While progress has been made in diversifying export destinations, diversification of Uruguayan goods exports has declined over the past two decades, reflecting an increasing role of agricultural commodities and a shrinking manufacturing sector. Export recommendations generated from machine learning algorithms suggest that diversification efforts could usefully rebalance away from commodities and focus on the higher value-added manufacturing categories that support agricultural and raw material production.

Abstract

While progress has been made in diversifying export destinations, diversification of Uruguayan goods exports has declined over the past two decades, reflecting an increasing role of agricultural commodities and a shrinking manufacturing sector. Export recommendations generated from machine learning algorithms suggest that diversification efforts could usefully rebalance away from commodities and focus on the higher value-added manufacturing categories that support agricultural and raw material production.

Export Product Diversification Analysis for Uruguay with Machine Learning1

While progress has been made in diversifying export destinations, diversification of Uruguayan goods exports has declined over the past two decades, reflecting an increasing role of agricultural commodities and a shrinking manufacturing sector. Export recommendations generated from machine learning algorithms suggest that diversification efforts could usefully rebalance away from commodities and focus on the higher value-added manufacturing categories that support agricultural and raw material production.

A. Introduction

1. A diversified export portfolio can foster sustainable growth and economic stability. The relationship between export diversification and countries’ economic performance has been extensively studied in the economic literature. Overall, the existing research indicates that export diversification is a key element in the process of economic development, particularly for developing and emerging market countries trying to catch up with advanced economies. Numerous studies provide evidence of a positive association between export diversification and economic growth and stability (e.g. Imbs and Wacziarg, 2003; Klinger and Lederman 2004 and 2011; Cadot et al., 2011), as a wider range of profitable export products makes growth more sustainable, and reduces the volatility of growth.

2. The economic literature shows that countries with export structures dominated by commodity exports based on natural resources tend to have lower long-term growth and stability. A vast number of economic studies is devoted to studying the impact of natural endowments, e.g. oil and gas, precious metal, and abundant agricultural land, on economic growth. The general conclusion is that, contrary to intuition, natural endowments, though providing economic advantages in the short term, are not necessarily a blessing to long-term growth. See for example, Frankel (2010), Bahar & Santos (2018), Bacha & Fishlow (2011).

3. Past studies reveal multiple reasons why a dominance of commodity exports may not be conducive to long-term growth. Specifically,2

  • Notwithstanding large cyclical swings, agricultural commodity prices have been on long-term declining trend.

  • Countries with a history of dependence on commodity exports tend to have a more concentrated export structure and a less developed industrial sector, although the latter is the source of most productivity increases and technology innovation.

  • The higher volatility of commodity exports, especially agricultural exports, arising from fluctuations in world prices and weather shocks, usually translates into greater macroeconomic volatility.

  • High reliance on commodity exports often correlates with under investment in human capital and other reforms.

For these reasons, commodity exporters’ economies are often plagued by high macro volatility, undiversified export structure, slow developing industrial sector, high fiscal deficit and weak governance and human capital accumulation. Diversifying away from commodities may be desirable for most economies.

4. However, not all types of export diversification are equal, and a growth-friendly diversification strategy needs to be consistent with a country’s comparative advantages. Diversification for its own sake is hardly a recipe for sustainable growth. A foundational idea of the classical international trade theory is that under free trade, countries will tend to export what they have a comparative advantage in. In fact, industrial policies that do not favor the most efficient use of a country’s factor endowment often lead to negative growth outcomes (see, for example, Lin, 2009). On the other hand, delayed industrialization can also lead to poor growth outcomes, as the experience of many resource-rich countries that are entrenched in their over-dependence on commodity exports has shown (e.g., Frankel, 2010).3 Thus, well-targeted industrial policies can be beneficial to growth. Designing growth-friendly industrial policies, however, requires identifying areas of comparative advantage where there is untapped potential for diversification. This paper uses machine learning algorithms for collaborative filtering to explore potential areas of export diversification for Uruguay.

B. Measuring Export Diversification

5. Throughout this paper, export diversification is measured by the number of export products a country has with high “revealed comparative advantage” (RCA). The RCA score, first introduced by Balassa & Noland (1965), is a popular measure in the economic literature for calculating the relative importance of a product in a country’s export basket. Formally, the RCA score of country i in product j can be calculated as:

RCApc=Epc/EcEp/ΣpPEp

where Epc is the export value of product p from country c, Ec is the total export values of country c, Ep is the total exports of product p from all countries around the world, and p'∈PEp' is the total world exports.

