A CONCEPTUAL FRAMEWORK FOR THINKING ABOUT ECONOMIC DEVELOPMENT AND DEVELOPMENT POLICIES. INFORMAL NOTES Marco Missaglia and Lucia Corno January 2005 CONTENTS 0. GDP ACCOUNTING 1. THE LEWIS MODEL. AN INFORMAL EXPOSITION 2. SOME REFINEMENTS OF THE LEWIS FRAMEWORK. POLICY ISSUES 3. MIGRATIONS: THE HARRIS-TODARO MODEL 4. LAND REFORM APPENDIX 0. GDP ACCOUNTING Measure of Economic Development: GDP per capita (material well-being….) where : Agriculture, Industry and Service where: Yj Lj Lj sector J L N= Population L= Labour Force A, I, S = = labour productivity in sector J = proportion of labour force in L = participation Ratio N From (1) we can see the potential sources of the gap in per capita income between different countries. a. A difference in the participation ratio. An economy with many children and/or unemployed will have a low ( L N ) . b. A different occupational distribution; since (YI L I ) >> (Y A L A ) , it is convenient to have ( LI L) as high as possible. 1 c. (Y A L A ) in the developed countries is roughly 40 times higher than in the developing countries d. (YI L I ) in the developed countries is roughly 10 times higher than in the developing countries. Thus, there is considerable scope for increasing labour productivity in agriculture in developing countries. However, statistically speaking, reason b. is by far the most important in explaining the gap (income gap) between developed and developing countries, which provides the central justification for an industrialization–led development strategy: the problem of development may be seen as the problem of moving people from agriculture to industry, to increase the relative size of the industrial sector. How can this objective be achieved? We need a broad conceptual framework to think about this transition, and this conceptual framework lies at the core of development economics. 1. THE LEWIS MODEL (1954). AN INFORMAL EXPOSITION If development is structural change (Agriculture→ Industry→ Service), the main roles of agriculture are: 1. 2. Supplier of labour to industry (people move) Supplier of a surplus of food. If international trade is not an option (and it is not to the extent that countries want to achieve self sufficiency in this is extremely important field), then a non-agricultural sector can develop only if agricultural produces more food than its producers need for their own consumption: this is the agricultural surplus. There is also a third role played by the agricultural sector: 3. Source of demand for industrial products, given the relevant size of agriculture in a developing economy, Lewis is mainly concentrated on roles (1) and (2). Lewis model has in mind a dual economy Traditional Sector Modern Sector Basically, but not only not only rural/agricultural Basically, but urban/industrial 2 What is,economically speaking, the main feature of the traditional sector? It is the existence of surplus labour or disguised unemployment. What does it mean? Be La the amount of labour employed in agriculture; the production function with surplus labour is: Y La B A Figure 1: the production function in agriculture In the background of this production function: a very limited quantity of capital (we are basically in a non-capitalistic, pre-modern sector) and a fixed amount of land (from which diminishing returns to labour). For any LA ≥ B, the marginal product of labour in agriculture in nil. Then, if you reduce labour from A to B total output is unchanged ≡ LABOUR SURPLUS For a capitalist, the profit maximizing rule is Real wage = Marginal product of Labour ( MPL ) But what happens in a family farm with labour surplus, where MPL = 0 ? People are not paid their marginal product, because the objective of the family farm is not profit maximization, but to guarantee an income to each of its members (an egalitarian, or pre-capitalistic rule, as opposed to a capitalistic rule). People are paid their average product w (this happens in the informal sector as well, in the cities. A cub driver might share his driving with a friend) 3 Y w La B A Figure 2: the average product of labour in agriculture Clearly, in La = A, the average product w > marginal product = 0 If somewhere in the economy there is another sector where MPL > 0 , then moving people from the traditional sector to this “modern” sector would increase total output of the economy, there would be an efficiency gain. But now imagine that, Y b w La C A Figure 3: disguised unemployment in agriculture In A MPL = b > 0 , but still w > b . Again, people are paid more than their marginal product. If somewhere in the economy there is a modern sector where MPL > b , the economy 4 would still enjoy an efficiency gain by moving people from the traditional to the modern sector, even if the marginal product in the traditional sector is higher than zero. How much labour would a profit maximizing, capitalist firm employ at a wage rate = w ? It would hire C units of labour ( MPL = w ) CA is the amount of disguised unemployment. Disguised unemployed are those who have a job, but are less productive than they would be should they be employed in the modern sector. In other and hopefully more transparent words: in a fully capitalistic world these people would be unemployed, without a job. We are now ready to understand the logic of the Lewis model of economic development (extended by Fei and Ranis in 1961). The logic of the model is illustrated in Figure 4. Consider a situation where the entire labour force is employed in agriculture. This is the starting point of a development process, people are so poorly productive that they must be all engaged in producing their own subsistence. We are in the labour surplus phase, since MPL = 0. Now consider a small reduction in the labour force employed in agriculture, LA: what happens? If the wage rate in agriculture stays at w (so that workers are now paid a bit less than their average product1) then, given that total output stays constant, an agricultural surplus opens up (AS in the lowest diagram). The average agricultural surplus, AAS, is defined as: Agricultural Surplus = AAS Number of transferred workers Clearly, AAS = w . And clearly AAS stays constant at w in the labour surplus phase (as long as the marginal product of labour is nil in agriculture), as depicted in the middle diagram. Now, a process of industrialization is possible if and only if the industrial wage paid to the transferred workers is such that they are able to buy (at least) the same quantity of food they received when they were employed in agriculture, w ; otherwise they will not move from agriculture (unless they are forced to). Well, what is the industrial wage (by “industrial wage” here we mean the wage paid by the industrial sector and expressed in units of industrial good) wI needed to buy w units of food? It must be true that PI ≥w PA wI 1 (2) The meaning of this assumption will be made clear later on, when we will focus on agricultural taxation 5 wI Li x y AAS w Li Y T AS w La A B C O Figure 4: the Lewis-Fei-Ranis model of economic development 6 However, industrialists want to make as much money as they can, and therefore (2) will hold as a strict equality (they do not want to pay more than is strictly needed to attract people from the rural/informal sector):2 P wI = w A = w p PI (3) From (3) we can see that, for any given level of the terms of trade PA PI (henceforth simply p for brevity), the industrial wage is constant as far as the agricultural wage w stays constant as well, i.e. in the labour surplus phase. This is shown in the upper diagram. Let’s now analyse the disguised unemployment phase (BC), i.e. the phase where in agriculture MPL > 0 but, assuming that the agricultural wage keeps staying constant (see footnote 1), w > MPL ? The average agricultural surplus begins to decline (since total output goes down); this can be seen by inspection of the lowest diagram and is shown in the middle one. One can therefore guess that, since less agricultural output is marketed (each worker that moves away from agriculture has to buy food on the market, but on average less food is available), food (agriculture) prices start to rise. But look at equation (3): the increase in p will push the industrial wage (w*) up, in order to make industrial workers able to buy w units of food and hence give them the appropriate incentive to move to industry. However, at a closer inspection, we can see that even if the industrial workers are getting more than w*, it is simply impossible for them to buy w units of food, because there is not enough to go around. To see why, just consider that by assumption people employed in agriculture are getting w ; if each industrial worker bought w as well, total agriculture production should be equal to OT, which is not the case. It follows that, at this stage, industrial workers will have to consume a mix of industrial and agricultural products. Under these conditions, will the potential “migrants” accept to move to industry? Clearly, they will if and only if w is not too close to the subsistence level (if it were, it could not be reduced!). This is a very important result, it says that agriculture must be sufficiently productive to favour an industrialisation process. In any case, potential “migrants” will only accept an industrial wage higher than w*: this is the reason why in the upper diagram the industrial wage becomes an increasing function at the beginning of the disguised-unemployment phase. When the transfer of labour reaches point C the disguised unemployment phase comes to an end. In the CO region MPL > w : agriculture is likely to become a fully capitalistic sector where wages are set according to the profit-maximisation rule (real wage = marginal product of labour). It follows that as labour moves from agriculture to industry, agricultural wage goes up (in other words, the wage bill in agriculture falls more slowly than before as labour moves from agriculture to industry) and the industrial wage (in the topmost panel) must increase even faster since it must not only compensate for higher terms of trade (p), but also for higher incomes in the agricultural sector. We are now ready to read properly the topmost panel where we have drawn a family of labour demand curves. For easiness of exposition, let us redraw that diagram: 2 For more realistic cases, a mark-up for higher costs of living in the urban sector or a lager degree of unionization in the industrial sector as well as some discount factor for the uncertainty of getting a job in the city are all factors that could be added to this simple model. However, they would not change the main conclusions in any relevant aspect. 