WORLD BANK GROUP Raising Agricultural Productivity November 2021 Kosovo Country Economic Memorandum  1 Kosovo Country Economic Memorandum Raising Agricultural Productivity Table of Contents Executive summary  7 Context 11 Productivity Analysis 21 Input-Specific Efficiency 31 Public Spending on Agriculture and Rural Development 35 Distribution of Farm Support 41 Policy Priorities and Institutional Environment 45 The Role of Agribusiness in Kosovo’s Economy  49 Policy Reform Options 55 Annex A: Theoretical Underpinnings and Methods 59 Annex B: Data 65 Annex C: Analytical findings of efficiency and productivity analysis  75 References 82 2  Figures Fig 1. Performance of Agriculture in Fig 11. Mean TE and SE scores Fig 21. Agriculture Spending by Kosovo and Peers, 2008–19 according to farm location, Function, Kosovo, Million Euros p.12 Kosovo, 2017 p.24 at Current Prices, 2014–19 p.37 Fig 2. Growth in Labor Productivity, Fig 12. Mean TE Scores by Farm Size Fig 22. Direct Payments by Type, 2009–2019 p.13 and Subsidization, Kosovo, Kosovo, 2013–2019 p.38 Fig 3. Agricultural Value-Added and 2017 p.25 Fig 23. Types of Rural Development Employment Compared to Fig 13. Mean TE Scores by Farm Support, Kosovo, Million Euros structural and aspirational Products and Subsidization, at Current Prices, 2013–19 p.39 peers p.13 Kosovo, 2017 p.26 Fig 24. Direct Payments to Farms by Fig 4. Labor Productivity, Kosovo Fig 14. Mean TE Scores by Farm Economic Size, Kosovo, 2015, and Regional and Aspirational Location and Subsidization Percent p.42 Peers, Agricultural Value Added Kosovo, 2017 p.26 Fig 25. Direct Payments to Farms by per Worker, Constant 2010 Fig 15. Distribution of TFP and its Economic Size, Kosovo, 2017, US$, 2018 p.15 Components, Kosovo, 2015–17 Percent p.43 Fig 5. Cereal Yields Compared, p.27 Fig 26. Agribusiness Firms by Product Regional and Aspirational Fig 16. Distribution of TFP Growth by Category, Kosovo, 2018, Peers, kg per ha, average 2017- Time Period, Kosovo, 2015–17 Percent p.50 2018 p.16 p.28 Fig 27. Agri-food Trade, Kosovo, Fig 6. Farm Households by Age, Fig 17. Distribution of TFP Growth by 2014–19, Percent of Goods Regional and Aspirational Farm Size Kosovo, 2015–17, Trade p.51 Peers, 2016, Percent p.16 Percent p.28 Fig 28. Agri-food Trade Compared, Fig 7. TE and SE in Agriculture, Fig 18. Farm Input-specific Efficiency, 2019, Percent of Goods Trade Kosovo, 2017 p.22 Kosovo, 2017 p.33 p.52 Fig 8. Mean TE and SE Scores by Farm Fig 19. Total Public and Agriculture Fig 29. Agriculture Trade Balance, Size, Kosovo, 2017 p.23 Spending, Millions Euros in Kosovo, Million Euros, 2019 Fig 9. GDP and Commercial Loan 2011 Values, 2011–2019 p.36 p.52 Shares by Sector, 2018 p.23 Fig 20. Agricultural Value Added Fig 30. Food Trade Balance, Kosovo, Fig 10. Mean TE and SE Scores by Farm (Percent of GDP) and Spending Million Euros, 2019 p.53 Specialization, Kosovo, 2017 (Percent of Public Spending) p.24 Compared with Peers, 2017 p.36 Tables Table 1. Farms in Kosovo and Comparators by Size, Percent Share p.17 Table 2. Farms with Input Slacks, Kosovo, 2017 p.32 Table 3. Probability of a Farm with Specific Number of Slacks, Kosovo, 2017 p.32 Table 4. Estimated and Actual Input Slack, Kosovo, 2017 p.32 Table 5. Farms with Input-Specific Slacks, Kosovo, 2017 p.33 Table 6. Probability of Input-specific Farm Inefficiency, Kosovo, 2017 p.33 Table 7. Agribusiness Enterprises in Kosovo, 2014–18 p.50  3 © 2021 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. 4  Acknowledgements This note was prepared by Silvia Mauri (Senior Agriculture Specialist), Demetris Psaltopoulos (Professor, Aristotle University of Thessaloniki), and Kostas Tsekouras (Professor, University of Patras), with contributions from Besart Myderrizi (Research Analyst) and Enrique Blanco Armas (Lead Economist), as part of the Kosovo Country Economic Memorandum team, which was led by Aslı Şenkal (Senior Economist, Task Team Leader) and Edith Kikoni (Senior Economist). The work was overseen by Linda Van Gelder (Country Director, Western Balkans), Massimiliano Paolucci (Country Manager, Kosovo and North Macedonia), Jasmin Chakeri (Practice Manager, Macro, Trade, and Investment Global Practice), and Frauke Jungbluth (Practice Manager, Agriculture and Food Global Practice). The analysis benefitted from comments from Ulrich Schmitt (Lead Agriculture Economist), Daniela Topirceanu (Program Manager, EU Delegation), Naser Krasniqi (National Team Leader, FAO), Mustafa Kastrati (Adviser, GIZ) and Fatmir Selimi (Chief of Party, USAID AGRO Project Kosovo). The team is grateful to Delvina Hana (Head of Economic Analysis Division), Edona Mekuli (Statistical Officer for Farm Accounting Data Network, FADN) and the Ministry of Agriculture, Forestry and Rural Development for providing access to the FADN database.  5 Acronyms ARDP Agriculture and Rural Development Program AKIS Agricultural Knowledge and Innovation Systems CAP Common Agricultural Policy, EU CEFTA Central European Free Trade Agreement DEA Data Envelopment Analysis ERP Economic Reform Program EU European Union FADN Farm Accounting Data Network FAO Food and Agriculture Organization FDI Foreign direct investment GDP Gross domestic product ha hectares IFC International Finance Corporation IPA Instrument for Pre-Accession Assistance (EU) IPARD Instrument for Pre-Accession Assistance for Rural Development KAS Kosovo Agency of Statistics KCGF Kosovo Credit Guarantee Fund KVFA Kosovo Veterinary and Food Agency LPIS Land Parcel Identification System MAFRD Ministry of Agriculture, Forestry and Rural Development SE Scale efficiency SEC Scale efficiency change TC Technical change TE Technical efficiency TEC Technical efficiency change TFP Total factor productivity UAA Utilized agricultural area USAID United States Agency for International Development WDI World Development Indicators 6  Executive summary Igniting farm productivity can support growth and job creation in Kosovo. Agricultural production, in real terms, has been decreasing in Kosovo since 2009 but employment has not changed much. Agriculture continues to be important, though its contribution to growth and job creation has been shrinking. Agricultural value-added as a share of GDP and agricultural employment are much lower than in other countries at a similar development level, while both are closer to that of high- income countries. This finding may at first suggest a faster structural transformation than expected based on Kosovo’s GDP per capita but in fact also stems from low agricultural productivity. This note examines drivers of agricultural productivity and its growth in Kosovo, and implied constraints on growth of agriculture, using farm-level data. To assess the productivity dynamics over time, implied constraints have been identified by estimating technical and scale efficiency and changes in total factor productivity (TFP). Technical efficiency (TE) and scale efficiency (SE) measure the implied loss in terms of production caused by not adopting the best production techniques and the optimal scale of production. The analysis also looks into total factor productivity changes over time driven by change in technical efficiency and scale efficiency as well, technical change (TC), which measures change in the amount of output produced from the same amount of inputs. The note also discusses the role of state subsidies in driving efficiency and productivity changes over time. The micro analysis is complemented by analysis of structural data to gauge the contribution of agriculture and agribusiness in Kosovo’s economy in terms of growth, jobs, and the external trade balance. The results of the productivity analysis suggest that in Kosovo there is a considerable misallocation of resources that if remedied could boost growth and job creation. Findings in this note suggest that improving the efficiency of production processes over time has contributed to growth. However, growth in agriculture is constrained because most Kosovar farms still do not efficiently transform their inputs into output, in terms of either technical or scale efficiency. In Kosovo, which suffers from low TE, an average farm could produce the same amount of output using 72.8 percent less inputs. Farms are more efficient with regard to their operating SE, but that finding is driven by larger farms. Smaller farms, which constitute more than 70 percent of Kosovan farms, do not benefit from returns to scale, most likely due to credit and other constraints.1 Because unemployment is high in Kosovo, the non-farm economy cannot absorb agriculture’s surplus labor. Farms are unable to achieve both TE and SE at the same time. For farms whose TE is higher, SE is low. Farms specializing in horticulture and wine grapes seem to possess high managerial competence but are unable to explore scale economies, due among other reasons to cash-flow and credit constraints, which seem to impede investment in higher-value products. Farms specializing in field crops, milk, grazing livestock show to be exploiting economies of scale relatively well (perhaps due to their comparatively high farm support granted by the state, which also could be used for investments) but suffer from technical inefficiencies. Younger farmers in Kosovo do not seem to farm more efficiently than their older peers. This is a surprising variation from regional experience. It is also an important finding, because the proportion of younger farm managers in total farmers in Kosovo is higher than in structural peers. On the positive side, driven by large farms, total factor productivity—the main driver of agricultural growth in Kosovo—is improving. Productivity analysis found a healthy 9.3 percent increase between 2015 and 2017, with TE improvements the foundation of productivity growth. However, in about 60 percent of micro farms productivity growth is negative, although it is positive in 75 percent of large farms. Similarly, improvements in farm management and productivity have been recorded in recent years but not all Kosovo farms exploit new technologies. Although most of its farms have attained positive technical change, in the 2015–17 period numerous farms could still be characterized by technical regression. There is also room for major improvements in land use. Input-specific efficiency analysis found that most Kosovo farms have been using land inefficiently. Agricultural land exceeds the optimal level of use by 9.1 percent, which thus applies to about 40 percent of farms. Land might thus be considered a quasi-fixed input with low adaptability to market changes or its inefficient use might imply lack of a dynamic land market. 1 Agriculture is one of the most underserved sectors in terms of credit.  7 Public spending on agriculture is high compared to other countries with similar income per capita but is lower than in EU member countries. The current types of farm support in Kosovo seem to be an additional constraint on TE and hence agricultural growth. TE is higher in farms that do not receive subsidies than in those that do, and this seems to hold for all types of subsidies except for those related to livestock. Direct payments in particular seem to have little capacity to improve farm economic performance. Current types of subsidies improve productivity but not managerial competence. Productivity growth is higher for farms that receive subsidies, but non-subsidized farms achieve much higher TE overtime. Subsidies seem to induce productivity growth especially in larger farms. It seems that the resources, competences, and capabilities of larger farms allow them to seek and obtain subsidies without the coordination and transaction costs becoming a major burden on their businesses. For micro and small farms, the current design of farm support does not facilitate income smoothing. The distribution of farm subsidies in Kosovo has been increasingly associated with farm size; larger farms seem to be increasing their share of support. Further, the shares of smaller farms in total support not only seem to be lower than those of farms of equivalent size in other Western Balkan countries but are also in decline. This trend raises concerns about the capacity of farm support to facilitate income smoothing for those most in need. On the other hand, agribusiness, mainly food processing, has been growing steadily in terms of number of firms, annual turnover, and employment. Interviews suggest that companies that were able to vertically integrate were able to grow and generate jobs, which also supported primary farms. Despite considerable investment during the last decade, challenges constrain the export potential of food processing, among them limited capacity to deliver products that meet international quality and safety standards, unfair competition, and minimal farmer aggregation and cooperation, which results in high costs and inconsistent quality. Processed foods continue to account for most of the agri-food trade deficit, indicating the importance of promoting investment in agribusiness. Finally, the impacts of COVID-19 on Kosovo agriculture have been multiple and so have been policy responses. The retail, hotels and restaurants sector reduced their demand quite substantially and as result, many farmers had to offload their produce for very low prices. Agricultural production did not decline substantially. Input prices declined. Policy response to the COVID-19 crisis in the country included emergency support packages targeting individuals, firms, and municipalities, while also the Economic Recovery Programme, is expected to allocate EUR 365 million in funds to support businesses, create jobs and stimulate aggregate demand. In the case of sectoral support, the MAFRD has allocated 31 million euros for grants and subsidies to increase agricultural production and rural employment. The analysis provides the basis for the following policy recommendations: 1. Provide incentives to encourage aggregation of farmers and other food chain actors. Farms in Kosovo become competitive mostly by enlarging their size. Incentives for aggregation could promote capital investments that lead to improvements in productivity and more efficient use of inputs. Producer associations and cooperatives could link smallholder farmers to finance and input and output markets. 2. Facilitate farm competitiveness by modifying current types of farm support. Current farm subsidies in Kosovo tend to promote less productive and technically inefficient farms. The main types of farm support negatively affect farm efficiency in Kosovo, and even the positive effects of subsidies on technical change seem to be dominated by factor misallocation. In view of the country’s EU accession path, a shift to decoupled farm support should be considered. 3. Reallocate public resources to farm activities with higher rates of return. Continued support of low-value crops reduces the potential of agriculture to add value and generate income opportunities in rural areas. Support for investments in high-value crops could considerably improve the sector’s trade deficit and also enhance farm incomes. 4. Facilitate the modernization of smaller farms through better targeted support. Lower eligibility thresholds for direct support and simpler rural development measures could benefit smaller farms and encourage their uptake of innovative and more efficient practices. 8  5. Promote an enabling environment for small and medium-sized farms. These farms considerably trail large ones in efficiency. This points to underlying structural difficulties, such as access to finance, technology, and markets, that make it difficult for smaller producers to transform into more efficient larger units. Current support policies do not seem capable of accelerating that transformation. The best way forward would seem to be public investment in the provision of public goods (e.g., advisory, training, technical, and information support; agricultural R&D; infrastructure; and storage capacities). The Kosovo Credit Guarantee Fund could facilitate financial deepening in the agriculture sector. 6. Use rural development measures to respond to farm needs. Such measures should provide incentives for technical change and innovation. They could also differentiate eligibility and selection criteria and support rates to take into account regional disparities and to encourage younger farmers with entrepreneurial potential. They could provide special incentives for medium-sized farms to pursue enlargement and technological and managerial modernization. The incentives could be complemented by measures to improve access to credit and enrich managerial skills. 