6. A high-RCA export product for country c is defined as a product with its RCApc > 1. This is the case when a product’s share in the country’s total exports is greater than the share of the same product in world exports, indicating that the country has a comparative advantage in the product relative to the rest of the world. The interest of the paper is not in recommending products that a country can export a little of, but rather in identifying products that the country can potentially have a high RCA in, in other words, those products that highly align with a country’s latent comparative advantages. Using data for 2007–17,4 Figure 1 shows that controlling for country size, GDP growth rate is positively related to the number of high-RCA exports a country has, and growth volatility is negatively related to the number of high-RCA exports.

Figure 1.
Figure 1.

Uruguay: Number of High-RCA Exports vs GDP Growth and Growth Volatility

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

7. Compared to countries of similar sizes, Uruguay’s level of diversification in goods exports is about average. Unsurprisingly, the number of high-RCA exports is positively correlated with country size as many industries need a minimum scale to be sustainable. Thus, smaller economies—with less capital, labor, and other production endowments—tend to produce fewer product categories than larger countries. Currently the number of high-RCA exports in Uruguay is about the level that would be predicted by its population size. Comparing Uruguay’s number of high RCA exports with those of other countries of smilar size also shows that Uruguay is in the middle of the pack (Figure 2).

Figure 2.
Figure 2.

Uruguay: Export Diversification in Countries of Similar Sizes to Uruguay

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

8. However, over the past 20 years, Uruguay’s goods exports have become less diversified. In late 1990s, Uruguay had over 150 high-RCA exports, placing the country in the “most diversified” category for its size group at the time. Over the past 20 years, though, the number of high-RCA exports has been consistently dropping, to about 2/3 of its peak level. It is worth noting that the analysis focuses on goods exports only. Service exports’ share in total exports of Uruguay has been relatively constant over the past three decades, at around 35 percent of total exports. These consist of mainly tourism and select other services such as information technology.

C. Why Uruguay’s Goods Exports Have Become Less Diversified

9. Some of the general predicaments of commodity exporters are apparent in Uruguay. Agricultural commodities have been increasingly dominating Uruguayan exports since the late 1990s, and currently represent 80 percent of the country’s goods exports (40 percent food, 40 percent raw materials).

Figure 3.
Figure 3.

Uruguay: The Number of High-RCA Exports from Uruguay, 1962–2018

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

uA003fig02

Export Shares by Industry (in percent of total goods exports, SITC 1-digit categories)

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

uA003fig03

Export Shares by Industry

(in percent of GDP, SITC 1-digit categories)

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

10. Agricultural exports concentrate in meat, soybeans, and more recently, forestry products. Meat products have been a staple in Uruguayan exports for decades, with a relatively stable share in total exports of about 20 percent. The increase in commodity exports since the early 2000s has been mostly driven by the growth of the soybean and forestry sectors, reflecting improved prices and large foreign investments. With the price boom ending in 2014, soybean exports declined in importance in both quantity and value, as investors in the sector increasingly moved to neighboring countries with lower production cost such as Paraguay and Brazil. In contrast, forestry exports, including wood and paper pulp, are poised to grow further with the new large foreign direct investment.

uA003fig04

Agricultural and Raw Material Exports

(in SITC 2-digit categories, percent of total exports)

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

11. In the meantime, there has been signs of de-industrialization. Material manufacturing (primarily textile and leather) and some other categories of basic manufacturing (apparel and footwear) once occupied a significant place in Uruguay’s export structure. “Material manufacturing” and “miscellaneous manufacturing” combined was over 35 percent of total goods exports (about 7 percent of GDP) at the beginning of the 1990s. By 2018 their shares had declined to less than 10 percent of total exports. This reflects both the rapid advance of commodity exports and stagnating manufacturing exports.

uA003fig05

Export Volume by Sector

(2005 = 100, simple average across products within a sector)

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

Sources: UN Comtrade and staff calculation.

