Portal de Periódicos da CAPES

Do rent‐seeking and interregional transfers contribute to urban primacy in Sub‐Saharan Africa?

2011; Elsevier BV; Volume: 92; Issue: 1 Linguagem: Inglês

10.1111/j.1435-5957.2011.00400.x

1435-5957

Kristian Behrens, Alain Pholo Bala,

Regional Economics and Spatial Analysis

We develop an economic geography model where mobile skilled workers choose to either work in a production sector or to become part of an unproductive elite. The elite sets income tax rates to maximize its own welfare by extracting rents, thereby influencing the spatial structure of the economy and changing the available range of consumption goods. We show that either unskilled labour mobility, or rent-seeking behaviour, or both, are likely to favour the occurence of agglomeration and of urban primacy. In equilibrium, the elite may tax the unskilled workers but does not tax the skilled workers, and there are rural-urban transfers towards the agglomeration. The size of the elite and the magnitude of the tax burden that falls on the unskilled decrease with product differentiation and with the expenditure share for manufacturing goods. All these results are broadly in line with observed patterns of urban primacy and economic development in Sub-Saharan African countries. Desarrollamos un modelo de geografía económica donde los trabajadores cualificados móviles pueden eligir entre trabajar en un sector productivo o formar parte de una élite no productiva. La élite fija las tasas de impuesto sobre la renta a fin de maximizar su propio bienestar mediante la extracción de rentas, lo que influye en la estructura espacial de la economía y produce cambios en la gama disponible de bienes de consumo. Mostramos que es probable que tanto la movilidad laboral no cualificada, como los comportamientos de búsqueda de rentas, o ambos, favorezcan la aparición de aglomeración y de primacía urbana. En el equilibrio, la élite podría gravar a la mano de obra no cualificada, pero no grava a los trabajadores cualificados, y existen transferencias rurales-urbanas encaminadas a la aglomeración. El tamaño de la élite y la magnitud de la carga fiscal que recae sobre la mano de obra no cualificada disminuyen con la diferenciación de productos y con el reparto del gasto en bienes manufacturados. Todos estos resultados están en consonancia con los patrones observados de primacía urbana y desarrollo económico en los países del África subsahariana. It seems to be the thinking that Africans obtain an education in order to enter the government bureaucracy so as to be able to share the wealth of the nation rather than create wealth for the nation (Bassey 1999, p. 106). Several explanations have been put forward to explain the urban growth in Sub-Saharan Africa (henceforth, ssa). On the one hand, there are push factors linked to rural population growth, environmental degradation, climate change (Barrios et al. 2006), and armed civil conflicts (Lozano-Gracia et al. 2010). On the other hand, there are pull factors which are essentially linked to the attraction exerted by higher incomes and productivity, as well as a larger range of available goods and services, in the urban centres on rural migrants. Productivity improvements and higher incomes may stem from agglomeration economies arising from the diversity of intermediate goods, the matching process in the local labour market, from knowledge spillovers across firms, or any combination of these. There is an extensive literature on the mechanisms that generate aggregate increasing returns to scale at the urban level (see Fujita and Thisse 2002; Duranton and Puga 2004). While evidence for agglomeration economies in developed countries and in Asia is fairly well documented, this is not the case for ssa. Yet, agglomeration economies may play an important role in African urban development. According to McGranahan et al. (2009), careful analyses of the relationship between economic change and urbanization show that in Africa, as elsewhere, urbanization has been associated with economic growth. The potential role that African cities might play in the development of their continent is even clearly praised in the 2009 World Development report which claims that "urbanization, done right, can help development more in Africa than elsewhere" (World Bank 2008, p. 285). However, agglomeration economies in ssa seem less important than those in Asia and in OECD countries (Collier 2006). Indeed, because countries in that region are too small and not integrated enough, many African cities tend to be too small compared to the optimum. As shown by Au and Henderson (2006) for the case of Asia, this may have serious impacts in terms of foregone growth. Research on agglomeration economies and international competitiveness further suggests that latecomers to industrialization, such as Africa's natural resource exporters, face a disadvantage linked to the spatial distribution of the global industry (Page 2008). As the overall productivity of Sub-Saharan Africa is fairly