7. Target special measures to youth employment. These measures could both facilitate knowledge and innovation and provide special incentives for investments both on- and off-farm. 8. Expedite cadastral reconstruction to cover the entire territory of Kosovo by prioritizing the more economically active agricultural land and cadastral zones. This would not only promote access to finance but would also facilitate protection of fertile agricultural land from illegal construction. Enforcement of the unused agricultural land tax and introduction of market- based valuation of properties could facilitate use of agricultural land for productive purposes by creating a more dynamic land market. 9. Finally, direct development programs to rural economic diversification and sustainable management of natural resources. Kosovo, identified as a “water-stressed” nation, is one of the countries with the least development and storage of water resources, which heightens the vulnerability of its agriculture to climate change. Putting into place an irrigation master plan based on irrigation infrastructure would make agriculture more productive and better use scarce water resources.  9 01 Context Agriculture has an important but shrinking role in the economy. The performance of the sector post-independence has been weaker than in other sectors in the national economy and in peer countries. In 2008–19, agricultural value-added in Kosovo performed worse than in any comparable economy, even though in this period, average growth of the country’s GDP was one of the highest among comparators (Figure 1). In the 2009–19 period, agricultural value-added fell by a cumulative 6.8 percent in real terms (Figure 2). Furthermore, in 2019 the ratios of agriculture to GDP (6.9 percent) and to employment (5.2 percent) were closer to aspirational peers (Figure 3).2 Though still important in Kosovo, agriculture has been shrinking in terms of its contribution to both growth and job creation. In 2009–19, the share of agriculture in GDP nearly halved. Figure 1. Performance of Agriculture in Kosovo and Peers, 2008–19 8.0 7.0 . 6.0 5.0 4.0 . . . . . . 3.0 . . . . . 2.0 . . . . . . . 1.0 . . . 0.0 KOS - . MKD ALB KGZ ARM MDA LVA URY EST SVN LTU CZE - . -1.0 -2.0 Agriculture, forestry, and fishing, value added (annual % growth) GDP growth (annual %) Source: World Bank staff calculations based on data from World Development Indicators (WDI). Note: ALB: Albania; ARM: Armenia; CZE: Czech Republic; EST: Estonia; KOS: Kosovo; KGZ: Kyrgyz Republic; LVA: Latvia; LTU: Lithuania; MDA: Moldova; MKD: North Macedonia; SVN: Slovenia; URY: Uruguay. 2 According to a Labor Force Survey conducted by the Millennium Challenge Corporation in 2017, agriculture accounted for 21.7 percent of employment, but the methodology used varied considerably. 12 Context Figure 2. Growth in Labor Productivity, 2009–2019 . % % rGDP growth rVA growth VA per worker % growth (RHS) % % % . % . % % -% -. % - . % - . % - % -% - . % - % Source: KAS data; World Bank staff calculations. Figure 3. Agricultural Value-Added and Employment Compared to structural and aspirational peers3 40.00 ALB 35.00 30.00 ARM Agriculture Employment share (%,2019) 25.00 MDA KGZ 20.00 MKD 15.00 URY 10.00 LTU KOSOVO SVN LVA 5.00 CZE EST 0 0 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 20.00 Agricultural Value Added (% of GDP) (2019) Source: WDI. 3 The countries considered as aspirational peers for benchmarking Kosovo (XKX) performance in this analysis are Albania (ALB); Armenia (ARM); CZE: Czech Republic (CZE); Estonia (EST); KGZ: Kyrgyz Republic (KGZ); Latvia (LVA); Lithuania (LTU); MDA: Moldova (MDA); MKD: North Macedonia (MKD); Slovenia (SVN); Uruguay (URY). Context 13 Agriculture’s contribution to growth has been declining due not only to structural transformation but also to low productivity. In recent years decreasing labor productivity in general, limited change in agricultural employment, and the lower contribution of agriculture to the economy compared to peers suggest limited productivity growth (Figure 1). To identify what drives and limits agricultural growth in Kosovo, this note analyses dispersion of scale and technical efficiency across farms. Furthermore, the note also decomposes change in outputs into change in inputs and growth in total factor productivity (TFP). The TFP changes are further deconstructed into technical and scale efficiency changes and technical change over time. The note also assesses the role of state support in improving agricultural productivity.4 The micro-level analysis is complemented by an analysis of structural data, to assess the role of agriculture in Kosovo’s economy in terms of its contribution to growth, jobs, and external trade balance. Box 1. Methodological Approach The study, which analyses productivity dynamics and drivers in Kosovo, is based on the 2015, 2016, and 20175 Farm Accountancy Data Network (FADN) individual farm database provided by the Ministry of Agriculture, Forestry and Rural Development (MAFRD). The dataset contains farm level information about size, location, type of farm, farm manager’s age, value of crops, number of hours worked, capital, cost of intermediate inputs, and other costs. Efficiency of the production processes are assessed by estimating technical efficiency (TE) and scale efficiency (SE). TE and SE show the loss in terms of production due to failure to adopt the best production techniques and the optimal scale of production. Technical change (TC) captures the ability of the farms to introduce new technologies— innovation as it becomes available—and push the frontier outward, i.e., the ability to produce more output given the same amount of inputs. Conceptually, TE captures managerial competence, effectiveness of organizational routines, and adjustment to business environment and regulation. Product, organizational, and marketing innovation all matter. SE reflects the influence of technology compatibility and indivisibility, market size, scale decisions, and irreversibility of investments (for more details see Annex A). TE analysis was carried out using data envelopment analysis (DEA), which is based on estimation of a production frontier defined by the most technically efficient farms in the 2017 FADN sample. Scores for different types of farm are derived based on the distance of each farm from this “optimal” frontier; the performance of farms associated with specific types of public support is emphasized. Multivariate analysis is also used to identify drivers of TE and SE. The FADN dataset for 2015–17 was used in analyzing drivers of total factor productivity (TFP) in agriculture over time. Constraints on agricultural growth in Kosovo are assessed by estimating TE, SE changes, and TC, which in turn translate into changes in TFP, overtime. TFP was estimated using the Malmquist TFP index (MPI), and changes in TFP. The time evolution of productive efficiency is captured by the MPI, which is defined for time t as: MPIt=(ΔΤΕ)t×(ΔSΕ)t×(ΤC)t Finally, input-specific efficiency for agriculture is assessed by decomposing agricultural output into inputs. Annex A presents the theories and methods used in this study. Equity in the distribution of agriculture support is assessed from individual farm data. FADN data on agriculture spending in 2015 and 2017 is used to assess the distribution of different types of farm subsidies by economic size. Average subsidy rates per farm of different farm economic sizes facilitate assessment of the pro-poor (small farms) or pro-rich (larger farms) distribution of subsidies. Among data sources were the MAFRD, the Agriculture Development Agency (ADA), the Ministry of Finance (MoF), Kosovo Agency of Statistics (KAS) and the World Bank World Development Indicators (WDI). Annex B presents details on the data used and Annex C details the results of both efficiency and productivity analysis. 4 Here it is worth noting that agricultural policy is not the only means to any problems that agriculture faces; and that the role of the market as well as other policies (e.g. education) is often very critical. 5 Despite the fact that the 2018 FADN data base was available, it was decided to carry out the productivity analysis for 2015-2017. This was because in 2018, nearly 50 percent of FADN farms were recorded for the first time. Hence, if 2018 was used, the size of the 2015-2018 data panel would have been very small. 14 Context The productivity of Kosovo’s farm labor and land is low compared to peers.6 Small farm size and low utilization of farm capital have led to low labor productivity (Figure 4) and underemployment in agriculture. Kosovo ranks considerably behind similar economies in terms of labor productivity and cereal yields (Figure 5). Low labor productivity in Kosovo is evident despite the high share of younger farmers (Figure 6), whose operations—unlike in neighboring countries—are not more efficient than those of older farmers. Structural constraints, such as minimal education and training and high levels of informality, deter these younger farmers and the sector itself from reaching their potential. Among similar economies, Kosovo outperforms only Armenia and the Kyrgyz Republic. Figure 4. Labor Productivity, Kosovo and Regional and Aspirational Peers, Agricultural Value Added per Worker, Constant 2010 US$, 2018 30,000 24,987 25,296 25,000 22,598 20,000 18,628 17,008 15,000 12,819 10,000 6,631 5,827 3,798 5,000 , 1,813 0 KOS MKD ALB KGZ MDA LVA URY EST SVN LTU CZE Source: KAS and WDI data; World Bank staff calculations. Note 1: ALB: Albania; CZE: Czech Republic; EST: Estonia; KOS: Kosovo; KGZ: Kyrgyz Republic; LVA: Latvia; LTU: Lithuania; MDA: Moldova; MKD: North Macedonia; SVN: Slovenia; URY: Uruguay. Note 2: Comparative data was not available for a latter year. 6 KAS data shows no move of labor from agriculture; in contrast, between 2012 and 2019, there was been a slight increase in primary employment in agriculture. Context 15 Figure 5. Cereal Yields Compared, Regional and Aspirational Peers, kg per ha, average 2017-2018 7,000 5,796 6,000 5,349 4,827 4,813 5,000 4,000 , 3,668 3,638 3,706 3,128 3,261 3,296 3,000 2,299 2,000 1,000 0 KOS KGZ ARM ALB MKD MDA LVA LTU URY EST SVN CZE Source: KAS and WDI data; World Bank staff calculations. Note 1: ALB: Albania; ARM: Armenia; CZE: Czech Republic; EST: Estonia; KOS: Kosovo; KGZ: Kyrgyz Republic; LVA: Latvia; LTU: Lithuania; MDA: Moldova; MKD: North Macedonia; SVN: Slovenia; URY: Uruguay. Note 2: Comparative data was not available for a latter year; data for Kosovo is for 2018-2019. Figure 6. Farm Households by Age, Regional and Aspirational Peers, 2016, Percent Up to 25 years 25-34 35-44 45 - 54 55 - 64 65 years and over KOS 2.4 8.0 20.8 28.3 22.1 18.3 MKD 11.2 13.5 16.1 20.6 20.3 18.4 CZE 0.4 4.1 14.7 22.1 32.0 26.8 EST 1.0 7.7 17.5 24.0 22.3 27.5 LVA 0.6 4.2 12.7 25.6 26.7 30.2 LTU 1.0 6.3 12.5 25.6 23.8 30.8 SVN 0.5 4.1 12.2 25.9 28.8 28.5 Source: KAS and Eurostat data, World Bank staff calculations. Note 1: CZE: Czech Republic; EST: Estonia; KOS: Kosovo (2014); LVA: Latvia; LTU: Lithuania; MKD: North Macedonia (2013); SVN: Slovenia. Note 2: Comparative data was not available for a latter year. 16 Context Although the number of farms has increased marginally, there has been an alarming decrease in farm size. Agricultural production is carried out by farm households applying mixed and extensive low-productivity methods (especially in cereals), with low cash flows, and weak integration with markets. Very small, fragmented holdings dominate agriculture in Kosovo, with an average utilized agricultural area per holding of 3.2 ha, fragmented into seven plots (Miftari et al. 2016). About 70 per cent of farms operate on less than 2 ha, and 93 per cent on less than 5 ha (Table 1); only 1.6 percent of farms are 10 ha or larger. Shares of very small farms are higher than in all comparable economies except Albania. Since 2017, there has been a 24 percent decrease in the number of farms over 30 ha, while the total number of farms has decreased by 2.5 percent. Structurally, this development is very alarming, and there is no clear evidence of the reasons for it. However, uncontrolled expansion of construction, legal and illegal, has likely contributed to the phenomenon. Also, Kosovo’s agricultural land is threatened by industrial contamination of soil, water, and air and by plants and vegetation that grow spontaneously and serve as pasture. Also very close to agricultural land, partially controlled or uncontrolled landfills present a permanent source of pollution. 7 Table 1.  Farms in Kosovo and Comparators by Size, Percent Share < 2 ha 2–5 ha 5–10 ha 10–20 ha > 20 ha Albania (2012) 86.0 Czech Republic (2013) 10.4 7.2 19.0 17.8 45.5 Estonia (2013) 9.4 22.1 21.2 17.8 29.5 Latvia (2013) 21.8 20.0 19.9 19.6 18.7 Lithuania (2013) 14.1 39.1 22.4 11.7 12.7 Slovenia (2013) 25.4 34.3 23.9 11.3 5.1 < 1 ha 1–2 ha 2–5 ha 5–10 ha > 10 ha Kosovo (2019) 47.8 21.9 23.0 5.7 1.6 < 1 ha 1–3 ha 3–5 ha 5–8 ha > 8 ha North Macedonia (2013) 58.2 29.4 7.3 2.9 2.2 Source: Agriculture Census data from KAS and Eurostat, World Bank staff calculations. Note 1: Comparative data was not available for a latter year. Low investment in irrigation and climate change adaptation, a dysfunctional land market, and lack of investment in technology affect agricultural productivity. Kosovo, which is identified as a water-stressed nation, is among countries with the least developed water resources and storage. Irrigation accounts for 41 percent of water use in Kosovo and the lack of modern irrigation infrastructure limits agricultural productivity growth and also heightens water stress.8 The lengthy drought in 2019 and the flood in 2020 in Gjilan municipality, which otherwise has real agricultural potential, caused major losses in agricultural output. The new irrigation master plan provides a roadmap to improve the infrastructure and estimates costs at 590M euros in the medium term, equivalent to 8.5 percent of GDP.9 An underdeveloped Agricultural Knowledge and Innovation System (AKIS) is also a deterrent to competitiveness (World Bank 2018). Among other impediments to productivity are minimal use of modern technology, low financial liquidity, shortage of capital for investment (especially for smallholders), outdated production management practices, lack of market opportunities and of aggregators of products and storage facilities, and limited value chains.10 Most farms are operating at a subsistence or semi-subsistence level; commercially oriented farmers are confronted by expansion obstacles. Between 2007 and 2019, agriculture in Kosovo accounted for only 1 percent of total FDI inflows.  Kosovo’s agricultural sector offers an opportunity for international investors since 53 percent of Kosovo’s land is considered to be arable. There is also potential for high-value crops due to a beneficial climate and soil quality, low production costs, and open access to the CEFTA and EU markets. However, as yet agriculture sector has attracted very little FDI due to issues related to land tenure and the fragmentation of plots.11  7 Kosovo Irrigation Master Plan, September 2020. 8 Kosovo Water Policy Note, February 2020. 9 As a share of projected GDP in 2021. 10 Miftari et al. 2016; IFC 2018; Bicoku et al. 2018. 