Planeación y Desarrollo Volumen XXXI, Números 3 y 4 (2000) 379-452
Geography and Economic Development: A Municipal Approach for Colombia*
Fabio Sánchez1 Jairo Núñez'1'
1 Universidad de los Andes b Universidad de los Andes
Abstract
The objective of this document is to determine the relationship between the geographical variables and per capita income, per capita income growth, population density and population growth of colombian municipalities. In order to carry out econometric cstimations at a municipal level we constructed a sct of geographical variables based on soils, climates and roads maps. Additionally, we extracted some other geographical variables from the physical homogeneous zonc statistics of the colombian Institute of Geography (IGAC).
* This documcnt vvas presentad to Research Network Project, Office of the Chief Econotnist, Inter-American Development Bank.The authors thank the National Planning Department of Colombia for financing the digitalization of the maps used in this studv. Scveral people collaborated in the construction and gathering of the data. We thank to Jaime Jiménez, Jenny Mendoza, Martha Parra, Jorge Alberto Sánchez, Gabriela Suárez and to the publie offlcials of the Cadastral offices of Bogotá, Medellín and Antioquia. We also thank the comments and insights of Alejandro Gaviria, John Gallup, Eduardo Lora from the IADB and to theparticipants in the seminar held in Cuernavaca, México. We benefited from comments from and discussions with I.uz Helena Chamorro, Javier Birchenall, Angela Cordi, Cristina Fernández, Samuel Jaramillo, José Leibovich,Santiago Montenegro, Juan Ricardo Ortega, Humberto Mora, Beatriz Barona, Ulpiano Avala, Juan Luis Londoño, Alberto Carrasquilla, Armando Rodríguez, Roberto Steincr, Adolfo Meisel, Juan Zapata and the rest of the participants in the seminars held at DNP, Fedesarrollo and Universidad de Los Andes. We also thank Beatriz Barona, Vivian Barrios and Ana María Fernández who helped us with the graphs and tables. We are grateful to Moritz Kreamer from the IADB for the detail comments to earlier versión ot this document. All errors and omissions are solé of the authors. 'E-mail: [email protected] y jnunezlituniandes.edu.co
379 1. Sánchez y |. Núñez | Planeación y Desarrollo XXXI i Julio-diciembre 201)11 (379-452)
VCe found that geography weighs both for the level of municipal per capita income and íor its growth. Thus, geography explains betwecn 36% and 47% of the municipal income per capita variance and between 35% and 40% of the municipal income per capita grovth variance. It was established that, among the geographical variables, distance to domestic markets and soil suitability have the largest influcnces both on per capita income and on its growth. Other important finding was that geographical variables sccm to be significant for the poor municipalities than for the rich ones. Thus, in the poor municipalities, geography explains betvveen 25% and 32% of per capita income variance and betvveen 24% and 27% of per capita income growth variance. In contrast, in the rich municipalities, geography matters less. It explains betvveen 18% and 25% of per capita income variance and betvveen 16% and 17% of its rate of growth variance. Thus, geography affeets income level and its growth through the productivity of land, the availability of natural resources such water and rivers, the presence of tropical diseases, and agglomcration. Although geography influences the fate of a región that is not theend. There is an important role for human action either through publie policy or prívate intervention. Kducation, infrastructure and better institutions can boost economic regional economic growth and help poor regions to overeóme the poverty trap of lovv income and lovv economic growth.
I. | Introduction
The recent economic debate has retaken the oíd question about the vvealth and poverty of nations. The ansvver to this question has involved every aspect of human Ufe: education, religión, institutions, technology and diffusion of knowledge, and more recently geography. The latter, it is said, contributed to shape the destiny of nations and people (although not inescapably). "Hisroryfollowed different courses for different peoples because of differences among people's environments" points out Diamond (1999). The idea that geography and environment influenced social development is not new. A leading historian as Landes (1999) argües that european economic advantage come, in part, from its more favorable rain regimes and mild seasons, which allowed europeans to raise better and bigger animáis than those of other lands. Larger animal meant an advantage in agricultural vvork, transport and war, advantage that surely helped to strengthen the \X estern Kurope economic leadership for several centuries. Geography clearly has some effect in shapingthe economic history of countries and regions. Hovvever, hovv much geography accounts for the current differences in per capita income and, in general, in cconomic development is a question that begins to be answered. In this document we will attempt to ansvver such question for the colombian inter-regional per capita income differences.