7 wI Yx w* O πx πy x y Figure 5: The process of industrialization LI Initially, the amount of industrial labour is x. Realised profits are equal to Πx3 and Lewis assumes they are automatically invested back into the industrial activities (there is not an investment function in the Lewis – Fei – Ranis model). With more capital, labour demand (the marginal product of labour) shifts up. The new level of industrial employment is y. Note that in this labour surplus phase new industrial labour is forthcoming at the fixed real wage w*. Despite the fact that the marginal productivity of labour in the industrial sector is going up, the real wage earned by each worker stays constant. In the labour surplus phase the fruits of industrial expansion are not equally distributed: labour is becoming more and more productive but this only translates into higher profits (Πy + Πx > Πx) So, one of the implications of Lewis’ view of economic development is that income distribution (at least functional income distribution) worsens in the early phases of industrialisation (something reflected in the well known Kuznets curve). This important issue of the link between industrialisation and income distribution will be further investigated below with the help of a formal model developed by Bourgignon (1990). But it is in any case worth stressing that this unpleasant distributive result is seen by Lewis as an element that helps the economy grow and industrialise faster: industrialisation comes from capital accumulation and capital accumulation comes from profits (the propensity to save out of profits is higher than the propensity to save out of wages). And, do not forget our initial remarks, per capita income growth is facilitated by industrialisation. 3 By definition of marginal product, total industrial production is equal to the area OXYx; the industrial wage bill equals the rectangle (OX)w* and what is left, the triangle Πx, goes to profit earners. 8 As is should be clear by our diagrams, once the labour surplus phase is exhausted, industrial employment can only rise at the price of an increasing wage and, not surprisingly, the very pace of industrialisation is likely to slow down. To summarise, the basic ideas of economic development ( = industrialization) behind the Lewis – Fei – Ranis model are: a. b. c. the engine of growth is capital accumulation (no problem of demand, saving are automatically reinvested. Keynes’ preoccupation about the level of effective demand was considered something relevant for rich economies only). As development proceeds, there is a process of rural – urban migration and urbanisation. As development proceeds, the terms of trade p increase. Food prices rise because a smaller and smaller number of farmers must support an increasing number of industrial workers. Briefly: “development” is driven by capital accumulation but limited by the ability of the economy to produce a surplus of food (the lower the surplus → the higher p → the higher the industrial wage → the weaker the incentive to invest in the industrial sector). The policy implications of such a view are very much controversial and hotly debated. Consider for instance the role of agriculture. Even if we accept the residual role given to agriculture in the Lewis framework (agriculture as a source of cheap labour and supplier of a food surplus), the question is: how these potentialities of agriculture are best exploited? By taxing agriculture, which would expand industrial labour supply (it is easier to convince people to move away from agriculture when agriculture is taxed), or by subsidising agriculture (for instance helping farmers buy relevant inputs like water, fertilisers, etc.), which would expand agricultural production and the available surplus of food? And what happens when agriculture does not coincide with food production alone, but it includes the production of non-food items as well? Again: provided that technical progress in agriculture is good for growth and industrialisation (since it raises the surplus of food), are we sure that in a poor economy there are the appropriate incentives to introduce better agricultural techniques of production? In this respect, what is the role of land reforms and land redistribution? How is the Lewis picture modified by the introduction of international trade and globalisation? Is the kind of development process depicted in the model necessarily associated with a worsening income distribution (growth for whom?) or some more pleasant outcome may be envisaged? In the following sections we will complicate a bit the Lewis framework in order to address some (not all) of these policy issues. 2. SOME REFINEMENTS OF THE LEWIS FRAMEWORK. POLICY ISSUES. Technical progress in agriculture and agricultural taxation According to the "accumulationist" view of Lewis, the process of industrialization is clearly helped by maintaining for a sufficiently long period of time a low level of the industrial wage. The longer the industrial wage stays at w*, the quicker the industrialization process will be, since more profits will be available for (automatic) re-investment. But look at equation (3). It is quite clear that the level of the industrial wage is ultimately determined by the wage prevailing in agriculture and by the agricultural terms of trade, p. It follows that the policy question is: how can these two variables be kept at a low level so as to ease the industrialization process? In principle, there are several possibilities, but each of them entails serious difficulties. 9 • AGRICULTURAL TAXATION. As people move from the rural to the manufacturing sector and the economy is in the labor surplus phase, the average product of labor in agriculture goes up. It follows that if farmers are paid their average product, the agricultural wage should go up, which is bad for industrialization. One way of escaping this trap is to tax agriculture: the net payment accruing to the farmer stays at w (see footnote 1) because the excess of the average product over w goes to the government. However, by discouraging agricultural production, such a measure could provoke a (faster) rise in the agricultural terms of trade, which in turn could slow the industrialization process. It has often been argued that Africa was a typical example of excessive taxation on agriculture, and this is the reason why several African countries implemented market-oriented reforms over the last ten years. The case of Africa, for obvious reasons, deserves a closer scrutiny. We should first try to understand whether agricultural has suffered from an excess of taxation. Then we should have a look at the behaviour of the agricultural terms of trade over the last decades and finally try to study the relation between agricultural production and real producer prices. Of course, these points are closely related to each other, but let us investigate them separately. Is it true that Africa agriculture used to be too heavily taxed? To answer this question it is important to distinguish among three kinds of agricultural products: exportable (cocoa, cotton, coffee, tobacco, etc.), importable (food crops such as cereals) and nontradable (cassava, plantain, millet, sorghum, white maize and other typically “African” food staples, which are not consumed outside Africa). As to the exportable, a way of addressing the question of their taxation is to look at the margin between export prices (i.e. the world price of the crop expressed in dollars multiplied by the nominal exchange rate) and prices actually received by farmers. Of course we are discussing the case when producers and exporters are different entities (and not when producers export directly, as in the case of plantation and agribusiness based on transnational corporations). In several African countries, from the days of the independence to the early 1990s, public marketing boardswere the principal exporting entity. The difference between the price earned in the international markets by a public marketing board and the price paid by the marketing board to the farmer is to be considered a “tax” paid by the latter to the government (in light of the public nature of the marketing board). Sure, such a difference is a crude approximation of the degree of taxation since no allowance is made for marketing and transportation costs. As a consequence, we have to take the margin as an overstatement of the true degree of taxation. Moreover, it must be stressed that the degree of taxation is linked to the exchange rate. A devaluation (i.e. a higher nominal exchange rate, you need more pesos to buy a dollar) would raise the domestic currency price received by the exporters (by the marketing board). If prices paid to farmers remain unchanged (or are raised by less than the rate of currency devaluation), the tax rate will rise4. With all this caveats in mind, the evidence presented by UNCTAD (TDR 1998, pp. 156-159) does not support the widespread view that African producers have always been more heavily taxed than those in other developing countries. Of the five export crops studied (coffe, cocoa, tea, cotton, tobacco) with reference to the period 1970-1994, it is only for cocoa and tobacco that the margin between border and producer prices was significantly higher in Africa than in the other major exporters. It must be added that this kind of exportables are not consumed by the farmers as “food” and therefore, even if they were excessively taxed (which is not true, as we have just seen), this would not contribute to slow the migration of people from the agricultural to the manufacturing sector. On the contrary, it could be argued 4 However, when devaluations lead to a widening of the margin but the price paid to the export producers goes up at least a bit, this tends to raise real producers prices of export crops vis-à-vis nontradables, thus providing incentives for export. 