11 Foreign direct investment note, Kosovo country economic memorandum. Context 17 The land market in Kosovo is dysfunctional, with issues related to property rights and land access. Land plots are often held without clear title or registration and many landowners are absent. Consolidation of physically separate land parcels has been slow. Additionally, “private investor interest in Socially Owned Land could assist in stimulating the land market, but is limited by several factors, including a lack of clarity over possession and/or ownership; the confused regulatory environment; continued debate about restitution; and ambiguity about how to treat land that has a public interest”.12 “Socially owned arable agricultural land possessed by socially owned enterprises under the former regime were transformed through a 99-year lease rather than as a right of ownership. Such leases are not commonly perceived as providing security of tenure, reducing investment to increase agricultural productivity (USAID and Republic of Kosovo 2016).” Further, the inadequacy of Kosovo’s cadaster and concepts of socially-owned property from the past that are still present in the property rights laws limit access to land for investors (USAID and Republic of Kosovo 2016). Further, foreign citizens and legal persons have been finding it difficult to register property rights: cadastral legislation has not defined who can register these rights, and what is required for registration. Not having clear instructions, municipal cadastral offices interpret the laws inconsistently. In recent years, the relative importance of EU accession ambitions to the Kosovo economy has led to active public intervention. Public support for agriculture in Kosovo jumped from about ¤22 million in 2013 to more than ¤55 million in 2019.13 Despite progress on aligning Kosovo’s agricultural support system with EU Common Agricultural Policy (CAP) principles, there are doubts about whether current types of support can induce optimal usage of production resources in agriculture and thus address the competitiveness14 problems of the sector and respond to sector reform needs. The impact of the current Kosovan interventions is not clear, and empirical analysis of how spending is linked to specific results is very scarce. The impacts of COVID-19 on Kosovo agriculture have been multiple and so have been policy responses. The retail sector reduced its demand quite substantially and a significant challenge which small farmers faced was in accessing markets and selling their goods. Open-air markets were closed for many weeks and only large supermarkets were able to operate. However, chain stores don’t sell food that comes from small-scale farmers. As a result, many farmers had to offload their produce for very low prices (GAFSP, 2020). Impacts also included a decline in demand for agricultural and food products by hotels and restaurants and also by the diaspora who was not able to visit the country (Hyseni, 2020). Agricultural production did not decline as at the time of the closure of free movement, main operations to crops had already been performed; in fact, favorable weather conditions led to a slight increase in output and yields (Hyseni, 2020). In terms of inputs, prices of seeds and planting material increased in March 2020, but declined thereafter; fertilizer and plant protection prices declined, as well as the price of energy. Despite the problems with inputs supply and restrictions on movement, larger areas were sawn than 2018 and 2019. Agricultural products prices in 2020 declined by an average of 0.3 percent per month, compared to 2019; however, the food consumer price index increased slightly, by 1 percent. Policy response to the COVID-19 crisis in the country included fiscal policy measures, including an emergency support package of about EUR 180 million (3% of GDP) to support individuals, firms, and municipalities affected by the COVID-19 crisis, and EUR 10 million to fund efforts to contain the spread of the virus and reinforce the healthcare system. In August 2020, the government approved the Plan for the Implementation of the Economic Recovery Programme, which will allocate EUR 365 million in funds to support businesses, create jobs and stimulate aggregate demand. This plan includes measures facilitating loan access for businesses and farms, provides targeted tax relief and rental subsidies for firms, stimulates employment by subsidising worker salaries, and incentivises capital investments (OECD, 2020). Further, the government allocated EUR 67 million to action promoting employment, especially for groups of workers with low probability of finding a job during the crisis. Support to the population was granted in the form of additional wages, higher social benefits, suspension of loan repayments and public utility payments, extension of tax and pension liabilities, suspension of interest on unpaid property taxes. Support to firms included salary subsidies, credit guarantees, a subsidization of firm rental expenses and a postponement of firm’s tax obligations. Finally, in the case of sectoral support, the MAFRD has allocated EUR 5 million for grants and subsidies to increase agricultural production during the crisis. In August 2020, the government allocated an additional EUR 26 million to increase domestic agricultural production and rural employment. Also, EUR 46 million were allocated to subsidize wages for new employees hired in specific sectors, enable manufacturing and service firms to access modern equipment and machinery and to the support of publicly owned enterprises business operations and capital investments (OECD, 2020). 12 USAID and Republic of Kosovo 2016. 13 Including all types of support: direct payments, rural development, and general services. 14 Alishani 2019; World Bank 2018; IFC 2018 18 Context Context 19 02 Productivity Analysis 2.1 Static Productivity Analysis There is considerable factor misallocation in agriculture. Findings based on the farm level analysis are aligned with low labor productivity measured using macro data, which suggests that farms have little capacity to efficiently transform inputs into output. Farms in Kosovo are characterized by technical inefficiency but operate more satisfactorily with respect to the scale side of technology. The mean efficiency score is 0.272 (Figure 7), which shows that the farms examined could on average produce the same output using 72.8 percent less inputs. The minimum value of TE is 0.027 and the maximum 0.780. The mean SE is higher, equal to 0.721, which means that the farms examined could produce the same output using 27.9 percent less in inputs if they could move toward the most-efficient scale size (though this result is driven by highly productive large farms).15 Figure 7. TE and SE in Agriculture, Kosovo, 2017 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 TE 0.272 SE 0.721 Source: FADN data, World Bank staff calculations. Inefficient production limits the growth of most farms in Kosovo. Efficiency analysis indicates a degree of polarization in the productive performance of agriculture. Most farms in Kosovo have TE scores between 0.25 and 0.30, which shows that they lose about 70 percent of their potential output. However, a considerable number of farms do have TE of about 0.65. Farms with different efficiency performance have different economic and technological characteristics, which relate to their size, technological competence, the degree to which they have adopted more effective business practices, and the mix of outputs produced. These factors are investigated below, where the drivers of efficiency scores are explored. Smaller farms, which comprise nearly 75 percent of holdings in the country, seem to be in a rather grave situation. They are under-utilized and have very low SE scores. In other words, due to credit and other constraints, small farms do not seem able to become scale-efficient and benefit from returns to scale. However, unlike in other countries in the region (World Bank 2019b, 2019c, 2019d), because their TE scores are lower, they also seem unable to exploit their managerial competences and capabilities in order to survive (Figure 8). Thus, they are trapped in poverty, and the non-farm economy does not have the capacity to absorb surplus labor from agriculture. Even medium-sized farms have inefficiencies that inhibit their potential to grow. Both small and medium farms have experienced losses in both TE and SE. It appears that constraints associated with management practices and structural difficulties (e.g., limited access to technology, markets, etc.) deter transformation to more efficient units. However, farms that are comparatively large perform quite well in both types of efficiency. Hence, it seems that in Kosovo, the only way for farms to become more efficient is to enlarge their size, possibly through common investments. 15 The most efficient scale size coincides with the scale of the minimum average costs. The minimum and maximum values of SE are 0.0028 and 1. 22 Productivity Analysis Figure 8. Mean TE and SE Scores by Farm Size, Kosovo, 2017 0.327 Large 0.889 0.264 Medium 0.843 0.239 Small 0.726 0.27 TE Micro 0.503 SE 0 0.2 0.4 0.6 0.8 1 Source: FADN data; World Bank staff calculations. Farms are not able to achieve both TE and SE at the same time, probably because of liquidity constraints due to limited financial deepening and lack of a functional land market. For types of farms where higher TE has been recorded, low SE limits economic performance. With respect to their production specialization, most farm types have low scores for either TE or SE. Farms specializing in high-value crops (horticulture, wine inputs, other permanent crops) seem to have high managerial competence but are unable to explore scale economies, possibly due to cash-flow and credit constraints.16 Because agriculture is one of the sectors least represented in commercial credit in Kosovo (Figure 9), it is likely that credit constraints impede growth of higher- value products. Other farm types (field crops, milk, other grazing livestock) seem to be exploring economies of scale relatively well (perhaps due to their comparatively high support, which could also be used for investments) but suffer from technical inefficiencies (Figure 10). Granivore and mixed farms, which are very important in Kosovo, are the exception, the former recording high scores and the latter low ones in both TE and SE. Figure 9. GDP and Commercial Loan Shares by Sector, 2018 20 Sector GDP Share (%) 15 Manufacturing Trade 10 Construction RE Ag 5 Trans Min ICT 0 0 10 20 30 40 50 60 Sector Commercial Loans Share (%) Source: KAS and CBK data, World Bank staff calculations. Note 1: Top eight sectors selected represented 77 percent of total commercial loans issued by value and 70 percent of production GDP in 2018. Size of circles is correlated with number of sector commercial loans. Note 2: Comparative data was not available for a latter year. 16 Though this has to be further explored. Productivity Analysis 23 Figure 10. Mean TE and SE Scores by Farm Specialization, Kosovo, 2017 0.224 TE Mixed farms SE 0.666 0.440 Granivores 0.861 0.253 Other grazing livestock 0.835 0.246 Milk 0.750 0.350 Other permanent crops 0.472 0.370 Horticulture and wine 0.647 0.260 Fieldcrops 0.750 0 0.2 0.4 0.6 0.8 1 Source: FADN data, World Bank staff calculations. On a more positive note, agricultural growth does not seem to be limited in all districts. There are significant differences between districts with regard to TE but not SE. Farms in the Peja (TE: 0.316), Ferizaj (TE: 0.314), and Gjilan (TE: 0.308) districts seem to have a comparative advantage in terms of TE (Figure 11); but the opposite holds true for farms in Prizren (TE: 0.219) and Gjakova (TE: 0.220). The differences are not due to differences in farm size (see Annex B – Table B6). Figure 11. Mean TE and SE scores according to farm location, Kosovo, 2017 TE Mitrovica 0.288 0.736 SE Prizren 0.219 0.728 Peja 0.316 0.704 Gjilan 0.308 0.722 Gjakova 0.22 0.706 Ferizaj 0.314 0.749 Pristina 0.262 0.721 0 0.2 0.4 0.6 0.8 1 1.2 Source: FADN data, World Bank staff calculations. Unlike in the rest of the region, the presence of younger farmers does not seem to help dismantle constraints to agriculture growth. Again, this finding is in contrast with similar analysis in the region as a whole (World Bank 2019b. 2019c. 2019d) where farms managed by young farmers were found to be more efficient. Taking into account the high share of younger farmers in Kosovo, this finding suggests that structural constraints faced by agriculture and by farmers themselves, such as those linked to education and training, prevent younger farmers from realizing their potential. 24 Productivity Analysis The current types of farm support in Kosovo also seem to diminish TE. Subsidized farms are associated with lower TE but higher SE than non-subsidized ones. Farms not receiving subsidies (TE: 0.284) seem to outperform those that do (TE: 0.264), and this seems to hold for most types of subsidy except those that are livestock-related. The pattern is much different for SE estimates, in which all types of subsidy, except crops-only and other livestock subsidies, are associated with farms that have higher SE scores. The capacity of direct payments to improve farm economic performance in Kosovo has been limited. Farms receiving area payments do not seem to be more efficient. This is consistent with findings of several studies (Latruffe et al. 2011; Rizov et al. 2013; Latruffe and Desjeux 2016; Bokusheva and Cechura 2017; World Bank 2018, 2019d). Rizov et al. (2013) argue that the negative impacts of farm support could be attributed to market imperfections in agriculture (e.g., credit problems) or partial decoupling. The limited capacity of area/headage payments to improve TE is further documented by the performance of farms not receiving any coupled subsidies, which have a higher TE. The exception is farms that receive only livestock subsidies and those receiving any type of livestock subsidy except those for milk and ruminants. However, subsidies do seem to help improve SE. Farm support seems to be relaxing liquidity constraints; farms that receive direct payments exploit scale economies better than those that do not. As noted, perhaps some of these farms have been using direct support for investment purposes. Medium and large farms are considerably more technically efficient if not subsidized, but the TE of micro and small farms is not affected by subsidies. However, micro and small farms (Figure 12) have higher SE if they receive subsidies, suggesting a relaxation of credit and liquidity constraints. In terms of output specialization (Figure 13), it is again confirmed that more highly supported farm types perform better if subsidized, while for farms offering high value-added products, TE improves if the farm is not subsidized. Higher TE seems to be associated with non-subsidized farms managed by more experienced farmers; SE is higher if farms are subsidized, regardless of the manager’s age. Finally, in terms of location, subsidized farms outscore non-subsidized ones in SE, but subsidization does not affect TE scores, except in Pristina and Mitrovica, where TE is higher if farms are not subsidized (Figure 14). Multivariate analysis shows that subsidies have a very small negative influence on TE and a similarly small positive influence on SE. Up to a certain point farm size increases have a negative influence on both TE and SE, but when farm size exceeds a certain limit (that of medium farms), it benefits both. Figure 12. Mean TE Scores by Farm Size and Subsidization, Kosovo, 2017 0.50 0.43 0.40 0.34 0.31 0.30 0.26 0.27 0.24 0.25 0.23 0.20 0.10 0 Micro - Micro - Small - Small - Medium - Medium - Large - Large - S NS S NS S NS S NS Source: FADN data, World Bank staff calculations. Note: S: subsidized; NS: non-subsidized. Productivity Analysis 25 Figure 13. Mean TE Scores by Farm Products and Subsidization, Kosovo, 2017 0.50 0.44 0.40 0.35 0.30 0.27 0.28 0.28 0.26 0.26 0.23 0.20 0.10 0 Fieldcrops - Fieldcrops - Hortic/Wine - Hortic/Wine - Dairying - Dairying - Exper. Farmers - Exper. Farmers - S NS S NS S NS S NS Source: FADN data, World Bank staff calculations. Note: S: subsidized; NS: non-subsidized. Figure 14. Mean TE Scores by Farm Location and Subsidization Kosovo, 2017 0.40 0.36 0.30 0.30 0.26 0.24 0.20 0.10 0 Pristina - S Pristina - NS Mitrovice - S Mitrovice - NS Source: FADN data, World Bank staff calculations. Note: S: subsidized; NS: non-subsidized. 