12. The decline of basic manufacturing is not necessarily a concern by itself, but the issue is it is not replaced by other non-commodity exports of higher value added. Most countries go through evolutionary shifts in their industrial and export structures accompanying their economic development. For example, decades ago, labor-intensive, low-value-added manufacturing products (e.g. textile and footwear) were a large share of exports from countries like Japan and South Korea. As these countries developed, domestic labor costs grew, their exports of basic manufacturing were increasingly priced out of the international markets, leading to a decline in the share of basic manufacturing in total exports. But in their place emerged other manufacturing exports of higher sophistication and value added—e.g. electronics, vehicles, specialized instruments—that align with these countries’ increased physical and human capital and technology know-how. In contrast, the decline in basic manufacturing in Uruguay does not appear to have been followed by industrial upgrades as indicated by the country’s decline in ranking (from 44 to 62) in the export product complexity index.

uA003fig06

Export Product Complexity Ranking

(higher ranking = more complex)

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

Sources: Atlas of Economic Complexity

13. The commodity price boom of 2003–14 likely contributed to the decline in non-commodity exports and the increased concentration in fewer products. The impact of the commodity price boom on export diversification was twofold. First, the boom in the commodity sector attracted capital, labor, and entrepreneurial resources away from the non-commodity tradable sector to the commodity sector, as well as to the non-tradable sector, due to the increased demand and prices of the latter. Secondly, the exchange rate appreciation and real wage increase prompted by a commodity price boom made the non-commodity exporting sector less competitive. The result was a more concentrated export structure in primary commodities.

uA003fig07

Long-Term Soybean Price (USD per metric ton) A

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

14. To summarize, Uruguay’s exports have become more concentrated over the years due to a multitude of factors. These include increased dominance of commodity exports, real exchange rate appreciation and domestic wage increases that reduced the competitiveness of the more traditional non-commodity exports, and the fact that few new non-commodity export categories have emerged to replace the ones that lost traction. Reducing the disruptive impact of commodity price cycles on the normal industrialization process while maintaining the comparative advantage in commodity exports is an important policy challenge.

D. Machine Learning Algorithms to Explore Export Diversification Options

15. A collaborative-filtering based methodology is applied to identify export diversification opportunities for Uruguay. The methodology follows Che (2020) and Che & Zhang (2021). The intuition behind the methodology is that (i) insights about a country’s latent comparative advantages can be gained by looking at other countries with similar comparative advantages, and (ii) a country is likely to have a latent comparative advantage in products that are closely related to existing high RCA products that the country already exports.

16. The methodology employs three main algorithms. These, already widely used in online collaborative filtering recommender systems, are product-based and country-based K Nearest Neighbors (KNN), and Singular Value Decomposition (SVD). These are so-called “top-N recommendations,” i.e., the goal of the algorithms is to generate a list of N product categories that a country can have the largest comparative advantages in. The algorithms are used to predict the RCA scores of products by SITC 4-digit categories for the country under study, using the training dataset of export values by country and export volumes in each category. The top-N recommendations are the N products with the highest predicted RCA scores.5

17. The methodology accurately predicts the export structure and its evolution for several high growth countries. Che (2020) shows that all three algorithms produce diversification recommendations that closely align with countries’ current export structure and comparative advantage profiles, for high-growth countries including China, India, Poland, and Chile. Che and Zhang (2021) further apply the product-based KNN algorithm to close to 200 countries and show that countries whose export structure closely aligns with the algorithm-recommended structure have higher growth and lower growth volatility.

18. The data used in the recommendation system can be represented as a m x n matrix R, where m is the number of countries in the database, and n is the total number of SITC 4-digit products. Each element of R (i.e., rij) is country i’s RCA score in product). R is a sparse matrix due to the fact that each country only exports a subset of the products in the SITC universe. In the case that country i does not export any product),rij = 0. If an implementation uses multiple years of export data, then each country-year is a row in R, i.e., m = c x y, where c is the number of countries in the dataset, and y is the number of years included. In most versions of implementations discussed below, y=1, i.e., if the task is to generate export recommendations for country i in 2017, only the cross-country export data for 2017 is included in the training set.6

Neighborhood-Based Algorithms

Product-Based KNN

19. KNN is one of the most frequently used methods in solving classification and pattern recognition problems and is a popular approach in constructing recommender systems. The basic idea of KNN is learning by analogy-classifying the test sample by comparing it to the set of training samples most similar to it. Different KNN implementations vary in terms of their choices of how the similarity between input vectors is calculated. In the present paper, the cosine similarity score is used as the similarity measure.