low compared to international standards, the economic drive does not seem that important in explaining city growth in that region. Indeed, the economic content of African cities "lacks the dynamism, specialisation, diversity and economies of scale normally associated with urban life" (Bryceson and Potts 2006, p. 324). Hence, complementary explanations are required to understand the urban growth spurt occuring in that region. One cause for Africa's rapid urban growth may be the political role of its primate cities. The location of the central administrations in African primate cities has favoured their development well beyond what is economically feasible and expected. The extent of the bureaucracy is often the most characteristic aspect of city development in ssa. Employment is predominantly generated in the government sector and parastatals (Bryceson and Potts 2006). Twenty one out of the 27 African cities belonging to the 100 largest cities in the world that had the fastest population growth between 1950 and 2000 are capitals. They are the principal location of state administration and public entreprises. The preponderant share of capitals among large and fast growing cities is an African specificity. Indeed, this ratio is much lower in other regions of the world – 10 out of 34 in Asia, four out of 21 in Latin America & Carribean, and 0 out of 4 in North America (Satterthwaite 2005). The dominant role of capitals within the national urban hierarchy is further strengthened by the fact that a disproportionate part of the national budget is spent within them – which explains why they are usually the subject of massive in-migration by those seeking economic and political opportunities not available elsewhere. This is where African political economy kicks in: political leaders allocate a potentially economically indefensible share of resources to primate cities to satisfy the urban mob because they fear the pressure of city elites (Lipton 1983; Bairoch 1985). Spatial proximity is likely to increase political influence in non-democratic countries since the leadership is more sensitive to the claims of the urban elites than to those of people living in the countryside. Hence, as empirically shown by Ades and Glaeser (1995), politics directly affects urban concentration as rent-seeking agents have to be spatially close to the political power. The rents they reap, together with the government's net transfer of resources from the countryside to the cities to run the state, further raise city population by attracting economic activity to the main centres of purchasing power.1 The net result is that still more population is attracted by these transfers, which draw ressources from the hinterland and make cities in the developing world absorb a disproportionate fraction of overall economic activity. Eventually, the process becomes self-reinforcing and leads to the formation of large urbanized core regions, as emphasized by the new economic geography (see Fujita et al. 1999; Fujita and Thisse 2002; Baldwin et al. 2003). The agglomeration process described in the foregoing highlights that economical, political, and spatial factors should be jointly taken into consideration when analysing the formation of large urban agglomerations in the developing world.2 Stated differently, political rent-seeking, rural-urban transfers, migration, increasing returns, and geography should be important ingredients of the whole story. The main objective of this paper is to present a new economic geography (henceforth, NEG) framework that combines these ingredients to shed some light on agglomeration, regional imbalances, and urban primacy in developing countries. Although the NEG literature has been rapidly growing these last years there are, to the best of our knowledge, only a few contributions dealing with political factors. Robert-Nicoud and Sbergami (2004) present a model that analyses the impacts of political factors, increasing returns, and economic integration on agglomeration. Their main objective is to explain how and why the regional integration process of the European Union leads to important transfers of resources from the 'core regions' to the 'peripheral regions'. Using a probabilistic voting model, the authors endogenize regional policy and transfers that are determined by the swing voters of the ideologically most homogeneous group. As the latter is predominantly located in the countryside, the peripheral regions can obtain transfers because of their relative political homogeneity. By contrast, the large region will keep the 'core' only if its relative economic size overcomes its political weakness due to ideological fragmentation. Robert-Nicoud and Sbergami's (2004) main result, namely that the political process leads to a more even distribution of economic activity than the market mechanism, is quite opposite to what we seem to observe in many developing countries, namely transfers of resources from the hinterland to the cities, which increases agglomeration. Furthermore, a