26 Productivity Analysis 2.2 Dynamic Productivity Analysis On a more positive note, total factor productivity (TFP)—currently the most important barrier to agricultural growth in Kosovo—seems to be improving. Between 2015 and 2017, TFP in Kosovo went up by 9.3 per cent. TFP growth was quite high in 2015–16 and 2016–17. In general, higher TE has been the foundation of TFP growth in Kosovo (Figures 15 and 16). Farm managerial competence and productivity have improved in recent years but Kosovo farms do not seem to be exploiting new technologies. TFP growth has been positive in 60 percent of Kosovo farms. The majority of farms attain positive TEC and SEC, which implies better management, but the pattern for TC17 is very different. In 2015–17, technical regression characterized the majority of Kosovo farms. It may be that Kosovo farms are not innovative enough to exploit technological progress. Hence, they seem to be improving TE and SE, which mainly reflects managerial competence, but are not capable of exploiting the potential of new technologies. In other words, the TC component offsets the positive impact of ΔTE and ΔSE components on TFP growth. Figure 15. Distribution of TFP and its Components, Kosovo, 2015–17 <0.6 Number of Farms 500 [0.6-0.8] (0.8-1.0] 440 (1.0-1.2] (1.2-1.4] (1.4-1.6] 379 400 >1.6 358 326 304 279 300 230 219 208 174 200 168 158 128 124 115 113 111 100 91 100 74 68 32 25 13 11 0 0 0 0 TFP and components TFP TC TEC SEC Source: FADN data, World Bank staff calculations. Note: All variables are normalized to 1 at 2015. Values lower than 1 indicate a negative growth rate for the variables, on average between 2015-2017. 17 Technical change captures the ability of farms to introduce new technologies— innovation that becomes available and pushes the frontier “outward.” Productivity Analysis 27 Figure 16. Distribution of TFP Growth by Time Period, Kosovo, 2015–17 1.107 TFP 1.075 1.093 1.057 SE change 0.969 1.013 1.071 TE change 1.023 1.047 2015-16 0.978 2016-17 Technical Change 1.084 Annual average 1.031 0.900 0.950 1.000 1.050 1.100 1.150 Source: FADN data, World Bank staff calculations. Productivity has not improved for many Kosovo farms. About 60 percent of micro farms had negative TFP growth for the whole period examined, but large farms are the “champions,” with 74.4 percent achieving positive TFP growth (Figure 17). Only about 40 percent of small and medium- sized farms saw TFP grow. Figure 17. Distribution of TFP Growth by Farm Size Kosovo, 2015–17, Percent 10 % of Farms Micro Small 7.8% Medium 8 Large 7.1% 6.0% 5.9% 5.7% 5.3% 6 5.2% 5.0% 5.0% 5.0% 4.8% 3.9% 3.9% 3.2% 3.2% 3.2% 4 3.0% 3.0% 2.7% 2.3% 1.8% 1.8% 1.8% 1.2% 2 1.1% 0.7% 0.4% 0.2% 0 >1.6 (1.4-1.6] (1.2-1.4] (1.0-1.2] (0.8-1.0] [0.6-0.8] <0.6 TFP growth Source: FADN data; World Bank staff calculations. Note: All variables are normalized to 1 at 2015. Values greater lower than 1 indicate a negative growth rate for the variables, on average between 2015-2017. 28 Productivity Analysis Current types of subsidies lead to productivity improvements but do not improve management. TFP growth is higher for farms that receive subsidies, but non-subsidized farms perform better in terms of improving their TE (see Annex C). Perhaps this could be attributed to efforts of farmers to access subsidy grants, which could generate internal costs (search, coordination, transaction) and deplete resources and ability to manage farms. Subsidies seem to be inducing productivity growth for larger farms. Smaller farms suffer productivity losses (Annex C) because they are below the minimum size for efficiency. The effect of subsidies on TFP growth are positive only for larger farms, showing that subsidies favor larger farms. The resources, competence, and capabilities of larger farms allow them to obtain subsidies without the search, coordination, and transaction costs becoming a major burden on their business operation. Larger farms have become more efficient, but over time smaller ones seem to be making more use of technology. Regression analysis results (Annex C) show that micro and small farms suffer TE losses, but technology advances are higher for smaller farms, which indicates a catch-up process. These results should be carefully interpreted; there are differences in the initial knowledge and technology conditions between small and large farms. Farms that are technologically mature are expected to innovate less than those that lag in technological terms. Farm location positively affects TE, SE, and TC in the districts of Gjakove, Gjilan and Prizren. Finally, farm manager age does not have a statistically significant influence on any of the TFP components examined. Productivity Analysis 29 03 Input-Specific Efficiency Slack-based inefficiency18 estimates indicate that most Kosovo farms have been using some inputs in quantities that exceed the optimal level of use. In fact, 828 farms (71.63 percent of the 2017 FADN sample) are inefficient with regard to input use (Table 2). About 40 percent of farms are inefficient in the use of one input and another 27 percent in the use of two (Table 3). In the literature, slacks are closely related to input indivisibilities and technology lumpiness, which may reduce productive performance and contribute to product proliferation. Table 2.  Farms with Input Slacks, Kosovo, 2017 At least one input with slack Number of Farms Percentage (%) No 328 28.37 Yes 828 71.63 Total 1,156 100.00 Source: World Bank staff calculations. Table 3.  Probability of a Farm with Specific Number of Slacks, Kosovo, 2017 Number of Inputs with Slacks Number of Farms Probability (%) 0 328 28.37 1 463 40.05 2 316 27.34 3 47 4.07 4 2 0.17 Source: World Bank staff calculations. Land is the farm input used most inefficiently in Kosovo. Agricultural land in Kosovo exceeds optimal use by 9.1 percent (Table 4) for about 40 percent of sampled farms (Table 5). In many cases, land is considered to be a quasi-fixed input not readily adaptable to market changes. Inefficiency is also comparatively high for overheads, which exceed optimal use by nearly 5 percent for about 30 percent of farms. For all other farm inputs, inefficiency is marginal. Table 4.  Estimated and Actual Input Slack, Kosovo, 2017 Input Mean (std. dev.) Max (Min) Percentage (std dev) Number of farms Labor 65.45 7,716 1.46 n=828 (415.67 (0.00) (0.07) Land 1.36 139 9.08 (6.76) (0.00) (0.14) Capital 39,466 8,459,652 1.95 (363,729) (0.00) (0.08) Intermediates 1,455 236,974 0.94 (11,811) (0.00) (0.23) Overheads 189.73 46,575 4.77 (1,725) (0.00) (0.71) Source: FADN data; World Bank staff calculations. 18 Slack-based inefficiency is defined as the potential for further increases in output, or reduction of input, that could be gained beyond that implied by the radial projection. Slack-based inefficiency results from the piece- wise linear form of the non-parametric frontier in DEA and may be considered an indication for necessary improvements of the input-output mix. which are measured by movements on the frontier. 32 Input-Specific Efficiency Farms with a slack in intermediates have a very low probability of also having a slack on land. However, the probability of a slack pair is higher for the combination of overheads and intermediates (Table 6). When use of intermediates is excessive, the quantity of land not used is relatively small. A farmer who has unused land will produce additional output using the excess intermediates. That is, the higher the probability that a farmer uses excessive quantities of intermediate inputs, the lower the probability the same farmer would also use the land in excess. For example, it is highly possible for a farmer who has some land available to use all available intermediates to produce additional output,  reducing the probability of slack in intermediates. In economic terms, for a farmer with unused land, the marginal cost of intermediates is low enough to approach zero.   Table 5.  Farms with Input-Specific Slacks, Kosovo, 2017 Number of Farms Input Input slack=Yes Input slack=No Labor 70 1,086 (6.06 %) (93.94%) Land 461 695 (39.88%) (60.12%) Capital 102 1,054 (8.82%) (91.18%) Intermediates 269 887 (23.27) (76.73 Overheads 342 814 (29.58) (70.42) Source: FADN data; World Bank staff calculations. Table 6.  Probability of Input-specific Farm Inefficiency, Kosovo, 2017 Variables Labor Land Capital Intermediates Over heads Labor 1.000 Land -0.051 1.000 Capital -0.002 0.046 1.000 Intermediates 0.058 -0.315 0.060 1.000 Overheads 0.058 -0.083 -0.015 0.253 1.000 Source: FADN data; World Bank staff calculations. There seems to be room for considerable improvement in farm use of land in Kosovo. Input- specific inefficiency of land use (Table B9; Figure 18) is much higher than general technical inefficiency. Inefficiency is also evident in the use of overheads, but at a much lower level. In contrast, use of labor and intermediates is comparatively efficient. Figure 18. Farm Input-specific Efficiency, Kosovo, 2017 0.35 0.32 Total 0.31 0.27 0.26 Di erential 0.25 0.23 0.15 0.05 0.05 0.04 0 0.00 -0.05 -0.02 -0.04 Labor Land Intermediates Capital Overhead Source: FADN data; World Bank staff calculations. Input-Specific Efficiency 33 04 Public Spending on Agriculture and Rural Development Public spending on agriculture in Kosovo is high compared to similar economies but low compared to the EU average. In 2011–19, it averaged 2.3 percent of total government spending (Figure 19); in real terms its annual average rate of change (22 percent) was much higher than that of total public spending (4.3 percent), driven by a major increase during 2015-18, followed by a decrease in 2019. In 2019, total public support to agriculture reached 8.6 percent of agricultural value added—considerably higher than other Western Balkan countries except North Macedonia. In relation to comparators, this share was among the highest as a share of agricultural value-added (see also Figure 20). Figure 19. Total Public and Agriculture Spending, Millions Euros in 2011 Values, 2011–2019 1,871.3 4.00 1,760.2 1,611.2 3.50 1,546.5 1,453.0 3.56 1,399.2 1,417.5 1,377.7 1,346.0 3.00 3.16 2.84 2.86 2.50 2.62 2.00 2.01 1.50 1.57 1.52 1.00 0.94 0.50 21.9 21.6 27.6 41.3 48.9 46.1 62.7 49.0 12.7 0.00 2011 2012 2013 2014 2015 2016 2017 2018 2019 Total government spending Agriculture spending % share of agriculture spending Source: MoF; APM database; MAFRD. Figure 20. Agricultural Value Added (Percent of GDP) and Spending (Percent of Public Spending) Compared with Peers, 2017 6 LTU MKD 5 (% total public spending, 2017) LVA 4 EST KOSOVO 3 ALB CZE ARM 2 ARD Spending SVN KGZ 1 0 0 2 4 6 8 10 12 14 16 18 20 Agriculture VA (% GDP, 2017) Source: World Development Indicators; FAOSTAT. Note: 2015 for North Macedonia; comparative data was not available for a latter year. 36 Public Spending on Agriculture and Rural Development Farm subsidies dominate agricultural spending in Kosovo; the capital budget is limited. The share of wages and salaries in the agricultural budget (on average, 4.4 percent for 2014–19) is only a fraction of the equivalent in the national budget (31.2 per cent). Also, the average share of capital spending in agriculture (8.5 percent) is much lower than in the national budget (28.7 percent), though investment in the sector is sorely needed. The new irrigation master plan specifies infrastructure improvements and estimates medium-term costs for irrigation infrastructure alone at ¤590 million, equivalent to 8.5 percent of GDP. Agriculture received its highest direct support in Kosovo before the COVID-19 pandemic. Direct payments are the most important category of support (Figure 21), averaging 49.7 percent of total agriculture and rural development funds in 2014–19 (Figure 21), but over the period their share has increased from 51 percent in 2014 to 55 percent in 2019. Direct payments cover all farm subsectors and their range has expanded since 2015 (Ilic et al. 2019). For each farm, minimum area/ headage thresholds are applied. Support rates are lower for cereals (¤150 and ¤100 per ha) than for vegetables (¤300), fruit (¤400) and vineyards (¤1,000). Currently, direct payments for livestock account for around 34 percent of total direct payments, followed by 32 percent for cereals and 25 percent for other crops. Since 2014, direct payments have increased by an annual average of 20 percent, though the increase was much faster between 2014 and 2016 (Figure 22) due to use of new instruments and higher outlays for high-value crops and livestock. Coupled support in the form of a milk quality premium (6.7 percent in 2019), seedling support (0.7 per cent), and wine support (1.3 percent) make up only a small part of direct payments. According to MAFRD and Agriculture Development Agency data, since 2014 direct payment absorption rates have on average been at 108 percent of planned funds, an overspending of budgeted support. Figure 21. Agriculture Spending by Function, Kosovo, Million Euros at Current Prices, 2014–19 80 Total spending 69.42 Direct payments 70 Rural development General services 60 54.84 52.83 50.57 50 44.49 40 29.70 30.63 29.65 30 26.13 27.03 21.44 30.97 20 15.19 22.50 19.55 15.45 16.07 10 11.10 8.09 8.80 8.14 0 3.40 3.50 4.20 2014 2015 2016 2017 2018 2019 Source: MAFRD data, World Bank staff calculations. Public Spending on Agriculture and Rural Development 37 Figure 22. Direct Payments by Type, Kosovo, 2013–2019 12.00 Direct payments for livestock Direct payments for cereals Direct payments for other crops 10.00 Milk quality premium Support for seedlings Wine coupled support 8.00 6.00 4.00 2.00 0.00 2013 2014 2015 2016 2017 2018 2019 Source: MAFRD data, World Bank staff calculations. Rural development spending in Kosovo emphasizes agri-food by supporting investments and introducing food safety standards. On average, rural development accounts for 38 percent of support funds, which in 2014-2018 increased by an annual average rate of 27.5 percent, before decreasing by almost 50 percent in 2019. In fact, among Western Balkan countries, Kosovo seems to have the highest relative rural development support (Ilic et al. 2019). In 2019 farm competitiveness measures accounted for 85 percent of total rural development funds, with 15 percent allocated to diversification, business development, and rural infrastructure—a share that has gone up substantially in the last few years (Figure 23). So far there has been very little spending on environmental measures.19 For 2013–19 farm competitiveness funds were of three types: farm investment plans (58.2 percent of rural development funds), investments in food processing and marketing (21.2 percent) and investment in irrigation (6.1 percent). Diversification funds concentrated on farm diversification and business development (7.2 percent), local development (1 percent) and a special program for less developed areas and rural infrastructure (no funds in 2013–17 and 2019; 30 percent of rural development funds in 2018). The increasing share of diversification measures is attributed to the 2018 Special Program; the shares of other measures in total rural development funds changed very little. In general, this strong sectoral focus seems incompatible with the country’s strategic goals on sustainable resource management and quality of life and poverty reduction in rural areas. Nor does it do much to shed surplus labor from agriculture and facilitate agricultural adjustment. Since 2014, however, the absorption rates of rural development measures seem to have improved. 19 In North Macedonia these account for 7 percent of rural development spending and in Serbia for 5 percent. In 2014–20 the relevant share in the EU for was about 53 percent. 38 Public Spending on Agriculture and Rural Development Figure 23. Types of Rural Development Support, Kosovo, Million Euros at Current Prices, 2013–19 16 Farm investments Investments in food 14 processing & marketing Farm diversification & business development 12 Local Development Strategies (LEADER) 10 Irrigation projects Special Program (Less dev. areas; 8 rural infrastructure) 6 4 2 0 2013 2014 2015 2016 2017 2018 2019 Source: MAFRD data, World Bank staff calculations. Support for general services has doubled since 2017. General services—almost solely food safety and quality—account for 11 per cent of total funds, with a notable drop in their share in 2015 and 2016, followed by an increase thereafter. Only a small part supports advice and extension. Agricultural R&D and education are under-developed (Daci-Zejnullahi 2014). Public Spending on Agriculture and Rural Development 39 05 Distribution of Farm Support20 20 Due to lack of data on the population of farms, the 2015 and 2017 FADN individual farm dataset was used to assess the distribution of direct payments by farm economic size. Farm support in Kosovo does not seem to be distributed equitably. The distribution of farm subsidies is increasingly associated with farm economic size.21 In 2015, large farms accounted for 68.3 percent of total subsidies and in 2017 for 77.3 percent (Figures 24 and 25). As large farms in the sample constitute about 20 percent, this pattern indicates a convergence with the EU, where 20 percent of farms receive 82 percent of subsidies) Between 2015 and 2017 all other farm types experienced losses in their shares of support. The share of micro farms (about 28 percent of the FADN sample) fell from 2.2 percent in 2015 to 1.2 percent in 2017. This is much smaller than support for farms of equivalent size elsewhere in the Western Balkans (World Bank 2019b, 2019c). A noticeable decrease in their total support share is also seen for small farms (30 percent of the sample), which in 2015–17 went from 12 to 7.7 percent. The drop was smaller for medium-sized farms, from 17.7 to 13.8 percent). These developments raise doubts about the capacity of farm support to help smooth income for the smaller farms that are most in need. Figure 24. Direct Payments to Farms by Economic Size, Kosovo, 2015, Percent 80 68.3 70 60 50 40 31.1 27.6 30 20.9 20.4 17.7 20 11.8 10 2.2 0 Micro Small Medium Large Farms (% of total) Subsidies (% of total) Source: FADN data, World Bank staff calculations. 