The purpose of this paper is to examine the impact of the geographical and other social and economic variables (human capital, infrastructure and living standards) on per capita income, per capita income growth, density of population and population growth at a municipal level. We use observations at a municipal level forseveral reasons, among them a) the geographic differences vvithin the colombian
cidepartments are quite large which makes the departmental averages of some variables meaningless. Ior instance, the temperature of Cundinamarca, department located right in the center of the country (and one of the richest regions) varíes from ven- hot in some towns to freezing cold in others. b) The great variance in the provisión of social and infrastructure services across town and cities in the same
w> I. Sánchez y J. Núñez ! Planeación y Desarrollo XXXI | |ulio-diciembrc 20U0 (379-452)
department. c) The localization of particular type of businesses (manufacturing, mining coffee, etc.) in some towns and not in others, which may have important effeets on income level differences among municipalities of the same department or región.
Moreover, given the considerable number of colombian municipalities it will be possible to introduce in the economctnc exercises a large set of explanaton- variables, which have been found in the economic literature, related to income and growth. Many of these variables are generally absent of the colombian regional studies of growth, based on departments, because of the small sample (22to 30 observations) and degrees of freedom problems.
This document is anattempt to explain the root of the differences both in the level and rate of growth of per eapita income, and population density of the colombian municipalities during 1973-1995. The paper is divided in five sections. The first is this introduction. Section two describes some characteristics of colombia's geography. Section three presents the sourecs of information and the constructionprocess of the geographical variables. Section four analyzes the econometric exercises of unconditional growth rates of income and population. Sections five examines the effeets of geography on municipal per eapita income, section six its effect on municipal per eapita income growth and section seven its importance for population density and growth. Section eight determines the role ofgeography in Ínter regional income differences and municipal income inequality. Section nine is dedicated to conclusions.
II. | Some Colombian Geographical Features and Descriptive Statistics
A. Geography and Fragmented Markets
Colombia is a country with great geographical contrasts, deep differences in the economic and social development among its regions and even within each región, and regional markets somehow developed around the so-called intermedíate cities. The economic history of Colombia has been marked by the existence ot geographical barriere, some of them not overeóme until very late in the 20th century. These barriere made very difficult inter-regional trade, brought about a scarce or nuil integration of some regions to the world markets and gave origin to a ven' fragmented domestic market (Bushnell (1996)). Moreover, the roads and railroads built at the end of the 19th century were aimed at connecting towns and villages within the same región. They did not connect regions due to the high costs of construction on and acrossthe colombian cordilleras, reinforcing the regional economic fragmentation of thecountry.
At the beginning of the 19th century, according to the travelers of the time, the trip from the both Sea and River Port of Barranquilla, on the Caribbean, to the Inland River Port of Honda was made on the Magdalena River and lasted around 80 days. Reaching the capital of the country (Bogotá) located on the so-called Bogotá plateau, at two thousand and six hundred meters above sea level, took about 8 more days by mulé. The trip from the Caribbean Coast was almost impossible by land because of the
3X1 I-'. Sánchez v ]. Núñez | Planeación y Desarrollo XXXI ! Julio-diciembre 20(10 (379-452)
mountains, the lack of roads and the high probability ot acquiring a tropicaldisease. In 1930, eighty years after the steamboats were introduced, the trip from Barranquilla to Bogotá was reduced to 12 days1. The lack of roads in Colombia still persists and it is not surprising it is one of the Latin American countries with the lowest road density2.