democratic process, as embodied in the probabilistic voting model, does not adequately characterize the political environment of many developing countries, especially in regions like ssa. We thus propose an alternative model in which the 'political process' consists in agents deciding on whether or not to enter a political elite in order to extract rents to maximize their own welfare. In such a setting, rent-seeking leads to rural-urban transfers that exacerbate regional imbalances, thus showing that the nature of the political process matters for the direction of net transfers and the degree of agglomeration. This illustrates that "when farmers form a majority of the population, they tend to subsidize the urban minority. When farmers form a minority, the urban majority subsidizes them" (Friedman, as quoted by Mbeki 2005, p. 5). The remainder of the paper is organized as follows. Section 2 presents some facts about urban primacy and its links with corruption in Sub-Saharan Africa. In Section 3, we develop the model and discuss the market outcome. Section 4 then investigates the spatial equilibrium and shows that rural-urban transfers make the emergence of agglomeration more likely. Section 5 deals with the issues of elite formation and tax setting, and derives the equilibrium taxes and the equilibrium size of the elite. Section 6 concludes. The urbanization process in ssa has led to a major population redistribution between rural and urban areas. Even among the urban centres, this redistribution has been very skewed towards the very large cities and metropolitan regions (Mabogunje 1994). ssa countries have, as a result, become increasingly affected by regional imbalances and urban primacy.3 Table 1 illustrates the prevalence of urban primacy in ssa. As one can see, there are only few countries in that region with primacy levels below 15 percent – South Africa and Nigeria being among those few. While South Africa, with its comparatively high level of economic development, is characterized by a more balanced urban system, Nigeria departs from the typical regional pattern because its large population is hardly compatible with the existence of a single primate city. We can further see from Table 1 that while primacy is typical of small countries like Gambia, Burundi, Rwanda and Togo, where the capitals dominate the economic landscape, it is also characteristic of larger countries like Gabon, Angola, Congo, Ivory Coast and Senegal. Another noteworthy feature is that the biggest cities in Africa experienced the fastest growth rates. The share of ssa urban population living in cities of more than 500,000 inhabitants rose from 6 percent to 41 percent between 1960 and 1980, and the number of such centres increased from three to 28.4 Moreover, the number of agglomerations with at least one million inhabitants jumped from two in 1950 to 21 in 1990 (Max Miller and Singh 1994). Smaller cities grew at a much slower pace. According to the United Nations (2002), while the average population growth rate of cities with fewer than 500,000 inhabitants was of 3.8 percent, those of cities with between one and five million inhabitants and of cities with more than five million inhabitants were of 6.67 percent and 5.35 percent, respectively. Figure 1 reveals that urban primacy is significantly correlated with central government expenditures and with corruption.5 Put differently, corrupt countries with large central government expenditures on goods, services, and compensation of employees, generally have larger primate cities than the remaining countries. As can be further seen from Figure 1, ssa countries (depicted in red) are more strongly affected by these two factors. Hence, more corruption and larger central government expenditures are more strongly associated with urban primacy in ssa countries than in other parts of the world.6 Corruption, central government expenditures on goods and services and compensation of employees, and urban primacy (ssa countries in red, the other countries in black; simple ols regression lines) To further assess the link between urban primacy and corruption in a multivariate regression setting with more controls, we also make use of panel data. Our econometric model that follows is essentially based on Davis and Henderson (2003), who derived their specification from a theoretical framework modeling rural-urban allocation and primacy. Two comments are in order. First, Equation (1) does not explicitly include fixed effects because their presence would entail an identification issue for the time invariant variables. However, since country fixed effects allow to control for unobserved cross-country time invariant factors that may impact urban primacy, it is important to include them (Davis and Henderson 2003). In Appendix C, we present a method that allows to estimate consistently time invariant variables in a panel setting with fixed effects. Second, specification (1) is prone to endogeneity issues as discussed at length in Davis and