Moldova Agricultural Sector Note: Land Module (2005)

1 Moldova Agricultural Sector Note: Land Module Zvi Lerman Revised August 12, 2005 Contribution to the World Bank’s Agricultural Sector Note for Moldova (2005-2006) Table of Contents Executive Summary and Policy Recommendations Introduction: The Status of Land Reform in Moldova A Note on Terminology and Data Sources 1. Changes in Land Ownership and Land Use Since 1990 2. Reorganization of the Farm Sector Creation of Private Farms Creation of New Corporate Farms 3. Land Use and Farm Structure: Moldova and EU-15 Average Farm Sizes Land Concentration 4. Productivity and Efficiency of Small and Large Farms in Moldova Partial Productivity Measures: Changes in Productivity of Land and Labor over Time Total Factor Productivity Technical Efficiency of Small and Large Farms Evidence of Increasing Returns to Scale among Corporate Farms 5. Development of Land Markets in Moldova Land Prices Transaction Costs Land Leasing 6. Land Leasing Patterns and Equity Rural Households as Suppliers of Land Households’ Land Disposal Strategy Motivation for Leasing Out Land Legal Arrangements and Payment Discipline Leasing and Family Income 2 Executive Summary and Policy Recommendations Moldova has made some very impressive achievements in land reform since the introduction of the USAID-sponsored National Land Program (NLP) in 1998-99. These include a dramatic increase in private (non-state) land ownership, which rose from practically zero in 1989 to 67% of all agricultural land (and to an even more impressive 80% of agricultural land used by producers), and a virtually completed allocation of physical plots to more than one million rural people. These highly positive developments appear to have led to the tentative signs of recovery in agriculture that we observe since 2000, when the steep decline in agricultural production was arrested and both output and productivity resumed growth. The growth in agriculture has been very slight so far, especially due to the intervening drought year in 2003, but if it is indeed associated with the progress in reforms, as we believe, more robust growth can be expected in the immediate future. Our main recommendation is therefore that Moldova stays the course of its reforms and desists from experimenting with major reversals of strategy until the achievements made so far have had time to produce their full impact. The progress with land privatization has not been fully matched by progress with individualization of agriculture – the second main facet of transition to market. Fully 50% of agricultural land in Moldova is still controlled by large-scale corporate farms. In itself, this is also a huge achievement, far surpassing the reform outcomes in Russia and Ukraine (where large corporate farms still control about 80% of agricultural land). Yet this is not satisfactory compared to land use patterns in market economies, where corporate farms control less than 2% of agricultural land. Of course market agriculture supports a wide spectrum of organizational forms, ranging from very small part-time family units (equivalent to household plots in Moldova) to fairly large corporate farms. Two salient points must be borne in mind, however: • • Market agriculture is agriculture of family farms, not corporate farms: corporate farms are few in number and control a very small share of agricultural land. Corporate farms in market agriculture are on average much smaller than corporate farms in Moldova: they fall in the range of 100-300 hectares rather than 1,000-3,000 hectares as is often the case in Moldova. The large corporate farms in Moldova are a carryover from the Soviet era. The Soviet agricultural ideology was driven, among other factors, by expectations of economies of scale. This ideology is still deeply implanted in the minds of all agricultural decision makers, regardless of their obvious devotion to market economy principles. This ideology is at the root of the persistent complaints on fragmentation of agricultural holdings and the need to achieve consolidation by transition to large cooperatives or corporations. Yet the policy makers in Moldova should realize that all attempts to preserve large-scale corporate structures in former Soviet republics (whether as agricultural cooperatives or as new corporations with marketsounding names) have not produced any positive results. The Russian and Ukrainian dream of “horizontal transformation”, making persistently inefficient corporate farms suddenly efficient, apparently does not work. On the contrary, it is the three small countries that resolutely abandoned the large-scale structures and made a clean shift to small-scale individual agriculture – Armenia, Georgia, and Azerbaijan – that demonstrate the most impressive recovery record 3 among the CIS countries in recent years. Moldova has much in common with these three small, densely populated countries, much more than with Russia and Ukraine, and the Moldovan policy makers are strongly advised to study the experience of Armenia, Georgia, and Azerbaijan. The farm structure conundrum as formulated in the two bullets above has two dimensions: (a) the organizational-form dimension – individual farms versus corporate farms; and (b) the size dimension – small farms versus large farms. With regard to organizational form, one thing is clear: agricultural production cooperatives are less efficient than individual farms and market-oriented corporate farms. This is suggested by the well-developed theory of cooperatives, but more importantly, this is proved by the almost total nonexistence of production cooperatives in market economies. The Israeli kibbutz is often cited as a counterexample to this proposition. This, however, is a total fallacy: (a) in all the history of Israel there have never been more than 200 kibbutzim – which is really zero in a world perspective; (b) the Israeli kibbutz has been undergoing radical organizational changes since the late 1980s, which include privatization of common property and individualization of common activities; (c) during the decades until the 1970s the kibbutz as a cooperative was held together by deep ideological commitment of its members and by its ability to ensure a higher standard of living than in the city – neither of which is valid today in Israel or has ever been valid for cooperatives in Moldova. We cannot make the same statement regarding market-oriented corporate farms. The plain fact is that corporate farms do exist in market economies (especially in the United States and Canada, much less so in the EU), which proves that they are able to compete with individual farms. Furthermore, the small number of corporate farms that do exist in market economies appear to be even more efficient than individual farms as a group: in the United States, corporate farms control 2% of agricultural land and generate 20% of output (in Moldova, on the other hand, the relation is reversed: corporate farms control 50% of land and generate less than 30% of output; in Russia and Ukraine, corporate farms generate 40% of output on 80% of agricultural land). The market economies have achieved an equilibrium farm structure, which includes a mix of individual farms (the dominant majority) and corporate farms (a small minority) determined by resource availability, managerial capacity, and personal preferences of farmers and investors. A similar process can unfold in Moldova, but the development of corporate farms must be left to market forces, free from government intervention and programming. The second dimension of the farm structure conundrum involves farm sizes – small versus large. There is a voluminous literature on the farm-size effect in developed and developing countries. The results are inconclusive: there is no clear proof that large farms are more productive and more efficient than small farms. A similar result is generally obtained for the transition countries, where studies do not detect any advantage of large corporate farms relative to small individual farms (the best that can be said is that large farms are not inferior to small farms in transition countries). Our analytical results for Moldova based on several surveys demonstrate with considerable confidence that small farms in Moldova are more productive and more efficient than large farms (as measured by both total factor productivity and technical efficiency; the small farms include peasant farms, but do not include household plots). This finding for Moldova receives unexpected support from a recent study of U.S. farms, which has found that an increase 4 of farm size reduces, rather than increases, agricultural productivity (as measured by TFP). The discussion of the farm-structure issue suggests that the government of Moldova should abandon its preference for large-scale corporate farms and concentrate on improving the operating conditions for small individual farms. At the very least, the government should ensure a level playing field for farms of all sizes and organizational forms, and desist from biasing its policies in favor of large farms. The comparative analysis of farm structure in Moldova and the EU countries (as representatives of a market economy) shows that Moldova is characterized by much greater land concentration in large farms than any of the EU countries. Even in countries closest to Moldova, such as Portugal and Greece, the large-farm sector controls a much smaller proportion of land and small farms achieve much greater dominance. To move closer to the farm-structure pattern typical of market economies, Moldova should allow land to flow from large corporate farms to small individual farms, rather than in the opposite direction. This will reduce the concentration of land in large farms, while at the same time increasing the share of land controlled by the small individual farms, and thus bring Moldova in closer conformity with the market pattern of land concentration. At the same time it may correct, at least partially, one of the two manifestations of land fragmentation in Moldova: the average size of the very small individual farms will increase somewhat as they acquire more land at the expense of large corporate farms. This, however, will not affect the other dimension of land fragmentation, which involves fragmentation of land ownership (rather than farm size). We did not have data in this study to examine how fragmentation of small holdings into several (even smaller) parcels affects productivity. In the absence of official data on this subject, a special survey needs to be conducted to capture both fragmentation and production variables for the same sample of farms. Unless such a special survey is conducted, we will not be able to establish with any degree of authority whether fragmentation of holdings into small parcels has a negative effect on productivity or not. It is universally believed that consolidation of small disjointed parcels into contiguous holdings is preferred by farmers and landowners. There can be no conceivable objection to this kind of a consolidation program. It is important to stress, however, that land consolidation should be carried out on a strictly voluntary basis in accordance with clear market principles (as proposed by the team from the Danish Ministry of Agriculture, for instance). Land consolidation programs should supplement market-driven consolidation through buying and selling of land by private entrepreneurs, not replace it. Correction of land concentration and land fragmentation requires development of land markets. Land sale transactions and land leasing have developed rapidly since 1999, mainly thanks to the organizational efforts in the wake of the National Land Program. Although no full statistical information is available, it seems that land sales and land leasing are having a significant effect on consolidation of holdings (but not necessarily consolidation of land ownership). The government should make every possible effort to facilitate this process. Immediate action is required with the objective of facilitating ownership transfers and encouraging consolidation: 5 • • • Simplify the administrative procedures for transfer of ownership (paperwork, number of trips to regional cadastre office, etc.) – this can be accomplished, for instance, by authorizing the primaria secretary to handle the paperwork and to interface with the cadastre office on behalf of the landowner. Reduce transaction costs – by reducing the minimum fee charged by notaries and by devising a system whereby multiple parcels may be treated as one consolidated transaction under appropriate circumstances. Improve availability of market information – the official attitude toward publication and dissemination of information on land transactions must change fundamentally: detailed transaction information must be made available to the public through web sites and regular publications; the Cadastre Agency should collect and publish information on actual land prices; systems should be devised to ensure that information on lease contracts is available centrally for all leases, and not only those for more than 3 years. Finally, a study such as this should primarily rely on official statistical data. It is only in the absence of official data that we should have to turn to various surveys for additional information and insights. Moldovan agricultural statistics, in the form that is available to the public, was unfortunately judged to be inadequate for our purposes, which explains our heavy reliance on private survey data. Moreover, the Department of Statistics proved to be less than willing to help us in the performance of our task by providing access to additional information not included in its official publications. We sincerely hope that in the future the Department of Statistics will show greater understanding to the data needs of the World Bank and more readiness to help. The professional experts in the Department of Statistics are strongly advised to study this report in order to identify the data blocks that should be strengthened (or added) in official statistics. This study would not have been possible without devoted assistance from a group of counterparts in Chisinau: Alexandr Muravschi, Victor Moroz, Anatol Bucatca, Felicia Izman, Valeriu Ginju, and Valery Chodsky. The support from the staff of the World Bank Resident Mission in Chisinau is also gratefully acknowledged. 6 Introduction: The Status of Land Reform in Moldova Land reform in all former Soviet republics, including Moldova, involves the transfer of land from state to private ownership, followed by allocation of individual entitlements to land. Ideally, it also involves restructuring of the traditional large-scale enterprises into market-oriented farming units. The new farming units may represent a wide range of organizational forms, including various associative or corporate structures as well as individual farms. The principles of land reform were developed and formulated while Moldova was still a Soviet republic, but the implementation of these principles was made feasible only with the adoption of the new Land Code and the Law on Peasant Farms (December 1991−January 1992). The Land Code set out the mechanisms for the privatization of land, while the Law on Peasant Farms provided the legal tools for the establishment of individual private farms through the process of exits from collective farm enterprises. Despite an early start, the process of land reform in Moldova advanced very slowly until 1996. The government and parliament lacked political resolve to follow through forcefully on this issue. As a result, administrative support for land privatization and farm restructuring was relatively weak at the beginning of reform in 1992, and the managers of former collective and state farms were reluctant participants in the process of change. The reforms came to a virtual standstill in late 1994 with the passage of laws which imposed additional bureaucratic and technical constraints to the process. Given the lack of clear commitment by the political leadership, the process of land privatization and farm reorganization in Moldova saw only minimum changes, and Moldovan agriculture retained many of the inherited Soviet-era structures. The pace of reforms accelerated after the intervention of the constitutional court, which led to the removal, in February 1996, of the previous legislative constraints. The constitutional court ruling provided an impetus for significant, fundamental changes in the organization of the agricultural sector. The National Land Program (NLP) launched in 1997 with USAID support focused on the assignment of individual titles to land plots carved out from the large collective fields and distribution of collective non-land assets after first resolving the issue of outstanding farm debt. Moldova today is unique among the Commonwealth of Independent States (CIS) countries in its comprehensive approach to land reform, which has combined the processes of land and property distribution with a radical procedure for resolution of the farm debt overhang without resorting to courts. 1 Largely through the NLP initiative, two dimensions of the land reform process in Moldova are essentially completed. These include the distribution of certificates of entitlement to the rural population (formal land privatization) and the subsequent allocation of physical plots to individuals (actual privatization). Considerable progress has also been achieved with farm restructuring, although the share of large corporate farms in Moldova is still much larger than in market economies. The sections that follow explore in detail the progress of land reform and its impacts on farm productivity. The report is organized as follows. We first examine the changes in land 1 For a discussion of farm debt resolution in Moldova see C. Csaki, Z. Lerman, and S. Sotnikov, Farm Debt in the CIS: A Multi-Country Study of the Major Causes and Proposed Solutions, World Bank Discussion Paper 424, World Bank, Washington, DC (2001), pp. 81-113. 