Since the colony, the population of Colombia concentrated in the western and northern parts of the country, where, late in the 19th century, thecoffee production would flourish and the early manufacturing would be born. Throughout the 19th century until the 197()'s most of the population lived in rural áreas scattered around small town and villages, whose primarily source of income was agriculture and livestock. In the second half of the 19th century coffee consolidated in the center mountainous part of the country as the main agricultural and exporting product of Colombia. The urban population established in the main four cities, mentioned below, of the country . The main sea and river port of Colombia, Barranquilla had its golden age during the late 19th century and through the first half of the 20th century. She became the largest city in the colombian Caribbean with an intense commercial activity, prompted in part by the foreign immigration (Posada, 1998). With the construction of the seaport of Buenaventura on the Pacific Coast, Barranquilla lost a great deal of its economic Ímportance. The new port was located nearby Cali, in the middleof the so-called "Cauca Yalley", which became the largest and richest city in the western part of the country. The city of Medelkn became the most important city of central Colombia and has been the heart of the so-called "coffee axis". During the 19th century she was the core of the gold commerce andat the beginning of the 2()th century turned into a manufacturing center. Finally, Bogotá, the capital of the country, located on the eastern cordillera, has historically been the colombian largest city of the country, although not necessarily the most dynamic. In fact, some authors have argued that the urban colombian development had, until the 1980's, four heads (Cuervo and González, 1998; Goueset, V., 1998). However, starting the1980's the economic growthand development of Bogotá left behind the rest of the main colombian cities. This trend continued in the 1990's in spite the opening and trade liberalization processes that were suppose the to prompt economic development around ports and borders (Fernández, 1999).
As said Colombia's economic development has been often characterized as having "four heads". Clearly, the colombian physical geography, that gave birthto a fractionated territory, explains a great deal of such pattern. Thus, the flows of rural population during the process of urbanization were as well fragmented and headed for the cities core of the regional markets, in particular the four mentioned aboye. The poor road networks contributed to curb as well the early concentration of the population and to fraction economic activity and markets (Gouset, Y, 1998).
The colombian soils and dimites are highly heterogeneous. For instance, in the so-called "warm thermal floor" there are eighteen types of physical homogenous zones, in the "médium thermal floor" four and the "cold thermal floor" seven. In each one of these zones interact the physical characteristics of land with the humidity lcvcl, the temperature and the quality of soils.
'• A history of the navigation on the Magdalena Rivcr is presented in Poveda (1998). J The impact of roads (or the lack of them) on the per capita income, per capita income grouth, their differences as well on GDP per arca and population distribution among the Colombian municipalities will also be analyzed in this document.
}82 !■'. Sánchez y ). Núñez ■ Plantación y Desarrollo XXX] ! (ulio-diciembre 2(100 (379-452)
How and through what channels the geographical endowment determine the course of development of the difterent regions in the same country? The channels of transmission from geography to economic development are related to land productivity and technology, extent of tropical diseases, access to markets,urbanization and geographical fragmentation (Lora, K., 1999). The influence of these geographical features, tor the colombian municipalities, on the income level and its growth, on the GDP per land and on the population distribution and its growth will be examined.
B. Colombian Regional Economic Differences
There are important differences in the economic and social development among colombian regions. Some of these differences have widened whereas others have decreased considerably3. Thus, the departmental income per capitacoefficientof variation was 0.22 in 1975. Increase to 0.28 in 1988 and 0.38 in 1995. The dramatic increase in the coefficient of variation may imply that the income per capita differences among the departments have increased. This has lead many authors to conclude that the colombian regions have not experienced income per capita convergence (Meisel et al. (1999); Birchenall etal. (1998); Rocha et al. (1998)4.Just togive anexample, the 1960 Bogotá per capita income was 1.8 times the Caribbean region's and difference that increased to 2.6 in 1995.
The departmental population growth has also been ven- different among departments. Bogotá is the región whose population has grown the most. The density of population increased 4.3 times, from 824 people per square kilomcters to 3,597 between 1960 to 1995. The Pacific Región density population grew 2.4 times while Antioquia's 2.3. The population density of the Central (1.86 times) and Caribbean regions (1.42 times) have had the lowest rate of growth.
In Colombia, there have been no attempts to explain municipal income level and its growth (income, taxes or any other proxy). This study will use a large set of municipal variables, some available in publie sector statistics and other constructed by the authors from maps (soil maps, erosión maps, road maps, etc) and other priman sources, which may explain the differences in per capita income level and income growth. The municipal per capita income varies a great deal in Colombia. In 1995, the non- weighted municipal average per capita income was, at 1975 prices, $12,731 (around SUS 1,500) the median per capita income was $7,139 and $18,929 standard deviation, figures that show the great income level dispersión among municipalities. The rate of growth of the municipal income per capita has some dispersión. As well any of the per capita income proxies used show that it grew at (non- weighted) an annual rate cióse to 1 % (using taxes as a proxy)or around 0% (usingthe per capita GDP proxy). The median statistics show, however, that the income growth rate was positive for most of the municipalities and fluctuated between -5% and 5% for most of the municipalities.