Henderson (2003). We also describe in Appendix C the two-step procedure that allows to obtain consistent three stage least squares (3SLS) estimates of time varying variables. Table 2 presents estimation results for Equation (1). Column (1) gives pooled ordinary least squares (OLS) results with time invariant covariates. The only time-varying covariate coefficient that is significant is the coefficient of the log of urban population. It is negative as expected: a greater urban scale entails that the economy can support more cities, reducing the population share of each of them. Column (1) yields other interesting results: primacy increases with the capital city dummy, which suggests that the extent of bureaucracy allows capital cities to achieve higher size than other primate cities. Urban primacy also decreases with the landlocked dummy. This supports the claim that long distances, deep divisions, and poor integration of transportation systems badly affect the size of African primate cities (Collier 2006; World Bank 2008). Results regarding the impact of transportation infrastructure are informative as well. While primacy is unaffected by railways and waterways densities, it is positively affected by the density of roadways. However, it decreases with the share of paved roads. In words, a larger stock of roads raises urban primacy, but the higher the quality of the roadways network, the lower the urban concentration. One may also notice that, in the pooled OLS regression, the non-corruption index has a positive and significant impact on urban primacy. Its interaction term with the ssa dummy is negative, however, thus suggesting that there is a ssa-specific effect. Since the balance of the two terms is negative (−0.0016), more corrupt ssa countries have higher urban primacy. Our pooled OLS estimates are prone to two kinds of problems: the first one is the neglect of unobserved heterogeneity, and the second is endogeneity. To mitigate the first issue, we present fixed effects estimates in Column (2). Time invariant regressors, estimated in a second step, are displayed in blue. Concerning time-varying covariates, the estimated coefficients of urban population and its square suggest that primacy varies non-monotonically with urban population. However, primacy declines with urban population throughout the sample range of the urban population variable.10 Observe that most of the fixed effects results for time invariant covariates are in line with pooled OLS estimates: primacy increases with the capital dummy, and decreases with the share of paved roads. Moreover, urban primacy increases significantly with the non-corruption index but declines with the interaction term. As with pooled OLS, the balance of the two terms is negative (−0.0014). Conversely to pooled OLS results, the fixed effects estimate of the landlocked dummy is no longer significant. Moreover, urban primacy now declines significantly with the logarithm of land area. Fixed effects estimation implies a substantial loss of degrees of freedom due to the inclusion of country dummies. Multivariate regression estimates from the first difference of (1) allows us to mitigate that issue and to obtain more efficient estimates. Column (3) gives results for this regression procedure. It shows that primacy declines with urban population throughout the sample range of the urban population variable. It also indicates that primacy increases as income rises, peaks, and then declines. The peak is at an income of about 2,120 US$, in line with the Williamson hypothesis. Concerning the road infrastructures covariates, we find the same kind of results in Column (3) than in the pooled OLS specification (1): primacy decreases with the share of paved roads while it increases with roadways density. Findings concerning the capital city dummy are similar to previous ones, confirming that being the seat of central administration and of political institutions increases the size of primate cities. Results related to the non-corruption index are also similar to those of Columns (1) and (2): the non-corruption index and the interaction term have opposite signs, with the same negative balance as in Column (2). Our 3SLS estimates in Columns (4) and (5) allow us to handle endogeneity. They yield coefficients of the ln(urban population) and square of ln(urban population) that have higher absolute values than in previous findings. Concerning time invariant covariates, most of the results in Column (4) are in line with previous results. Yet, small differences can be noticed: the interaction of primacy with the ssa dummy is still negative, as in previous results, but the coefficient of the non-corruption index is no longer significant. Hence, corruption favours urban primacy in ssa, whereas it has apparently no significant impact in other regions of the world. In Column (5), we finally add the current value of the democracy index as a time-varying covariate. We obtain a negative coefficient that is significant at the 10 percent level. This