20. The intuition behind the current papers product-based KNN implementation is simple. First the algorithm looks at what products a country already has a revealed comparative advantage in and, then, recommends other products that are similar to the former products (similar in the sense that they share common exporters). To explain the approach in more details, let’s first rewrite the RCA score matrix R as:

R = [p1,p2,...,pn]

where pj is a vector of length m that represents the RCA scores of product j for all the m countries in the sample:

pj=[rijr2j..rmj]

21. In machine learning terminology, each product in the sample has m features. The cosine similarity between products i and j is equal to (pi · pl)/(|| pj || pl||), which ranges from -1, when the two vectors are the exact opposite, to 1, when the two are the same. The intuition is that by comparing the two sets of countries that export i and j, and how important the products are in the countries’ export baskets, information can be inferred regarding how closely related the two products are.

22. The implementation of the product-based KNN recommender for country i involves the following steps:

  • a. Represent each product in the SITC 4-digit product space as a vector of RCA scores, pj.

  • b. Select the set of K products in which country i has a revealed comparative advantage, i.e. rij > 1. Let’s call it the high-RCA product set of country i.

  • c. For each j ∈ [1,n], calculate the predicted value of rij as the weighted average RCA score of the high-RCA product set, weighted by the cosine similarity between product j and the products in the country’s high-RCA set.

  • d. The recommended products for country i are the N products with the highest predicted rij values.

Country-Based KNN

23. The exercise can also be thought of as “recommending” other countries similar to the country in question. In other words, instead of recommending product categories related to a countries’ existing export products, the algorithm can be thought as finding a group of countries that are similar to country i. And because these countries have similar comparative advantages, the products they export, beyond the ones country i is already exporting, can be good candidates for diversification for country i. More specifically, the RCA score matrix R can be represented as:

pj=[qiq2..qm]

where qi is a vector of length n that represent country i ‘s RCA scores for the n product categories in the SITC 4-digit product space.

24. The execution of the country-based KNN algorithm for country / can then be broadly described as follows:

  • a. Calculate the cosine similarity score between qt and qj, where 1 ≤ j ≤ m, and j ≠ i.

  • b. Select a set of K countries with the highest similarity scores to country i.

  • c. For each I[I, n], calculate the predicted value of rij as the weighted average RCA score of product I across the K countries, weighted by the similarity score between each country and country i.

  • d. The recommended products for country i are the N products with the highest predicted rij values.

25. It is important to note that although the product-based and country-based KNN recommenders apply similar algorithmic logic, the differences in the perspectives of the two methods lead to different recommendation results, as will be demonstrated in the next section. Generally speaking, since in the data sample n > m,7 it may be easier to identify the relatedness between products with more accuracy than to identify similar countries, which make the product-based KNN a superior approach. The results presented in the next section confirm this hypothesis.

Matrix Factorization Algorithm

26. The KNN algorithms, though intuitive and easy to implement, suffer from some significant drawbacks. First, these algorithms have limited scalability. As the sizes of m and n increase, the amount of computation required to calculate the similarity scores increases at 0(ri) time, reducing the performance of the algorithm on larger data sets. Another disadvantage of the KNN algorithms is their problem with sparse data. Since the KNN algorithms require explicitly calculating similarities among vectors, the calculation becomes increasingly inaccurate when there’s a lot of missing data in matrix R. This problem is exacerbated by the fact that the algorithm essentially treats each row of the product vector (or the country vector) as independent features of equal importance, which is not the most efficient use of information in the data, and also makes missing rows generally more damaging, compared to algorithms that impose some discretion on the relative importance of different data points. For the current use case, the first drawback is not a big concern, as the m and n of the country-product space are relatively small, especially when we do not include multiple years in the calculation. The second drawback is more problematic, as it implies that the KNN algorithms would perform worse on countries that are significantly under-diversified, i.e. lots of missing entries in R for these countries. This would potentially defeat the purpose of the exercise, as under-diversified countries are arguably the ones that are most in need of diversification recommendations.