7 ownership and land use since 1990; we then describe the results of farm reorganization efforts, including some comparisons with farm structure in market economies (as represented by EU-15); this is followed by an analysis of productivity for large and small farms; the report concludes with a preliminary discussion of land markets based on extremely limited information available at the time of writing. A Note on Terminology and Data Sources The terminology for agriculture in transition is as fluid as the transition itself. Here we try to systematize the terminology used in this study, which largely follows the conventions that are being gradually adopted in the literature. The term private is used to characterize ownership in the legal sense of the word. “Private land” is the opposite of “state-owned land”. It includes land owned by “private” individuals and also land owned by private corporate farms, i.e., all farms that are not state farms. In this sense, privatization is transfer of land ownership from the state to private individuals and private corporations. Privatization of land ownership does not necessarily lead to individual farming. The process that transfers land to individual use is termed individualization. The farm structure is dichotomized by organizational form into individual farms and corporate farms. Individual farms are roughly equivalent to family farms in market economies. In these farms, the farmer (the head of household or the head of family) is both the owner and the manager. Individual farms rely largely on family labor, which is supplemented by hired help as needed. They use mainly own land, which may be supplemented by leased land in growth-oriented farms. In transition economies, the individual farm sector is further subdivided into very small household plots and somewhat larger “peasant farms”. In this study we always make it clear when individual farms include both household plots and peasant farms and when the discussion of the individual sector is limited to peasant farms only. Corporate farms are legal bodies, corporations in the standard sense of the word. They are also called “farm enterprises” or “farming organizations” – terms inherited from Soviet statistics. Corporate farms come in a variety of organizational subforms, which are specified in the Civil Code and in the Law of Enterprises and Entrepreneurship. They are subdivided by ownership into private corporate farms and state owned enterprises. The main organizational forms among private corporate farms are partnerships, limited liability companies, joint-stock (shareholding) societies, and agricultural production cooperatives. These organizational forms are usually referred to as “new corporate forms” because they did not exist under Soviet legislation and began to emerge only after 1991. Traditional corporate forms include state farms as well as kolkhozes – collective farms inherited from the Soviet era and all. Legally, kolkhozes are private corporate farms because they are owned by the members of the collective, not by the state. Traditional corporate farms still exist in Moldova, but their role in agriculture has shrunk from total dominance to almost nil. Corporate farms are owned by shareholders and managed by hired professional managers. In transition countries, including Moldova, the shareholders are typically the local village residents who were formerly members of the local collective farm and received shares in its land and assets. In principle, outside investors may also purchase shares in corporate farms. Corporate farms rely on hired labor. Some of the workers may be shareholders, but they receive a wage for their work, like all hired workers. All shareholders are entitled to dividends from the corporate farm. An important feature of farm restructuring introduced after 1997 is that the new units are not committed to continue employing all the members who have formerly worked on the land assigned to the unit (i.e., the original shareholders). The new units can shed member labor, as long as they continue paying dividends to shareholders or rent to the owners of land that they cultivate. Finally, a few words concerning the terms fragmentation and consolidation, which are at the center of the ongoing policy debate in Moldova. These terms are used in two basically different senses (although consolidation is always the opposite of fragmentation). In the legal sense, fragmentation reflects the fact that the land owned by an individual is split into several parcels in different locations. The process of land privatization in Moldova, because of its equity-driven design, produced fragmentation of land ownership: each individual received on average 8 0.6 hectares of agricultural land divided into three parcels (a parcel of arable land, a parcel of orchards, and a parcel of vineyards). This is fragmentation of land ownership, and in this sense consolidation should involve exchanging original parcels for contiguous parcels (possibly giving up some vineyards and acquiring more arable land instead). In the operational sense, fragmentation reflects the existence of a large number of very small farms, which control a disproportionately small share of agricultural land. In Moldova today, the land use structure is fragmented in both senses. There is a huge proportion of very small farms, and each of these farms is split into three or four parcels. However, it should be made perfectly clear that by correcting fragmentation of land ownership (i.e., consolidating three 0.2 hectare parcels into one contiguous parcel of 0.6 hectares) we will have done nothing to correct fragmentation of farm structure: Moldova will remain with a huge number of very small farms using a disproportionately small share of agricultural land (compared to the situation in established market economies). To correct fragmentation of farm structure and thus ensure that the farm size distribution moves closer to the market pattern, agricultural land policies in Moldova should encourage downsizing of large corporate farms (where the bulk of agricultural land is concentrated) and transfer of land resources to smaller individual farms (thus increasing their average size and reducing the proportion of very small units). Our concern with fragmentation of farm structure (in the second sense of this term) has led to another dichotomization of farms in the empirical part of this study. In parallel with organizational-form dichotomy (individual vs. corporate farms), we use a dichotomy based on farm size: small farms (up to 10-50 hectares) and large farms (in principle more than 50 hectares, but in practice more than 100 hectares). There is considerable overlap between the organizational-form dichotomy and the farm-size dichotomy, and they are often used interchangeably as proxies for one another. Yet the two dichotomies are not identical. Individual farms are typically small farms, but some individual farms fall in the large-size category. Corporate farms are typically large, but some fall in the small-size category. In presenting the analytical results based on the small-large dichotomy we are driven by our interpretation of the fragmentation-consolidation debate in Moldova. The focus among policy makers seems to be on the advisability of eliminating individual farms and creating instead large corporate farms. Nobody is examining the productivity implications of enlarging (even slightly) the small individual farms by transferring some of the land from large corporate farms. Our small-large dichotomy is intended to focus the debate on this direction of farm-size adjustment. The data used in this study are based on official statistics and on various farm surveys conducted since 1997. Official statistics: Statistical yearbooks of Moldova (Department of Statistics) Agriculture in Moldova 2004 (Department of Statistics) Agricultural Activity of Households and Farms – Results of the Statistical Survey, 2002 and 2003 (Department of Statistics) Land balance tables (State Cadastre Agency) Farm surveys: World Bank/ARA survey of farm managers and households (Lerman, Moroz, Csaki), 1967 World Bank baseline survey – preparation of Moldova Agricultural Strategy ( Lerman, Moroz, Izman, Kim), 2000 World Bank survey – cross-country study of reform impacts (Dudwick, Fock, Sedik, Moroz), 2003 USAID/PFAP surveys of peasant farms and farm enterprises (Muravschi, Bucatca), 2003 9 1. Changes in Land Ownership and Land Use Since 1990 The main features of the process of land reform in Moldova are dramatic reduction in state ownership of land, virtually complete distribution of land entitlements (“land shares”) to the rural population, and rapid acceleration in the physical allocation of land plots to rural families. During the Soviet era all agricultural land in Moldova was state-owned (including the “private” household plots cultivated by the rural population). As of January 2004, some 12 years after the beginning of land reform, fully two-thirds of agricultural land is formally classified in private ownership (Figure 1.1). The rest is owned by the state and the municipalities. Private ownership, however, does not mean individual use of land. Half the privately owned land is managed by large-scale corporate farms, which are basically corporate shareholder structures with joint, not individual, cultivation of land (joint stock societies, agricultural production cooperatives, limited liability companies, etc.). Only the other half is used by the individual sector (peasant farms and household plots). The structure of land use as of January 2004 is shown in Figure 1.2. The ownership structure differs from the structure of land use because farms (mainly in the individual sector) cultivate land owned by the state and the municipalities (in addition to privately owned land). Thus, two-thirds of the municipally owned land (Figure 1.1) is in fact allocated to household plots for family farming (Figure 1.2) and only one-third is retained for municipal uses. The process of land privatization has been accompanied by a continuous increase in the share of land in individual use. Since January 2001 the individual sector (independent peasant farms and household plots) manages over 40% of agricultural land in Moldova, double its share in 1997, and the corporate farm sector has lost its former dominance. Among individual farms, peasant farmers operating independently outside large corporate structures control about 30% of agricultural land, up from 8% only three years previously (Table 1.1). The subsidiary household plots also increased their share substantially as a result of the “small privatization” in 1991-1992 and subsequent distributions. The state as the “residual claimant” in the process of privatization retains direct control of 28% of agricultural land, most of it reserve land 2 (Table 1.1, Figure 1.2). The changes in the structure of land use since 1990 are presented in Table 1.1 and Figure 1.3. The state farms practically disappeared during the previous decade, as almost all of them transformed into collectives and received ownership of their land from the state. Many collectives, in turn, registered in new organizational forms. As a result, the structure of land tenure has undergone a dramatic change: in 1990, state farms controlled 30% of agricultural land and collectives another 60%; in 2002-2003, the traditional collectives had all but disappeared, the state farms controlled 9% of agricultural land (after an administratively driven rebound from nearly zero in 2001) 3, and the new 2 The state land reserve established in 1994 was intended to provide a pool of land for redistribution and future uses. The reserve was created by extracting a certain proportion of the land managed by corporate farms, and the first phase of contraction of state land ownership (1990-94) was entirely attributable to this process (see also Figure 1.3). 3 This “rebound” is the result of a large conversion in 2001 of corporate land previously classified as privately owned into state ownership. Nearly 200,000 hectares of privately owned land managed by corporate farms was reclassified as state-owned land. This process is reminiscent of the conversion of collective into state farms frequently practiced during the Soviet era, but we do not have information on the exact reasons for this reclassification. State farms today are mainly seed and livestock selection centers, experimental stations, and educational facilities: they do not engage in large-scale commercial production as in the past. 10 corporate farms that had emerged in the process of reform controlled 30% of agricultural land. The share of land cultivated in large-scale corporate farms (including state, collective, and other corporate farms) declined from 90% in 1990 to 40% in 2002, mainly due to the transfer of land to the individual sector (household plots and peasant farms) and the reserve fund. Private 67% State* 9% Reserve 16% Municipal 8% Figure 1.1. Structure of agricultural land ownership in Moldova (January 2004). Source: State Cadastre; total agricultural land 2.5 million ha. Note: The segment labeled “State” represents state-owned land allocated to agricultural users (both state and private). Corporate 32% State 11% Peasant farms 28% Reserve 16% Households 13% Figure 1.2. Structure of agricultural land use in Moldova (January 2004). Source: State Cadastre; total agricultural land 2.5 million ha. Note: State users include state farms, municipal uses, and “old style” collectives (0.5% of land). 11 Table 1.1. Structure of Land Use 1990-2003 (end of year data, percent of agricultural land)* 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 State sector State farms Reserve land Other state users 32.1 31.4 0.6 0.0 26.7 26.0 0.7 0.0 26.2 24.2 0.6 1.3 32.2 17.7 13.2 1.3 29.8 15.7 13.8 0.3 16.4 2.4 13.9 0.2 17.1 2.9 14.1 0.2 17.4 2.7 14.5 0.2 17.4 2.5 14.6 0.2 17.0 2.2 14.6 0.1 17.8 2.2 15.4 0.2 27.0 8.8 16.7 1.5 27.2 8.8 16.6 1.7 27.4 8.9 16.4 2.1 Corporate forms Collectives New corporate forms 59.5 59.5 0.0 63.0 63.0 0.0 63.0 63.0 0.0 56.5 56.1 0.4 55.4 45.2 10.3 68.0 39.3 28.7 64.7 23.1 41.6 61.4 15.1 46.2 57.2 13.2 44.0 46.0 10.8 35.3 37.5 5.3 32.2 28.8 0.5 28.2 31.0 0.5 30.4 32.5 0.5 31.9 14.8 2.1 12.7 100.0 2556.7 15.6 2.4 13.2 100.0 2556.3 18.1 4.8 13.4 100.0 2555.5 21.2 7.8 13.4 100.0 2555.7 25.5 12.0 13.4 100.0 2556.6 37.0 22.3 14.7 100.0 2550.3 44.7 29.8 14.9 100.0 2543.6 44.2 30.9 13.3 100.0 2538.7 41.9 28.9 12.9 100.0 2533.8 40.1 27.6 12.5 100.0 2528.3 Individual sector 8.5 10.3 10.8 11.3 Peasant farms --0.0 0.0 Household plots 8.5 10.3 10.8 11.3 Total agricultural land 100.0 100.0 100.0 100.0 ‘000 ha 2562.2 2563.6 2559.6 2557.3 *Data include Transnistria. Source: State Cadastre, land balance tables; transposed to end of year. 3000 '000 ha agricultural land 2500 Reserve Households Peasant farms New corporate Collectives State farms 2000 1500 1000 500 0 1990 1992 1994 1996 1998 2000 2002 Figure 1.3. Structure of agricultural land use in Moldova 1990-2003. Source: State Cadastre; transposed to en d of year. 12 “Privatization” of land, i.e., transfer of land to non-state ownership, has stabilized since 1996 (the date of the landmark resolution of the constitutional court). State land consistently accounts for around 35% of agricultural land resources, half of this in reserve land (see Table 1.1). “Privatization” has given way to ongoing changes in reallocation of land between different corporate farms, as these continue to re-register in new organizational forms, and fairly dramatic shifts of resources from corporate to individual farms. Corporate farms as a category lost over 700,000 ha (nearly 30% of all agricultural land) between 1996-2003. This land moved out of the corporate farm sector to the individual sector, and specifically to the subsector of peasant farms, which increased their total land holdings almost ten-fold since 1995. The individual sector continued to grow since 1995 at the expense of the corporate sector, which is clearly demonstrated in Figure 1.3. Creation of peasant farms was not accompanied by a reduction of the reserve fund, which has not changed markedly since its creation in 1994-95. Further reduction of state-owned land will have to come through distribution of reserve lands to individuals. 13 2. Reorganization of the Farm Sector Initially, prior to the launch of the National Land Program (NLP) in 1997-98, the land and asset shares used in privatization were paper certificates, representing the entitlement of each individual to a portion of total land and total assets of the collective farm. Individuals could elect to keep their shares in the restructured farm or leave the collective enterprise withdrawing their land and assets in physical form. NLP shifted the emphasis to physical allocation of land and assets to individuals, including legally binding titling of the land plots. It thus simplified and encouraged the various reconfiguring decisions, be it an exit from the old collective as a private farmer, or participation in a new entity with other members. Creation of Peasant Farms As of the end of 2003, some 650,000 holders of land shares, or about two-thirds of all beneficiaries, had withdrawn nearly 900,000 hectares of agricultural land from large-scale collectives. The growth of land withdrawals between 1991 and 2003 is shown in Figure 2.1. There has been a marked acceleration in the creation of new peasant farms since the launch of the National Land Program in 1998. An individual who withdraws a plot of land from a former collective in principle becomes an independent peasant farmer. In reality, there are about 300,000 registered peasant farms in Moldova today with slightly over 500,000 hectares of agricultural land, which gives an average farm size of 1.9 hectares. 