21 Micro farms have an economic size of less than EUR 8,000; small farms lie between EUR 8,000 and 20,000, medium farms between EUR 20,000 and 50,000, and large farms are over EUR 50,000. 42 Distribution of Farm Support Figure 25. Direct Payments to Farms by Economic Size, Kosovo, 2017, Percent 90 77.3 80 70 60 50 40 28.8 29.6 30 20.4 21.2 20 13.8 7.7 10 1.2 0 Micro Small Medium Large Farms (% of total) Subsidies (% of total) Source: FADN data, World Bank staff calculations. Distribution of Farm Support 43 06 Policy Priorities and Institutional Environment Kosovo has made considerable progress in aligning its agricultural policy and administrative infrastructure with EU requirements. Legal, strategic, and programming documents like the Agriculture and Rural Development Program (ARDP) 2014–20 (MAFRD 2013), the Medium-Term Expenditure Framework (MoF, various years) and the Economic Reform Program (ERP; Republic of Kosovo, various years) have made policy more stable and transparent.22 Short-term policies are detailed in the annual ARDP (MAFRD, various years). The main strategy document is the ARDP 2014–20 (MAFRD 2013), which addresses long-term policy goals that are compatible with the EU Common Agricultural Policy (CAP) and guides agriculture and rural development in Kosovo toward modernization and approximation to EU standards.23 Kosovo also adopted a 2018–21 action plan for organic agriculture. Finally, MAFRD has set up institutions (Monitoring Committee, Managing Authority, Payment Agency – ADA) that are responsible for carrying out aspects of the ARDP (Miftari and Hoxhaj 2014). The Ministry has also established a Department for Advisory Services and opened municipal information centers to provide advice and support to farmers (Ilic et al. 2019). There is considerable donor support for agriculture and rural development in Kosovo.24 The EU has been the main donor, through IPA II 2014–20, which has supported Kosovo institutions aligning with the CAP (EC 2018). IPA II also supports the Kosovo Veterinary and Food Agency (KVFA) to bring food safety standards in line with the acquis, and plans to support improvements in data collection and production of statistics. In total, planned IPA II funds for agriculture and rural development in 2014–20 were close to ¤79 million and disbursements from all donors amounted to ¤126.8 million (about 24 percent of primary sector 2019 GVA), of which about ¤112.6 million were grants and ¤14.2 million loans. The EU Office and the EC together contributed ¤61.6 million, 43 percent of total support. The EU has supported Kosovo institutions in aligning with the CAP (EC 2018) through IPARD, and IPA has supported KVFA. The United States Agency for International Development (USAID) has provided ¤38 million (27 percent of total support), which includes support to the small fruits sector and capacity building and infrastructure for KVFA. The World Bank has disbursed ¤13.52 million (10 percent of total support) for knowledge transfer and investments, mainly in livestock and horticulture. The Austrian Development Agency provided ¤5.93 million (4 percent of total support), mainly for Managing Authority capacity building. Finally, various other donors provided ¤22.93 million (16 percent of total support), among them the German government (capacity building, policy development and development assistance), the Swedish government (capacity building), the Swiss Agency for Development and Cooperation (horticulture promotion), Italian Cooperation (technical assistance; education and development strategy), the Food and Agriculture Organization (capacity building, policy development, and development assistance), and the Luxembourg government (remote areas development support). Further steps are required to align agricultural policy with the EU acquis on both agriculture and food safety. Better evaluation and monitoring of support should be emphasized, so that investments in the sector are more efficient (EC 2019). Currently, there is a lack of capacity for evaluation and monitoring and delays in application processing and execution of payments. Kosovo has made progress in establishing an Integrated Administration and Control System, an animal identification system, and the FADN and is currently upgrading its Land Parcel Identification System (LPIS).25 Farm advisory services need to be strengthened and progress is necessary in setting up Common Market Organizations and formulating a quality policy. There has been progress in drafting the laws governing food and feed safety and veterinary policy, and the capacity of KFVA has improved. However, there is still room for upgrading inspection mechanisms and infrastructures, animal identification and registration systems, and the operational capacity of institutions responsible for plant health (EC 2019). Agricultural support policy in Kosovo has been evolving rather slowly. Farm support is mostly in the form of direct payments to producers, based on current agricultural area and heads of livestock, and secondarily on coupled subsidies (milk quality and seedlings). Decoupled payments are not applied, despite the commitment in the ARDP 2014–20.26 Current types of subsidy are hardly compatible with the CAP and are not yet conditional on compliance with environmental standards. Rural development measures correspond to the 2014–20 CAP framework on rural development policy. However, spending is primarily directed to agri-food competitiveness (78.8 percent of planned funds) and promoting economic and social inclusion (16.6 percent); measures on preservation of the environment have not really been activated (1.4 percent) and, together with measures on knowledge 22 Kerolli-Mustafa et al. 2017. 23 Miftari et al. 2016. 24 Aid Management Platform, https://amp-mei.net/portal/. 25 An LPIS is an IT system based on photographs of agricultural parcels used to check direct payments made to farmers through the EU CAP. LPIS determines the eligibility of agricultural land for agricultural policy support. 26 Kerolli-Mustafa et al., 2017 46 Policy Priorities and Institutional Environment transfer and innovation (3.2 percent), are assigned very low shares of the budget. The Instrument for Pre-Accession Assistance for Rural Development (IPARD) II program has been prepared (Ilic et al. 2019) but institutional capacities are not yet up to speed. MAFRD cooperates with USAID and the European Bank for Reconstruction and Development to guarantee farm loans; the Kosovo Credit Guarantee Fund (KCGF) provides loan guarantees to small and medium agribusinesses (MAFRD 2019). Finally, in 2018 general support for food safety and quality control accounted for 13 percent of spending. There are concerns about the effectiveness of current support measures and on their capacity to restructure the sector in view of EU accession. The contribution of direct payments to improvements in farm productivity has been criticized (World Bank 2017a; Kerolli-Mustafa et al., 2017). Area and headage payments reduce the incentive to raise production and do nothing to accelerate structural change (World Bank 2017a; 2019b); and coupled subsidies distort the allocation of productive resources. Considering also that at least until 2013 relatively little funding was dedicated to rural development, including the provision of public goods (agricultural infrastructure, R&D, advisory services, etc.), Kosovo’s agricultural policy has hardly facilitated farm restructuring and modernization. Policy Priorities and Institutional Environment 47 07 The Role of Agribusiness in Kosovo’s Economy The important role of agribusiness in Kosovo’s economy has been growing steadily. According to MAFRD (2019) and IFC (2018), between 2014 and 2018, the number of active agribusiness enterprises went up by 43 percent (Table 7), annual turnover rose from ¤312.2 million to ¤461.6 million, and employment went from 8,044 to 13,156 employees. Table 7.  Agribusiness Enterprises in Kosovo, 2014–18 Turnover (million euros) Number of Employees Number of Active Firms 2014 312.2 8,004 2,055 2015 323.4 8,790 2,130 2016 360.5 10,024 2,314 2017 432.3 10,449 2,398 2018 461.6 13,156 2,942 Source: MAFRD 2019. Food processing currently accounts for 50 percent of both turnover and employment in agribusiness and for 43 percent of active firms (Figure 26). The most important subsectors are bread production and flour processing, dairy products, beverages, and meat processing (IFC 2018). Beverage firms are responsible for 23.4 percent of total turnover and seem to be larger and quite capital-intensive (IFC 2018). Plant and animal products account for nearly 10 percent of turnover but for 18 percent of sector jobs and 32 percent of firms. Despite investments over the last decade in establishing and modernizing firms, the export potential of food processing is limited. Currently, the sector has limited ability, through certification and skills upgrades, to deliver products that meet international quality and safety standards. Other limitations are unfair competition from a large number of unregistered firms, and the minimal aggregation and cooperation of farmers, which results in high costs and a lack of uniform quality standards. According to the IFC (2018), Kosovo seems to have a dual structure with a large number of SMEs and small number of large modern firms. Privatized firms with significant capacity are profitable but suffer from capacity underutilization, deficient market knowledge, and inadequate managerial capacity. Newly built modern facilities are mostly small; most have excess capacity and suffer from inability to access financing and insufficient turnover to generate investments. Finally, there are not enough cold storage facilities (mainly for soft fruit) and collection centers, inadequate food safety certification, and limited human and working capital. Figure 26. Agribusiness Firms by Product Category, Kosovo, 2018, Percent Active Firms (%) Paper & paper products Employees (%) Wood, wood products & cork Turnover (%) Leather & its products Tobacco products Beverages Food processing Fishing and aquaculture Forestry and wood harvesting Plant & animal products, hunting & related services 0.0 20.0 40.0 60.0 Source: MAFRD 2019. 50 The Role of Agribusiness in Kosovo’s Economy In recent years the country’s trade deficit in agri-food products has worsened. Between 2010 and 2019 it rose from about ¤390 million to ¤694 ml, even though the average annual rate of increase for exports (14 percent) was much higher than for imports (5.5 percent). High producer prices, difficulties in developing food distribution chains; adopting food marketing, quality, and safety standards, and reducing and removing tariffs have all contributed to the trade deficit. In 2019, food exports represented nearly 17.1 percent of Kosovo’s merchandise exports, compared to 12.1 percent in 2014, while agri-food imports were declining from 24.3 percent in 2014 to 21.7 percent in 2019 (Figure 27). In both imports and exports, Kosovo ranks high among comparators (Figure 28), perhaps because its economy is so small. Kosovo is a net importer in all agri-food categories, but processed food accounts for most of Kosovo’s trade deficit, which indicates the importance of promoting investment in food processing. In 2019 food products accounted for 59 percent of the country’s agri-food trade deficit, and agricultural products for 41 percent (Figures 29 and 30). Categories with a considerable deficit include miscellaneous edible preparations; meat products; tobacco products; preparations of cereals, flour, starch, and milk; beverages; dairy products, eggs, and honey; and cereals. Most agri-food trade flows are associated with the Central European Free Trade Agreement (CEFTA) countries (61 percent of exports and 34 percent of imports) though flows with EU countries have been increasing (33 percent of exports and 42 percent of imports). Figure 27. Agri-food Trade, Kosovo, 2014–19, Percent of Goods Trade 30.0 24.3 24.1 23.6 25.0 22.8 21.3 21.7 20.0 17.4 17.1 16.2 14.6 15.0 12.1 12.8 10.0 5.0 0.0 2014 2015 2016 2017 2018 2019 Food exports (% of merchandise exports) Food imports (% of merchandise imports) Source: KAS. The Role of Agribusiness in Kosovo’s Economy 51 Figure 28. Agri-food Trade Compared, 2019, Percent of Goods Trade 70 65.4 60 56.1 50 41.1 40 31.7 30 . 20.8 20 . 16.5 16.3 17.3 13.5 13.3 15.3 12.8 13.3 12.2 9.4 9.6 9.0 10.2 10 7.2 5.7 4.4 4.2 0 KOS MKD ALB KGZ ARM MDA LVA URY EST SVN LTU CZE Food exports (% of merchandise exports) Food imports (% of merchandise imports) Source: WDI; KAS. Note: ALB: Albania; ARM: Armenia; CZE: Czech Republic; EST: Estonia; KOS: Kosovo; KGZ: Kyrgyz Republic; LVA: Latvia; LTU: Lithuania; MDA: Moldova; MKD: North Macedonia; SVN: Slovenia; URY: Uruguay. Figure 29. Agriculture Trade Balance, Kosovo, Million Euros, 2019 -80 -70 -60 -50 -40 -30 -20 -10 0 -19.67 01 Live animals -74.05 02 Meat & edible meat o al -3.50 03 Fish -49.04 04 Dairy products, eggs, honey, etc. -1.18 05 Products of animal origin, nes. -3.64 06 Live trees & other plants -22.88 07 Vegetables -28.88 08 Fruit, citrus, nuts -25.73 09 Co ee, tea, spices -39.16 10 Cereals -9.65 11 Flour, malt, starches, etc. -7.17 12 Oilseeds & oleaginous frutis, etc. -0.45 13 Lac, gums, resins, etc. 0.00 14 Vegetable planting materials, etc. Source: KAS. 52 The Role of Agribusiness in Kosovo’s Economy Figure 30. Food Trade Balance, Kosovo, Million Euros, 2019 -80 -70 -60 -50 -40 -30 -20 -10 0 -28.69 15 Animal & vegetable fats -28.98 16 Preparation of meat, fish, etc. -29.73 17 Sugars & sugar confectionary -22.83 18 Cocoa & cocoa preparations -64.47 19 Preparations of cereals, flour, starch, milk -24.33 20 Preparations of vegetables, fruit, nuts -74.03 21 Misc. edible preparations -52.08 22 Beverages & spirits -15.54 23 Residues & waste from food -68.67 24 Tobacco & tobacco products Source: KAS. The Role of Agribusiness in Kosovo’s Economy 53 08 Policy Reform Options Incentives are needed to actively encourage aggregation of farmers in the food chain. Farms in Kosovo become competitive mostly by becoming larger. Incentives for aggregation could promote capital investments to improve productivity through more efficient use of inputs. Small producer associations and cooperatives could link smallholder farmers to finance and input and output markets. The KCGF agriculture window could also help micro and small farms to access financing. Farm competitiveness could also be facilitated by modifying current types of farm support. Kosovo’s current subsidies seem to be supporting too many unproductive and technically inefficient farms, so the net effect on farm efficiency is negative because of factor misallocation. In view of the country’s EU accession path, farm support should be decoupled. Decoupling (see Box 2) would encourage farmers to make production decisions on the basis of competitive advantage, increase farm investment and production specialization, and shift land use to high-value production.27 Combined with the cross-compliance of support to environmental good practices, which should be also introduced, this will promote the adoption of sustainable farming practices, make Kosovo agri-food products more competitive, and facilitate the transfer of farmland to more efficient and innovative farmers. Ultimately, there could be benefits for rural incomes and job creation. Box 2. Decoupling Decoupling is a process introduced through reform of the EU Common Agricultural Policy (CAP) in 2003. It represented a change from supporting farmers with direct payments linked to the type and volume of output produced or to areas cultivated. Decoupled support allows farmers to produce in response to market demand.  