27. The Singular Value Decomposition (SVD) algorithm provides a possible remedy to the problem. SVD is a matrix factorization technique widely used in dimensionality reduction and principal component analysis. The basic idea is that matrix R can be decomposed into three matrices:

R = USV'

where U and V are two orthogonal matrices of size m x r and n x r respectively. r is the rank of R. And S is a r x r diagonal matrix, with the singular values of R as its diagonal elements, sorted in the order of decreasing magnitude.

28. The main purpose of the decomposition is to represent the products and countries as combinations of the latent factors in the data, which are implicit, orthogonal features that can be used to characterize the entire country-product space. U represents the relationship between countries and the latent factors, while V represents the similarity between products and latent factors. The diagonal elements of S can be thought of as the relative scaling values assigned to various latent factors.

29. To illustrate the intuition behind the algorithm, here is a simplified example. Suppose the matrix R can be summarized by three independent latent factors: labor, land, and knowledge. Row i of matrix U represents the comparative advantage of country i as a combination of the latent factors. ut = [.55,.4,.05] would mean that country i’s profile can be described as 50% labor, 40% land, and 5% knowledge- a resource-rich, developing country. Column j of matrix V represents the characteristics of product; as a combination of latent factors. Thus vi = [. 15,.05,.80]’ means that the production of product; can be characterized as 15% labor, 5% land, and 80% knowledge- a technology-product that requires mostly intangible inputs. rij = ui vJ, scaled by the appropriate diagonal element in S. It’s not difficult to see that rij would be relatively small, i.e. country i does not have a comparative advantage in producing product;. This is, of course, a very hypothetical example. In practice, the latent factors computed by the optimization algorithm are not human-interpretable, and only serve as features that more efficiently characterize data.

30. The goal of the SVD algorithm is essentially to find the best estimations of U and V', and then produce recommendations based on estimated rij^=ui^v^j. In practice, because R is already sparse, observing orthogonality constraints for U and V' becomes computationally untenable. The execution of the algorithm thus centers on solving the following optimization problem:

minui,vjΣrijR(rijuivj)2+λ(|ui2+|vj2)

where λ is a regularization factor. The minimization is performed with stochastic gradient descent, using python Surprise library for building recommender systems. The recommended products for country i are the products with the highest predicted rij^ value.

E. The Application to Uruguay

31. Export recommendations for Uruguay are derived using SITC 4-digit export data for 2017, consisting of 786 product categories and around 260 countries. Table 1 presents a few summary statistics of the results. The three algorithms identify 144, 155, and 189 products from the SITC 4-digit categories as products that Uruguay could have a comparative advantage in. The hit rate measures the percentage of recommendations with an actual RCA score greater than 1, that is, those Uruguay is already exporting a lot of. The top-100 hit rate measures the percentage of the top 100 recommendations from each algorithm with an actual RCA score greater than 1.

Table 1.

Uruguay: Summary Statistics of the Algorithm Recommendations

article image

32. Table 2 compares the sectoral composition of actual high-RCA exports in 2017 with sectoral composition recommended by the models, by SITC 1-digit category. Specifically, it shows the numbers of 4-digit high-RCA exports contained in each 1-digit category, for actual and recommended exports. The shares are the numbers of 4-digit products in each category as percent of the total number of high RCA exports in actual and recommended export portfolio. Figure 4 compares the actual sectoral composition for high-RCA exports with the weighted-average composition from the three algorithms.8

Table 2.

Uruguay: Actual and Recommended Export Composition for High-RCA Exports

article image
article image
Figure 4.
Figure 4.

Uruguay: Actual and Recommended Export Composition for High-RCA Exports

Citation: IMF Staff Country Reports 2022, 017; 10.5089/9798400200335.002.A003

Sources: staff calculation

33. Overall, the results recommend a more balanced export portfolio, shifting somewhat away from agricultural commodities and towards non-commodity manufacturing exports. Currently the largest number of high-RCA exports from Uruguay is in SITC category 0 (food and live animals) and the models pick up that Uruguay has strong comparative advantages in this category, as indicated by high numbers of recommendations in this category from all three algorithms. However, the potential to diversify further within this category appears to be limited. To increase the total number of high-RCA exports, the models indicate that Uruguay would need to expand its export diversity further in categories 2 (inedible crude materials), 6 (basic manufacturing), and 8 (other manufacturing). In addition, though Uruguay currently does not have any high-RCA exports in category 7 (machinery and transport equipment), the models suggest that the country could have a comparative advantage in some products in this category (see Table 5 for some of the examples).