10 along the same lines followed in the illustration of the Lewis model that this could even accelerate the industrialisation process. To see why, just go back to equation (3) and read it from the point of view of a producer of an exportable good like cocoa or coffee. The terms of trade p must be interpreted as the ratio between the price of food (not coffee or cocoa) and the price of the manufactured good; w I is the wage paid by the industrial sector expressed in units of industrial good and, finally, w is the real wage earned by the coffe or cocoa producer expressed in units of food. w can thus be thought of as the net nominal earning of the coffee or cocoa producer divided by the price of food. Well, even admitting (in a counterfactual exercice) that the taxation on the exportable good is somehow excessive, this should decrease w and therefore stimulate the transfer of people to the urban areas. Fresh workforce available for the industrialisation process. Still, the general belief that these exportables were too heavily taxed lead at the end of the 1980s to a major reform process. In most African countries the marketing boards were simply abolished and devaluations were applied to several currencies. The idea was to eliminate the tax paid by the exportable producers by eliminating the tax collector (the marketing board) and to provide further incentives to the exporters through a currency devaluation. Well, it may be surprising to see that from the beginning of the reform process until mid-90s the ratio between producer and export prices has declined for all the five products considered except coffee. Why is it that the elimination of the tax collector actually lead to a raise in the tax paid by the producers? Because the reform process was based on ideology rather than governed by pragmatism. Indeed, eliminating a marketing board does not imply that a producer (and, let us repeat, we are not talking about transnational companies and plantations) suddenly becomes able to sell to the international market. She still needs a trader, and if a public trader (a marketing board) is not there, a private one will enter into the picture. No doubt, this private trader will try to maximise his profits and, when faced with a currency devaluation, what will he do? He will increase the price paid to the producer by less than the devaluation rate and therefore there should not be any surprise to see that the reform process was actually accompanied by an increase in the degree of taxation of the major export crops5. What about the importable goods? Here we are talking about goods such as rice, wheat and maize. A local producer can be said to be taxed when the net price she gets is lower than the price paid (in domestic currency) to import the same good. On the contrary, a local producer is said to be subsidized when the net price she receives is higher than the domestic currency import cost6. Well, from 1970 to 1994 the producer price for this importable good had been systematically higher than the unit import cost, which indicates a positive rate of subsidization (UNCTAD, TDR 1998, Chart 16). In principle, this should stimulate the production of these food items, lower the terms of trade p (look again at equation 3) and thus favour the industrialisation process for any given level of the real (food) wage earned by farmers. 5 It must be stressed that a further drawback of this reform was the disappearance of the public money formerly in the pockets of the marketing board. Even when it is heavily corrupted, you can ask a state to build some roads or develop some utilities. But if the “tax” collector is a private trader, you can’t ask anything. 6 A numerical example may be useful. Imagine that the world price of a unit (a kilo, a ton, a trunk, whatever you like) of rice is 200 US$. In the country we are considering the exchange rate is 10 pesos per dollar and a tariff of 15% is applied to the import of rice. Thus the overall price paid by the locals to buy a unit of rice produced abroad is 200x10x1.15 = 2,300 pesos. If some kind of competition is at work (and usually it is), the price applied by local producers to sell their own rice will be exactly the same, 2,300 pesos. In this case, local producers are said to be (implicitly) subsidized since the price they get (2,300) is higher than the import cost net of tariff (2,000). A tariff is not the only way of protecting a local farmer. Another possibiity is to subsidize the purchase os some relevant input (water, fertilizers, etc.). In this case the consumers will pay less than the domestic currency world price to buy rice, but the net price (inclusive of the subsidy) received by the local producers will be higher than the domestic currency world price. 11 As to the non-tradables, lack of data (to our knowledge) makes it impossible to say something relevant, but the former analysis should have made clear that it is hard to say that agriculture in Africa has suffered from an excess of taxation and this has slowed or even blocked the industrialisation process. All the more so after the reforms and the dismantling of the marketing boards. At this point, a question naturally arises: what are the likely reasons of the missing African agricultural development? There is no doubt that a serious industrialisation process cannot even start without a significant growth of agricultural productivity, but the foregoing analysis should have convinced the reader that the main problem does not lie in perverse (insufficient) price incentives. The main problem for African agriculture is the weak response to price incentives: even when prices are high or go up, this is not enough to stimulate more production. There are a host of institutional and structural factors that may explain this feature. Let us see the most important. In the short run, aggregate supply response to price incentives can occur through two basic processes: either idle land is brought into use or more variable inputs (labour, fertilizers, etc.) are put at work on a given unit of land. In the long run, aggregate supply response is mainly determined by investment and productivity growth. Well, even when there are community land resources available (which is not always the case), poorer farmers simply cannot farm extra land because they cannot mobilize the necessary complementary inputs. On top of this, high levels of poverty means that farmers cannot afford to keep either their labour or land idle even at very unattractive prices. As to the intensification of the agricultural process, i.e. the use of more variable inputs on a given unit of land, one of the rason why after the reform process the supply response to higher output prices7 was weak lies in the behaviour of the input prices. Indeed, together with the abolition of the marketing boards, African farmers had to face the removal of the subsidies to buy such items like fertilizers. There is something more: even in the presence of an appropriate incentive for the farmers to put more variable inputs at work, it must be recognized that buying these extra inputs requires some credit. Well, “the marketing boards had offered an institutional response to the problem of missing private credit markets. As they had a legal monopsony over marketed output, they could provide seasonal inputs on credit against the potential crop as collateral......With privatization, this system of seasonal credit has broken down” (UNCTAD, TDR 1998, p.169). Sure, the role once played by the marketing boards is now played by private traders, but obviously the latter, being moved by the sacred principle of profit maximization, apply tougher conditions to the farmers8. And, last but not least, the money earned by a private trader cannot be used (unless you are able to impose a tax) to finance the provision of public goods. The issue of pubblic goods (or in any case private goods whose production entails strong positive externalities) is strictly connected to the long run response of agricultural output to price incentives. There cannot be any significantly positive response unless better infrastructure is available, especially transport9. Again: together with the dismantling of the marketing boards and the removal of the subsidies on inputs, “reforms” usually asked for a cut of public expenditures... 7 Remember that the basic idea behind the dismantling of the marketing boards was to increase the price directly received by the producer. 