suggests that the less democratic a country is, the higher its primacy. As indicated in Table 3, ssa countries are more autocratic than countries in other regions or groupings, except those of the Middle East and North Africa. This further drives up average urban primacy in ssa. To sum up, our panel data analysis suggest that corruption, autocracy, landlockedness and central administration in capital cities raise urban primacy in ssa countries. Therefore, urban primacy in ssa may stem from other factors than in the rest of the developing world: high corruption which leads to a redistribution of purchasing power from the country-side to the cities, where it is spent by an essentially unproductive elite. ssa may be the developing region where politics has the strongest impact on regional imbalances and primacy. In what follows, we propose a 'simple model of economic geography' that captures these features. We extend the analytically solvable NEG model of Forslid and Ottaviano (2003) to formalize the interactions between an elite ('corruption'), the redistribution of purchasing power across regions ('interregional income transfers'), and agglomeration ('urban primacy'). We consider a country with two regions, labelled 1 and 2. Variables associated with each region will be subscripted accordingly. The population consists of exogenously given masses of unskilled and of skilled workers. All workers are geographically mobile. The skilled may be either production workers or part of an unproductive elite – on top of being geographically mobile they are thus also socially mobile. The unskilled are socially immobile and are always production workers. Yet, contrary to the skilled, they can choose between working in the formal or in the informal sector of the economy (e.g., because they are harder to monitor and to tax than the skilled).11 All agents spend their incomes locally and work in the region they live in. In what follows, we denote by S the mass of the skilled in the production sector and by E the mass of unproductive skilled constituting the political elite. Both S and E are endogenously determined, with . We denote by 0 ≤λS≤ 1 and 0 ≤λU≤ 1 the share of productive skilled S and unskilled workers living in region 1. We assume that the political elite is clustered into a historically determined center of power, which we henceforth refer to as the capital of the country (e.g., the historical capital). We do not attempt to endogenously determine where this centre is located. Although this is an interesting question, it is secondary to the aspects we are interested in. While skilled workers are thus a priori mobile across regions, they become de facto immobile if they want to be part of the elite. In other words, social mobility of the skilled can only occur in the capital and once they are part of the elite they become regionally immobile.12 Without loss of generality, we assume that region 1 is the capital. Our model may be viewed as a game with four stages: (i) the elite sets the tax rates tU and tS for the unskilled and the skilled; (ii) skilled workers decide whether or not to enter the elite; (iii) skilled production workers choose the region they live and work in; (iv) firms maximize profits, production and consumption take place. We solve that game by backward induction.13 There are two production factors, skilled and unskilled labour, and two sectors, manufacturing and agriculture. The agricultural sector produces a homogeneous good using unskilled labour only. We assume that this good is costlessly tradable across regions. Without loss of generality, we normalize the unit input coefficient in that sector to one. Perfect competition and costless trade then imply that the unskilled wages wU are equalized across regions: , where the last equality comes from our choice of numéraire.15 All unskilled workers are a priori free to work in agriculture or in manufacturing. Hence, the wages for unskilled in the two sectors will be equalized. Skilled workers can choose to remain in the production sector to earn a wage wi in region i, or to get involved in politics and become part of the elite in the capital. The benefit of belonging to the elite is to participate in running the country and to extract rents for personal consumption. The elite determines the tax rates for the different groups of agents. To keep things simple, we suppose that the elite levies proportional income tax rates tS and tU on the incomes of skilled and unskilled workers in the manufacturing sector.16 The agricultural sector is assumed to be informal and untaxed because it is hard to monitor. The unskilled hence face the choice of working either in the formal manufacturing sector and to pay taxes, or to work in the agricultural shadow economy, hence evading taxation. Put differently, the unskilled tax base is not perfectly inelastic but generally shrinks with the level tU of taxation. This observation fits well with the empirical fact that all African countries have large shadow economies (fiscal evasion seems to be fairly easy due to the lack of enforcement). Recent estimates using a sample of 37 African countries indeed reveal that the average share of the shadow economy is about 43.2 percent of GDP (Schneider 2004, Figure 3.1.1). The skilled do not work in the shadow economy and can thus evade taxation only by becoming part of the elite. for any given value of S, the price index in a region decreases with the share of firms located in that region ('regional market crowding effect'); and for any given distribution of firms, the price indices in both regions decrease with the mass S of productive skilled workers ('global market crowding effect'). This second effect, which implies that a smaller mass of productive skilled workers decreases welfare by reducing product diversity and by increasing consumer prices, will be important in the subsequent analysis of the elite's behaviour. We first analyse the market outcome for any given allocation of the skilled between the production sector and the elite, and for any given spatial distribution (λS, λU) of skilled and unskilled workers across regions. Furthermore, tU and tS are considered fixed at this stage. Some comments are in order. First, both and are increasing in the unskilled-to-skilled ratio , which will itself be endogenously determined later in our analysis. The reason for this is that, on top of standard endowment effects, a larger supply of skilled production workers increases the mass of competing firms, which leads to global product market crowding and, therefore, lower equilibrium wages. Such an effect does not arise in standard NEG models where the mass of firms is usually proportional to the exogenously fixed mass of skilled workers (Krugman 1991; Ottaviano et al. 2002). Second, since the denominator of does not depend on λU, we clearly have and . In other words, there is 'complementarity in agglomeration' as the clustering of the unskilled in one region raises skilled workers' wages there and reduces skilled workers' wages in the other region. Hence, agglomeration forces in our setup are stronger than in traditional models where the unskilled are immobile and assumed to be evenly spread across regions. Finally, standard but longer calculations show that is decreasing in tS, whereas is increasing in tS. Stated differently, increasing taxation of the skilled shifts nominal wages in favour of the capital region and away from the periphery. The reason is that as tS increases, the elite spends proportionally more tax revenues on varieties produced in the capital region, thereby raising demand and wages there. As we show later, the widening interregional wage gap induced by taxation increases the tendency for agglomeration of the mobile sector. Having a unique and well-behaved solution λU(λS) – as well as complementarity between λU and λS– allows us to reduce an a priori complex problem with mobility of both the skilled and the unskilled to a simpler problem where we can only focus on the mobility of the skilled, but take into account the implied distribution of the unskilled. In words, the essentially two-dimensional problem boils down to a one-dimensional one, which will make the analysis of the equilibrium (and its stability) easier. A spatial equilibrium is such that no worker (neither skilled nor unskilled) has an incentive to change location, conditional upon the fact that product markets clear at the equilibrium prices and factor markets clear at the equilibrium wages. Formally, a spatial equilibrium arises at λS*∈ (0,1) and λU*=λU(λS*) when ΔUS*(λS*) = 0; or at λS*= 0 and λU*=λU(0) if ΔUS*(0) ≤ 0; or at λS*= 1 and λU*=λU(1) if ΔUS*(1) ≥ 0. Following Fujita et al. (1999), an interior equilibrium is said to be stable if and only if the slope of the indirect utility differential ΔUS* is negative in a neighbourhood of the equilibrium, whereas the two agglomerated equilibria are always stable whenever they exist (recall that ΔVU* is always downward-sloping in λU for any given value of λS, i.e., there is never locational instability because of the migration incentives of the unskilled). Appendix A.3 presents the benchmark case without taxes and with symmetrically distributed immobile unskilled labour. This case has been previously analysed by Forslid and Ottaviano (2003) and we report their results as a benchmark only. Note that our model boils down to theirs when (tU, tS) = 0 and β→∞, in which case λU= 1/2 irrespective of the spatial distribution of the skilled. The foregoing results can be summarized as follows: Proposition 1 (mobility of the unskilled and agglomeration). When the unskilled are mobile across regions: (i) the capital hosts a larger share of the unskilled population; (ii) full agglomeration in the capital is more likely and can occur for higher values of transport costs than in the case where the unskilled are immobile. Proof. See Appendix A.4. ■ Proposition 2 (taxation and agglomeration). When compared with the no-tax case, (i) in case of an interior equilibrium, the