Table 3.

Uruguay: Top 10 SITC 2-digit Categories with Increased Shares in Total High-RCA Exports1/

article image

These are results from the product-based KNN algorithm.

Table 4.

Uruguay: Top 10 SITC 2-Digit Categories with Reduced Shares in Total High-RCA Exports1/

article image

These are results from the product-based KNN algorithm.

Table 5.

Uruguay: Top 20 SITC 4-Digit Product Recommendations with Actual RCA <11/

article image

According to the results from the product-based KNN.

34. Tables 3 and 4 present the biggest “gainers” and “losers” by SITC 2-digit categories, when comparing the recommended export structure with the existing one. These results are based on the product-based KNN algorithm, which is the best performing one among the three algorithms. Column A in both tables report the current number of SITC 4-digit high-RCA exports in percent of the total number of high RCA exports. Column B shows the number of high-RCA exports, as percentage of the total, from the recommendation results. Table 3 suggests that the categories showing the largest diversification potentials are select categories in manufacturing, as well as some categories in crude materials.

35. Table 5 provides greater granularity by focusing on the top 20 recommendations of the SITC 4-digit categories which Uruguay currently does not export much of, i.e., RCA < 1, according to the product-based KNN algorithm. Most of these top items are in the categories of machinery and equipment, as well as material manufacturing.

36. The main identified export gaps (between recommended and actual export shares) relate to higher value-added categories connected to the agricultural sector. Many of the top recommended items in Table 5—e.g., harvesting machines, dairy machinery, textile fabrics for machinery—are production tools that support the agricultural exporting sector. In fact, the top-20 products recommended in Table 5 have an average Product Complexity Index (PCI)—an index measuring the relative knowledge intensity of products—that is higher than 70 percent of the SITC 4-digit product categories, according to the ranking by the Observatory of Economic Complexity. In contrast, Uruguay’s existing export portfolio is at the 55 percentile in the product complexity ranking.9

37. Many of these recommendations are relatively more knowledge-intensive than Uruguay’s current export profile, although still intimately related to the country’s comparative advantages in agriculture. The fact that the country is a significant agricultural commodity exporter means that many of the industrial products serving agricultural production recommended in Table 5 could leverage domestic product and customer networks, opportunities for knowledge exchange, and feedback loops for product research & development. Moving in the recommended direction would allow the country to take advantage of its existing knowledge base, human capital, and production networks in agricultural products, while shielding the economy from the volatility of commodity prices to some extent, since the prices of agriculture-related manufacturing products are less variable than those of raw commodities.

38. More in-depth analysis is needed on how to foster product diversification towards these higher value-added sectors. Given that Uruguay has relatively limited experience in exporting products in SITC categories 7 and 8, diversifying into these categories may require targeted efforts. Doing so requires identifying the main constraints and market frictions that prevent the emergence of these recommended products and sectors, and the related policy remedies to support diversification.

F. Conclusion

39. Uruguay’s goods exports have become increasingly concentrated in the agricultural sector over the past 20 years. In late 1990s, Uruguay was one of the most diversified countries in its size group. Since then, however, the number of high-RCA export products in SITC 4-digit categories has dropped by close to 40 percent. The agricultural commodity price boom has been a key contributing factor to this trend, attracting resources away from manufacturing and the normal industrialization process.

40. The diversification recommendation models point to significant diversification potential in higher valued-added manufacturing linked to agriculture and other traditional exports. While the models confirm that Uruguay has strong a comparative advantage in agricultural commodities, they also indicate that there is potential for diversification by rebalancing towards non-commodity export categories such as material manufacturing and machinery & equipment. In particular, manufacturing categories that provide the production means for the agricultural and raw material sectors can be a promising diversification area.

41. Greater efforts to identify the key barriers to diversification are needed. While Uruguay appears to have latent comparative advantages in the above-mentioned categories—including due to the country’s relative abundance in human capital and physical capital (although both need maintenance and upgrades)—the limited growth, so far, of such categories suggests that there may be barriers to their development. Further analyses are needed to identify such barriers and design policies to lift them.