8 See the Appendix for a formal analysis of the relation between a farmer-borrower and a private traderlender. 9 The rural transport bottleneck is so important because it prompts two negative effects. First, it reduces the real return to a given investment (let us say that the cost of bringing to the market the goods produced thanks to the investment goes up) and, second, it is a source of product market imperfections (if the transport network is underdeveloped, a small farmer must sell to a trader and this lowers her market power) 12 • TECHNICAL PROGRESS IN AGRICULTURE. Technical progress is a way of producing more output with the same quantity of labor (better: this is labor saving technical progress, but, as we are going to see, there are also other kinds of technical change). In principle, this will increase the average agricultural surplus and moderate the upward pressure on the agricultural terms of trade associated with the process of industrialisation. Note that a larger agricultural surplus will come out of technical advancement in any case, even if the fruits of technical progress should be entirely consumed by the farmers. The following diagram should make the point clearer: Y’ Y Lo La Figure 6: technical change in agriculture In the diagram two very standard production functions for the agricultural sectors are represented. One (Y’) is higher than the other (Y) because of technical change (the same amount of labour produces more output). Formally, we can write Y = ALα and Y’ = A’Lα, with α positive but less than one and A’ > A. La is the available labour force in the economy. The agricultural average product of labor when everyone is employed in agriculture is Y/L = AL(α-1) with the less productive technology and Y’/L = A’L(α-1) with the more advanced. Assume, as we did before, that at least in the labour surplus phase farmers get w ( w ' with the more advanced technique) even when some of them start moving to the manufacturing sector (and now we know that this may happen because of agricultural taxation). To be more precise, the wage earned by the farmers is −1 w = ALα a and −1 w ' = A ' Lα a depending on the kind of technology in use. Clearly, w ' > w : the fruits of technical progress are by assumption enjoyed by the farmers. Well, what is the agricultural surplus corresponding to a situation where Lo people are employed in agriculture? It’s the difference between total production and the wage bill. Formally, with the less productive technique we will have: 13 α −1 α −1 α AS = ALα o − ALa Lo = A Lo − La L0 , and with the more advanced technique ' α −1 ' α α −1 AS ' = A ' Lα o − A La Lo = A Lo − La L0 . Since A’ > A, we will have AS’ > AS. So, even when the fruits of technical change are fully enjoyed by the farmers, it still remains a higher agricultural surplus, which is likely to compensate the upward pressure on the agricultural terms of trade provoked by the transfer of people in the course of the industrialisation process. As a consequence, there is no doubt that labour saving technical progress in agriculture is potentially very good for industrialization and economic development. However, there are two major problems to be emphasised. First, agricultural technical advancements are not necessarily labour saving. On the contrary, they could be labour using and, if they are, there will be a weak incentive at least for the smallhholder producers to introduce them. To see why, one has to think of Africa and look at Table 1: Table 1: Non-agricultural to agricultural income ratio. Developing regions 1950-60 1960-70 1970-80 1980-1990 Africa 7.05 8.33 8.74 7.79 Asia 1.87 3.37 3.31 3.57 Latin America 2.42 3.00 2.81 2.51 Other 1.88 2.17 2.15 2.25 Source: UNCTAD, TDR 1998, p.140 It is clear from the table that the ratio of non-agricultural to agricultural value-added per worker is much higher in Africa than elsewhere in the world. This differential is one of the key indicators of “urban bias” in Africa (agriculture too heavily taxed? If it is – but go back to our previous analysis – the farmers will have too low an incentive to invest and increase productivity and a disruptive process of rural-urban migrations might take off), but it is ultimately based on lack of investment in African agriculture and agricultural infrastructures, features that will be examined below. This differential underlies the attractiveness to farm households of “straddling” between the agricultural and nonagricultural sectors and may explain why labour using technical progress is not adopted: “..to the extent that off-farm employment opportunities are available, there is a continual pressure for productive labour to be diverted from agriculture. Under these conditions, there may be little incentive to adopt high-yielding crop varieties, which can require greater labour inputs [Italics is ours]. Rather, the types of innovation which are attractive are those which save households labour time and thus enable the diversion of labour from the farm” (UNCTAD, Trade and Development Report, 1998)10. The second problem with agricultural technical progress is due to one of the most pervasive interlinked contracts in poor countries, that between a farmer (sharecropper) who is also a borrower and a landlord who is also a lender. The issue was first formally raised by Amit Bhaduri (1973), who suggested that a landlord-lender may discourage a technical advancement by the sharecropper-borrower since the former may lose more in interest income, even though his rental income goes up, as the latter becomes better off and therefore less dependent on the landlord for consumption credit. In such a case, the 10 Section 3 will present the Harris-Todaro model, one of the most powerful tool to understand the nature of rural-urban migrations at least in the poor countries. 14 semi-feudal relation between the two subjects is the reason why technical advancements are not implemented. In other words: technical progress is not a (purely) technical issue; it’s a socio-political story. However, another Indian economist, T.N. Srinivasan (1979) tried to moderate Bhaduri’s pessimism and show that under certain conditions technical advancements may be actually implemented despite the prevailing semi-feudal rural framework in many poor countries of the world. Here we are not going to present Srinivasan’s model, but to build a counterexample inspired to that model to show that it is possible that the introduction of technical progress may increase the interest income earned by the land-lord borrower. The story goes as follows. There are two periods. In the first, the slack season, the sharecropper borrows an amount B from the landlord for consumption purposes (assume she has no other source of income and credit, which is often quite realistic), and will give this money back to the landlord in the second period together with an interest rate i. Output is harvested in the second period. The actual output is θx , where θ is a random variable with expected value equal to unity and which takes the value θ L with probability p and θ H with probability (1-p), where θ H > θ L and pθ L + (1 − p )θ H = 1 . The sharecropper gets a fraction α of the harvest, and the landlord gets (1-α). If the harvest turns out to be sufficiently high, the sharecropper will pay the rent, repay the loan and consume the rest. Formally, if θ = θ H the sharecropper in the second period will pay (1 − α )θ H x + (1 + i) B and consume θ H x − (1 − α )θ H x − (1 + i) B = αθ H x − (1 + i ) B . On the contrary, if the harvest turns out to be too small to cover the sharecropper’s minimum subsistence consumption, cs, and repay the loan as well, the sharecropper will repay the balance at a stipulated price of working γ days of labour per unit of the loan amount owed to the landlord-borrower. Forrmally, if θ = θ L the sharecropper will consume cs and will have to work (obligatory work) l = γ [ c s + (1 + i) B − αθ L x] days to repay the loan11. This obligatory work provokes disutility to the sharecropper, denoted by v(l) with v’ > 0 and v’’ > 0. In each period the overall utility of the sharecropper is u(c) – v(l), with u’ > o, u’’ < 0 and, by normalisation, u(cs) =0. The sharecropper chooses her level of borrowing B in the first period in order to maximise her lifetime expected utility: EU = u ( B) + β { p[ − v(γ (c s + (1 + i ) B − αθ L x))] + (1 − p)u (αθ H x − (1 + i) B)} , where β is the discount factor. In order to study the behaviour of the sharecropper we have to calculate the derivative of the expected utility with respect to B and then set it equal to zero: { } u ' ( B) + β − pv ' (γ (c s + (1 + i) B − αθ L x))γ (1 + i) − (1 − p)u ' (αθ H x − (1 + i) B)(1 + i) = 0 . Now, both the value of u’ and v’ depend on B and x and therefore we can study how B is affected by a change of x. By totally differentiating the first order condition we get {u' '+ β [− pv' ' γ 2 } (1 + i ) 2 + (1 − p )u ' ' (1 + i ) 2 ] dB + β [ pv' ' αθ L γ (1 + i ) − (1 − p )u ' ' αθ H (1 + i )]dx = 0 It follows that 11 It should be clear that the framework we have just described makes sense if and only if θ L < [c s + (1 + i ) B ] αx . Equivalently, it must be θ H > [c s + (1 + i ) B ] αx 15 dB β [(1 − p )u' 'αθ H (1 + i) − pv' 'αθ L γ (1 + i )] − = = >0 dx u ' '+β [ − pv' ' γ 2 (1 + i) 2 + (1 − p )u ' ' (1 + i ) 2 ] − The logic driving the outcome of this specific example is straightforward: the sharecropper knows that, thanks to technical progress, in the next period she will get more on average. Therefore, she borrows (and consumes) a bit more in the first period because, even if the harvest turns out to be bad in the second period, she will have to work less days for the glory of the landlord12. But