In parallel, decoupled support is granted provisional to the introduction of environmental and animal welfare requirements (cross-compliance). Through decoupling, direct payments are made as compensation for public goods provided by EU farmers, such as clean air or landscape, and as compensation for competitive disadvantages due to tighter EU quality and environmental standards. For member states, decoupled support per ha is determined according to a variety of models, ranging from historical entitlements to a flat rate per ha. Decoupled payments are production-neutral: as pure lump sum transfers, they have no effect on production. Because they are seen as an agricultural policy instrument that does not distort production, consumption, and international trade flows, decoupled payments are consistent with the requirements for domestic support of the WTO Green Box. Decoupled payments represent the vast majority of support provided by the CAP through Pillar I.  A different type of support is provided through CAP Pillar II, which provides support to individual farmers for both on-farm and off-farm investments and for environmentally friendly land management. It also provides support to rural residents investing in economic diversification and to rural areas for various types of infrastructure. 27 Lattruffe and Desjeux 2016; Mary 2013; Rizov et al. 2013; World Bank 2019b, 2019c; Zhu and Lansink 2010. 56 Policy Reform Options Public resources could be more effective if allocated to farm activities with higher rates of return. Continued support of low-value crops reduces the potential of agriculture to add value and generate income opportunities in rural areas. Support for investments in high-value crops could usefully relieve the sector’s trade deficit and improve farm incomes. The fact that small and medium farms trail far behind large ones in terms of efficiency implies underlying structural difficulties. Problems of access to finance, technology, and markets) make it difficult for smaller producers to become more efficient larger units. Current support does not seem to facilitate such a transformation. The best way forward would be public investment in such public goods for farmers as advice, training, and technical and information support; R&D; and infrastructure and storage capacities. Farm support could consider facilitating the modernization of smaller farms. This might be achieved by lower eligibility thresholds for direct support and design of simpler rural development measures that promote adoption of innovative practices that heighten efficiency. Rural development should target the current needs of Kosovo farms by providing incentives to induce technical change and innovation in agriculture. Such measures might differentiate eligibility and selection criteria and support rates that recognize regional disparities and/or target younger farmers with entrepreneurial potential. They might also offer special incentives for medium-sized farms to pursue enlargement and modernized management. Such options could be complemented by measures to improve access to credit and enrich managerial skills. Special measures could also be directed to youth and employment. These measures could both facilitate knowledge and innovation and provide incentives for off- as well as on-farm investments. Use of agricultural land can be protected by creating a more dynamic land market. It will be crucial to expedite cadastral reconstruction to cover the entire country by giving priority to economically active agricultural land and cadastral zones. This would not only promote access to finance but would also help protect fertile agricultural land from illegal construction. Enforcing an unused agricultural land tax and introducing market-based valuation of properties could facilitate use of agricultural land for more productive purposes. Guided by the CAP, rural development programs should also focus on economic diversification and sustainable management of natural resources. Encouragement for creating rural businesses would accelerate the shedding of surplus labor from agriculture and boost rural incomes. Combined with promotion of sustainable farming practices which would speed agri-food commodification28 of the country’s rich natural resources, business creation would widen the rural economic base and enable the transfer of resources currently underemployed in agriculture to other economic activities. Furthermore, water-stressed Kosovo would benefit greatly from prompt application of the irrigation master plan, which would make agriculture more productive in agriculture from better use of scarce water resources. 28 Adding value to the products by utilizing intangible assets. Policy Reform Options 57 Annex A: Theoretical Underpinnings and Methods Annex A.1 The Static View The key concept of the analysis is Productive Performance (PP) which is placed in the microeconomic context of production theory and depicts the quantity of output (Y) produced per input unit (X), or: Produced Outputs PP= Used Inputs The transformation of inputs to outputs is ruled, in economic and technological terms, by the notion of production function . That is, for each i examined farm the following holds: Yi=f(Xi)PPi In terms of measurement, productive performance is captured by the distance of each examined production entity (farm) from the production frontier which is defined by the subset (the locus) of farms which transform the available inputs (X) to outputs (Y) optimally, that is in the best possible way. Therefore, all farms examined are compared to their best counterparts and the analysis becomes of the benchmarking type which is facilitated by the corresponding distance functions. Two significant issues related to this type of analysis are the orientation of the distance function and the characteristics of the employed technology with respect to returns to scale. Regarding the orientation the alternatives include the input orientation version where the question posed is, “how much the farm can reduce the used inputs bundle keeping the produced output constant” while the output orientation is reflected on the question “how much more can we produce with given level of inputs?”. In the current analysis we have adopted the input orientation approach. Regarding returns to scale, the Variable Returns to Scale (VRS) technology is the adopted one where the n% increase (decrease) of inputs results in m% (m≠n) increase (decrease) of outputs. However, we have also tested for the Constant Returns to Scale (CRS) technology option, where the n% increase (decrease) of inputs results in n% increase (decrease) of the produced outputs. Productive Performance may be considered as a construct of several productive efficiency composites, namely technical, scale and allocative. Briefly, technical efficiency (TE) depicts the potentials for input quantity savings producing the same level of output, while Scale efficiency (SE) reflects the level of exploitation of scale economies where the Most Productive Scale Size (MPSS) is the scale in which CRS is the prevailing technology . Finally, allocative efficiency signals the extent to which the production inputs are employed (optimally) according to their relative prices and the available technology. In the current analysis allocative efficiency is not measured because the required data on prices of inputs and outputs are not available. In addition, strict assumptions required in the context of allocative efficiency, about the competitive structure of the inputs and outputs markets are questionable. Conceptually, Technical Efficiency captures managerial competencies, effectiveness of organizational routines and adjustment to business environment and regulatory framework. Product, organizational and marketing innovation matters. On the other hand, Scale Efficiency reflects the influence of technology compatibility and lumpiness, market size, scale decisions, and irreversibility of investments. Process innovation is crucial. 60 Annex A:Theoretical Underpinnings and Methods Figure A1: Technical efficiency measurement Y Frontier Output orientation: TEO=DA/DB Input orientation: TEI=EC/EA B C E A D X A graphical presentation of Technical Efficiency in input (TEI) and output orientation (TEO) context is presented in Figure A1. Farms B and C are fully technical efficient, but farm A is technical inefficient. In Figure A2, there is a graphical presentation of the Scale Efficiency measurement. It is noticeable that the measurement of scale efficiency requires the estimation of technical efficiency under both CRS and VRS technologies. Farm B although is technical efficient is scale inefficient. Figure A2: Scale efficiency measurement Y CRS Frontier VRS Frontier TEVRS = DB/DA TECRS = DC/DA SE = DC/DB = TECRS/TEVRS D B A X Given the above, all t examined farms are ranked according to their productive performance and its efficiency components. Farms with technical efficiency score equal to 1 (C and B in Figures A1 and A2) define the frontier. Farms which are distant to the frontier (A Figures A1 and A2) are assigned with efficiency scores smaller than 1 and exhibit inefficiency losses. Furthermore, farms which do not operate on the CRS frontier (B in fig. 2) suffer additional losses and exhibit SE<1. Distance measures defined and presented above, are estimated using parametric, named as Stochastic Frontiers (SFA), and non-parametric, named Data Envelopment Analysis (DEA), methods. DEA excels SFA in the non-requirement of functional form for the frontier and/or distributional forms for the errors, the accommodation of multiple outputs and inputs, and the effective handling of the statistical noise in the bootstrapped version. On the other hand, SFA accounts for data noise and facilitates the direct hypothesis testing. In the analysis presented here the bootstrapped DEA approach is adopted. Annex A:Theoretical Underpinnings and Methods 61 Using DEA implies the following: • Each Decision-Making Unit (DMU) is compared against all its peers • At least one DMU defines the frontier • Usually multiple DMUs define the frontier in different production scales • Using Bootstrapped version of DEA we estimate statistical properties of the efficiency scores (such as its variance) by measuring those properties when sampling from an approximating distribution (resampling). • One standard choice for an approximating distribution is the  empirical distribution function of the observed data. In the DEA context, the input-specific efficiencies are also estimable. In particular, the efficiency of each input of all the examined farms, the overall productive performance is decomposed into the employed inputs taking into consideration both the radial and the slack based inefficiency facets.29 The efficiency of input xi in time t of agriculture production process () is estimated as: 0 < Effxit = 1 - { Input Slacksit + Radical Improvementit Actual Use of Inputit } <1 Fully efficient use of the input corresponds to a value of equal to 1. The inefficiency of the input use increases as the values of the parameter gets closer to zero. In a second stage efficiency analysis is extended towards the identification of efficiency drivers. The efficiency scores are regressed on farm specific characteristics which include farm size, farm type, location, subsidies status, and farmer’s age. The econometric approach employed is that of truncated regression and tobit with endogenous regressors. Considerations in the Dynamic Context When the attention shifts towards the dynamics of productive performance, the productivity analysis framework is introduced. Actually, the extension of efficiency analysis in time dimension results in productivity analysis which portraits the dynamics of efficiency. We avoid partial productivity measures – output per person, output per hour worked, output per hectare etc- and instead we opt for Total Factor Productivity (TFP) measures. TFP measures of productivity, consider not only the contribution of each production factor but also the role of interaction of inputs within the production process. In this line, the time evolution of productive efficiency is captured by Malmquist TFP index (MPI) which is defined for time t as: MPIt=(ΔΤΕ)t×(ΔSΕ)t×(ΤC)t Where TC is the so-called technical change and Δ denotes change the corresponding term between periods t and t-1. Technical Change captures the ability of the farms to introduce new technologies – innovation which become available and push the frontier “outwards”. In this line, although General Purpose Technologies and disruptive innovations are of primary importance for the movement of the frontier, the exploitation of the technological progress by each farm may be realized based on incremental farm specific innovation. The ΔTE and ΔSE terms are defined as the ratio of the farm’s efficiency (technical or scale) in time t to the corresponding score in period t-1. Overall, values of the MPI index greater than one indicate improvement of productivity, while values less than one indicate deterioration. The same applies for each one of the three components of the Malmquist index. A second stage analysis applies also in the case of productivity analysis which is aiming at the 29 Radial inefficiency is defined as the distance of the DMU from the frontier measured on the ray, which connects the DMU with the axis origin. That is, radial efficiency is related to the potential of equal increase in all outputs (output orientation) or decrease in all inputs (input orientation) which results in efficiency improvement. In the same context, slack based inefficiency is defined as the potential of further increases in output (or reduction in input) that could be gained beyond that implied by the radial projection. Slack based inefficiency results from the piece-wise linear form of the non-parametric frontier in DEA, and may be considered as an indication for input- output mix necessary improvements which are measured by movements on the frontier. 