It is often argued that returns to R&D are low in developing countries, making imported technologies a better path to growth. Yet technologies designed for frontier nations may not fit local conditions, limiting their productivity gains. This column studies Brazil’s Embrapa, a large public R&D effort to create agricultural innovation suited to Brazil’s ecology. Embrapa shifted research toward local needs and raised agricultural productivity by 110%, far outweighing its costs. Sustained, locally targeted R&D can be a key part of development policy.
Can targeted public investments in R&D drive productivity growth in developing countries? A common presumption in economics is that innovation takes place exclusively in a set of frontier nations, where returns to R&D are high, and countries outside of this set should focus on enacting policies that reduce any barriers to foreign technology adoption (e.g. Acemoglu et al. 2006, Parente and Prescott 1994). Underlying this view is the belief that returns to R&D are low in low- and middle-income countries – perhaps due to more limited human or physical capital – and that foreign technologies can generate productivity and income growth regardless of where they are applied.
Opportunities for growth without local innovation, however, may be more limited in practice. Technology developed in and for frontier nations may be less suited to local conditions elsewhere in the world and therefore generate limited productivity benefits. Mounting evidence suggests that advances in medical and agricultural technology, for example, are tailored to the environmental and ecological conditions of rich countries and, as a result, are much less effective elsewhere in the world (Kremer and Glennerster 2004, Moscona and Sastry 2025).
From this perspective, there could be large returns to developing locally-appropriate innovation, and funding public R&D is the main policy lever that developing country governments can use to achieve this goal. However, little is currently known about the cost-effectiveness of R&D programmes in developing regions or whether they allow local countries to escape this ‘technology mismatch trap.’
In a recent paper (Akerman et al. 2025), we study one of the most prominent examples of public R&D investment in a developing country: the Brazilian Agricultural Research Corporation (Embrapa). Embrapa was established in 1973 to catalyse agricultural science and technology development in Brazil. An explicit goal of the programme was overcoming the existing dearth of agricultural innovation relevant to Brazil’s unique ecological conditions.
In the words of one of Embrapa’s founders, Eliseu Alves:
[T]he major problem in Brazilian agriculture was not a lack of potential. The potential existed, but there was no science capable of generating technology suited to what we needed. To address this, we needed an institution capable of focusing high-level science on solving the concrete problems of Brazilian agriculture (Duarte 2018).
Embrapa combined large-scale investments in R&D – comparable as a share of GDP with that of the US – with researcher training programmes and established a network of research centres covering all parts of the country. Figure 1 shows every Embrapa research lab, colour coded by founding year.
Figure 1 Location of all Embrapa research centres, by year of establishment
Notes: This map displays the locations of all Embrapa research centres by year of creation, overlaid on Brazil’s major biomes and state boundaries. Data sourced from Embrapa and the Instituto Brasileiro de Geografia e Estatistica.
Over decades, Embrapa expanded to all parts of the country, covering every state and every biome. Researchers at these centres developed soil modification tools adapted to the acidic and nutrient-poor soils of central Brazil, identified genetic traits that conferred crops resistance against Brazil’s pest threats, and released hundreds of new crop varieties, including the first that allowed for soy to be produced in tropical latitudes (Monteiro et al. 2012, Correa and Schmidt 2014).
Over the same period, Brazil transitioned from being a major recipient of US food aid to one of the most productive agricultural exporters in the world, achieving the fastest growth in agricultural total factor productivity of any country with over one million inhabitants. Figure 2 displays national productivity trends for three important staple crops: soy, maize, and rice.
Figure 2 Productivity trends for staples: soy, maize, and rice
Notes: This graph shows the evolution of physical yields (tons of output per hectare) for three crops in Brazil, using data from the Brazilian Census of Agriculture (1950 to 2017). Each dot indicates a separate observation. In each panel, we normalize the yield in 1970 to one. Data for soybeans are not available before 1970.
Productivity in these crops was stagnant prior to the 1970s, and a major productivity take-off coincides with the founding of Embrapa. But beyond a handful of case studies, did Embrapa systematically fuel the development of locally-appropriate innovation? If so, to what extent did this technology development explain Brazil’s agricultural productivity take-off? And perhaps most important to policymakers, did these productivity benefits justify the large investment costs required to build a local R&D ecosystem?
Embrapa shifted agricultural research towards more ‘Brazilian’ topics
To address these questions, we first investigate the impact of Embrapa on agricultural research output. We construct a detailed record of agricultural science in Brazil using information from the near universe of resumes of agricultural scientists. We compile this data set using Brazil’s Lattes platform, a government-run database in which all researchers are required to have an up-to-date CV to apply for any form of public funding. We then use this information to construct detailed employment and publication histories for each scientist and use the title of each publication alongside keyword searches to determine the topic of each paper.
We find that Embrapa facilitated research on local ‘Brazilian’ topics: articles written by Embrapa’s researchers are much more likely to mention a Brazilian biome, a major Brazilian crop-affecting pest or pathogen, or one of the staple crops singled out in Embrapa’s founding to address the nation’s food insecurity. This was largely due to the location of the research labs and the fact that Embrapa spread its research efforts throughout the country. For example, researchers were much more likely to study a particular biome or pest if their lab is located in that biome or where that pest is present, and this fact drives a large part of Embrapa’s disproportionate focus on these topics.
