of scenario results. Third, the future expandability of agricultural land supply and the contribution that by-products of biofuel production can make to livestock feed balances are likely to be crucial factors that determine the price impacts of model-based projections, and should be carefully studied. Fourth, the on-going research that is being undertaken by various groups in projecting long-term biofuel impacts on agriculture can benefit greatly from more coordinated modeling efforts, so as to improve the sharing of knowledge and generate a better understanding of the key factors that may offset or aggravate the effects that biofuels can have on market dynamics, environmental quality and, ultimately, human welfare.