if B increases with x (with technical progress) it is not true, as claimed by Bhaduri, that the interest income earned by the lender-landlord falls: she has an incentive to introduce technical advancements because both her interest and rental income goes up. Of course it is possible to build a different example and show that there are cases where B decreases with x (see Bardhan and Udry, p. 121). But the point to be stressed is that, however extremely interesting, the outcome proposed by Bhaduri is just a possibility, it does not necessarily hold. Lewis-growth and inequality Is the process of development and structural change as depicted by Arthur Lewis inevitably associated with an increase in the degree of inequality? We are going to see, through the help of a simple model developed by Bourgignon (1990) that it’s difficult to answer this question without ambiguities. And we will also stress that this issue, however important, should not be over-emphasised: it’s not that important. Consider again the Lewis framework, where we can see three types of social groups: 1) peasants, who are self-employed in the agricultural sector (a) and own their family farm. Their per capita income is ya; 2) workers, who are employed in the manufacturing sector (m) and earn a fixed wage, w; 3) capitalists, who employ the workers in the manufacturing sector and earn a per capita profit equal to π. Let’s introduce a (very reasonable) hierarchy of incomes: π > w > ya. This assumption (which could be removed without altering the basic message) implies that there is an excess supply of labour willing to come to the manufacturing sector, something very close to the “labour surplus” idea of Arthur Lewis. There are La peasants, Lm workers and n capitalist. To simplify things, their sum is normalised to one (so that absolute values coincide with shares) and the number of capitalists is held constant. The numéraire of the model is the manufactured good (pa/pm = pa = p as before). The manufactured good is produced with a Leontief technology, so that output Qm = AK and employment Lm = BK, where A and B are fixed technical coefficients. Hence, the income of each single capitalist is π = (Qm − wLm ) n = ( A − Bw) K n (4) The per capita income of peasants is, as before, equal to the average product of labour in agricultural, let’s call it Va. Expressed in units of manufactured good, this income is equal to ya = pVa. In this framework every peasants would like to move to industry, which means that the number of peasants is determined as a residual, all those who are not absorbed elsewhere: La = (1 – n – BK). The mean income of the overall population (equal to one, remember) is 12 Just to get the basic intuition, imagine that without technical progress the sharecropper borrowed 5 in the first period, and for simplicity assume a zero interest rate. In the second the harvest was bad, 2 only (2 is the part going to the sharecropper). Subsustence consumption is 1. Under3γthese circumstances, the sharecropper must work 4γ days for repaying her debt. Now introduce technical progress. The sharecropper could borrow 6 and, even if the harvest was bad, say 4, he would have to work only days for the glory of the landlord. 16 y = Qm + pQa = Qm + pLaV a In such a simplified framework a Lorenz curve is easy to build. % of income % of population 1-n-BK 1-n 1 Figure 7: the Lorenz curve Thanks to the hierarchy of incomes and population normalisation, the cumulated percentages of population on the horizontal axis of the diagram correspond to, respectively, the number of peasants (1 – n – BK), the number of peasants plus the number of workers (1 – n – BK + BK = 1 – n) and 1. Now, before trying to understand whether inequality increases or decreases with capital accumulation (increase in K), let us calculate the slope of the three segments of the Lorenz curve, segment A, segment B and segment C. To this purpose, let us redraw the diagram using a more general notation: Ni stands for the number of people belonging to the i-th fractile of the distribution (in Bourgignon model, i = 1, 2, 3, where 1 denotes peasants, 2 denotes workers and 3 denotes capitalists); yi is the income of each single person belonging to fractile i and N is total population: 17 % of income C By B Ay A % of population Ax Bx 1 Figure 8: the different segments of the Lorenz curve As can be see from the diagram, the calculation of the slopes is straightforward. Slope of A = Slope of B = Ay y1N1 yN y1 = = N1 N y Ax By [( y1 N1 + y 2 N 2) yN ] − ( y1 N1 yN ) y 2 = = [( N1 + N 2 ) N ] − ( N1 N ) y Bx Using the same argument, y Slope of C = 3 y In words: the slope of each “piece” of the Lorenz curve is equal to the ratio of the income of the corresponding fractile of the distribution to the mean income of the whole population. In the Bourgignon’s framework we have y1 = ya, y2 = w, y3 = π. Now, in such a framework, look again at the horizontal coordinates of the Lorenz curve and try to see what happens with capital accumulation and structural change13. Clearly, La = 1 – n – BK decreases with K, but the second kink is constant at (1 – n). This implies that if the Lorenz curve were to shift upward (i.e. the distribution were to become unambiguously more egalitarian), the following necessary and sufficient conditions have to be satisfied (look at Figure 7 to fully understand the rationale of these conditions): 13 In a Lewis-type model capital accumulation is basically the same thing as structural change. When capitalists invest in the manufaturing sector people are withdrawn from agriculture. 18 ∂( y a y ) ≥0 ∂K and ∂(π y) ≤0 ∂K (5) In words: the slope of the first “piece” of the Lorenz curve must increase (or stay constant) with capital accumulation, whilst the slope of the last “piece” must decrease (or stay constant). The economic meaning of this condition is obvious: if income distribution has to improve, the income of the poorest segment of the population must approach the mean income from below and the income of the richest segment of the population must approach the mean income from above. Well, to check whether those conditions hold we have to write down an explicit formula for the ratio of the income of both the peasants and the capitalists to the mean income of the whole population. Let’s call βa the share of agriculture in national income and αm the profit share in sector m (which is a constant equal to (1 – Bw/A)); by definition we will have: βa = pQ a pL aV a y a (1− n − BK ) = = y y y from which we get ya βa = y 1 − n − BK (6) As to the profits of the single capitalist, it can be written as Total Profits Q Qm α m m y Qm α (1− β a ) y y π= = = m n n n from which we get π α m (1 − β a ) = y n (7). From (6) and (7) we can see that conditions (5) are satisfied with certainty if the share of agriculture in national income increases with K, with capital accumulation and growth. But historically this has never occurred; well on the contrary, the share of agriculture in national income declines with economic growth. So, the reduction of βa will make π y increase, as can be seen from (7). What will happen to y a y ? As one can see by (6), a priori we can’t say that much, since the changes in the numerator and the denominator push in opposite directions. But something less vague can be said if we rewrite (6) more explicitly: y a βa = y La Now we can say the following: if, during the process of economic growth and structural change, the decline in the share of agriculture in national income is faster than that of share 19 of agricultural employment in total employment, then y a y will fall. For this condition to be met, average labour productivity must increase faster than labour productivity in agriculture, which is almost always the case. So, historically, the likely changes to be considered are: ∂( y a y ) <0 ∂K ∂(π y) >0 ∂K and With similar slopes changes we cannot be unambiguous about an increase in inequality with economic growth and structural change. The two following diagram show one ambiguous case with intersecting Lorenz curve and a case with unambiguous increase in inequality, both consistent with the slope changes we have just described. % of income % of population 1-n-BK 1-n 1 Figure 9: the ambiguous case 20 % of income % of population 1-n-BK 1-n 1 Figure 10: the increase of inequality To sum up: growth and structural change, the transition from an agrarian to a modern, industrial economy are not inevitably associated with a worsening income distribution as reflected by a deterioration of the Gini index. That said, two important qualifications must be added. First: even if inequality increases in the early phase of economic development, is this a serious problem? Consider carefully the stylised framework described by Bourgignon, in particular look at the case of unambiguous increase in inequality (the Lorenz curve moves outward, figure.....). Why is inequality higher than before? Compared to the initial situation, the new, “more unequal” Lorenz curve depicts a case where: a) there are more workers than before. Those who were already employed in the manufacturing sector are as well off as before, since the real wage is constant by assumption. The new employed are better off than before, since in the agricultural sector they got less; b) there is the same number of capitalists, and each of them is richer than before (total profits are higher than before since the share of profits in the manufacturing sector is constant and the share of manufacturing in total income is higher than before) ; c) there are less peasants than before. Are they worse or better off? We know that their per capita income is pVa. Well, p, as we saw in