1000 Thousands 800 600 land, ha recipients 400 200 0 1991 1993 1995 1997 1999 2001 2003 Figure 2.1. The potential for peasant farms: allocation of land against land shares 1991-2003. Source: 1991-1998 from State Cadastre land balances; 1999-2003 from Agriculture in Moldova 2004, Table 5.1, p. 120 Land withdrawals and creation of peasant farms between 1999 and 2003 are presented in Table 2.1, which shows that cumulatively 60% of the land withdrawn by individual landowners from corporate farms has been used to establish (registered) peasant farms and the number of registered peasant farmers is about 40% of the number of people allocated land. The remaining landowners in part used their land endowment to augment the household plots or create “unregistered” peasant farms, and in part leased their land to peasant farmers or “leaders”, i.e., managers of new corporate farms. 14 Table 2.1. Land withdrawal from corporate farms and creation of peasant farms 1999 2000 Number of people allocated physical plots against land shares, 429.0 502.7 ‘000 Total land allocated against land shares, ‘000 ha 590.8 701.8 Number of registered peasant farms, ‘000 131.6 201.5 Total land in registered peasant farms, ‘000 ha 285.4 364.1 Average size of peasant farm, ha 2.17 1.81 Ratio of land in registered peasant farms to allocated land 0.48 0.52 Ratio of the number of registered peasant farmers to number of 0.31 0.40 people allocated land Source: Agriculture in Moldova 2004, Table 5.1, p. 120; data transposed to end of year. 2001 565.8 2002 617.0 2003 645.3 805.4 248.3 448.5 1.81 0.56 0.44 836.6 268.4 513.6 1.91 0.61 0.44 867.9 283.2 526.0 1.86 0.61 0.44 There is unfortunately no precise quantitative information on these processes and official sources give widely conflicting numbers on the area of agricultural land in peasant farms. In Agriculture in Moldova 2004, one table (p. 120) gives 526,000 hectares in 283,200 registered peasant farms, while another table on the next page gives 706,700 hectares (34% more!) in an unspecified number of peasant farms (presumably both registered and unregistered). Another official source (Agricultural Activity of Households and Farms in the Republic of Moldova, 2003 issue) gives an even larger figure for land in peasant farms (around 750,000 hectares). Cadastral data summarized in Table 1.2 above correspond to 700,000 hectares in 558,000 peasant farms, which implies an average farm size of 1.3 hectares. Accuracy and consistency of the peasant farm statistics obviously leaves much to be desired. Limited information on the size distribution of peasant farms is available for the three years 20012003 only. This information is summarized in Table 2.2. About 80% of agricultural land is in farms smaller than 50 hectares on average. The rest is controlled by just 300-400 farms (out of the total count of 283,200 registered farms). Table 2.2. Share of land in peasant farms smaller than 50 and 10 hectares 2001 2002 Agricultural land in peasant farms, ‘000 ha 791.7 796.5 % of land in farms smaller than 50 ha 81.9 75.7 % of land in farms smaller than 10 ha n.a. 74.7 Source: Agricultural Activity of Households and Farms in the Republic of Moldova, 2002 and 2003 issues. 2003 743.5 81.7 80.8 Creation of New Corporate Farms Out of more than one million beneficiaries of the privatization process, about 700,000 decided not to switch to independent farming. These shareholders entrusted their shares to “leaders,” i.e., enterprising persons who are willing to manage the land and assets of a whole group of individuals. Some of these “leaders” are truly new rural entrepreneurs, many others are former managers of traditional collective farms. The “leaders” today manage about 1,500 large-scale farms registered mainly as limited liability companies, but also joint-stock societies, agricultural cooperatives, and farmers associations. The traditional forms – collective and state farms – have all but disappeared, and the large-scale farms today are basically represented by new organizational forms. According to the Department of Statistics, over 95% of corporate farms in Moldova are non-traditional structures 15 (Agricultural Yearbook of Moldova 2004, Table 4.13). A full census of different organizational forms in Moldova as of January 2004 is presented in Table 2.3. Table 2.3. Large-Farm Reorganization Number of units as of Jan. 2004 Traditional forms: State farms 75 Collective farms 4 Interfarm cooperatives 7 New forms: Joint stock societies 95 Limited liability companies 1,188 Agricultural cooperatives 111 Farmers associations 44 Total large-scale forms 1,527 Agricultural land, ‘000 ha 994 Average size, ha 650 Source: Land balance tables, State Cadastre Agency. Number of units in 1991 389 600 96 B B B B 1,085 2,274 2,096 The new corporate farms are certainly new in the legal sense. These organizational forms simply did not exist in the Soviet legislation. They were introduced into Moldovan “company law” in January 1992 by the Law on Entrepreneurship and Enterprises and subsequently entered the Civil Code. It is by no means certain however that they are also “new” in terms of their genesis and management style. Many of these “new” corporate farms have been created by straightforward reorganization (and sometimes formal mechanical re-registration) of former traditional collectives. Others are “splinters” of a former collective that divided into two or three smaller components in the process of reform. Certainly some of them are totally new, having been created by new “leaders” enterprisingly configuring the land of new owners. Yet we have no statistics to show what percentage of the new corporate farms are indeed new in the sense that they are not reincarnations of some previous organizations. Moreover, there have been no studies of the comparative performance of the new corporate farms and the traditional collectives in Moldova. The total number of corporate farms in Moldova today is about 1,500, which is 40% more than the number of collective and state farms before the reform. Yet these farms control 40% of the land that they controlled previously, which implies that the process of land reform has produced a noticeable downsizing of the corporate (“leader-managed”) farms in Moldova. The average corporate farm today manages 650 hectares, compared with 2,100 hectares in 1991. According to NLP statistics for September 2000, farms larger than 1,000 hectares manage less than 20% of agricultural land, whereas 35% of land has shifted to a new category of medium-sized corporate farms with up to 500 ha (Figure 2.2). The recent land reform efforts in Moldova have somewhat smoothed out the sharply dual farm structure that characterized socialist agriculture. Contrary to the Soviet era, when the control of agricultural land was polarized between very small household plots and very large collectives with more than 1,000 ha, there is now a mix of organizational forms in the middle range of farm sizes between 100 and 1,000 ha that did not exist previously. 16 30 percent of NLP land 25 20 15 10 5 0 Individuals <100 ha 100-500 500-1000 1000+ Figure 2.2. Distribution of land in “leader-managed” new corporate farms: September 2000. Source: Unpublished performance rep ort, NLP, Chisinau 2001. 17 3. Land Use and Farm Structure: Moldova and EU-15 Prior to 1990, the farm structure in Moldova, as throughout all parts of the Soviet Union, was characterized by extreme duality, with very large collective and state farms at the upper end of the distribution and very small household plots at the other extreme. The changes since 1990 have substantially reduced the size of the large corporate farms (see TABLE 4), while the household plots have practically doubled in size and a new category of medium-sized peasant farms has emerged to smooth out the formerly sharp dual structure. To examine the extent of the of the adjustment in farm structure during transition, it is instructive to compare the emerging farm size distribution in Moldova with that observed in market economies. For purposes of this comparison we have to rely on data provided by various surveys conducted in Moldova since 2000, as official Moldovan statistics do not include information on distribution of farm sizes. Average Farm Sizes Table 3.1 summarizes the available survey results on sizes of farms of different organizational forms (corporate and individual) in Moldova. The corporate farms surveyed range in size mainly from 500 to 1,000 hectares, averaging 700-900 hectares. Peasant farms that fall in the conventional definition of up to 50 hectares have mean sizes of 4-8 hectares; the mean size of the smallest peasant farms (up to 10 ha) is around 2 hectares. The surveys have also captured hundreds of relatively large individual farms with more than 50 hectares. These large peasant farms control hundreds of hectares, but their average size (400 hectares) is about one-half the size of corporate farms. Table 3.1. Mean farm sizes according to various surveys, hectares Corporate farms Mean size Range WB 2000 survey 701 11-6000 Individual farms Mean size Range 8.0 0.8-50 411 50-1800 WB 2003 survey 971 50-5600 4.5 0.6-20 PFAP 2003 survey 917 14-7300 4.0 0.3-50 397 50-2300* Department of Statistics 650-800** 1.85# < 10 ha *While these farms are sampled with the corporate farms, they are formally registered as peasant farms. **The lower number is based on land balance tables and includes Transnistria; the higher number is calculated from Agricultural Yearbook of Moldova 2004 and is biased due to the fact that agricultural land includes Transnistria, while the number of farms is without Transnistria. #Agricultural Activity of Households and Farms in the Republic of Moldova in 2003 – Results of the Statistical Survey, Department of Statistics, Chisinau (2004), p. 12. Land economists always maintain that there is no one optimal farm size for all countries, and that the average farm size is necessarily a product of national land endowment and rural population density. In this sense, the relatively densely populated and land-poor Europe is a much more appropriate comparison for Moldova than North America, with its sparsely populated huge expanses of agricultural land (similar to Russia). We accordingly compare the average farm sizes and farm size distribution in Moldova to those in the 15 original member countries of the European Union (EU-15). 18 In EU-15 taken as a group, the average farm size is 18.7 hectares, but 58% of the farms have less than 5 hectares (2000 data). Considered country by country, the EU-15 show considerable variability in average farm sizes (Table 3.2). In three of the 15 countries (Greece, Italy, and Portugal) the average farm size is less than 10 hectares; in five other countries the average size is between 10 and 30 hectares; in six countries it is between 30 and 50 hectares; and only one country (UK) has farms with nearly 70 hectares on average. Overall, we get the distinct impression that, despite the cross-country variability, the average farm in EU-15 is not much larger than the average peasant farm in Moldova, while on the other hand it is much smaller than the average corporate farm in Moldova. Within Europe, the relevant comparison for Moldova is Greece, Italy, and Portugal, not the UK or France. Given these benchmarks, the average size of individual farms is Moldova does not appear too small at all compared with farms in Greece, Italy, and Portugal. Table 3.2. Average farm size in EU-15 in 2000 Hectares EU-15 (weighted mean) 18.7 Greece 4.4 Italy 6.1 Portugal 9.3 Austria 17.0 Netherlands 20.0 Spain 20.3 Belgium 22.6 Finland 27.3 Ireland 31.4 Germany 36.3 Sweden 37.7 France 42.0 Luxembourg 45.4 Denmark 45.7 UK 67.7 Unweighted mean for EU-15 28.9 Source: Agriculture in the European Union: Statistical and Economic Information 2002, EU Directorate-General for Agriculture (February 2003). Data on the distribution of farm sizes by country for the EU-15 are published in five size groups ranging from the smallest farms with up to 5 hectares to the largest farms with more than 50 hectares (unfortunately, no such data are available for Moldova from official sources). In the context of the fragmentation concerns in Moldova, the farm size distribution can be conveniently characterized by the percentage of the smallest farms (up to 5 hectares). Figure 3.1 shows the percentage of the smallest farms (those of up to 5 hectares) in the total number of farms in each country and in EU-15 as a group. In four of the EU-15 countries, between 60% and 80% of all farms fall in the smallest category (2000 data). These countries are Portugal, Italy, Greece, and Spain. In seven other countries the smallest farms account for between 20% and 40% of all farms. Finally, in the last four countries (Sweden, Finland, Ireland, and Denmark) the percentage of the smallest farms is 10% and less. 19 The distribution of farm sizes in EU-15 did not change much from 1993 to 2000. The changes in the weight of this category between 1993 and 2000 shown in Figure 3.1 were generally very slight. A certain tendency toward consolidation (i.e., elimination of the smallest farms) is observed only in Germany, Luxembourg, and Belgium, and to a smaller extent in Austria, the Netherlands, and Ireland. In the UK – the country with the largest average farm size – the percentage of the smallest farms actually increased substantially, rising from 15% in 1993 to 23% in 2000. EU-15 Portugal Italy Greece Spain Austria Netherlands Belgium France Germany UK Luxembourg Sweden Finland Ireland Denmark 2000 1993 0 20 40 60 80 100 % of farms <5 ha Figure 3.1. Percent of smallest farms with up to 5 hectares in EU-15 countries 1993-2000. Source: see Table 3.2. Land Concentration The sharp difference in average sizes between individual and corporate farms is reflected in fairly strong concentration of land in the largest farms in the sample – a feature inherited from the sharply dual farm structure of the Soviet period. The Lorenz curve provides a standard tool for visualizing inequality of land distribution between large and small farms. Plotting the cumulative percent of the number of farms (from smallest to largest) on the horizontal axis and the cumulative percent of agricultural land used by farms on the vertical axis, we obtain a curve whose downward bulge below the diagonal provides a measure of inequality or concentration. In the absence of a country-wide size distribution for all farms in Moldova, we produced a “sample” Lorenz curve ordering by size the farms in all three 2003 surveys (the WB 2003 surveys and two PFAP surveys, n = 1885). The Lorenz curve for Moldova is shown in Figure 3.2 (black curve), where we see that 70% of the smallest farms (mostly individual farms) account for just under 1% of land holdings while the remaining 30% of larger farms (corporate farms plus a substantial number of relatively large individual farms) account for 99% of land holdings. At the top end of the distribution, just 5% of the largest farms (practically all of them large corporate farms) control more than 50% of land. 20 Figure 3.2 and the supporting Table 3.3 contain the only available information on land concentration in Moldova. It is not entirely appropriate to compare the somewhat sporadic sample results for Moldova with the situation in the EU countries, where the corresponding data are systematically collected by Eurostat for whole countries. Bearing this caveat in mind, we have nevertheless superimposed on Figure 3.2 the aggregated land concentration curve for the 15 countries of the EU (gray curve). In the 15 countries of the EU combined, 10% of the largest farms control 64% of agricultural land compared with as much as 73% in Moldova (Table 3.3). On the other hand, the small-farm tail in EU-15 is much thicker than in Moldova, with 80% of the smallest farms controlling 17.5% of agricultural land compared with only 6.4% in Moldova 100 percent of ag land 80 60 40 EU-15 20 Moldova 0 0 20 40 60 80 100 percent of farms Figure 3.2. Land concentration curves for Moldova (2003) and EU-15 (2000). Source: Moldova based on three 2003 surveys; EU-15 see Table 3.2. Table 3.3. Land concentration in Moldova compared with Portugal and EU-15: cumulative percent of the number of farms and of agricultural land by increasing farm size EU-15 Moldova Portugal Number of farms Agricultural land Number of farms Agricultural land Number of farms Agricultural land 60 0.5 60 5.2 70 0.9 70 9.8 75 2.3 80 13.8 80 6.4 80 17.5 85 14.0 90 21.4 90 26.7 90 36.1 95 29.6 95 47.3 97.5 39.9 97.5 64.5 100 100.0 100 100.0 100 100.0 Source: Moldova from three 2003 surveys; Portugal and EU-15 see Table 3.2. 21 Figure 3.3 presents the corresponding curves for three EU countries with different levels of land concentration: low for Ireland, medium for Belgium, and high for Portugal. Comparing the two figures we immediately see that even in Portugal, the country with the highest land concentration in the EU, the concentration of land in the largest farms is much less extreme than in Moldova. In Moldova 80% of the smaller farms control 7% of agricultural land, whereas in Portugal the same percentage of small farms control fully 14% of agricultural land (Table 3.3). As a result, 20% of the largest farms control 93% of land in Moldova and 86% of land in Portugal. 100 % land 80 60 Ireland 40 Belgium 20 Portugal 0 0 20 40 60 80 100 % farms Figure 3.3. Land concentration curves for three EU countries 2000. Source: see Table 3.2. The observed results for Moldova fall somewhere between the market model and the former Soviet model: the land concentration is not as extreme as in Russia and Ukraine, which are still very close to the former Soviet model characterized by sharply dual farm structure, but it is substantially more pronounced than in the EU (and also in the US and Canada). To move closer to the market pattern, Moldova has to undergo further farm size adjustment. The prescription for this adjustment, however, is totally different from that advocated by the Ministry of Agriculture: our analysis clearly demonstrates that continued progress toward the market pattern of farm sizes in Moldova requires further significant downsizing of the large corporate farms and further transfer of land to small individual farms. Land should be allowed to flow from the largest corporate farms to the smaller individual farms, shifting the relative weights of the two farm sectors in the direction of the equilibrium observed in established market economies. 