Moreover, we find no evidence that Embrapa’s research crowded out research on these topics that might have happened in its absence; instead, Embrapa shifted the overall direction of research toward more ‘Brazilian’ topics, including research by scientists not directly employed by Embrapa.
One potential concern is that this shift in research topics may have reduced researcher’s overall productivity. To cover Brazil’s heterogeneous ecology, Embrapa established centres in potentially research-unproductive areas with low pre-existing human capital, research capacity, and researcher agglomeration; this might have come with a cost to research productivity. However, we find that, if anything, researchers become more productive when they move to an Embrapa lab. This is especially true in more remote regions of the country and, quantitatively, more than compensates for the direct negative productivity effect of working in these regions.
Thus, consistent with qualitative accounts, Embrapa seems to have spurred the development of productive agricultural research labs that focused on environmental and ecological conditions specific to Brazil.
Embrapa had positive effects on agricultural productivity
We next investigate the impact of Embrapa on agricultural productivity, using municipality-level panel data on agricultural outcomes from the Brazilian Census of Agriculture (1960–2017). To estimate the effect of Embrapa on productivity, we develop a measure of ‘Embrapa exposure’ that exploits both the staggered rollout of Embrapa centres over time (see Figure 1) and heterogeneous local suitability of Embrapa’s innovation.
Specifically, motivated by our results documenting the local ecological focus of agricultural research, we measure the suitability of research developed in each Embrapa centre for all other municipalities using an environmental-similarity index based on several characteristics of climate, topography, and soil conditions (as in Moscona and Sastry 2025, Bazzi et al. 2016). We then exploit the fact that the staggered opening of new Embrapa centres shifts the extent to which Embrapa is developing new technology that is appropriate for every other region of Brazil.
We find that Embrapa had large, positive effects on agricultural productivity. This effect accumulates over time, consistent with the time lags in technology development and diffusion, and is similar for various measures of local agricultural output and productivity.
We then turn to the mechanisms underlying this baseline effect. First, we show that exposure to Embrapa was associated with greater spending on intermediate inputs like seeds, chemicals, and fertilisers that were the main focus of Embrapa’s technology development. Land devoted to crop production also increased (while land devoted to pasture declined), as did the use of mechanical technologies, suggesting that productivity growth and intensification went hand in hand.
Second, separating the effect of Embrapa by crop, we find that the productivity effects are concentrated exclusively in the staple crops that were the focus of Embrapa’s R&D investment. These findings are consistent with directed innovation as the driving mechanism for our main findings.
Embrapa was a cost-effective investment
Finally, we combine these estimates with a quantitative model and detailed data on Embrapa’s research investment costs to analyse the aggregate effect of Embrapa on Brazilian agricultural productivity growth and the cost-effectiveness of this investment. The results are displayed in Figure 3. We find that Embrapa increased aggregate agricultural productivity by 110% (Figure 3 panel (a), first bar). This corresponds to 39% of the total agricultural productivity growth in Brazil between 1970 and the present (see Fuglie 2015).
Figure 3 Aggregate effect of Embrapa on Brazilian agricultural productivity growth and the cost-effectiveness of investment
Notes: Panel (a) displays the impact of Embrapa’s research on aggregate agricultural productivity. Panel (b) shows the benefit-cost ratio implied by the productivity gains presented in panel (a) and data on the costs of research. In both panels, the first bar shows the actual impact of Embrapa and the second bar shows the impact in a counterfactual scenario in which all of Embrapa’s resources were invested in a single research centre. Error bars are 95% confidence intervals computed based on the delta method and p-values are based on the null hypothesis of no difference between columns.
Combining this estimate with Embrapa’s expenditures, all converted into common units, implies a benefit-cost ratio of 17 (Figure 3 panel (b), first bar). Thus, while the cost of Embrapa was considerable – about 1% of Brazil’s total agricultural GDP at its peak, comparable to the scale of investment in the US (Correa and Schmidt 2014) – our analysis suggests that the benefits were considerably larger.
Moreover, we find that a large share of these benefits come from the fact that Embrapa spread its investment across ecological zones rather than concentrate it all in one location. While there may be benefits from concentrating researchers in a single location (e.g. research spillovers or learning from others), our results suggest that the overall impact of Embrapa and its benefit-cost ratio would be less than half as large if all investment were made in a single, large centre (Figure 3 panels (a) and (b), second bar).
Together, these results suggest that investment in public R&D can be an important catalyst for productivity growth. Global R&D is concentrated in a handful of rich countries and the resulting lack of locally suitable technology can constrain productivity growth in large parts of the world. Targeted but sustained R&D investment in developing countries could be an important component of development policy.
While our results suggest that Brazil’s Embrapa was a clear success story – and more than paid for itself in social benefits – important questions remain. First, how was investment in Embrapa sustained over several decades, during which there were major political and economic swings that might have undermined support for science? Second, what aspects of Embrapa’s design are applicable to other sectors or other parts of the world, where technology mismatch may also be constraining productivity growth? Understanding the political economy of support for R&D investment and portability of R&D programmes across contexts strike us as important areas for further investigation.
Source : VOXeu