the previous section, is very likely to be higher than before. What about Va, the average product in agriculture? If we postulate for agriculture a standard production function with decreasing return to labour, the average product will be higher than before as well. In sum, with capital accumulation and structural change everyone gets at least the same income as before. The overall cake produced by the economy is larger than before (because labour productivity in the manufacturing sector is higher than in the agricultural sector) and the only reason why inequality has increased is that the new slices are more unequally distributed. Still, everyone has a larger slice than before and therefore no one would prefer the preaccumulation, more egalitarian world. All this is to say that inequality is more a political than a strictly economic problem, since too much inequality could threaten political stability. For 21 instance, “agricultural policy has been used in Africa to promote a pattern of income distribution which is regarded as legitimate and which therefore does not threaten political stability. This is an extremely delicate problem in nation-state building in Africa. Some aspects of agricultural pricing policy, particularly the practice of providing uniform guaranteed prices countrywide, have been part of an implicit social contract designed to redress colonial imbalances and ensure that certain ethnic groups with less fertile land and limited access to markets are not totally excluded” (UNCTAD, Trade and Development Report 1998). The second consideration relates to what we observed in the previous section: even in the labour surplus phase the industrial wage is unlikely to be constant because of the pressure exerted by a rising p. It follows that the profit share in the manufacturing sector, instead of being constant as postulated by Bourgignon, is likely to decrease, which in turn produces a move toward more equality. 3. MIGRATIONS: THE HARRIS-TODARO MODEL (1970) According to the Lewis model the process of industrialisation entails an “automatic”, someway harmonious migration of people from the rural areas to the cities. Can we say more on this migration process? Can we add, on top of the agricultural and the manufacturing sector, an urban informal sector to the picture? After all, in many poor countries there is a large urban population engaged in an extremely diverse set of activities outside the direct scrutiny of the state and not covered by labour unions. And is creating new employment opportunities in the city always a good idea? Or is there the risk of providing people the incentive to move too fast to the city, so as to create all the problems inevitably associated with the concentration of a large mass of people in a relatively small area? After all, many cities in Africa, Latin America and Asia are growing at 5-7 per cent per year, which is likely to be above any realistic possibility of giving these people a job. These questions can be addresses through the help of the model developed in 1970 by Harris and Todaro (for the subsequent changes to the original framework, see Bardhan and Udry, 1999). The key institutional assumptions of the model accord pretty well with many highly visible features of some developing countries: - the rural labour market is competitive - the wage paid by modern firms in the city is fixed above the market clearing level, either because unions’ activities or governmental legislation (for instance minimum wage regulations) or efficiency wage considerations - there is an informal sector in which urban residents not otherwise employed can earn their living out of activities outside the control of the state and performed using heir labour force alone (petty trade, craft production, urban agriculture). Let Lr be the rural population, employed in agriculture on a fixed amount of land. Agricultural output is determined by the standard production function g(Lr) and sold on a world market at a price normalised to unity. Since the rural labour market is assumed to be competitive, rural wages will be equal to the marginal productivity of labour: wr = g ' ( Lr ) . This relation is represented in Figure 11 by the decreasing right-to-left curve. The urban population is either employed in manufacturing (Lm) or working in the informal sector (Lu). Total population is normalised to 1, so that Lr + Lm + Lu = 1. To simplify the calculations we put the wage paid in the informal urban sector equal to zero (which is nothing but a way of claiming that people living in the city always prefer to get a job in the formal, modern sector. 22 wm = f ' ( Lm ) wr = g ' (1 − Lm − Lr ) e E w*m E’ wr* * wr e’ L*m 1 * * L m + Lu Figure 11: the Harris-Todaro model The manufacturing wage, wm, is institutionally fixed. Since manufacturing firms maximise profits, their demand for labour is implicitly determined by wm = f ' ( Lm ) , where f is the manufacturing production function. The probability for an urban resident of getting a job in the manufacturing sector is equal to the number of jobs divided by the number of urban residents, and her expected income in the city will be equal to this probability multiplied by the institutionally fixed manufacturing wage (remember that the wage of people employed in the informal urban sector is zero). Of course, migration will occurs to equalise the expected wage of an urban resident with the wage that the resident could earn in the rural areas: Lm (wm ) wr = w Lu + L m ( wm ) m (EC). The meaning of this equality can be better understood by describing what happens when it does not hold. Imagine for instance that Lm (wm ) wr < w . Lu + Lm (wm ) m 23 People living in the rural areas will decide to migrate to the city: But since the manufacturing wage stays constant, Lm will not change and the new urban residents will increase Lu. At the same time, the reduction of the rural labour force will increase the agricultural marginal product and therefore the agricultural wage. At the end of the story the equality will be restored. Let’s call the fixed manufacturing wage wm* . The implicitly determined level of manufacturing employment will be L*m . The equilibrium condition can be rewritten as wr ( Lu + L*m ) = L*m w*m In words: the rural wage multiplied by the urban labour force must be equal to a constant. This is the equation of a rectangular hyperbola (like yx = constant): in the diagram, the curve ee’ represents such an equilibrium locus (of course the hyperbola must pass trough the point ( w*m , L*m ) ). At points E and E’ there is an informal urban sector of size L*u , a rural population of 1 − L*m − L*u and thus a rural wage of w*r . Since E’ lies on ee’, wr ( Lu + L*m ) = L*m w*m and expected wages are equalised in the urban and rural sectors. For an even fuller understanding of the model, let us see diagrammatically what would happen should the rural wage be less than w*r . wm = f ' ( Lm ) wr = g ' (1 − Lm − Lr ) e w*m E E’ wr* e’ wr0 * Lm L*m + L0u L*m + L*u Figure 12: disequilibrium in the Harris-Todaro model As can be seen from Figure 12, when the rural wage rate is at wr0 , below its equilibrium value, there are more people employed in agriculture and less people employed in the 24 informal sector ( L*m + L0u < L*m + L*u ). But the point ( wr0 , L*m + L0u ) lies below the equilibrium rectangular hyperbola, which means that wr0 ( L*m + L0u ) < w*m L*m , i.e. the agricultural wage is lower than the expected urban wage: people will move to the city and, due to the increase of both the agricultural wage and the pool of informal workers in the city, the equilibrium condition (EC) will be finally restored (the rural wage line shifts up). In this model the informal urban sector with a very low living standard serves to equilibrate the migration process. If the fixed manufacturing wage is very high (substantially above its perfect competition level), very few people will be employed in the manufacturing sector. In the absence of an informal sector (Lu = 0), all the remainder of the population will be employed in agriculture, where as a consequence the marginal product of labour and hence the wage will be extremely low. At the same time the probability of employment would be one (since Lu = 0) and therefore the expected urban wage extremely high. Peasants will migrate but, faced with the rigidity of the manufacturing wage, will become unemployed and earn a living only thanks to the informal sector. So, according to the Harris-Todaro view, the informal sector serves essentially to: 1) provide a subsistence to the urban unemployed (what would happen with the introduction of a safety net for poor people, something like a welfare state?) and 2) guarantee the persistence of the gap in living standards between rural residents and the urban residents employed in the formal sector. However, one must recognise that the existence of an informal sector has also some negative, sometimes tremendously negative implication for the urban life (congestion, slums degradation, lack of any kind of rights’ respect, a high crime rate, etc.). For this reason, it may happen that a government tries to favour the creation of job in the urban formal sector, and to this purpose it can implement such measures as tax holidays or better treatment in the credit market for the urban manufacturing firm. This way, a governement hopes to reduce the size of the informal sector and increase that of the formal sector. But what is (could be) the final outcome of such a policy? It might happen that, because of migrations from the countryside, the pool of informal workers in the city actually increases despite the governmental policy. This can be