62 Annex A:Theoretical Underpinnings and Methods identification of the drivers of productivity growth and its components. The analysis of this type is based on the econometric estimation of a set of equation(s): t) Prodti = f (xt1 , ... , xk + uit where . k indicates the number of the farms under investigation, t denotes the corresponding period and is the econometric error term. In this line, and taking into considerations the available data, the productivity drivers that is the Xs variables, the following can be investigated: • Do subsidies matter? • The role of farm size • The influence of knowledge Conditions - Innovation • The distinction between Experienced and young farmers? • The role of the farms’ type • The existence of regional differentials • The influence of economies of scope • The impact of Breadth of product mix • The impact of multi-plant production As in any regression type analysis the estimated coefficients of the Xs variables convey a two-fold information. First, they indicate if the driver under investigation exerts any significant influence on the production variable(s) and then reveals the magnitude of this influence. All other variables are held constant at a predetermined level. Finally it should be noted that the above described type of econometric estimations are grounded on a dataset with both time series and cross – section dimensions, which usually are called panel or longitudinal data and require tests regarding the structure if the data, i.e. fixed-effects, random effects and pooled cross section. Annex A:Theoretical Underpinnings and Methods 63 Annex B: Data The employed dataset for efficiency analysis is of the cross-section type for year 2017 and is grounded on the corresponding information derived from FADN30. The dataset initially consisted of 1,192 observations (farms) and after the exclusion of 36 cases which exhibit: • negative values either for the crops output and/or livestock production • zero values for both outputs under consideration • zero or extreme low values for the inputs of land, total specific costs and total farming overheads • the final dataset employed contains 1,156 observations. From 1,192 farms of the sample, 48 do not report any produced crops output and 245 any produced livestock output. These farms are of the single-output type, which are examined in a multi-output context. This is not new in the relevant literature and may offer some opportunities with respect to the multidimensionality of the analysis. In the output side, two outputs are considered: (i) total output of crops and crop products (FADN variable SE135) and (ii) total output of livestock and livestock products (SE206). In the input side, five production factors are considered and in particular: (i) labor input (SE011), (ii) total Utilized Agricultural Area (SE025), (iii) total specific costs (SE281), (iv) total farming overheads (SE336) and (v) total assets (SE436). The descriptive statistics of the employed input and output for year 2017 variables are given in Table B1 below. FADN is an instrument for evaluating the income of agricultural holdings and the impacts of the agricultural policy support. The FADN is a representative sample, drawn annually, of the farm population. Three criteria for sample stratification are used: region, economic size and type of farming. The number of farms in each stratum is derived from the Farm Structure Survey (FSS). FADN defines a threshold based on economic size and draws the sample for farms over this threshold, which for Kosovo is 2,000 euros. Hence, FADN sample is representative of all farms above this threshold (and only farms above this threshold define the FADN population in every given year). Table B1: Descriptive statistics of employed outputs and inputs variables, Kosovo, 2017 (N=1,156)31 Descriptive Statistics Variable Mean (St. Dev.) Min (Max) Outputs Value of fieldcrops production 13,865.7 0 (58,144.4) (1,562,560) Value of Livestock production 1,478.2 0 (42,420.0) (774,375) Inputs Labour 3,622.24 50.0 (5,121.97) (98,500) Land 11.05 1.0 (29.80) (651.0) Capital 3,227,268 4,485.7 (1,031,728) (2,700,000) Intermediates 15,772.62 70.0 (71,431.81) (1,943,596) Overheads 2,425.92 0.0 (11,707.51) 345,800.0 Source: World Bank staff calculations based on data from FADN 30 Farm Accounting Data Network 31 All variables, except labor and land inputs are measured in monetary values. Labor is measured in time worked in hours by total labor input on holding, and the land input is measured in hectares. Monetary values have been deflated using Kosovo GDP deflator (World Bank, National Accounts data). 66 Annex B:Data Additional information from FADN on years 2016 and 2015 to carry out the TFP analysis. Following the same data cleaning process described above for 2017 for years 2016 and 2015, the number of farms for each one of the three years is presented in Table B2. Table B2: Number of FADN farms, Kosovo, 2015-2017 Year Number of Farms Percentage (%) 2015 1,186 33.65 2016 1,182 33.54 2017 1,156 32.8 Total 3,524 100 Source: World Bank staff calculations based on data from FADN Descriptive Statistics of input-output variables for the three years (period 2015-2017) are presented in Table B3. Comparing tables B1 and B3, it is easily seen that the time variation of inputs-outputs is small. On the contary, the within-year variance is quite high and indicates the significant heterogeneity of the examined farms. Table B3: Descriptive statistics of employed outputs and inputs variables, Kosovo, 2015-2017 Descriptive Statistics Variable Obs Mean (St. Dev.) Min (Max) Outputs Value of fieldcrops production (FADN Variable SE135: Value of total 3,524 13,206.1 0.0 output of crops and crop products) (57,043.9) (1,665,000) Value of Livestock production 3,524 13,716.2 0.0 (SE206: Value of total output of livestock and livestock products) (36,796.7) (774,375.0) Inputs Labour (SE011: time worked in hours by total labour input on 3,524 3,516.1 10.0 holding) (6,401.0) (195,150.0) Land (SE025: total utilized agricultural area of holding_ 3,524 11.3 0.1 (29.9) (651.0) Capital (SE436: Total assets of farms (current and fixed) closing 3,524 313,655.1 1,660.0 valuation (983,069.9) (20,700,000) Intermediates (SE281: cost of crop- and livestock-specific inputs) 3,524 13,656.5 0.0 (50,277.7) (1,943,596) Overheads (SE336: Supply costs linked to production activity but 3,524 2,307.2 0 .0 not to specific lines of production) (8,768.1) (345,800) Source: World Bank staff calculations based on data from FADN The availability of information for three successive years permits us to conduct the dynamic part of the analysis, that is the estimation of TFP growth and the corresponding decomposition in major components. It should be noted that the dynamic analysis requires the employed dataset should be of the balanced panel type. Table B4 presents the distribution of the number of farms per year. Based on this distribution in the productivity analysis we have incorporated 1,078 farms and 3,234 observations in total. Annex B:Data 67 Table B4: Number of FADN farms per year, Kosovo, 2015-2017 Years of Appearance Number of Farms Percentage (%) Only 2015 9 0.26 Only 2016 3 0.09 Only 2017 60 1.70 All years 3,234 91.77 2015 & 2016 183 5.19 2016 & 2017 19 0.54 2015 & 2017 16 0.45 Source: World Bank staff calculations based on data from FADN Having estimated the frontiers for each one of the three examined years, we have droped 48 additional obseravtions which violated the convexity conditions for the frontiers and resulted in outliers values for TFP and its components. Therefore, 1,062 farms were finally incorporated in the dynamic part of the analysis. It should be noted that the dynamic analysis is realized in terms of change from year to year. Hence, the final number of observations in the dataset of the dynamic analysis is 2,132. 1.1 A picture of the FADN sample of farms in Kosovo Size, location and type of examined farms Figures B1-B6 and Tables B5-B8 present the size, district location, and type of farms distributions (also according to farm managers’ age), as well as the joint distributions of (i) farm size with respect to type, (ii) farm size with respect to location district and (iii) farm type according to location distrct. Figure B1: Farm size distribution, Kosovo FADN, 2017 328 346 236 246 Micro Small Medium Large 0 200 400 600 800 1000 1200 Source: World Bank staff calculations based on data from FADN Figure B2: Farm location distribution, Kosovo FADN, 2017 0 50 100 150 200 250 300 Pristina 282 Ferizaj 103 Gjakova 183 Gjilan 122 Peja 189 Prizren 138 Mitrovica 139 Source: World Bank staff calculations based on data from FADN 68 Annex B:Data Figure B3: Farm types distribution, Kosovo FADN, 2017 Horticulture and wine 30 368 Fieldcrops 334 Mixed farms 246 Milk Granivores 49 78 51 Other Other grazing permanent livestock crops Source: World Bank staff calculations based on data from FADN Table B5: Farm size and type joint distribution, Kosovo FADN, 2017 Number of Farms (column %) Size Farm type Micro Small Medium Large Total Fieldcrops 77 69 89 133 368 (23.48) (19.94) ( 37.71) (54.07) (31.83) Horticulture and Wine 6 5 2 17 30 (1.83) (1.45) ( 0.85) ( 6.91) (2.60) Other Permanent Crops 14 19 9 9 51 (4.27) (5.49) (3.81) (3.66) (4.41) Milk 90 93 49 14 246 (27.44) (26.88) (20.76) (5.69) (21.28) Other Grazing Livestock 12 26 31 9 78 (3.66) (7.51) (13.14) (3.66) (6.75) Granivores 4 1 6 38 49 (1.22) (0.29) (2.54) (15.45) (4.24) Mixed Farms 125 133 50 26 334 (38.11) (38.44) (21.19) (10.57) (28.89) Total 328 346 236 246 1,156 Source: World Bank staff calculations based on data from FADN Annex B:Data 69 Figure B4: Farm size and type joint distribution, Kosovo FADN, 2017 0 20 40 60 80 100 120 140 Fieldcrops 77 Micro 69 Small 89 133 Medium Horticulture and wine 6 Large 5 2 17 Other permanent crops 14 19 9 9 Milk 90 93 49 14 Other grazing livestock 12 26 31 9 Granivores 4 1 6 38 Mixed farms 125 133 50 26 Source: World Bank staff calculations based on data from FADN Table B6: Farm size and district joint distribution, Kosovo FADN, 2017 Number of Farms (column %) Size District Micro Snall Medim Large Total Pristina 82 80 56 64 282 (25.00) (23.12) (23.73) (26.02) (24.39) Ferizaj 27 39 16 21 103 (8.23) (11.27) (6.78) (8.54) (8.91) Gjakova 41 57 44 41 183 (12.50) (16.47) (18.64) (16.67) (15.83) Gjilan 33 35 25 29 122 (10.06) (10.12) (10.59) (11.79) (10.55) Peja 56 61 38 34 189 (17.07) (17.63) (16.10) (13.82) (16.35) Prizren 49 35 30 24 138 (14.94) (10.12) (12.71) (9.76) (11.94) Mitrovica 40 39 27 33 139 (12.20) (11.27) (11.44) (13.41) (12.02) Total 328 346 236 246 1,156 (100.00) (100.00) (100.00) (100.00) (100.00) Source: World Bank staff calculations based on data from FADN 70 Annex B:Data Figure B5: Farm size and district joint distribution, Kosovo FADN, 2017 90 Number of Farms Micro Small 82 Medium 80 80 Large 70 64 61 60 57 56 56 49 50 44 41 41 40 39 39 40 38 35 35 34 33 33 30 29 30 27 27 25 24 21 20 16 10 0 Pristina Ferizaj Gjakova Gjilan Peja Prizren Mitrovica Source: World Bank staff calculations based on data from FADN. Table B7: District and farm joint distribution, Kosovo FADN, 2017 District Farm Type Pristina Ferizaj Gjakova Gjilan Peja Prizren Mitrovica Total Fieldcrops 91 36 61 39 46 50 45 368 (32.27) (34.95) (33.33) (31.97) (24.34) (36.23) (32.37) (31.83) Horticulture and Wine 4 0 11 6 2 4 3 30 (1.42) (0.00) (6.01) (4.92) (1.06) (2.90) (2.16) (2.60) Other Perm. Crops 21 6 0 3 15 1 5 51 (7.45) (5.83) (0.00) (2.46) (7.94) (0.72) (3.60) (4.41) Milk 63 12 40 29 53 24 25 246 (22.34) (11.65) (21.86) (23.77) (28.04) (17.39) (17.99) (21.28) Other Grazing 14 6 14 9 12 15 8 78 Livestock (4.96) (5.83) (7.65) (7.38) (6.35) (10.87) (5.76) (6.75) Granivores 12 4 9 6 8 4 6 49 (4.26) (3.88) (4.92) (4.92) (4.23) (2.90) (4.32) (4.24) Mixed Farms 77 39 48 30 53 40 47 334 (27.3) (37.86) (26.23) (24.59) (28.04) (28.99) (33.81) (28.89) Total 282 103 183 122 189 138 139 1,156 Source: World Bank staff calculations based on data from FADN. Annex B:Data 71 Table B8: Basic descriptive statistics of farm managers’ age, Kosovo FADN, 2017 Mean 51.53 (std dev) (12.77) Max 86 (Min) (17) Under 40 years 227 (% ) (19.64) Over 40 years 929 (% ) (80.36) Source: World Bank staff calculations based on data from FADN. Figure B6: Farm managers’ age distribution, Kosovo FADN, 2017 Density .03 The Static View .02 .01 0 Age 20 40 60 80 100 Source: World Bank staff calculations based on data from FADN 72 Annex B:Data Annex B:Data 73 Annex C: Analytical findings of efficiency and productivity analysis Farms in Kosovo are characterized by significant technical inefficiency but operate satisfactorily with respect to returns to scale side of technology. The bias-corrected scores under variable returns to scale (bcTE) are rather low, and also present significant variation (Table C1). The mean efficiency score is 0.272 with standard deviation 0.158. The minimum value of TE is 0.027 while the maximum is 0.780. The mean Scale Efficiency32 (SE) is rather high, equal to 0.721, with standard deviation equal to 0.234, while the minimum and maximum values of SE are 0.0028 and 1 respectively. Efficiency analysis confirms that the agriculture sector is highly polarized in Kosovo. One group of farms seems to perform quite well and attain a bcTE score which exceeds 0.6 (Table C2). This cluster consists of 85 farms (7.3 percent of the total sample) and indicates a degree of polarization in terms of productive performance, within agriculture in Kosovo. On the other hand, 480 farms (41.5 percent of the sample) perform very poorly. The high performance of Kosovo farms with respect to scale economies should be interpreted in the context of learning by doing which allow the optimal decisions with respect to scale to be taken, grounded on their repetitive character, and the quasi- competitive character of agricultural markets. Even then, around 15.6 percent of sample farms perform poorly with respect to scale economies (Table C3). Table C1: Technical and Scale Efficiency, Kosovo, 2017 Descriptive Statistics Mean Max Efficiency (Std. Dev.) (Min) TE vrs bias corrected 2017 0.272 0.780 (0.158) (0.027) TE vrs 2017 0.356 1.000 (0.237) (0.037) TE crs 2017 0.266 1.000 (0.225) (0.002) SE 2017 0.721 1.000 (0.234) (0.028) Source: World Bank staff calculations. Table C2: Technical Efficiency intervals, Kosovo, 2017 TE vrs bias corrected interval No of Farms Percent (%) Cumulative (%) (0.0-0.2] 480 41.52 41.52 (0.2-0.4] 468 40.48 82.01 (0.4-0.6] 123 10.64 92.65 (0.6-0.8] 85 7.35 100.00 Total 1,156 100.00 Source: World Bank staff calculations. 32 SE is defined as the ratio of the TE under constant returns to scale (crs) to TE under variable returns to scale (vrs). When the bootstrapped values of TE are employed for both technologies, SE becomes greater than one for 102 farms. This is due to the unequal-width confidence intervals which result from the bootstrap procedure for estimating TE under vrs and crs technologies. Since SE values which exceed one are not economically meaningful, we opt for the use of the SE defined in terms of bias non-corrected frontiers. One should expect that this SE measure does not convey any bias since any influence of data noise is included both in the numerator and the denominator, thus offsetting each other. 76 Annex C:Analytical findings of efficiency and productivity analysis Table C3: Scale Efficiency intervals, Kosovo, 2017 SE interval No of Farms Percent Cumulative (%) (%) (0.0-0.2] 29 2.51 2.51 (0.2-0.4] 123 10.64 13.15 (0.4-0.6] 195 16.87 30.02 (0.6-0.8] 296 25.61 55.62 (0.8-1.0] 513 44.38 100.00 Total 1,156 100.00 Source: World Bank staff calculations. Multivariate analysis of efficiency drivers rather confirms findings already presented and enriches them with a causal relationship. Subsidies (Table C4) seem to exert a very small negative influence on TE and a very marginal positive influence on SE. Size increases have a negative influence on both TE and SE. These negative effects of size on TE can be attributed to the fact that most farms in the sample belong to the micro and small size groups, and hence an increase in size would likely lead to efficiency losses. However, when farm size exceeds a certain limit (that of medium farms), this benefits TE and SE. In other words, subsidies seem to be associated with a misallocation of productive resources. With field crops as the reference group, all other types of farm33 enjoy a comparative advantage in terms of both technical and scale efficiency. With Pristina farms as the reference group, farms located in Peja/Pec, Gjilan and Mitrovica are characterized by significant performance inferiority associated with technical inefficiency. The opposite is observed in the case of farms located in Gjakove and Prizren. Table C4: Drivers of Technical and Scale Efficiency – Multivariate Analysis, Kosovo, 2017 TE SE Drivers Coefficient Coefficient (std. err.) (std. err.) Size and Subsidies Size -0.256*** -1.055*** (0.041) (0.079) Size squared 0.027*** 0.116*** (0.005) (0.009) (Size*subsidies) 0.000***+ 0.000+ (0.000)+ (0.000)+ Subsidies -0.001*** 0.000+ (0.000)+ (0.000)+ Age 0.000+ -0.002*** (0.000)+ (0.001) 33 Estimates on mixed farms are omitted due to multicollinearity. Annex C:Analytical findings of efficiency and productivity analysis 77 TE SE Drivers Coefficient Coefficient (std. err.) (std. err.) District Pristina Baseline Ferizaj 0.060 0.038* (0.037) (0.021) Gjakova -0.041* -0.021 (0.025) (0.027) Gjilan 0.043*** -0.017 (0.009) (0.029) Peja 0.057*** -0.012 (0.020) (0.019) Prizren -0.036*** -0.006 (0.008) (0.022) Mitrovica 0.032*** 0.014 (0.011) (0.035) Fieldcrops Baseline Farm Type Horticulture and Wine 0.295*** 0.803*** (0.032) (0.054) Other Permanent Crops 0.392*** 1.147*** (0.057) (0.112) Milk 0.315*** 1.435*** (0.055) (0.107) Other Grazing Livestock 0.283*** 1.303*** (0.046) (0.088) Granivores 0.359*** 0.881*** (0.028) (0.055) Mixed Farms Omitted due to multicolliearity Constant 0.535*** 1.808*** (0.032) (0.068) Fit Stats -0.244 0.469 Pseudo R-squared AIC -1,210.34 -100.974 BIC -1,180.03 -700.663 Three, two and one asterisk indicate statistically significance at 1%, 5%, and 10% level respectively ++ Actually smaller than 0.001 Source: World Bank staff calculations. Multivariate analysis of TFP growth drivers shows that farm structural characteristics explain TFP growth well. In fact, fourteen out of the eighteen explanatory variables exert statistically significant influence on TFP growth (Table C5). Farm size exerts an inverted U-shape influence on TFP growth. Micro and small farms appear to suffer productivity losses as being below the minimum efficiency size. On average, the minimum efficient size is approximately between EUR 18,000 and 22,000 Subsidies affect TFP growth but not considerably. However, the influence of subsidies on TFP growth becomes positive when allowance for farm size is introduced into the analysis. This finding, although of very small magnitude, reveals that the subsidies mechanism is in favor of larger farms. The available resources, competencies and capabilities of larger farms allow them to seek and obtain subsidies without search, coordination and transaction costs becoming a major burden on their other business operations. 