22 4. Productivity and Efficiency of Small and Large Farms in Moldova Fragmentation of holdings due to land privatization and the advisability of implementing administrative measures to encourage consolidation and re-creation of large-scale corporate farms are at the center of the ongoing policy debate in Moldova. In this section we present empirical evidence to show that small individual farms achieve higher productivity and efficiency than large corporate farms. The productivity analysis is carried out on several data sets. First, we use official national-level statistics to calculate partial and total productivity measures of individual and corporate farms. Then we analyze four surveys carried out by the World Bank and PFAP in 2000 and 2003, where the data can be dichotomized by farm size (small vs. large farms). Nationally, corporate farms are large farms, whereas individual farms (including peasant farms and household plots) are small farms. The organizational form dichotomy in the national-level analysis is therefore a good proxy for the farm-size dichotomy in the analysis of the surveys. The different datasets all consistently show that the productivity and efficiency of small individual farms is significantly higher than the productivity of large corporate farms. Partial productivity measures: changes in productivity of land and labor over time 4 The continuing shift of agricultural land from corporate to individual farms has produced a dramatic change in the structure of land use by agricultural producers. Back in the early 1990s, corporate farms controlled 90% of the agricultural land used by agricultural producers (excluding various components of reserve land). The individual sector managed the remaining 10% (Table 4.1). Since 1999-2000, the agricultural land resources are evenly divided between corporate and individual farms, with each sector controlling about 50% of the total (excluding reserve land). The changes in land use by agricultural producers since 1990 are illustrated in Figure 4.1. This figure incidentally highlights a possible reversal of the previous trend: although the two sectors of corporate and individual farms each control about 50% of agricultural land, there has been a slight decrease in the share of individual farms and a slight increase in the share of corporate farms since 2000. This is in clear opposition to the trend observed between 1990 and 1999, but we have to wait for more observations in the future to determine if there indeed has occurred such trend reversal in land use or we are simply witnessing stabilization at a new level. The significant changes in land use have naturally affected the production structure of agriculture (Table 4.1). While the output of large collective and corporate farms declined through a complex combination of factors that included loss of land and disruption of the old economic order, the output of the individual sector (including peasant farms and household plots) has been growing (Figure 4.2). In 1998, the individual sector overtook the collective and corporate sector by volume of production. As of 2003, the individual sector, with about 50% of total agricultural land, produces more than 70% of agricultural output (Table 4.1). 4 The analysis in this section is based on time-series data from official statistical publications (Statistical Yearbook of Moldova for various years and Agriculture in Moldova 2004). 23 Table 4.1. Land, output, and labor by farm type 1990-2003 Agricultural land used by farms* Gross Agricultural Output Employed in agriculture Million lei, 2000 Corporate, Individual, Corporate, Individual, ‘000 Corporate, Individual, ‘000 ha prices % % % % workers % % 1990 2545.8 91.5 8.5 17757 79.6 20.4 678 83.2 16.8 1991 2544.9 89.6 10.4 15749 76.3 23.7 743 75.8 24.2 1992 2509.5 89.0 11.0 13311 70.6 29.4 749 74.1 25.9 1993 2187.3 86.8 13.2 14647 62.5 37.5 730 73.1 26.9 1994 2196.6 82.7 17.3 11086 58.1 41.9 767 69.6 30.4 1995 2196.4 81.9 18.1 10293 55.5 44.5 771 69.2 30.8 1996 2191.3 78.9 21.1 9071 53.8 46.2 711 67.4 32.6 1997 2181.2 75.1 24.9 10108 54.4 45.6 684 63.2 36.8 1998 2177.8 70.1 29.9 8935 42.8 57.2 750 48.5 51.5 1999 2173.8 56.6 43.4 8184 32.8 67.2 731 33.8 66.2 2000 2146.7 47.1 52.9 7917 29.0 71.0 766 23.1 76.9 2001 2076.0 46.0 54.0 8427 28.0 72.0 764 20.7 79.3 2002 2069.2 48.7 51.3 8717 29.0 71.0 747 20.6 79.4 2003 2059.8 50.7 49.3 7535 25.0 75.0 583 23.9 76.1 *End of year data; land used by farms is agricultural land excluding the areas not allocated to agricultural producers (the state reserve, miscellaneous state lands, and part of the municipal land not allocated to agricultural producers). 100 % of land used in farms 80 60 Corporate Individual 40 20 0 1990 1992 1994 1996 1998 2000 2002 Figure 4.1. Land use by agricultural producers in Moldova 1990-2003. Source: State Cadastre; end of year data. 24 16 million lei, constant prices (2000) 14 12 10 Individual Corporate 8 6 4 2 0 1990 1992 1994 1996 1998 2000 2002 2004 Figure 4.2. Gross agricultural product for individual and corporate farms 1990-2004. Source: Statistical Yearbooks of Moldova 1999, 2004. The discrepant shares of the individual sector in land and output can be applied to calculate the socalled relative productivity of individual farms. Taking the average sector productivity as 1.0 (with 100% of land producing 100% of output), we obtain 1.4 (=70%/50%) for the relative productivity of land used by individual farms, compared with only 0.6 (=30%/50%) for corporate farms. Already these rough calculations show that the small individual farms use their land more productively than the large corporate farms. This phenomenon has persisted since 1990, as the share of individual output has always been greater than the share of land in individual tenure (Table 4.1). Agricultural labor is the second main resource that affects performance of the agricultural sector. The total number of employed in agriculture (including hired labor, members of cooperatives and shareholder farms, and self-employed) remained fairly stable at 700,000-750,000 between 1990 and 2002 (Table 4.1; the reported number of employed in agriculture dropped by more than 20% in 2003, but the reasons for this may be purely technical and forthcoming revisions will hopefully clarify the situation). Yet the steady overall picture of agricultural employment hides dramatic changes in farms of different types (Figure 4.3). The agricultural labor in corporate farms decreased precipitously, especially between 1995 and 2001, while that in individual farms increased sharply, especially after 1996, following the influx of agricultural land into the individual sector. In farms of both types the changes in agricultural labor use are strongly correlated with the changes in agricultural land use (the coefficient of correlation is greater than 0.95 for 1990-2003). The opposite employment trends in corporate and individual farms have resulted in a sharp increase of the share of agricultural labor in the individual sector – from 25% in the early 1990s to more than 75% in 20002003 (see Table 4.1). Given the value of agricultural output in constant 2000 lei (Table 4.1), we can calculate the partial productivity of land and labor in absolute terms. The results are presented in numerical form in Table 4.2 and graphically in Figures 4.4 and 4.5. The productivity of land and the productivity of labor decrease over time in both corporate and individual farms. However, despite the similar trends, the productivity of individual farms is generally higher than the productivity of corporate farms. The land productivity of individual farms is higher in each and every year between 1990 and 2003. The 25 labor productivity is higher in 11 of the 14 years: the exception is the period 2000-2002, when the labor productivity of corporate farms increased due to signs of increasing output (Figure 4.2) combined with continuing decrease of labor in these years (Figure 4.3). 800 '000 workers percent individual 100 75 600 400 50 200 25 Corporate Individual %Individual 0 0 1990 1992 1994 1996 1998 2000 2002 2004 Figure 4.3. Agricultural employment in individual and corporate farms: thousands of workers (solid curves) and share of individual farms in total employment (bars). Source: Department of Statistics; number of employed in individual farms calculated as the difference between total number of employed and number of employed in corporate farms. Table 4.2. Land and labor productivity for corporate and individual farms Productivity of land, ‘000 lei/ha Productivity of labor, ‘000 lei/worker Year Corporate Individual Corporate Individual 1990 6.1 16.8 25.0 32.0 1991 5.3 14.2 21.3 20.8 1992 4.2 14.2 16.9 20.2 1993 4.8 18.9 17.2 27.9 1994 3.5 12.2 12.1 19.9 1995 3.2 11.5 10.7 19.3 1996 2.8 9.0 10.2 18.1 1997 3.4 8.5 12.7 18.3 1998 2.5 7.9 10.5 13.2 1999 2.2 5.8 10.9 11.4 2000 2.3 4.9 13.0 9.5 2001 2.5 5.4 14.9 10.0 2002 2.5 5.8 16.5 10.4 2003 1.8 5.5 13.5 12.7 Ave 1990-2003 3.4* 10.1* 14.7 17.4 Ave 1990-1996 4.3* 13.8* 16.2* 22.6* Ave 1997-2003 2.4* 6.3* 13.1 12.2 *The differences between corporate and individual farms significant at p < 0.1 by both parametric and non-parametric tests. Source: Calculated from Table 4.1. 26 The land productivity of individual farms is statistically significantly higher than that of corporate farms. 5 The difference in labor productivity, on the other hand, is not statistically significant, although the mean for the entire period 1990-2003 is observed to be higher for individual farms (Table 4.2). In other transition countries we also observe that the productivity of land is higher for individual farms, but the productivity of labor is actually higher for corporate farms. For Moldova the labor productivity of corporate farms is indeed higher in the later subperiod 1997-2003, but again the difference is not statistically significant (see Table 4.2). Thus, the two partial productivity measures for land and labor do not give a consistent picture: while land productivity is definitely higher for individual farms, the results for labor productivity are ambiguous (and furthermore do not fit the results in other transition countries, where labor productivity is typically lower for individual farms). To resolve the ambiguity, we have to calculate a measure of Total Factor Productivity (TFP), which is the ratio of total output produced (in money units) to the total bundle of inputs used (also in money units). TFP calculations using various databases are presented in the following sections. 20 '000 lei/ha (2000 prices) 15 Corporate Individual 10 5 0 1990 1992 1994 1996 1998 2000 2002 Figure 4.4. Land productivity for individual and corporate farms 1990-2003 (absolute values in constant 2000 prices). Source: Author’s calculations (Table 4.1). 5 A caveat is in order in connection with land productivity calculations: the land data cover all of Moldova, including Transnistria (for the entire period 1990-2003); the agricultural output data are reported for Moldova without Transnistria (since 1995). To correct for the resulting bias, we recalculated the land productivity using the agricultural land series without Transnistria since 1995 (a rough approximation due to lack of authoritative data for Transnistria). The new results did not differ much from the original numbers: the mean productivity of land for corporate farms increased from 3.4 to 3.7, and that for individual farms from 10.1 to 10.2 27 35 '000 lei/worker (2000 prices) 30 25 20 Corporate Individual 15 10 5 0 1990 1992 1994 1996 1998 2000 2002 Figure 4.5. Agricultural labor productivity for individual and corporate farms 1990-2003 (absolute values in constant 2000 prices). Source: Author’s calculations (Table 4.1). One of the features that clearly emerges from Figures 4.4 and 4.5 is the general decline of agricultural productivity since 1990 for farms of all types. The ongoing reforms have not produced significant productivity improvements after the initial shock. The labor productivity index constructed for the entire agricultural sector (combining both corporate and individual farms – Figure 4.6, thick black curve) shows signs of slight recovery after 2000, when the reforms accelerated with the introduction of the National Land Program. The labor productivity index bottomed out in 2000 at 40% of 1990, increasing to 50% of 1990 in 2003. 160 1990=100 140 120 100 Moldova CIS CEE 80 60 40 20 0 1990 1992 1994 1996 1998 2000 2002 Figure 4.6. Agricultural labor productivity for Moldova, CIS, and CEE (index numbers, percent of 1990). Source: Author’s calculations based on Table 4.2 for Moldova, Official Statistics of CIS Countries, CDROM 2004-9 for CIS, and official country statistics from statistical yearbooks for CEE. 28 This pattern for Moldova is not different from the pattern observed for the CIS countries as a group, where agricultural labor productivity declined up to 1998 and showed slight signs of recovery thereafter (Figure 4.6, thick gray curve). However, the productivity loss in Moldova was initially greater than the CIS average and the recent recovery still lags behind the average. The productivity recovery in Moldova is attributable to the reported gains in agricultural production since 2000 (see Table 4.1), which in turn appear to be associated with the intensification of land reform after the introduction of NLP. The behavior of agricultural labor productivity over time in Moldova and CIS is totally different from what we observe in Central Eastern Europe, where labor productivity has been increasing since 1994 through a combination of dramatic reduction of agricultural labor (at least in some countries) and a certain (though by no means dramatic) growth in agricultural output (Figure 4.6, thin black curve). In established market economies, such as the United States, Canada, and the European Union, agricultural labor productivity shows a steady growth over time as agricultural employment shrinks while output grows due to technological change. Thus, USDA data show that between 1990 and 1999 agricultural labor in U.S. farms decreased by 4% while agricultural output increased by 20%, producing an increase of 24% in agricultural labor productivity (U.S. Level Tables: Inputs, Outputs, and Productivity, 1948-99, www.ers.usda.gov/Data/AgProductivity/). The CEE countries in fact matched this productivity growth during the corresponding period, while CIS dropped to 60% of the 1990 level and Moldova’s productivity declined even more to an abysmal 40% of the 1990 level (see Figure 4.6). Total Factor Productivity Partial productivity measures reflect the use of a single input (land or labor) taken separately. They often present an ambiguous picture, as some farms may have a higher productivity of land (say) and a lower productivity of labor. The ambiguity is resolved by switching from partial productivity to total factor productivity (TFP), which is calculated as the ratio of the value of output to the aggregated cost of input use. In the absence of market prices for valuing the cost of inputs (such as the price of land), TFP is usually determined by estimating a production function and then using the estimated input coefficients as the weights to calculate the value of the bundle of inputs. The ratio of the observed output to the estimated bundle of inputs is the TFP. 6 The national-level database for Moldova contains information on the value of agricultural output (in constant 2000 lei) and the quantities of two main inputs: agricultural land and agricultural labor. These data are available for 14 years 1990-2003 for individual and corporate farms separately. Unfortunately, our attempt to estimate a two-input production function of Moldovan agriculture from these data has failed due to relative shortness of the time series: the coefficients 6 In principle, the production function can be estimated for any number of observed inputs. In the economic literature, however, TFP is typically calculated assuming two inputs: capital and labor. We have decided to follow the same approach from considerations of data reliability, which also suggested using only land as a proxy for capital (ignoring the extremely deficient data on farm machinery and buildings). The physical variables (land area and number of workers) were judged to be much more reliable and consistent than the accounting figures for other factors of production, such as the cost of purchased inputs and the value of fixed assets (especially for individual farms). For a calculation of TFP as the ratio of output to the reported cost of inputs see the World Bank draft report by N. Dudwick, K. Fock, and D. Sedik (February 2005). 29 of both land and labor turned out statistically not significant. To get a qualitative picture of TFP changes over time despite the estimation failure, we assumed a conventional Cobb-Douglas production function with stylized factor shares of 0.7 for land and 0.3 for labor (these are the factor shares that we consistently obtained in production functions estimated using various farm surveys in Moldova – see below). Figure 4.7 presents the TFP results calculated with these land and labor weights. The TFP for individual farms is higher than for corporate farms over the entire period 1990-2003. The respective means for 1990-2003 are 11.5 for individual farms and 4.4 for corporate farms (the difference is statistically significant). 