investigated by looking at Figure 13. First, a policy aimed at accelerating the rate of absorption of labour in the formal sector may be represented by an outward shift of the labour demand curve in the manufacturing sector (for any given market wage, manufacturing firms are willing to hire more people because of the incentive they are receiving). It can be seen that in the new equilibrium there will be more people employed in the urban formal sector (as expected) and less people working in the agricultural sector (because of migration) earning a higher rural wage than before (because of the increase n their marginal product). More precisely: in the initial equilibrium, before the implementation of the governmental policy, the manufacturing wage is m and manufacturing employment is a; in the rural sector there are 1 – c workers, each getting a rural wage equal to q; finally, in the urban informal sector there will be 1 – a – (1 – c) = c – a workers. The governmental policy shifts both the labour demand curve in the manufacturing sector and the equilibrium locus (the rectangular hyperbola) up, as indicated by the arrows. If the labour demand (marginal product) curve in the agricultural sector is LRd , in the new (after-policy) equilibrium there will be b manufacturing workers (with b > a, as expected); 1 – d agricultural workers, each getting a rural wage equal to r (it can be seen that 1 – d < 1 – c and r > q, as expected); as to the urban informal sector, the pool of workers is now equal to d – b. 25 LRd LRd ' m r s q o a b c d e 1 Figure 13: a pro-urban/formal sector policy The question is: is d – b < c – a? Or, to put it differently: does the governmental policy succeed in shrinking the pool of informal urban workers? A priori, we simply cannot answer to this question. But we can say something, notably that the answer ultimately depends on the slope of the labour demand curve in the agricultural sector. Indeed, ' imagine that the relevant agricultural demand curve is LRd , which is flatter than LRd . In this case the effect of the governmental policy is to reduce even more the number of agricultural workers (there will be more migrants to the city), so that in the new equilibrium there will be e – b informal urban workers, which is clearly greater than d – b. So, in general, the flatter the agricultural labour demand curve, the more likely is that the absolute size of the urban informal sector goes up despite the aim of the policy is to reinforce the urban formal sector. Basically, the free choice of the peasants to move to the city can render the governmental policy even counterproductive. There are two points that are worth stressing, the first related to the relative size of the urban informal sector, and the second to the economic meaning of the slope of the labour demand curve in the agricultural sector. As to the first point, from the equilibrium condition (EC) we can immediately infer that after the introduction of the governmental policy the relative size of the urban informal sector (its size measured as a fraction of the total urban sector) must have diminished, irrespective of what happens to its absolute size. However, what really matters in policy and social terms (congestion, crime rates, diffusion of diseases, etc.) is the absolute size, the relative being quite unimportant. As to the slope of the agricultural labour demand curve, it is clearly a measure of the elasticity of labour demand to the real wage: the flatter the curve, the more responsive labour demand. In the limit, with a horizontal curve, agricultural labour demand is perfectly elastic at a given wage rate and we are back to the Lewis case of surplus labour. In such a case, any increase in the 26 manufacturing, formal employment will be accompanied by an equivalent (in percentage terms) increase in the urban informal employment14. The city is larger than before, but the proportional expansion of the formal and informal sectors has compromised the realisation of the government’s objectives. This general principle can be applied to other policies as well, not necessarily linked to formal labour demand: “ ..policies aimed at directly reducing urban congestion (say, by building more roads), reducing pollution (say, by building a subway), or increasing the provision of health (say, by building new public hospitals) might all have the paradoxical effect of finally worsening these indicators......because fresh migrations in response to the improved conditions ends up exacerbating the very conditions that the initial policy attempted to ameliorate” (Ray, pp.381-382). Exercises: 1) What are the effects of technical progress in agriculture in the framework of the Harris-Todaro model? 2) What are the effects of introducing full flexibility of the manufacturing wages (instead of having them fixed)? 3) What are the effects of technical progress in the manufacturing sector? 4. LAND REFORM • • • Is the enormous inequality in land holdings bad for agriculture productivity? If it is, can land rental markets and/or land sale markets spontaneously redress the balance, reduce such inequality and therefore allow for a productivity increase? If not, what is the role of a land reform? Let’s start from the first question. Basically, we can contrast two opposite arguments (in that they produce contrasting conclusions), a technological argument and an economic argument. The technological argument rests on the notion of economies of scale and leads to the conclusion that large land holdings (and then a certain degree of inequality in land holding) are good for productivity growth in agriculture. Think of those techniques that allow farmers to achieves a high productivity level: Draft animals. A minimum size of the plot is needed to use them in an economically viable way. Imagine you have 2 bullocks, which can be productively employed only on a plot of at least 1 hectare. But you just have a plot of ½ hectare. There would be no problem if you could rent one of the bullocks out, but such a rental market is usually very thin, for two reason: • 14 If you rent the bullock out, it could be overworked or mistreated (and you would lose value, for you and your sons). This point may be better understood by rewriting the equilibrium condition (EC) as wr = 1 wm . ( Lu Lm ) + 1 Since neither the manufacturing nor the rural wage change under the labour surplus assumption, formal and informal labour in the city must grow at the same rate from one equilibrium to the other. 27 • Inside a village (the “natural” dimension of a market, especially when infrastructure and transport facilities are poorly developed) there is often an almost perfect correlation in the use of animal power. Machinery (tractors, pump sets….). Here the minimum size required for efficient ownership is even higher (despite the scope for a rental market is somehow better). So, from a strictly technological point of view, no discussion: large plots of land are more productive than small plots of land. Inequality in land holding is good for productivity growth. But now consider the “economic argument”, based on people’s incentives to put as much effort as possible in the production process. We will see that, from this perspective, small holdings (and then a certain degree of equality in land holding) are good for productivity growth. 2 agents Landowner Tenant risk-neutral risk–averse Let us specify the technology in agriculture: p G Y = B (1 − p ) with G>B There are two possible arrangements between the landowner and the tenant: Fixed Rent Contract: the tenant pays R to the landowner Sharecropping Contract : the tenant pays a fraction s (sY) to the landowner. In terms of efficiency (the effort put in the production by the tenant) a fixed rent contract is to be preferred to a sharecropping contract (the reason is easy to understand: if you are the tenant and have to decide whether to put some extra effort in jour job, what do you do in case of fixed rent contract? And what do you do in case of sharecropping?). So, why do we observe so many sharecropping contracts all around the world? Sharecropping P Tenant (1-s)G Landlord sG (1-p) (1-s)B sB 28 29 Fixed rent P Tenant G-R Landlord R (1-p) B-R R Let’s fix “s” in such a way that the expected return to the landlord is the same and, given his risk–neutrality, he is indifferent between the two contracts: psG + (1-p)sB = R, or s(pG+(1-p)B) = R, so that s* = R pG + (1 − p)B Tenant’s return in the good state Fixed rent (G-R) Sharecropping (1 − s*)G = 1 − R GR G = G − pG + (1 − p) B pG + (1 − p) B In order to compare these two returns we can calculate the difference between them: GR G <0 G − − ( G − R ) = R 1 − pG + (1 − p ) B pG + (1 − p ) B <G It follows that in the good state the tenant gets more with a fixed rent contract. Tenant’s return in the bad state Fixed rent Sharecropping (B-R) (1 − s*) B = 1 − R BR B = B − pG + (1 − p) B pG + (1 − p) B In order to compare these two returns we can calculate the difference between them: BR B >0 B − − ( B − R ) = R1 − pG + (1 − p ) B pG + (1 − p ) B >B 30 It follows that in the bad state the tenant gets more with a sharecropping arrangement. Overall, the tenant prefers sharecropping because of his risk–aversion. Indeed, if one is risk-averse she will prefer the option that is better in the bad state. Since, by construction, the landlord is indifferent, the negotiation between the two will come up to a sharecropping agreement (with s slightly above s*). So, despite its inefficiency (in terms of effort provision), sharecropping is the predominant agrarian agreement. What is the prevailing argument? The argument based on technology (large holdings are good for