References

  • Adeniyi, David Adedayo, Zhaoqiang Wei, and Y Yongquan. “Automated web usage data mining and recommendation system using K-Nearest Neighbor (KNN) classification method”. In: Applied Computing and Informatics 12.1 (2016), pp. 90108.

    • Search Google Scholar
    • Export Citation
  • Aiginger, Karl and Dani Rodrik. “Rebirth of Industrial Policy and an Agenda for the Twenty First Century”. In: Journal of Industry, Competition and Trade (2019), pp. 119.

    • Search Google Scholar
    • Export Citation
  • Bacha, Edmar L and Albert Fishlow. “The recent commodity price boom and Latin American growth: More than new bottles for an old wine?” In: The Oxford Handbook of Latin American Economics. 2011.

    • Search Google Scholar
    • Export Citation
  • Balaguer, Jacint and Manuel Cantavella-Jorda. “Structural change in exports and economic growth: cointegration and causality analysis for Spain (1961-2000)”. In: Applied Economics 36.5 (2004), pp. 473477.

    • Search Google Scholar
    • Export Citation
  • Cadot, Olivier, C’eline Carr’ere, and Vanessa Strauss-Kahn. “Export diversification: what’s behind the hump?In: Review of Economics and Statistics 93.2 (2011), pp. 590605.

    • Search Google Scholar
    • Export Citation
  • Che, Natasha. “Intelligent export diversification: an export recommendation system with machine learning”. IMF Working Paper, 2020.

  • Che, Natasha and Xuege Zhang. “High performance export portfolio: design growth-enhancing export structure with machine learning”. IMF Working Paper, 2021.

    • Search Google Scholar
    • Export Citation
  • Clickstreams, Multi-faceted Web. “Workshop Notes”, 2005.

  • Feenstra, Robert and Hiau Looi Kee. “Export variety and country productivity: Estimating the monopolistic competition model with endogenous productivity”. In: Journal of international Economics 74.2 (2008), pp. 500518.

    • Search Google Scholar
    • Export Citation
  • Frankel, Jeffrey A. The Natural Resource Curse: A Survey. Tech. rep. National Bureau of Economic Research, 2010.

  • Giri, Rahul, Mr Saad N Quayyum, and Rujun Yin. Understanding Export Diversification: Key Drivers and Policy Implications. International Monetary Fund, 2019.

    • Search Google Scholar
    • Export Citation
  • Hausmann, Ricardo, Jason Hwang, and Dani Rodrik. “What you export matters”. In: Journal of economic growth 12.1 (2007), pp. 125.

  • Hausmann, Ricardo and Bailey Klinger. “The structure of the product space and the evolution of comparative advantage”. In: CID Working Paper Series (2007).

    • Search Google Scholar
    • Export Citation
  • Herzer, Dierk and Felicitas Nowak-Lehnmann D. “What does export diversification do for growth? An econometric analysis”. In: Applied economics 38.15 (2006), pp. 18251838.

    • Search Google Scholar
    • Export Citation
  • Hidalgo, Cesar A and Ricardo Hausmann. “The building blocks of economic complexity”. In: Proceedings of the national academy of sciences 106.26 (2009), pp. 1057010575.

    • Search Google Scholar
    • Export Citation
  • Imbs, Jean and Romain Wacziarg. “Stages of diversification”. In: American Economic Review 93.1 (2003), pp. 6386.

  • IMF. “Sustaining Long-Run Growth and Macroeconomic Stability in Low-Income Countries-The Role of Structural Transformation and Diversification”. In: IMF Policy Paper (2014).

    • Search Google Scholar
    • Export Citation
  • IMF. “Real Exchange Rate And Sectoral Competitiveness in Uruguay”. In: Uruguay: Selected Issues (2019).

  • Harvey, David I., Neil M. Kellard, Jakob B. Madsen, and Mark E. Wohar. “The resource curse, commodity prices and economic growth.” In Global Commodity Markets and Development Economics, pp. 1649. Routledge, 2018.

    • Search Google Scholar
    • Export Citation
  • Klinger, Bailey and Daniel Lederman. Discovery and development: An empirical exploration of “new” products. The World Bank, 2004.

  • Klinger, Bailey and Daniel Lederman. “Export discoveries, diversification and barriers to entry”. In: Economic Systems 35.1 (2011), pp. 6483.