25 lei output per unit aggregated inputs 20 15 Individual Corporate 10 5 0 1990 1992 1994 1996 1998 2000 2002 Figure 4.7. Total factor productivity for individual and corporate farms 1990-2003 (inputs from Table 4.1 aggregated using hypothetical factor shares of 0.7 to land and 0.3 to labor). To substantiate these simulated findings, we turned to the data collected in various farm surveys in Moldova since 2000. These surveys provide much larger samples and allow fairly reliable estimation of production functions. Yet the survey data are inherently cross-sections observed at a certain point in time and do not generate the time-series perspective afforded by national-level statistics. Extensive data for small and large farms are available from four surveys: the World Bank 2003 survey conducted as part of a cross-country study of reform impacts; the PFAP 2003 survey of corporate farms; the PFAP 2003 survey of individual farms; and the World Bank 2000 baseline survey conducted as part of the preparation work for the Moldova Agricultural Strategy. 7 The sample structure of the four surveys is shown in Table 4.3. 7 General analyses of these surveys can be found in the following unpublished reports: N. Dudwick, K. Fock, and D. Sedik, A Stock-Taking of Land Reform and Farm Restructuring in Bulgaria, Moldova, Azerbaijan, and Kazakhstan (World Bank, February 2005) for the WB 2003 survey; A. Muravschi and others, Efficiency of the Agricultural Sector in the Post-Privatization Period (USAID/PFAP, Chisinau, 2004) for the two PFAP 2003 surveys; and Z. Lerman, Moldova: Baseline Farm Survey October-November 2000 (World Bank, April 2001). 30 Table 4.3. Structure of recent farm surveys in Moldova Small individual farms WB 2003 survey 176 PFAP 2003 surveys 1,166 WB 2000 baseline survey 170 Large corporate farms 22 521 84 Large individual farms --96 Table 4.4 presents the size characteristics and the partial productivity measures for small and large farms in the four surveys. While the large farms as a group are substantially larger than the small farms by all measures – output, land, and labor, the partial productivities show a mixed picture: • • • The partial productivity of land (output per hectare) is higher for small individual farms than for large corporate farms The partial productivity of labor (output per worker) is lower for small individual farms than for large corporate farms The number of workers per hectare is much higher in small individual farms than in large corporate farms (the “labor sink” effect of individual farms). The results are consistent with national-level findings (Table 4.2). We attempt to resolve the ambiguity in partial productivity measures by calculating total factor productivities (TFP). Table 4.4. Size characteristics and productivity measures for small and large farms: survey data WB 2003 survey PFAP 2003 surveys WB 2000 baseline survey Small farms Large Small farms Large Small farms Large farms (individual) farms (individual) farms (individual) (corporate (corporate) (corporate) and individual) Number of observations 176 22 1,166 521 170 180 Ag land (ha) 4.48 971 4.02 918 5.7 533 Workers 4.51 332 6.27 150 1.6 43.7 Ag output (‘000 lei) 25.8 3,230 25.3 2,038 75.4 1,642 Output/ha (lei) 6,765 2,745 9,535 2,085 6,414 3,145 Output/worker (lei) 6,857 17,135 5,145 17,824 55,304 54,393 Workers/ha 1.42 0.26 3.25 0.19 Note: All differences between small and large farms are statistically significant at p = 0.1 (except the differences in productivity of labor – output/worker – in the WB 2000 survey). TFP by dummy variable estimation Differences in TFP between categories of farms can be captured by estimating appropriate production functions with a dummy variable for different farm types. If the dummy coefficient for type A farms is found to be greater than for type B farms, this implies that type A farms produce a greater value of output at any given bundle of inputs and essentially means that type A farms have higher TFP than type B farms. This procedure enables us to assess differences in TFP without actually calculating the TFP in absolute values. Simple two-input Cobb-Douglas production functions, relate the aggregated value of output to agricultural land and agricultural labor, were estimated for two datasets: the WB 2003 survey on its own (198 farms classified into small and large) and the pooled dataset combining the WB 2003 survey with the PFAP survey of corporate farms (521 additional observations on corporate 31 farms). 8 The two-input production functions were first estimated for both datasets without dummy variables (Models 1 and 1P in Table 4.5). In both samples, the coefficients of the two factors of production (land and labor) summed to less than 1, and the difference from 1 was statistically significant at p = 0.10. The production function thus shows decreasing returns to scale: large (corporate) farms produce less per unit of inputs in the margin than small (individual) farms (and this result is statistically significant). Table 4.5. Estimation of Cobb-Douglas production function for large and small farms: WB 2003 survey and pooled sample Dependent variable: value of output Model 1 Model 1P Model 2 Model 2P (lei) coefficients coefficients coefficients coefficients Explanatory variables: Land (ha) 0.60 0.58 0.69 0.75 Labor (workers) 0.30 0.39 0.31 0.33 Farm type (dummy): large farms relative to small farms ---0.58 -0.84 Sum of input coefficients 0.90 0.97 n.a. n.a. R2 0.770 0.911 0.773 0.917 Number of observations 198 719 198 719 Estimation sample WB survey Pooled WB survey Pooled This conclusion is strengthened and quantified by estimating the same two-input production function with a dummy variable for large (corporate) and small (individual) farms (Models 2 and 2P in Table 4.5). The intercept for large farms (relative to small farms) is negative, which means that at each level of inputs (land and labor) large corporate farms attain lower output than small individual farms (the negative coefficient was statistically significant at p = 0.10). The mathematics of the Cobb-Douglas production function translates the negative dummy variable coefficient of -0.58 obtained in the 2003 survey into a difference of 45% in output between corporate and individual farms for each bundle of inputs (1 - exp(-0.58) = 1 - 0.55 = 0.45). In the pooled sample, the gap is even greater (57%). TFP calculated from production function coefficients The estimated production function provides another technique for calculating the TFP in absolute values for different groups of farms. As we move from the small individual farms to the large corporate farms, the agricultural product increases, but so do the land endowment and the labor force (see Table 4.4). The production function is a mathematical relationship that links the increase in agricultural product with the increase in aggregated input use. The inputs are aggregated by applying the weights (or factor shares) from the corresponding production function to specific values of the inputs (land and labor in our case). TFP is calculated as the aggregated value of output divided by the aggregated value of inputs. In this sense it is similar to the standard partial productivity measures, in which the aggregated value of output is divided by the quantity of a single input (land or labor). 8 We decided not to pool the 1,166 individual farms from the PFAP sample with the rest because their large number would overwhelm the much smaller WB 2003 sample. An alternative approach would be to pool the two PFAP samples – 1,166 individual farms with 512 corporate farms, leaving the WB 2003 sample aside. 32 Table 4.6 presents the estimated production function coefficients and the weights used in TFP calculations. Two features are worth highlighting in these numbers. First, in all two-input production functions agricultural land accounts for around 70% of input use and labor for 30% (see the rows for input weights). Second, mixed samples of individual and corporate farms (WB 2003 and WB 2000) as well as the sample of pure individual farms (PFAP) reveal decreasing returns to scale (the sum of the estimated coefficients is significantly less than 1). Corporate farms taken on their own (PFAP sample of corporate farms) reveal increasing returns to scale (the sum of the estimated coefficients is significantly greater than 1). These issues are discussed in some detail in a separate section. Table 4.6. Regression coefficients and input weights in production functions estimated for three samples WB 2003 survey PFAP individual PFAP corporate WB 2000 survey (n = 198) farms (n = 1166) farms (n = 521) (n = 268) Estimated coefficients: Ag land 0.6007 0.5247 0.8150 0.6305 Workers 0.2993 0.1865 0.3068 0.2325 Sum of coefficients 0.90 0.71 1.12 0.86 R2 0.77 0.40 0.84 0.89 Input weights: Ag land 0.67 0.74 0.73 0.73 Workers 0.33 0.26 0.27 0.37 Note: The estimated coefficients are significantly different from zero (p< 0.01); all sums of coefficients significantly different from 1. The mean TFP values obtained by this method for small and large farms in the four survey samples are presented in Table 4.7 and Figure 4.8. Small individual farms attain consistently higher TFPs than large corporate farms. The differences are statistically significant at p = 0.1. The TFP calculations thus eliminate the ambiguity between the partial productivities of land and labor for large and small farms and conclusively show that small farms use their resources more productively than large farms. 9 This finding is consistent with the indication of decreasing returns to scale in the production function. Table 4.7. TFP (lei per aggregated unit of inputs) Small (individual) farms WB 2003 survey 6,426 PFAP surveys 7,424 WB 2000 survey 8,420 9 Large (corporate) farms 4,745 3,464 4,010 Large-to-small ratio 0.74 0.47 0.48 Our results for the relative TFP of corporate and individual farms are not too far from the result of Dudwick, Fock, and Sedik (World Bank, February 2005), who calculate the TFP as the ratio of the value of output to the accounting value of total costs. The TFP of corporate farms in Dudwick et al. (Table 5) is 30% of the TFP for individual farms, whereas our results give around 45% (by dummy variable analysis for the WB 2003 sample and by input aggregation for the pooled sample). 33 10 '000 lei 8 6 WB2003 PFAP WB2000 4 2 0 Small farms Large farms Figure 4.8. Total factor productivity for farms of different types. Source: Based on two World Bank surveys (2000, 2003) and two PFAP surveys (2003). Figure 4.9. Total factor productivity versus farm size (WB 2000 survey). The dichotomized productivity comparison between large and small farms in Table 4.7 and Figure 4.8 was strengthened by checking the relationship of TFP with farm size as a continuous variable (measured in hectares of agricultural land). This analysis was carried out only in the WB 2000 survey, where individual farms span the entire spectrum of farm sizes and are not limited to the small-size range as in the other samples. Consistently with the previous findings, TFP decreases with increasing farm size (see the regression line for TFP in Figure 4.9; the relationship is highly significant). These results corroborate the previous conclusion of decreasing returns to scale. 34 Our findings that TFP is higher for small farms (Figure 4.8) and that TFP decreases with increasing farm size (Figure 4.9) are reinforced by recent results for U.S. farms. 10 Using a time series of labor and capital in U.S. farms for 1978-1996, the researchers have found that an increase of farm size reduces, rather than increases, agricultural productivity (as measured by TFP). “This suggests that on average a type of diseconomies of size is operating [in U.S. farms]” (p. 20). Technical Efficiency of Small and Large Farms The efficiency of input use for a particular farm is measured in relation to the production frontier, which is the locus of “best attainable” points, i.e., points where the maximum output is achieved for every given bundle of inputs. Once the production frontier has been constructed, we can calculate the technical efficiency of each farm by measuring its relative distance from the frontier. Points on the frontier are technically efficient; their distance from the frontier is 0, and their technical efficiency (TE) score is 1. As the distance of a particular point from the frontier increases, its TE score decreases. Each TE score is a number indicating the output that a particular farm achieves with a given bundle of inputs as a fraction (or a percentage) achieved by the “best performer” with the same bundle of inputs. Stochastic Frontier Analysis (SFA) is a production frontier technique that is conceptually close to production function estimation. This is an econometric method that starts with the production function and then iteratively shifts it outward by a certain algorithm until a production frontier is obtained. The actual observed points generally fall below the frontier (in this sense they are inefficient). The TE scores are calculated by taking the ratio of the actual output of each farm (adjusted for random errors) to the stochastic frontier output for the corresponding bundle of inputs. Table 4.8. TE scores obtained by Stochastic Frontier Analysis (SFA) for 2003 surveys WB 2003 survey (n = 198) Pooled database (n = 719) Corporate 0.46* (n = 22) 0.67# (n = 543) Individual 0.64* (n = 176) 0.70# (n = 176) *Difference statistically significant at p = 0.10 by parametric t-test and nonparametric Wilcoxon test. #Difference statistically significant at p = 0.10 by nonparametric Wilcoxon test only. Table 4.8 presents the mean TE scores obtained for farms of different types in the WB 2003 sample and in the pooled sample augmented with 512 corporate farms from the PFAP survey. Small individual farms achieve higher TE scores than large corporate farms (the difference is statistically significant in both samples). This indicates that the small individual farms on average utilize the two inputs (land and labor) more efficiently than the large corporate farms: for any given bundle of inputs the small farms produce on average more than the large farms. These results are consistent with the TFP results obtained by production function analysis: small farms are more productive in the production function paradigm and more efficient in the production frontier paradigm. 10 M. Ahearn, J. Yee, and W. Huffman, “The Effect of Contracting and Consolidation on Farm Productivity,” paper presented at the Economics of Contracting in Agriculture Workshop, Annapolis, MD (July 2002). 35 Evidence of Increasing Returns to Scale among Corporate Farms So far we have been looking at datasets with two clearly differentiated groups of farms: small individual farms (generally farms with less than 50 hectares) and large corporate farms (technically farms with more than 50 hectares, but in practice managing hundreds and thousands of hectares on average). Given this dichotomy, we obtained evidence of decreasing returns to scale and clear proof of higher total factor productivity in small individual farms. The PFAP database taken on its own (without pooling with the WB 2003 survey) provides 512 observations of large corporate farms only. The coefficients of the production function estimated for this sample of large corporate farms are 0.81 for land and 0.31 for labor. They sum to more than 1, and the difference from 1 is statistically significant at p = 0.01. The production function thus shows increasing returns to scale within the sample of corporate farms. This result is consistent with previous findings for corporate farms in Russia, where several researchers have observed increasing returns to scale specifically among corporate farms (Uzun, 2005; Epshtein, 2003, 2005). The TFP calculations were repeated for the PFAP sample of corporate farms considered on its own. Here, we keep the farm type constant (corporate farms) and examine the TFP scores as a function of farm size. The coefficient of correlation between the TFP scores of corporate farms and land (taken as a continuous variable) is positive (0.2) and statistically significant at p < 0.01. This implies that for corporate farms TFP indeed increases with farm size. To visualize the results more clearly, we additionally classified the 521 corporate farms into three size groups (up to 500 hectares, between 500 and 2000 hectares, more than 2000 hectares). The productivity of land clearly increases with farm size, whereas the productivity of labor does not. Total factor productivity calculated by aggregating land and labor with appropriate weights from the production function shows a definite increase with farm size (all pairwise differences in TFP are statistically significant by standard simultaneous comparison tests). This behavior is illustrated in Table 4.9. Table 4.9. TFP of corporate farms by land size categories: PFAP 2003 survey <500 ha 500-2000 ha >2000 ha (1) (2) (3) Number of farms 238 225 58 Land productivity (output/ha, lei) 1,927 2,162 2,430 Labor productivity (output/worker, lei) 18,660 16,580 19,219 TFP (lei per unit of aggregated inputs) 3,162 3,603 4,167 The TFP results were verified by carrying out Stochastic Frontier Analysis for the PFAP sample of corporate farms and testing for significant differences in TE scores across the three size categories. The mean TE scores by size group are presented in Table 4.10. The lowest TE score is observed for the smallest corporate farms with up to 500 hectares; the highest TE score is observed for the largest corporate farms with more than 2,000 hectares. These results were found to be highly significant by standard simultaneous pairwise-comparison methods. 