agricultural productivity) or the argument based on economic incentives (small holdings are good for agricultural productivity because in a small plot land can be cultivated directly by the farmers and his families, no need to reach any agreement – necessarily: an inefficient sharecropping agreement - with external people. There is therefore a strong incentive to put the maximum possible effort)? The evidence suggests that the incentive argument is more relevant and productivity is higher on small plots of land (see Ray). However, this raises different policy questions: Pooling Land and Cooperatives Drawing on what we have just said, it is tempting to claim that small farmers (who do not need to respect any agreement with external agents) should pool their lands (basically, form a cooperative) to take advantage of economies of scale. The validity of this argument depends on whether the source of economies of scale lies at the cultivation (production) level or outside the cultivation process. If it lies outside the cultivation process, for instance because the advantage of pooling comes from marketing (a big, pooled subject is able to get better prices on the market), the cooperative works: land is cultivated separately in small plots and then the fixed cost of setting up a marketing group can be pooled and shared But when the source of economies of scale lies at the production level (say, through the use of tractors), then the incentive problem returns with full force: additional effort by one farmer leads to additional output, which is then shared among the team. If the farmers fail to internalise this positive externality (which requires a complete sense of altruism) effort will be undersupplied. So, do not expect to see successful wheat or rice cooperatives (sectors where the economies of scales originate at the marketing level), and do not be surprised to discover that collectivisation in China lead to a tremendous reduction in agricultural productivity. Land Rents and Land Sales So, a more egalitarian distribution of land would increase agricultural productivity and therefore help sustain the industrialization/development process. A quite natural question arises: why doesn’t a large landowner sell his land splitted into smaller plots to several small farmers? After all, this should be a very good deal for each part of the dealing. To see why: 31 Figure (14): a large plot subdivided into 4 small plots Well, we can expect the market value of a plot of land (even more generally: the value of any asset) to be equal to the discounted value of the stream of future profits generated on that land (why?). Consider the larger plot in Figure (14) and assume that its market value, when it is owned by a single landlord who will need to sign a sharecropping arrangement with one ore more tenants, is 80. In other words: per hectare productivity is such that the discounted flow of future profits generated on this large plot is 80. Now, what is likely to be the value of the smaller four plots when they are owned by four distinct smallholders? It will be, for instance, 25, the relevant point being that 25x4 = 100 > 80. The small plots can be cultivated by the family farmers without the need of any (inefficient) sharecropping tenancy arrangement and therefore per hectare productivity will be higher, which in turn increases the discounted stream of future profits and the market value of the land. So, why doesn’t the large landowner sell voluntarily his land to four smallholders (getting 25 x 4 = 100) instead of keeping it and getting profits (present value) equal to 80? Why, in other terms, is the land market so thin? In a world where the credit market is imperfect (asymmetric information, monopolies, etc.), the value of the land has two components: • • discounted flow of future incomes (as we saw). Let’s call D this component; land as collateral (measured by the profitability of the additional loans one can get in that he/she owns land). Let’s call C this component. Now, if a buyer must obtain a loan to buy the land (which is typically the case for a small buyer) and must mortgage that very piece of land for the loan, then he can’t reap the benefit of the land as a collateral (C) until the loan is paid off. It follows that For the Potential Seller: D+C D`>D15 Potential Buyer: D` If D + C > D` (and it could be), there will be no market for land. 15 In the previous example, D is 20 (80/4) and D’ is 25. 32 So: how to realise a more egalitarian distribution of land and reap the productivity gains that should follow? We are left with very few standards: • • “Revolution” (political upheavals in society: Cuba, Japan, Korea, Taiwan). Government Public Intervention (Land Reform): International Agencies (NGO`s) The role of public institutions – either the governments or some international agency or NGO – is, from a strictly financial point of view, to pay for the difference (D + C) – D’. APPENDIX: THE INTERLINKED CONTRACT BETWEEN A TRADER-LENDER AND A FARMER-BORROWER Suppose the output produced by the farmer depends only on the working capital she has to borrow: F(K) is the standard concave production function of the farmer. The farmer borrows this working capital from “her” trader, i.e. the person who will be in charge of marketing farmer’s output (the farmer is often too poor, and the rural infrastructure too poorly developed, to be able to go directly to the market). The opportunity cost of capital to the trader is the rate r (the rate he has to pay to borrow this money from a formal bank or the rate he gets from a deposit in the formal bank). Should the samll farmer borrow from an alternative source (for instance a formal bank), she would pay the rate r0 > r. Let i the interest rate charged by the trader-lender, with (1 + i) = α (1 + r), and for the moment α can be less than or equal to or greater than unity (the value of this parameter will emerge as an optimal and endogenous choice of the trader-lender). The market price of output is p, but the price offered by the trader to the farmer is q = βp, with the value of β to be determined endogenously. α and β are chosen by the trader. It follows that the farmer will take them as given and maximizes her income Y = βpF ( K ) − α (1 + r ) K (A1), by choosing K. The first order condition is βpF ' ( K ) = α (1 + r ) (A2) The minimum income the farmer can get even without entering the contract with the trader is Ymin = pF ( K min ) − K min (1 + r 0 ) (A3), where Kmin satisfies pF ' ( K min ) = (1 + r 0 ) . The trader will maximize his income 33 π (α , β ) = (1 − β ) pF ( K ) + (α − 1)(1 + r ) K (A4), by choosing α and β. The trader will have to take into consideration two constraints, known in the literature as the incentive-compatibility constraint and the participation constraint. The former coincides with (A2): it says that the trader will have to chose a value for α and β by taking into account that his choice will affect farmer’s choice of K and therefore his own profits. The latter says that the choice of α and β must be such that the farmer does participate to the transaction, instead of being pushed away (if the trader tries to make too much money, the farmer will simply go to a formal bank or to some other moneylender). So, formally the problem of the trader is to choose α and β in order to maximise (A4) subject to (A2) and Y ≥ Ymin. One way of solving this problem is to observe that (A4) can be rewritten (and this is obvious) as the difference between what the trader gets from the market (net of the cost of capital) and what is left to the farmer: π = [ pF ( K ) − (1 + r ) K ] − Y But one thing is sure: the trader will choose α and β so as to press down Y to Ymin (there are no reasons from the point of view of the trader to let the farmer make more money than what is sctrictly needed to induce her participation to the transaction). Since Ymin does not depend on α and β, it follows that trader will maximize his income when pF(K) – (1+r)K is maximized, that is to say when pF’(K) = (1+r). But compare this condition with (A2): it must be α = β. Let us call γ the common optimal value of these two parameters. From (A1) the value of farmer’s income can be rewritten as Y = γ [ pF ( K * ) − (1 + r ) K * ] (A5), where K* is the value of K such that pF’(K*) = 1 + r. Using this propety, (A5) can be rewritten as Y = γ [ pF ( K * ) − pF ' ( K * ) K * ] (A6). But, as we have already stressed, it must be Y = Ymin, and thus γ= F ( K min ) − F ' (K min )K min G(K min ) = F (K * ) − F ' (K * )K * G(K * ) (A7). To check whether γ = α = β is greater than or equal to or lower than unity, we have to understand the behaviour of the function G and keep in mind that Kmin < K* as r0 > r. Since by definition of G we have G’ = - F’’K > 0, it must be γ = α = β < 1. The economics is relatively simple: the farmer is given an interest discount (i < r), which is compensated by the underpayment in the output market (q < p). Exercises i. ii. What is the effect of a reduction of r0? In the limit, what would happen if r0 = r? Can you give a policy interpretation to this result? Imagine that your objective is to prevent underpayment to farmers in the output market: what is to be preferred in efficiency terms, an intervention directly in the output market or an intervention in the credit market? 34 iii. What would happen in this model if the government decided to launch a program of public works in the city? Main References Bhardan, P. and C. Udry, Development Microeconomics, Oxford University Press, 1999; UNCTAD (United Nations Conference on Trade and Development), Trade and Development Report 1998, Geneva, 1998; Ray, D., Development Economics, 1999; 35 36