    • Search Google Scholar
    • Export Citation
  • Koren, Yehuda, Robert Bell, and Chris Volinsky. “Matrix factorization techniques for recommender systems”. In: Computer 42.8 (2009), pp. 3037.

    • Search Google Scholar
    • Export Citation
  • Lathia, Neal, Stephen Hailes, and Licia Capra. “kNN CF: a temporal social network”. In: Proceedings of the 2008 ACM conference on Recommender systems. 2008, pp. 227234.

    • Search Google Scholar
    • Export Citation
  • Lin, Justin Yifu and Feiyue Li. Development strategy, viability, and economic distortions in developing countries. The World Bank, 2009.

  • Al-Marhubi, Fahim. “Export diversification and growth: an empirical investigation”. In: Applied economics letters 7.9 (2000), pp. 559562.

    • Search Google Scholar
    • Export Citation
  • Paterek, Arkadiusz. “Improving regularized singular value decomposition for collaborative filtering”. In: Proceedings of KDD cup and workshop. Vol. 2007. 2007, pp. 58.

    • Search Google Scholar
    • Export Citation
  • Prebisch, Raul. “The economic development of Latin America and its principal problems”. In: Economic Bulletin for Latin America (1962).

    • Search Google Scholar
    • Export Citation
  • Sarwar, Badrul et al. Application of dimensionality reduction in recommender system-a case study. Tech. rep. Minnesota Univ Minneapolis Dept of Computer Science, 2000.

    • Search Google Scholar
    • Export Citation
  • Sarwar, Badrul et al. “Incremental singular value decomposition algorithms for highly scalable recommender systems”. In: Citeseer.

  • Singer, Hans W.The distribution of gains between investing and borrowing countries”. In: The Strategy of International Development. Springer, 1975, pp. 4357.

    • Search Google Scholar
    • Export Citation
  • Singer, Hans W. The Distribution of Gains between Investing and Borrowing Countries”. In: The American Economic Review 40.2 (1950), pp. 473485.

    • Search Google Scholar
    • Export Citation
1

By Natasha Che.

2

The literature on the so-called “natural resource curse” documents these patterns for various countries and time periods. See surveys by Frankel (2010) and Harvey et al (2018).

3

Relatedly, Hausmann et al. (2007) finds that countries that export more sophisticated, or knowledge-intensive products, tend to grow faster, controlling for initial income levels.

4

The number of high RCA exports are calculated at SITC 4-digit product level.

5

For a detailed account of the methodology and rationales behind the algorithms, see Che (2020).

6

The paper experimented with including multiple years of data in the training set, but found no significant improvement in the evaluation metrics, while the model took longer to compute as the size of m increases.

7

There are close to 800 product categories in the SITC 4-digit product space, while there are just over 250 countries in the sample.

8

The weights are .6, .2, and .2, for product-based KNN, country-based KNN, and SVD respectively. These are subjective weights assigned based on the algorithms’ general performances according to the summary statistics. The product-based KNN model is given a higher weight because it is shown as more forward looking when the algorithms are applied to historical data from other countries.

9

See https://oec.world/en/profile/country/ury for details on the calculation of product complexity.

Uruguay: Selected Issues
Author: International Monetary Fund. Western Hemisphere Dept.
  • View in gallery

    Uruguay: Number of High-RCA Exports vs GDP Growth and Growth Volatility

  • View in gallery

    Uruguay: Export Diversification in Countries of Similar Sizes to Uruguay

  • View in gallery
  • View in gallery

    Uruguay: The Number of High-RCA Exports from Uruguay, 1962–2018

  • View in gallery

    Export Shares by Industry (in percent of total goods exports, SITC 1-digit categories)

  • View in gallery

    Export Shares by Industry

    (in percent of GDP, SITC 1-digit categories)

  • View in gallery

    Agricultural and Raw Material Exports

    (in SITC 2-digit categories, percent of total exports)

  • View in gallery

    Export Volume by Sector

    (2005 = 100, simple average across products within a sector)

  • View in gallery

    Export Product Complexity Ranking

    (higher ranking = more complex)

  • View in gallery

    Long-Term Soybean Price (USD per metric ton) A

  • View in gallery

    Uruguay: Actual and Recommended Export Composition for High-RCA Exports