36 Table 20. TE scores obtained by Stochastic Frontier Analysis for the PFAP sample of corporate farms (n= 521) Farm size category SFA scores <500 ha 0.64* 500-2000 ha 0.76* >2000 ha 0.84* On balance, we do find evidence of increasing returns to scale in the homogeneous sample of corporate farms, while farms in a larger mixed sample of different types (both individual and corporate) show decreasing returns to scale. The different behavior may be understood if we recall that in our samples scale is a proxy for farm type. Small farms are typically individual farms, and they do better than large corporate farms not necessarily because of a size effect, but because of an organizational form effect: individual farms outperform corporate farms. June 3, 2005 Addendum: Impact of fragmentation on farm performance (omitted from v.5 of my report, but included in Bill Sutton’s December 2005 note) The only results that we have on the “economics of fragmentation” are Blarel’s paper on Ghana and Rwanda and my analysis of Georgia. Since it is quite likely that these cases will not sit very well with the Moldovan officials, I have continued thinking (with Victor Moroz’s help) on ways to reach some conclusions from Nora’s 2003 survey. As I wrote to you some time ago, only the households’ survey contains a fragmentation variable (not the peasant farmers’ survey). Unfortunately, we do not have the value of output for the households’ survey, so we cannot calculate TFP, etc., as we did for the farmers’ survey. We thus have a dilemma: fragmentation without value of output for households; value of output without fragmentation for farmers. Victor Moroz has called my attention to section G of the households’ questionnaire, where respondents give their monthly income (in lei) and the structure of that income (in percent of various components). Based on these questions, I calculated the value of farm income (including cash revenue from sales of farm products and value of own consumption). This is very rough, of course, but still we can use it as a proxy for the value of farm output. In the next stage I calculated farm income per hectare and farm income per work day (including family workers and outsiders) as approximations to partial productivity of land and labor. Finally I looked at the relationship between these partial productivities and the number of parcels (as a fragmentation measure). And lo and behold! I get a strong negative relationship between productivity and the number of parcels. Household data in Moldova suggest that consolidation makes sense! The results are given below. The negative relationship is statistically significant by all standard parametric tests (ANOVA, regression). 37 Number of parcels Number of observations Farm income per ha 1 35 946 2 54 562 3 72 348 4 52 288 5 52 202 6 30 251 7 22 193 8 12 188 9 5 136 10 and more 6 97 Farm income per work day 6.3 3.0 3.4 4.6 2.1 3.6 1.7 1.6 1.3 0.9 To reinforce these results I ran a regression of farm income on three variables: land used, work days, and number of parcels. All three coefficients came out significant. Controlling for land and labor, the number of parcels has a negative effect on farm income and the coefficient is significant at p < 0.1. The table below shows the regression results. Dependent variable: farm income (calculated as described above) Independent variables Estimated coefficient Land used, ha 98.772 Labor, work days 0.444 Number of parcels -17.077 Intercept 90.210 R-square 0.13 Significance level 0.000 0.000 0.063 0.032 38 5. Development of Land Markets in Moldova 11 In all market economies, land markets provide the medium that enables agricultural land to flow from less productive to more productive farmers, thus contributing to productivity growth in the farm sector. Land markets, and not government regulation, are the main tool for adjustment of farm sizes toward greater productivity and efficiency. Titling and registration of ownership rights in a state cadastre is generally a prerequisite for normal functioning of land markets as it provides security for land transactions. The cadastre system in Moldova was set up in 1998-99 with extensive donor support in conjunction with the USAID-sponsored National Land Program, which initiated large-scale allocation of agricultural land plots against previously distributed land shares. The progress achieved with cadastral registration of agricultural land since 2000 is illustrated in Figure 5.1. The number of registered agricultural plots increased from about 1 million in 2000 to 3.2 million in 2004. The cumulative registered area grew from 600,000 hectares to 2.8 million hectares. The cumulative numbers for 2004 include some double counting as a result of repeated registrations during the five years, but it seems that virtually all agricultural and household plots have been registered in the cadastre. 3.5 mln plots mln ha 3 3.5 3 2.5 2.5 2 2 1.5 1.5 1 Area Plots 1 0.5 0.5 0 0 2000 2001 2002 2003 2004 Figure 5.1. Progress with cadastral registration of agricultural land 2000-2004. The agricultural land market in Moldova began to emerge in 1998, as an immediate outcome of the trail-blazing changes introduced by the National Land Program. The total number of 11 The State Cadastre Agency is the only source of information on agricultural land transactions, including buyingand-selling and lease contracts. However, neither the Cadastre Agency nor the First Cadastre Project distributes this information openly to the public: they do not have a web site and they do not publish statistical collections. In a special effort undertaken as part of the sector study, the technical staff of the Cadastre Project has collected authoritative and up-to-date information on the number of land transactions, the volume of agricultural land transacted annually, and land prices achieved. The only previous report that covers similar issues for 1999-2003 is the baseline study of the agricultural land market in Moldova carried out by the Consulting and Credit in Agriculture (CCA) NGO for DAI/USAID in Chisinau (November 2003). 39 transactions in agricultural land increased from virtually zero in 1999 to about 130,000 in 20032004 (Table 5.1, Figure 5.2). During the 5-year period 2000-2004, the Cadastre Agency reports nearly 400,000 agricultural land transactions, and by the end of 2005 the cumulative number of transactions is expected to exceed 550,000. Of the total number of recorded transactions, 36% involve buying and selling of land, 13% are leasing transactions, and the remaining 51% are transactions involving inheritance and gifts (the share of transactions of different types has remained fairly constant over time since 2001; see Table 5.1, Figure 5.3). Table 5.1. Number and structure of transactions in agricultural land 1999-2005 Number of transactions Percent of total Other Other ownership ownership Sale transfers Lease Total Sale Lease transfers 1999 147 51 16 214 69 24 7 2000 3,450 2,155 3,085 8,690 40 25 36 2001 15,308 17,735 6,332 39,375 39 45 16 2002 28,632 47,434 8,179 84,245 34 56 10 2003 51,352 64,371 13,129 128,852 40 50 10 2004 43,918 64,953 21,067 129,938 34 50 16 2005* 17,402 28,102 3,783 49,287 35 57 8 1999-2005 160,209 224,801 55,628 440,638 36 51 13 Source: First Cadastre Project. *Preliminary data for January-April 2005. Table 5.2. Transacted area of agricultural land 1999-2005 (excluding leasing) Transacted area, ha Percent of total Other Other ownership ownership Sale transfers Total Sale transfers 1999 74 28 102 73 27 2000 1879 1364 3243 58 42 2001 9,238 14,201 23,439 39 61 2002 17,599 28,825 46,424 38 62 2003 36,248 47,036 83,284 44 56 2004 53,818 40,421 94,239 57 43 2005* 32,363 38,952 71,215 45 55 1999-2005 151,121 170,825 321,946 47 53 Source: First Cadastre Project. *Preliminary data for January-April 2005. Total 100 100 100 100 100 100 100 100 Average transaction, ha Other ownership transfers Sale Total 0.68 0.48 0.57 0.61 0.61 0.61 0.62 0.74 0.65 0.63 0.59 0.61 0.66 0.68 0.67 1.15 0.58 0.73 1.62 0.89 1.06 0.88 0.68 0.72 The relatively small share of leasing transactions reported by the State Cadastre is surprising, given the observation that most small landowners tend to lease their land to other users. Thus, individual farm surveys conducted by the Department of Statistics indicate that 57% of respondents leased their entire land to other users, and many among the remaining 43% leased out at least part of their land (Agricultural Activity of Households and Farms in the Republic of Moldova in 2003 – Results of the Statistical Survey, Chisinau, 2004, p. 12). This apparent discrepancy arises because only lease contracts for a term of 3 years or longer are subject to registration in the regional cadastre office. Leases for less than 3 years are registered at the 40 village primaria, and no central record of these contracts exists. Local experts estimate that at least 70% of all lease contracts in Moldova are for a term shorter than 3 years and are therefore not reflected in State Cadastre records. If the lease figures in Table 5.1 represent only 30% of the total number of lease transactions, the actual number of lease contracts can be estimated at 185,000, or one-third of all land transactions in Moldova during 1999-2005. 160 Thousands 140 120 100 Lease Inherit+Gift Sale 80 60 40 20 0 1999 2000 2001 2002 2003 2004 2005 Figure 5.2. Number of transactions in agricultural land 1999-2005. Source: First Cadastre Project; data for 2005 extrapolated to a full year. 100% 80% 60% Lease Inherit+Gift Sale 40% 20% 0% 2000 2001 2002 2003 2004 2005 Figure 5.3. Structure of transactions in agricultural land 2000-2005. Source: First Cadastre Project. The transacted area increased sharply with the increase in the number of transactions. Cadastre records summarized in Tables 5.1 and 5.2 show that some 320,000 hectares of agricultural land changed ownership in 380,000 transactions between 1999 and 2005. To put the numbers in perspective, we have to note that there are 2.5 million hectares of agricultural land in Moldova and about 1.8 million hectares are in private ownership. The total area changing ownership 41 during the entire period 1999-2005 is thus 13% of all agricultural land or 18% of all privately owned agricultural land in the country. To compare land market activity in Moldova to international statistics, we need to estimate transfer rates, i.e., the ratio of the number of titles transferred per year to the total number of titles in the cadastral registry. The number of registered agricultural parcels in Moldova is about 3-4 million; the number of ownership transfers is around 100,000 per year (see Table 5.1; this includes ownership transfers through sale, inheritance, and gifts in 2003-2005). The transfer rate is thus 2.5%-3% in recent years. This is substantially lower than the EU average transfer rate of 7%, but relative to the more relevant benchmark of transition countries it is comparable to Hungary (2.5% in 1998) and is higher than the transfer rates in a number of other Central Eastern European countries (around 1% for the Czech Republic, Slovakia, Latvia, Poland, and Slovenia in 1998). 12 100 thousands ha per transaction 2.5 80 2 60 1.5 40 1 20 0.5 0 1999 2000 2001 2002 2003 2004 2005 Transactions Area Ave area 0 Figure 5.4. Number of land sale transactions (black curve), total transacted area of agricultural land (gray curve), and average area per transaction (bars). Source: First Cadastre Project; data for 2005 extrapolated to a full year. Cadastre records summarized in Tables 5.1 and 5.2 show that some 150,000 hectares of agricultural land were sold and bought in 160,000 transactions between 1999 and 2005. The average land sale transaction recorded in the national cadastre during 1999-2005 was thus less than 1 hectare (see Table 5.2). The average transaction size remained fairly constant at 0.6-0.7 hectares between 1999-2003, and then increased significantly to more than 1 hectare in 20042005 (Figure 5.4). This is the average size of a parcel recorded as a cadastral object in the system, reflecting the original fragmentation of the land shares in the process of privatization. Each physical transaction by one buyer could involve many such small parcels. An entrepreneur buying 120 hectares of land would have to register 200 average transactions to complete the transfer of ownership (more on transaction costs below). The increase in average transaction 12 The international comparisons are constructed from R. Baldwin, The Development of Land Markets in Central and Eastern Europe, ACE Project P2128R, Brussels (June 1998). See also Z. Lerman, C. Csaki, and G. Feder, Agriculture in Transition: Land Policies and Evolving Farm Structures in Post-Soviet Countries, Lexington Books, Lanham MD (2004), p. 81. 42 size between 1999-2003 and 2004-2005 may in fact reflect certain parcel consolidation trends in Moldova. Table 5.3 focuses on the regional distribution of the 160,000 land sale transactions reported during 1999-2005. There is considerable variability in the total number of transactions and total transacted area across the 12 territorial cadastre offices, which is mainly a function of the local conditions. The average transaction size, however, shows a fairly clear pattern: the smallest plots are bought and sold in the central regions of the country, those within 100 km around Chisinau (0.6 ha on average). The land sale transactions in the South and especially in the North involve larger plots (0.8 ha and 1.4 ha, respectively). Table 5.3. Sale transactions by distance from the capital 1999-2005 Territorial Distance from Number of sales Transacted Region cadastre office capital transactions area, ha All Moldova 160,209 151,121 Edinet North 202 8,968 37,543 Soroca North 160 20,257 13,590 Balti North 125 11,937 19,938 Ungheni Center 107 6,983 3,544 Orhei Center 46 19,332 10,813 Chisinau Capital 0 10,595 3,970 Straseni Center 23 24,800 27,535 Hincesti Center 36 14,925 8,724 Causeni Center 88 18,994 9,286 Comrat South 105 5,860 4,324 Taraclia South 165 2,584 2,249 Cahul South 173 14,974 9,607 Source: First Cadastre Project. Average transaction, ha 0.88 2.11 0.70 1.49 0.44 0.76 0.38 0.63 0.63 0.57 0.86 0.82 0.68 Regional average, ha 0.88 1.43 0.57 0.78 Land prices The median price of agricultural land in 2004-05 was 5,400 lei per hectare across 11 territorial cadastre offices excluding the capital Chisinau (Table 5.4). The land prices showed considerable regional variation. The median price ranged from less than 3,000 lei per hectare in Soroca (far north) to more than 10,000 lei per hectare in Balti and Comrat (north-center and south, respectively). The price of land within the municipal limits of Chisinau was an order of magnitude higher: 53,000 lei per hectare in 2004-05. Although strictly speaking this is agricultural land, the higher prices in the capital probably capture expectations of windfall profits from non-farming uses that will become possible through eventual rezoning. These prices were estimated for agricultural land outside the village limits. Household plots, i.e., small parcels of land within the village limits, generally sold for much higher prices per hectare. Thus, the median price for land in household plots in 2004-05 was 12,000 lei per hectare (outside Chisinau), more than double the median price for agricultural land outside village limits. The prices reported for household plots in all 12 territorial cadastre offices are shown separately in Table 5.4. 43 Table 5.4. Median land prices for buy-and-sell transactions (lei/ha) 2004-2005 1999-2005 (2005 prices)* TCO Agricultural land Household plots TCO Agricultural land Household plots Soroca (N) 2,823 6,290 Soroca (N) 3,111 8,366 Taraclia (S) 4,353 3,742 Edinet (N) 3,767 8,178 Edinet (N) 4,484 5,441 Hincesti (C) 4,407 9,015 Causeni (C) 4,526 5,650 Orhei (C) 5,458 19,703 Hincesti (C) 4,711 9,271 Taraclia (S) 5,688 10,028 Straseni (C) 5,179 57,697 Causeni (C) 5,862 5,341 Orhei (C) 5,458 12,789 Balti (N) 6,480 29,000 Cahul (S) 6,833 6,155 Straseni (C) 8,588 100,298 Ungheni (C) 8,873 11,842 Cahul (S) 8,877 6,155 Balti (N) 10,983 23,709 Ungheni (C) 10,002 13,350 Comrat (S) 12,984 15,620 Comrat (S) 11,354 22,024 All Moldova (excl. All Moldova Chisinau) 5,380 11,780 (excl. Chisinau) 5,540 14,850 Chisinau 52,636 264,204 Chisinau 61,547 285,341 Source: First Cadastre Agency. *Adjusted for inflation using the following CPI values: 2004-05=100; 2003=88.0; 2002=78.9; 2001=74.9; 2000=68.2; 1999=52.0 (IMF Country Report No. 05/53, February 2005). It is often conjectured that land prices are inversely related to the distance from the capital. However, our analysis failed to detect a statistical relationship between prices and distance from the capital once Chisinau itself had been removed from the analysis. Figure 5.5 plots the land price per hectare in the 12 cadastral regions (from Table 5.4) as a function of the average distance of each region from the capital in kilometers (the city of Chisinau is represented by zero distance). The agricultural land prices estimated for different years between 2001 and 2005 were adjusted for inflation to obtain “real prices” in constant 2005 lei and logged for better visualization. Without Chisinau (at point 0 on the horizontal axis) the scattergram of price versus distance is a random cloud of points showing no statistical relationship whatsoever. Alternative analyses carried out with distances to the cadastre regions expressed on an ordinal scale (using distance ranks) and on a nominal scale (dividing the country into four zones) did not detect any statistical relationship either. Another feature that emerges from Figure 5.5 is the constancy of land prices over time. The random mixing of the points for different years in each cadastre region (i.e., for each distance) shows that there was no systematic increase (or decrease) of real, inflation-adjusted land prices between 2001 and 2005. This probably indicates that the demand for agricultural land is still low and is not sufficient to drive the land prices up over time. The visual conclusions from Figure 5.5 are confirmed by regression analysis, which has failed to detect any statistically significant impact of distance and time on real land prices. In a separate set of analyses we did not find a statistically significant association between land prices and land supply factors (as proxied by the number of transactions or the transacted area across the cadastre regions). This finding probably indicates that at the present stage there is an excess supply of land that is more than sufficient to satisfy the limited demand at the given 44 prices. For the situation to change the productivity of land use and the profitability of farming should increase beyond their present level. Figure 5.5. Land prices versus distance from Chisinau 2001-2005. Source: First Cadastre Project. Transaction costs We have seen that the average price of agricultural land in Moldova is around 5,000 lei per hectare (excluding Chisinau; see 2004-2005 data in Table 5.4). Since the average sale transaction recorded in the State Cadastre is 0.9 hectares (see Table 5.2), purchasing one hectare of agricultural land involves practically one cadastral transactions (“parcels”). Our field visits yielded some estimates of the transaction costs associated with registration of transfer of ownership (Table 5.5)13. The cost per transaction is estimated at 300 lei. This is about 5%-10% of the price of land, which is not at all exorbitant by CIS standards (in Russia, for instance, the transactions costs may exceed the price of land by a large margin 14). This cost, however, does not include surveying and mapping, which were carried out with USAID funding as part of the National Land Program. Without surveying, the main cost component is the notary fee for authentication of documents. It is charged at 180 lei per transaction and thus accounts for 60% of the total transaction costs. In theory, notary fees are charged pro rata on a sliding downward scale. However, the sliding scale starts at 1.3%, but not less than 180 lei. The minimum fee is 1.3% of 14,000 lei, which is equivalent to a 3hectare transaction. This threshold is seldom reached in practice, because the average transaction is 0.9 hectares. If the minimum were eliminated, the notary fee for the average transaction would be 65 lei instead of 180 lei. The practice of charging a flat notary fee on small transactions seems to be universal in CIS and is contrary to the prevailing practice in the United States and the EU, where notary fees are a percentage of the transaction amount without a minimum. This shortcoming of the 13 For similar estimates see Agricultural Land Market in Moldova: Baseline Study, USAID/CCA, Chisinau (November 2003), p. 47. 