78 Annex C:Analytical findings of efficiency and productivity analysis Farmers managers’ age does not appear to be a significant driver of TFP growth. The characteristics of human capital, that may be possessed by younger farmers, that is creativity, level of education, and technological progressiveness, are offset by influence of experience and learning by doing effects. Table C5: Drivers of TFP growth, Kosovo, 2015-2017 TFP Growth34 Group of variables Variable Estimated Coefficient Group of variables Variable Estimated Coefficient (Std. Error) (Std. Error) Farm specific Size -3.267** Farm Type Fieldcrops 0.744 characteristics (1.609) (0.917) Size squared 0.356* Horticulture and Wine 2.858 (0.182) (1.753) Subsidies -0.090 Other Permanent crops35 - (0.198) (Size*Subsidies) 0.000**+ Milk 4.061* (0.000)+ (2.142) Age -0.001 Other Grazing Livestock 4.248* (0.006) (2.312) District Pristina36 - Granivores 4.887*** (1.854) Ferizaj 0.314*** Mixed Farms 3.839*** (0.031) (1.006) Gjakova 0.017 Constant 4.420*** (0.013) (1.473) Gjilan 0.975*** Fit Statistics N 2,134 (0.033) Peja 0.204*** BIC -158.724 (0.014) Prizren 1.131*** AIC -317.054 (0.044) Mitrovica 0.104*** (0.023) Three, two and one asterisks indicate statistical significance at the 1%, 5%, and 10% levels, respectively + Smaller than 0.001 Source: World Bank staff calculations. 34 Random effects panel data model results are presented. We have also estimated the fixed effects version of the same model. A Hausman test does not support the rejection of random effects rejection. Results are available upon request. 35 Other Permanent Crops is the reference farm type. 36 Pristina is the reference district. Annex C:Analytical findings of efficiency and productivity analysis 79 In terms of farms type, only fieldcrops and horticulture/wine farms do not present positive productivity differentials compared to the reference other permanent crops category. Also, most of the examined districts exhibit higher productivity growth than Pristina district. Regression analysis shows that farm size affects TEC and Technical Change. In the case of TEC, a U-shaped relationship is documented (Table C6). It is the same pattern identified in the case of the influence of size on overall productivity. Micro and small farms suffer TEC losses up to a limit and then further increase in size results in significant technical efficiency gains. The opposite is found for the case of Technical Change. The embodiment of technology advances reflected in the outward movement of the frontier is higher for smaller farms up to a certain level. Beyond this size threshold, farms show smaller differences in their innovation activities. These results should be interpreted carefully, considering the differences of the initial knowledge and technology conditions between small and large farms. Farms which are technologically mature are expected to innovate less than their counterparts which lag behind in technological terms. Therefore, the above difference between small and large farms with respect to technological change may also encompass differences in the current level of their technological status. It is also worth noticing that the declining rate of Technical Change beyond the size threshold is, though significant, quite small. That is, larger farms remain innovative, but at a lower level. Size variables do not exert a statistically significant effect on scale efficiency change. Farm types seem to be influencing TEC and Technical Change. All farm types outperform the reference type of “other permanent crops”, in terms of TEC. With the exception of mixed farms, all the remaining types seem to lag behind the Technical Change component of the “other permanent crops” farm group. No significant difference, between all farm types and the reference type, has been identified for the case of SEC. Farm location seems to affect differences in all TFP components. Farms located in Gjakove, Gjilan and Prizren districts possess positive differentials across all productivity components, compared to the reference district of Pristina. On the other hand, none of the examined districts underperforms on all the three components jointly, compared to the reference district. Finally, farm manager age does not exert any statistically significant influence on any of the examined TFP components Table C6: Drivers of TFP components growth, Kosovo, 2015-2017 TFP Components37 Group of variables Variable TE Change SE Change Technical Change Estimated Coefficient Estimated Coefficient Estimated Coefficient (Std. Error) (Std. Error) (Std. Error) Farm specific Size -2.409*** -0.189 0.260*** characteristics (0.771) (0.339) (0.095) Size squared 0.261*** 0.021 -0.028*** (0.087) (0.038) (0.011) Subsidies -0.185* 0.259*** 0.017* (0.096) (0.050) (0.008) (Size*subsidies) 0.000**+ 0.000+ -0.000**+ (0.000)+ (0.000)+ (0.000)+ Age 0.002 -0.001 0.001 (0.003) (0.003) (0.001) 37 Random effects panel data model results are presented. We have also estimated the fixed effects model. Hausman tests do not support the rejection of the random effects model version in all the three cases. Results are available upon request 80 Annex C:Analytical findings of efficiency and productivity analysis TFP Components37 Group of variables Variable TE Change SE Change Technical Change Estimated Coefficient Estimated Coefficient Estimated Coefficient (Std. Error) (Std. Error) (Std. Error) District Pristina38 - - - Ferizaj 0.306*** -0.007 -0.042*** (0.012) (0.007) (0.004) Gjakova 0.025*** 0.024*** -0.036*** (0.006) (0.008) (0.003) Gjilan 0.213*** 0.149*** 0.012*** (0.015) (0.010) (0.002) Peja 0.144*** -0.064*** 0.002 (0.009) (0.003) (0.002) Prizren 0.221*** 0.204*** 0.048*** (0.018) (0.011) (0.002) Mitrovica 0.110*** -0.030*** 0.003 (0.012) (0.005) (0.003) Farm Type Fieldcrops 2.480*** 0.192 -0.236** (0.881) (0.325) (0.107) Horticulture and Wine 3.571*** -0.130 -0.355*** (0.987) (0.444) (0.130) Other Permanent - - - crops39 Milk 3.009*** 0.446 -0.304** (1.075) (0.466) (0.123) Other Grazing 3.002*** 0.358 -0.259** Livestock (0.890) (0.440) (0.115) Granivores 2.894*** 0.171 -0.186** (0.496) (0.241) (0.080) Mixed Farms 1.042 0.218 0.108 (0.672) (0.555) (0.095) Constant 3.540*** 1.109*** 0.831*** (0.746) (0.409) (0.103) Fit Statisticts N 2,134 2,134 2,134 BIC -157.621 -18.750 -107.443 AIC -200.547 -64.551 -133.822 Three, two and one asterisks indicate statistical significance at the 1%, 5%, and 10% level respectively. + Smaller than 0.001 Source: World Bank staff calculations. 38 Pristina is the reference district. 39 Other Permanent Crops is the reference farm type. Annex C:Analytical findings of efficiency and productivity analysis 81 References Alishani, A. (2019). The Impact of Support Policies on Technical Efficiency of Farms in Kosovo. Review of Agricultural and Applied Economics, XXII: 46-54. Agriculture Development Agency (2019). ADA, Grants and Direct Payments 2009-2018. ADA. Pristina. APM Database for Western Balkan Countries (2019). Agricultural Policy Measures Database compiled under the SWG Projects. Available at: http://app.seerural.org/agricultural-statistics/ Bicoku, Y., Zeqiri, M. and Gjeci, G. (2018). Awareness about Livestock Standards – The Case of Dairy Farmers Albania and Kosovo. Albanian Journal of Agricultural Science, 17: 187-196. Bogetoft, P. and Otto, L. (2011). Benchmarking with DEA, SFA and R. Springer-Verlag. New York. Bokusheva, R. and Cechura, L. (2017). Evaluating Dynamics, Sources and Drivers of Productivity Growth at the Farm Level. OECD Food, Agriculture and Fisheries Papers, No. 106. OECD Publishing, Paris. Available at: http://dx.doi.org/10.1787/5f2d0601-en Daci-Zejnullahi, E. (2014). Agriculture Sector in Kosovo And Opportunities for Cooperation with Balkans Countries. Turkish Journal of Agricultural and Natural Sciences, 2: 1985-1992. Dixit, A. K. and Stiglitz, J. E. (1977). Monopolistic Competition and Optimum Product Diversity. The American Economic Review, 67: 297–308. European Commission (2014). Kosovo: Agriculture and Rural Development Support Programme V. IPA II 2014-2020. European Commission. Brussels. European Commission (2015). Kosovo: Further Support to Kosovo’s Development of the Agri-Rural Sector and Food Safety. IPA II 2014-2020. European Commission. Brussels. European Commission (2019). Economic Reform Programme of Kosovo: Commission Assessment. SWD (2019) 168 final. European Commission. Brussels. GAFSP (2020). From Farmers to Financial Institutions, the Impacts of COVID-19 in Kosovo are Widespread. Available at: https://www.gafspfund.org/news/farmers-financial-institutions- impacts-covid-19-kosovo-are-widespread Hyseni, B. (2020). The Impact of COVID-19 on Agriculture, Food and Rural Areas in Western Balkans. South Eastern Europe Regional Rural Development Standing Working Group. Agricultural Policy Forum 2020, 12-16 October. Ilic, B., Pavloska – Gjorgjieska, D. and Ciaian, P. (eds) (2019). Agricultural Policy Developments and EU Approximation Process in the Western Balkan Countries. IPTS-JRC. Seville. International Finance Corporation (2018). Agriculture Finance in Kosovo. IFC Kosovo Office. Pristina. Kerolli-Mustafa, M., Gjokaj, E. and Hoxha, B. (2017). Kosovo: Agricultural Policy Development and Assessment. In: Volk, T., Erjavec, E., Ciaian, P. and Gomez Y Paloma, S. (eds.). Monitoring of Agricultural Policy Developments in the Western Balkan Countries. IPTS-JRC. Seville. Kosovo Agency of Statistics (2020). Various Data on Agriculture and Trade. KAS. Pristina Latruffe, L., Bravo-Ureta, B.E., Moreira, V.H., Desjeux, J. and Dupraz, P. (2011). Productivity and Subsidies in EU Countries: An Analysis for Dairy Farms Using Input Distance Frontiers. Paper presented at the EAAE 2011 Congress, August 30 – September 2, Zurich. Latruffe, L. and Desjeux, Y. (2016). Common Agricultural Policy Support, Technical Efficiency, and Productivity Change in French Agriculture. Review of Agricultural, Food and Environmental Studies, 97: 15–28. MAFRD (2013). Agriculture and Rural Development Programme, 2014-2020. MAFRD. Pristina. MAFRD (2020; 2019; 2018; 2017; 2016; 2015; 2014). Kosovo Green Report. MAFRD. Pristina. MAFRD (various years). Annual Agriculture and Rural Development Programme. MAFRD. 82 Annex C:Analytical findings of efficiency and productivity analysis Pristina. MAFRD (2020). FADN Individual Farm Database. MAFRD. Pristina. Mary, S. (2013). Assessing the Impacts of Pillar 1 and 2 Subsidies on TFP in French Crop Farms. Journal of Agricultural Economics, 64: 133-144. Masterson, T. (2007). Productivity, Technical Efficiency and Farm Size in Paraguayan Agriculture. Working Paper No. 490, Levy Economics Institute, New York. Mellor, J. W. (1969). The Subsistence Farmer in Traditional Economies. In: Wharton, C.R. Jr. (editor). Subsistence Agriculture and Economic Development. Ed. Aldine, Chicago. Miftari, I. and Hoxhaj, H. (2014). Agriculture and Agricultural Policy in Kosovo. In: Volk, T., Erjavec, E. and Mortensen, K. (eds.). Agricultural Policy and European Integration in Southeastern Europe. FAO. Rome. Miftari, I., Hoxha, B. and Gjokaj, E. (2016). Kosovo: Agricultural Policy Brief. In: Volk, T., Erjavec, E., Ciaian, P. and Gomez Y Paloma, S. (eds.). Analysis of the Agricultural and Rural Development Policies of the Western Balkan Countries. IPTS-JRC. Seville. Milenkovic, M., Jakovljevic, J., Curcovic, V. and Mijacic, D. (2017). Development of Agriculture in Kosovo and Environmental Impact. Policy Brief. Institute for Territorial Economic Development. Pristina. MoF (various years). Medium Term Expenditure Framework. Ministry of Finance. Pristina. Ministry of Trade and Industry (2014). Sector Profile of Food Processing and Packaging Industry. Pristina. OECD (2020). The COVID-19 Crisis in Kosovo – 2020 Archive. Tackling Coronavirus (COVID-19) – Contributing to a Global Effort. OECD. Paris. Official Gazette of the Republic of Kosovo (various years). Law on the Budget of the Republic of Kosovo – various years. Pristina. Republic of Kosovo (various years). Economic Reform Programme. Pristina. Republic of Kosovo (2020a). Aid Management Platform. Available at: https://amp-mei.net/portal/. Pristina. Republic of Kosovo (2020b). National Programme for Implementation of Stabilisation and Association Agreement (NPISAA) 2020 – 2024. Pristina. Rizov, M., Pokrivcak, J. and Ciaian, P. (2013). CAP Subsidies and Productivity of EU Farms. Journal of Agricultural Economics, 64: 537-557. Simar, L. and Wilson, P.W. (1998). Sensitivity Analysis of Efficiency Scores: How to Bootstrap in Nonparametric Frontier Models. Management Science, 44: 49-61. Simar, L. and Wilson, P.W. (2000). Statistical Inference in Nonparametric Frontier Models: The State of the Art. Journal of Productivity Analysis, 13: 49-78. Sini, M.P. (2009). Evolution of Efficiency Factors, Different Ways by which it can be Reached and Competitiveness of Small Farms. Paper presented at 111th EAAE-IAAE Seminar ‘Small Farms: Decline or Persistence’, University of Kent, Canterbury, 26-27 June. Tsekouras, K.D., Skuras, D., Dimara, E. and Tzelepis, D. (2008). Back to the Basics: The Comanor- Wilson MES Index Revisited. Small Business Economics, 32: 111-120. USAID and Republic of Kosovo (2016). Kosovo National Strategy on Property Rights. USAID and Kosovo Ministry of Justice. Pristina. World Bank (2010). Kosovo: Unlocking Growth Potential: Strategies, Policies, Actions – A Country Economic Memorandum. World Bank Group. Washington DC. World Bank (2017a). Agriculture for Jobs and Growth in the Western Balkans. Report No. AUS19883. Agriculture Global Practice. The World Bank. Washington DC. Annex C:Analytical findings of efficiency and productivity analysis 83 World Bank (2017b). Republic of Kosovo – Systematic Country Diagnostic (P151814). Report No. 114618-XK. ECA Region. The World Bank. Washington DC. World Bank (2018). Exploring the Potential of Agriculture in the Western Balkans: A Regional Report. Agriculture Global Practice. The World Bank. Washington DC. World Bank (2019a). Albania – Agriculture Potential and Public Support: A Public Expenditure and Technical Efficiency Review. Report No: AUS0000951. The World Bank. Washington DC. World Bank (2019b). The Republic of Serbia – Public Expenditure Review: Efficiency, Effectiveness and Equity of Public Spending in Agriculture. Agriculture Global Practice. The World Bank. Washington DC. World Bank (2019c). Rebalancing Government Support to Improve Agriculture. In: Republic of North Macedonia Public Finance Review: Sowing the Seeds of a Sustainable Future. Report No. 135704 – MK. The World Bank. Washington DC. World Bank (2019d). Sector Diagnostic and Analysis of Public Spending in Agriculture and Rural Development. Croatia STARS RAS. Agriculture and Food Global Practice. The World Bank. Washington DC. World Bank (2020). World Development Indicators. Retrieved from WDI website: databank.worldbank.org/source/world-development-indicators. The World Bank. Washington DC. Zhu, X. and Lansink, A. O. (2010). Impacts of CAP Subsidies on Technical Efficiency of Crop Farms in Germany, the Netherlands and Sweden. Journal of Agricultural Economics, 61: 545-564. 84 Annex C:Analytical findings of efficiency and productivity analysis WORLD BANK GROUP Raising Agricultural Productivity Kosovo Country 88 November  2021 Economic Memorandum