14 For a detailed analysis of transaction costs in Russia see N. Shagaida, “Agricultural Land Market in Russia: Living with Constraints,” Comparative Economic Studies, 47(1): 127-140 (March 2005). 45 registration system in CIS has been repeatedly commented on by the World Bank and other donors. To reduce transaction costs under conditions of Moldova’s highly fragmented holdings, notary fees should be calculated pro rata and the unrealistically high minimum fee should be abolished. Table 5.5. Estimated land transaction costs according to the standard procedure and the consolidated option Standard procedure Consolidated procedure Extract from cadastre registry 25.50 n.a. Authentication of sales contracts by 180.00 25.00 notary State tax for authentication 15.00 n.a. New record of ownership 42.50 34.00* Two trips to cadastre office (one trip 37.00 n.a. to submit the paperwork, second trip to collect new title) Total 300.00 59.00 *The primaria receives a rebate or a grant of 8.50 lei per title from the government. Source: Field visit in Jora de Mijloc primaria, Orhei district (February 2005). There seems to be a possibility of reducing the transaction costs from the present level of 300 lei to somewhere around 60 lei by adopting a consolidated procedure, whereby the primaria secretary prepares a single list of all parcel sales in the village at a particular time and then travels alone to the district cadastre office on behalf of all the buyers and sellers (Table 5.5). The primaria secretary is legally empowered to act as a notary for the local residents, typically charging 25 lei for the services (compared with 180 lei notary fees). Furthermore, with his close knowledge of the local scene, the secretary can authenticate the sales contract without requiring a cadastral extract, thus eliminating another cost component. This procedure is being implemented on an experimental basis in the primaria of Jora de Mijloc (Orhei district) and, if officially approved, it may lead to substantial savings in transaction costs for buyers who are forced to assemble their holdings from a mosaic of small parcels. Land leasing The average lease contract is for 0.84 hectares (Table 5.6, 1999-2005 data). This is identical with the average size of plots in land sale transactions (0.88 hectares), which indicates that leasing is a practical alternative to buying and selling of land in Moldova. Consistently with the legal requirement to register leases for terms longer than 3 years in the state cadastre, less than 1% of the lease contracts in the database are for shorter terms. In fact, only 5% of the contracts are for terms of 3-5 years and fully 95% of the leases are for terms longer than 5 years (Table 5.6). A closer look at long-term leases (using a different cross-section of the data) reveals two frequency peaks: 55% of the contracts cluster around 5 years and another 35% cluster around 10 years. The number of contracts for terms longer than 10 years is negligible, similarly to the number of contracts for terms shorter than 3 years. We thus have a clear preference for lease terms of 5 and 10 years across all regions. The lease term structure is summarized in Figure 5.6. 46 Table 5.6. Average leased plot (in ha) for various lease terms Lease term Number of contracts (1999-2005) Less than 12 months 90 1-3 years 159 3-5 years 2,090 Longer than 5 years 41,470 Total 43,809 Source: First Cadastre Project. Percent of contracts 0.2 0.4 4.8 94.6 100.0 Average lease size, ha (1999-2005) 0.89 0.82 0.83 0.84 0.84 5-8 years 56% < 5 years 6% >10 years 1% 9-10 years 37% Figure 5.6. Distribution of agricultural land contracts by lease term 1999-2005. Source: First Cadastre Project. Unfortunately, the State Cadastre does not contain information on the monetary value of lease contracts and it is therefore impossible to compare the lease payments to the average sale prices of around 5,000-6,000 lei per hectare. This comparison can only be based on survey data, which estimate annual lease payments at 800-900 lei per hectare per year (see Table 6.7). The implied capitalization rate is thus around 15%, which is relatively high compared with the 10% rule of thumb rate. 47 6. Land Leasing Patterns and Equity Leasing is a basic component of land market transactions that supplements land purchase by providing an additional channel for transfer of land to more efficient users and for adjustment of farm sizes. The deficiencies of the official leasing statistics (see Chapter 5) usually force us to rely on survey data for the analysis of land leasing. The World Bank conducted several comprehensive rural surveys in Moldova (1967, 2000, and 2003), which were designed to cover a wide range of reform-related topics and included inter alia certain aspects of agricultural land leasing. To bring the information on land leasing (and land markets in general) up to date, a new specialized survey was conducted in May 2005, focusing on transactions in agricultural land and on issues of equity in relations between lessors (typically small landowners) and lessees (commercially oriented peasant farms and large corporate farms). 15 The WB 2005 survey covered two major groups of respondents: small rural landowners, representing the supply side of agricultural land markets; commercially oriented peasant farms and large corporate farms, representing the demand side of agricultural land markets (“land users”). The survey instruments naturally allowed for the possibility that landowners were also land users in the sense that they farmed at least some of their land, and that peasant farms and corporate farms, in turn, owned some land and could therefore contribute to the supply side of agricultural land markets. The survey has demonstrated, however, that the dichotomy between “landowners” and “land users” in rural Moldova is quite distinct and sharp. The “land users” were selected on the basis of village-level lists, covering all peasant farms and corporate farms in the area. The “landowners” were selected from among the rural households in the village not listed as peasant farmers. The “landowners” are accordingly referred to as “households” in what follows, while the “land users” are generally divided into peasant farms (an individual or familybased form of organization) and corporate farms. Table 6.1. Sample design for the World Bank May 2005 land survey Households Peasant farms Number surveyed 75 60 Number of districts 22 19 Number of villages 25 20 Number of villages per district 1-2 1-2 Number of respondents per village 3 3 (1-4) Geographical distribution (% of respondents) North 32 35 Center 36 37 South 32 28 Corporate farms 21 19 19 1 1 (1-2) 42 29 29 The sample design is shown in Table 6.1. The territorial coverage included most of Moldova’s 24 administrative districts. A two-stage random sampling procedure was used for households (“landowners”) and peasant farms (a component of “land users”): one or two villages were selected at random in each district, followed by random selection of the respondents in each 15 The WB 2005 survey was supervised by Alexandru Muravschi (PFAP) and Victor Moroz (UNDP); Anatol Bucata (PFAP) was responsible for data entry; Zvi Lerman provided inputs during the development of survey instruments and analyzed the survey data. 48 village. The selection of corporate farms was pre-determined by the random selection of the villages in each district: there was typically one (large) corporate farm per village, and that farm was included in the sample; two of the sampled villages had two corporate farms each, and both were included in the sample. In terms of organizational form, most corporate farms (17 of the 21) were limited liability companies or partnerships. Previous World Bank surveys have shown that land leasing indeed fulfills its role as a facilitator of farm size adjustment: peasant farms with leased land are, on average, much larger than farms based only on privately owned land. The size adjustment effect achieved by peasant farms through land leasing is demonstrated in Table 6.2. In the WB 2000 survey, all farms— both individual and corporate—relied heavily on land leased from outsiders to increase their size. A very strong correlation was observed between farm size and the amount of leased land: on average, an increase of 1 ha in land holdings was achieved entirely through leasing from nonmembers, i.e., outsiders (R2=0.85, p < 0.001). Table 6.2. Size of peasant farms with and without leased land WB 1967 WB 2003 Farms w/out Farms with Farms w/out leased land leased land leased land Percent of farms 94 6 79 Total land use 2.8 16.9 3.8 Private land 2.8 3.4 3.8 Leased land -13.5 -Source: World Bank surveys, 1967, 2003, and 2005. Farms with leased land 21 11.6 3.1 8.5 WB 2005 Farms w/out leased land 72 3.7 3.7 -- Farms with leased land 28 9.5 5.0* 4.5 The markets for land leasing evolved strongly over time: only 6% of peasant farmers reported leasing land in the 1997 survey, and this percentage increased to 28% in the 2005 survey (Table 6.2). Although no comprehensive official statistics on lease transactions are available to this day, the land lease market has definitely grown much stronger as leaders of the new corporate farms join private farmers in competing for additional land among inactive landowners. 16 Rural Households as Suppliers of Land Based on the findings of the latest WB 2005 survey, the average household owns 2.7 hectares of agricultural land, but actually farms less than 0.5 hectares, or only 18% of the total endowment. The farmed portion is the traditional household plot (0.13 hectares around the house and 0.32 hectares in the fields outside the village, typically split into two parcels). The remaining 2.2 hectares represents land shares received in the process of privatization, and practically everybody leases all this land to other producers. While all respondents lease out land, not a single respondent reported leasing in land to augment the family holdings. Nobody reported buying or selling land in the last three years. Table 6.3 compares the findings of the WB 2005 16 As noted in Chapter 5, only lease contracts for a term of 3 years or longer are subject to registration in the regional cadastre office (also see Table 5.6). Yet, even with this restriction on data availability, the number of lease transactions recorded in the State Cadastre increased from around 3,000 to more than 21,000 between the years 2000 and 2004 (see Table 5.1). 49 surveys with those of the WB 2000 survey. In both surveys, rural families lease out most of their privately owned land to operators. Table 6.3. Landownership and land use by rural households WB 2005 survey Area, ha Percent of owned land Plot around the house 0.13 5 Household plot in the fields 0.30 11 Land shares 2.19 82 Other land 0.04 2 Total owned 2.67 100 Total used 0.46 18 Leased in 0.00 0 Leased out 2.21 82 Source: World Bank surveys, 2005 and 2000. Area, ha 0.2 0.2 2.6 -3 1.2 -1.8 WB 2000 survey Percent of owned land 6.5 6.5 87 100 40 -60 The role of rural households as the suppliers of land in the agricultural sector emerges clearly when we analyze the structure of holdings of the other two cohorts in the WB 2005 survey – peasant farms and corporate farms. In farms of both these types, owned land is only a portion of used land (contrary to household plots, which use only a small fraction of owned land), and the difference is made up by leasing in land from outside sources. Leasing out is hardly practiced by farms of either type (Table 6.4). The reliance on leased land is particularly pronounced for corporate farms, where the component of owned land is very small. Table 6.4. Structure of land holdings in farms of different types (in percent of land used) Households Peasant farms Corporate farms Area, ha Structure of Area, ha Structure of Area, ha Structure of land use, % land use, % land use, % Total owned 2.7 540 5.7 85 14 1 Leased in 0.0 0 1.3 19 1006 100 Leased out 2.2 440 0.3 4 12 1 Total used 0.5 100 6.7 100 1008 100 Source: WB 2005 survey. Households lease the bulk of their land to corporate farms, which account for 90% of all land leased by the households in the WB 2005 survey (Table 6.5). The remaining 10% is leased to peasant farms. 17 Other households, pensioners, or entrepreneurs do not lease land from the small landowners. This in a way is consistent with the observation that the households in the WB 2005 survey do not lease in land. To the extent that peasant farms lease out land, it also goes to corporate farms. On the demand side, households are the main source of leased land for both peasant farms and corporate farms. Some land is leased internally, i.e., from farm members or shareholders, but fully 70% is leased from outsiders (Table 6.5). The previous surveys also showed that farmers leased land mainly from rural households. This source accounted for 80% of land leased by peasant farms in the 1967 survey and over 90% of leased land in the 2003 survey. The remainder was leased from local 17 This is very close to the results of the WB 2000 survey, where 86% of households’ leased land went to corporate farms and 11% to peasant farms. 50 authorities and to a certain extent from the local corporate farm (especially in 2003). Peasant farms acting as lessors provide 12% of the leased land in corporate farms. Table 6.5. Land leasing: who to and who from (percent of leased land) Lessors: supply side Households Peasant farms Farm members/shareholders --Households --Peasant farms 10 -Corporate farms 90 100 Others --Leased land, ha (mean per farm) 2.2 0.3 Source: WB 2005 survey. Lessees: demand side Peasant farms Corporate farms 33 12 67 70 -12 ---6 1.3 1006 There is a sharp dichotomy between household plots as supply side players and commercial producers (peasant farms and corporate farms) as agents of the demand side in land markets. Analysis of leasing participation rates in Table 6.6 shows that 96% of households lease out land and virtually none leases in land. At the other extreme, 100% of corporate farms lease in land and none leases out. Peasant farms occupy an intermediate position: they act as both lessees and lessors, yet their demand side role clearly predominates, as nearly 30% of peasant farms lease in land and only 8% lease out (Table 6.6). The stronger demand side role of peasant farms also emerges from Table 6.4, which shows that the average peasant farm leases in 1.3 hectares, while leasing out only 0.3 hectares. Table 6.6. Participation in land leasing (percent of respondents) Households Peasant farms Leasing out 96 8 Leasing in (overall participation) 0 28 From members NA 18 From non-members NA 13 Source: WB 2005 survey. Corporate farms 0 100 67 95 Both lessors (households) and lessees (peasant farms and corporate farms) provided consistent estimates for lease terms and lease payments. The average lease term is typically 3 years and the annual lease payments are around 800 lei per hectare for all categories of respondents (Table 6.7). Lease payments in kind are somewhat higher (around 900 lei per hectare as reported by the lessees), but the difference is not statistically significant (within each category of farms). The differences in lease payments across farm types are not significant either: corporate farms pay for leased land roughly the same as peasant farms. Table 6.7. Lease term and annual lease payments in the survey Households Peasant farms Corporate farms All sample (lessors) (lessees) (lessees) Lease term 3.2 (1-10) 2.6 (1-3) 3.1 (1-10) 3.1 (1-10)* Lease payments, lei/ha 761 846 900 820 In cash 790 756 753 773 In kind 750 924 967 844 * Only 8% of respondents report terms longer than 3 years (5-10 years) and 9% report short term leasing for 1 year. Source: WB 2005 survey. 51 Lease payments of 800-900 lei per hectare per year are relatively high compared with land prices of 5,000-6,000 lei per hectare as estimated from state cadastre data (see Table 5.4): the implied capitalization rate is close to 15%, which is substantially higher than the 10% rule of thumb rate. Households’ Land Disposal Strategy Physical distribution of land plots against paper shares as part of the National Land Program since 1998 was obviously the most momentous stage in Moldova’s land reform. Close to a million rural people received parcels of land in exchange for paper certificates of ownership that they had held since 1992. Yet the World Bank surveys (especially WB 2005, but also WB 2000 – see Table 6.3) show that they failed to take advantage of this program and simply turned around and leased their new land to other producers, mostly corporate farms. As a result, the rural people surveyed continue to cultivate the same small household plot that they cultivated before 1998, and someone else farms their “land shares”. It can be argued, of course, that this land use pattern reflects decisions made by relatively uninformed new landowners back in 1998 and that today they would make radically different decisions. However, when respondents in rural households were asked to reflect on their land disposal strategy with the benefit of hindsight, 90% answered that they would lease their land shares to others even if they were allowed to make new decisions. Still, the WB 2005 survey suggests that the existing situation – whereby rural families generally farm their small household plot and lease out their land shares to other producers – is not really what the people want. Asked about the optimal size of their farm, 50% of the respondents gave a desired size of 1.0 hectare, 25% indicated that they would like to farm more than 3.0 hectare, and 10% set the optimal farm size at 5.0 hectares or larger (Table 6.8). Compared to currently farmed land (about 0.5 hectares), the median augmentation desired by the households is by more than a factor of 3, with 25% of respondents seeking to augment their current plot size by a factor of 7.5 or more. 18 Currently owned land (about 2.7 hectares), on the other hand, appears to be quite sufficient to meet the desired augmentation for more than half the households. Table 6.8. Desired augmentation of farm size as reported by households* Mean Median Land used 0.46 0.44 Land owned 2.66 2.20 “Optimal” size 2.86 1.00 Percent used 0.23 0.18 Augmentation factor relative to land used 7.1 3.3 relative to land owned 1.3 0.8 *For respondents reporting nonzero “optimal” size (n = 73). Source: WB 2005 survey. 18 Upper quartile 0.55 3.38 3.00 0.26 7.7 1.0 The augmentation factors for lessees are much more moderate: the respective medians are 1.7 for peasant farms and 1.2 for corporate farms (relative to land used). 52 We are thus faced with an intriguing puzzle: people own enough land to meet their farming target, and yet they lease out the bulk of this land and keep a plot which is much smaller than the desired target. This situation raises serious doubts concerning the true voluntary nature of landshare leasing arrangements between rural families and local corporate farms. We cannot say, however, if this apparent distortion is the result of external pressures applied to landowners by “leaders”, or a natural outcome of existing constraints to normal functioning of the markets that interfere with efficient farming operations. Motivation for leasing out land What reasons do the households give for leasing out land? The main reason is insufficient labor (40% of respondents in the WB 2005 survey). Difficulties with access to purchased inputs and credit (or money in general) rank next. In aggregate, reasons associated with the functioning of normal markets are cited by 78% of the households as responsible for their decision to lease out land (Table 6.9) 19. It may be argued that these individuals would tend to farm the land on their own if the missing or distorted markets were corrected. This conjecture is supported by the observation that the desired augmentation factors are substantially greater for respondents who attribute leasing to market imperfections than for respondents who lease out because of physical deficiencies of their land. There is a small group of respondents (7%) who classify themselves as “passive co-owners” or “shareholders”. They are institutionally obliged to lease their land shares to the corporate farm (the “leader”), but they are obviously not entirely happy with the arrangement that leaves them with a very small plot to farm. For these respondents the median augmentation factor is much higher than for the two other groups. Table 6.9. Reasons to lease out land and relationship with augmentation factor for households Percent of Grouped Percent of Median augmentation factor lessors reasons lessors Relative to land Relative to land used owned Plot too far from house 1 Plot too small 3 Physical 15 1.0 0.2 Land of poor quality 0 Farming not profitable 11 Inputs not available 19 No money 15 Market 78 4.7 0.8 Insufficient labor 40 No marketing channels 3 Obliged to lease as 7 Institutional 7 50.0 7.3 member/shareholder Source: WB 2005 survey. 19 It is impossible to conduct a similar analysis for peasant farms, as only 5 respondents in this group report leasing out land (and none among corporate farms). These five respondents lease out 53% of their holdings and justify their decision mainly by market reasons – availability of inputs and shortage of money (3 out of 5; the other 2 give “other” reasons). 53 The decision to lease out or to farm on one’s own may also be determined by a whole range of household characteristics. Thus, it would be tempting to hypothesize that younger families with a relatively large number of children tend to keep the land in the family, whereas pensioners without backing of younger family members tend to lease most of their land to the local large farm. This hypothesis, however, is not borne out by the data. Table 6.10 presents the comparative profiles of the two groups of households in the WB 2000 survey (no comparable household data were collected in the much smaller WB 2005 survey). The two groups are statistically indistinguishable by family size, number of children, or age of the head of household and the spouse. The only significant difference is in land ownership and land use: households that lease out some of their land have a larger endowment (3.2 ha versus 2.8 ha in private ownership) but actually cultivate a much smaller plot (0.5 ha versus 2.8 ha). Table 6.10. Profiles of households that lease out or cultivate most of their private land Land leased out to other Land cultivated by the operators household Family size 3.11 3.10 Children under 15 0.54 0.67 Age: Head of household 54 55 Spouse 48 51 Full-time work on household plot: Head 54%* 75%* Spouse 57% 62% Land ownership 3.2* 2.8* Land use 0.5* 2.8* Households with commercial sales 51%* 77%* Total family income, lei 6,600* 8,900* Farm sales revenue, lei 3,100* 6,500* Income from lease payments 1,500 -Farm workers 3.2 3.8 Availability of farm machinery 6%* 30%* Use of machinery services 88%* 97%* Use of purchased inputs 61%* 89%* Planning to become private farmers 17%* 42%* Perceived standard of living adequate/comfortable 13% 15% Standard of living improved in recent years 7% 11% * Differences statistically significant by the appropriate test. Source: WB 2000 survey. All other variables that differ significantly between the two groups are in effect driven by the differences in land use. Thus, a larger plot requires a higher percentage of heads of the households to work full time on their plot. A larger plot leads to a higher frequency of commercial sales and a higher family income: the revenue from lease payments received by the small plots covers less than half the difference in revenue from farm sales earned by the larger plots. Larger plots require machinery and purchased inputs, and the corresponding frequencies are indeed significantly higher in households that cultivate most of their land than in households that lease out a large part of their land. In conclusion we can say that, although there are significant differences between the profiles of the two groups of households by several variables, these differences are directly attributable to 54 the difference in land use and do not explain why some households decide to cultivate less land than others. The available survey data do not shed light on the determinants of the decision of households to lease out land or keep it for cultivation in the family. Legal Arrangements and Payment Discipline Leasing out of land shares by households is generally formalized in the form of a contract with a legal entity (representing the corporate farm). This mode is reported by 80% of the households surveyed, while only 8% have signed a lease contract with the manager as an individual (WB 2005 survey). Investment of land shares in the equity capital of corporate farms is not reported by any of the households. While land leasing is mostly covered by formal contracts, the situation with asset shares is different: nearly 40% of asset share owners indicate that their shares are used by the corporate farm without their formal consent (the rest have formal or informal contracts). The contractual status of land and asset leasing by households in the WB 2005 survey is summarized in Table 6.11. Table 6.11. Disposition of land and asset shares owned by households Land shares (n = 75) Invested in equity capital of corporate or peasant farm 0 Formal lease contract with corporation 80 Formal lease contract with manager 8 Informal contract 5 Shares used without formal consent 7 Source: WB 2005 survey. Asset shares (n = 65) 9 29 9 14 38 Peasant farms and corporate farms always sign formal contracts with outside lessors. There are, however, also relations with insiders, i.e., shareholders in corporate farms or members in peasant farms (Table 6.12). In peasant farms, the insiders are typically co-owners who have invested their land in the equity capital (72% of farms); in another 26% of the cases the insiders lease their land to the farm, but mostly under an informal agreement. In corporate farms, on the other hand, the insiders have invested their land in the equity capital in only 10% of the cases and have signed a formal lease contract (generally with the corporation) in 90% of the cases. A similar pattern is observed for legal arrangements concerning the use of asset shares. In peasant farms, the members have mostly invested their asset shares in the equity capital (67% of farms using members’ asset shares) and formal lease contracts are practically not used. In corporate farms, investment of asset shares in the equity capital is reported in only 17% of the cases, while use of signed lease contracts is reported in 55% of the farms. Table 6.12. Legal arrangements for the use of insiders’ land and asset shares in corporate and peasant farms Peasant farms Corporate farms Land shares Asset shares Land shares Asset shares (n = 60) (n = 33) (n = 21) (n = 18) Invested in equity capital 72 67 10 17 Formal lease contract with corporation 2 0 85 50 Formal lease contract with manager 7 3 5 5 Informal contract 18 15 0 28 Shares used without formal consent 1 15 0 0 Source: WB 2005 survey. 55 Corporate farms do not use either land or assets without the formal consent of the owner. In peasant farms the rules are more relaxed, at least for asset shares, as 15% of farms report using members’ asset shares without formal consent. This is similar to the situation reported by households for asset shares in Table 6.11. Despite the existence of formal signed contracts, the lessees do not always honor their obligations to the lessors. In the WB 2005 survey, fully 17% of lessor households did not receive the payments that were due under the lease contracts and 45% received only partial payment. All lessees (corporate and peasant farms) pay for leased land, but only 37% of the respondents fully discharged their contractual obligations toward lessor households (no information was provided on internal lease payments to members and shareholders). Among the lessors who did receive some lease payments, two-thirds were paid in kind and one-third were paid in cash and in kind. Cash-only lease payments are not practiced. The data collected in the WB 2000 survey indicate that outgoing lease payments account for 20% of farm production costs. The share of lease payments increases sharply from 3% for farms with up to 10 hectares (relatively small peasant farms) to 27% for farms of more than 1,000 hectares (mostly large corporate farms). There is no way to determine if the lease payments are equitable in the sense of providing a fair market return to landowners. However, we clearly see that farms do not monopolistically “expropriate” the private land of rural residents and do not blatantly ignore their legal and moral obligations to pay for the use of private land to rural families. Leasing and Family Income Lease payments account for fully 12% of total family income in households participating in the WB 2005 survey (this result is very close to the WB 2000 survey, where the corresponding figure is 13%). In peasant farms they make a much smaller contribution, because farmers lease out land to a much smaller extent than households. Table 6.13 shows the structure of family income for households and peasant farms in the survey. Table 6.13. Structure of family income for households and peasant farmers in the survey (in percent) Households Peasant farmers Farm income (including value of consumption of own products) 29 64 Salary income 28 10 Income from non-farm activities 14 21 Pensions and social transfers 17 3 Lease payments and dividends 12 2 Total family income 100 100 Source: WB 2005 survey. To assess the contribution of land leasing to overall family well-being for rural households, we analyzed the relationship between lease payments and the family standard of living as perceived by the respondents in the WB 2005 survey. The respondents classified their standard of living on a four-level scale: below subsistence (family income not sufficient to buy food); subsistence level (family income sufficient to buy food and daily necessities); comfortable level (family income 56 sufficient to buy clothing and other consumer goods above and beyond daily necessities); satisfied level (no material difficulties). None of the respondents made the satisfied level; only 11% regarded their standard of living as comfortable; 70% were on the subsistence level; and 20% did not have enough food for their families. Figure 6.1 shows that overall peasant farmers enjoy a higher standard of living than rural households in the survey. 80 percent of respondents 70 60 50 Households Farmers 40 30 20 10 0 Below subsistence Subsistence Comfortable Figure 6.1. Perceived standard of living for households and peasant farmers. Source: WB 2005 survey. Lease payments have a significant impact on the standard of living of rural households. Thus, lease payments accounted for 20% of family income in families reporting a comfortable standard of living and only 9.5% in families on the sub-subsistence level (the difference between the two extreme standard-of-living levels was significant at p < 0.1). Curiously, no statistically significant relationship was observed between the standard of living and farm size for the households in the WB 2005 survey. This is in sharp contrast to the findings in other transition countries, where a higher standard of living typically goes with more land. For peasant farmers, on the other hand, a comfortable standard of living is indeed seen to be associated with a much larger farm size than the lower standards of living. Peasant farms reporting a comfortable standard of living in the WB 2005 survey have 11 hectares on average, compared with less than 5 hectares for farms in the two lower categories (the difference is statistically significant at p < 0.01). The standard of living of peasant farmers is thus an increasing function of farm size, as is commonly observed in farm surveys in CIS and other transition countries. The different pattern for households and peasant farmers may be attributable to the much greater importance of farm income for peasant farmers (see Table 6.13). It is specifically farm income that shows the greatest dependence on farm size, whereas rural households are characterized by strong reliance on components not related to land use, such as salaries, social transfers, and ultimately lease payments received, which are the antithesis of farm size. 57 The relationship between standard of living and farm size is illustrated in Figure 6.2. Here the probability of being in the highest standard of living (gray curve) increases with farm size, while the probability of being on the lowest subsubsistence or “poverty” level (thick black curve) sharply decreases with farm size. 20 These results provide the ultimate support for land consolidation policies and hence the need to encourage land market development. 1 probability 0.8 0.6 Poverty Subsistence Comfortable 0.4 0.2 0 0 10 20 30 40 50 land use, ha Figure 6.2. Probability of achieving a given standard of living as a function of farm size for peasant farmers. The lowest standard of living “Poverty” corresponds to “Below subsistence” in Figure 6.1. Source: WB 2005 survey. 20 The probabilities of achieving a given standard of living were obtained in a multinomial logistic regression with the three-level standard of living as the discrete dependent variable and farm size as the continuous covariate.