' Differences in access to social ancl publie services have decreased considerably. (See Sánchez et al., 1999). 'The paper by Cárdenas et al. (1992) was the tlrst to conclude that colombian departments have experienced a convergence process. However, this hypothesis was soon challenged by Meisel (1992), and Meisel et al. (1999), Birchenall et al. (1997), Rocha y Vivas (1998), Soto (1998) and Montenegro y Suárez (1998). F. Sánchez y J. Núñcz | Planeación y Desarrollo XXXI | julio-diciembre 2000 (379-452)
The regional differences of income per capita are also important. Thus, according to Table 1, the 1973 average per capita income of the Caribbean and the Andean regions were very similar (around$16,000 in 1975 pesos) and the Pacific Region's around 20°/» lower. However, the 1995 per capita income of the Caribbean región was similar to its 1973 income. In contrast, the Pacific Región per capita income grew faster than the national average (2.3% versus 1.92%) and got cióse to the Andean" Región income. Then according to these results there has been a convergence process between the Pacific and Andean regions while the Caribbean Región lagged behind. The income level dispersión is also important within regions (Table 1). Thus, the per capita income coefficient of variation in 1995 was 0.62 in the Andean Región, 0.75 in the Pacific and 0.76 in the Caribbean one.
Table 1 Weighted Descriptive Income Statistics
Total Countty Mean Standard Deviation
Per capita income growth, 1973-1995 0.019 0.025
Per capita taxes growth, 1973-1995 0.045 0.030
Income per capita 1973 1,6043.2 9,873.5
Income per capita 1995 2,4397.3 15,323.8
Taxes per capita 1973 138.6 13.7
Taxes per capita 1995 36".3 31".5
Andean Región Mean Standard Deviation
Per capita income growth, 1973-1995 0.018 0.107
Per capita taxes growth, 1973-1995 0.029 0.373
Income per capita 1973 16,402.6 10,701.5
Income per capita 1995 24,321.7 15,965.2
Taxes per capita 1973 141.9 117.2
Taxes per capita 1995 264.1 191.9
Caribbean Región Mean Standard Deviation
Per capita income growth, 1973-1995 -0.001 (1.030
Per capita taxes growth, 1973-1995 0.039 0.035
Income per capita 1973 16,272.5 12,561.6
Income per capita 1995 16,081.5 12,246.0
Taxes per capita 1973 59.1 32.6
Taxes per capita 1995 138.3 113.2
!\Yithout including Bogotá.
3X4 F Sánchez y J. Núñez ] Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
Table 1 (continuation) Weighted Descriptive Income Statistics
Pacific Región Mean Standard Deviation
Per capita income growth, 1973-1995 0.024 0.033
Per capita taxes growth, 1973-1995 0.042 0.081
Income per capita 1973 13,241.1 9,650.7
Income per capita 1995 22,258.9 16,142.2
Taxes per capita 1973 122.9 121.7
Taxes per capita 1995 305.3 262.0
Orinoquian Región Mean Standard Deviation
Per capita income growth. 1973-1995 0.023 0.032
Per capita taxes growth. 1973-1995 0.054 0.032
Income per capita 1973 23,732.7 9,843.0
Income per capita 1995 38,974.0 26,477.6
Taxes per capita 1973 65.3 37.0
Taxes per capita 1995 208.0 137.4
Amazonio Región Mean Standard Deviation
Per capita income growth. 1973-1995 -0.008 0.038
Per capita taxes growth. 1973-1995 0.035 0.044
Income per capita 1973 19,258.9 10,619.9
Income per capita 1995 16,199.9 14,397.0
Taxes per capita 1973 23.5 23.5
Taxes per capita 1995 50.0 32.8
Bogotá D.C. Mean
Per capita income growth. 1973-1995 0.011
Per capita taxes growth. 1973-1995 0.061
Income per capita 1973 27,561.6
Income per capita 1995 35,417.2
Taxes per capita 1973 286.1
Taxes per capita 1995 1,058.2
385 1:. Sánchez v |. Núñez | Plantación y Desarrollo XXXI |ulio-diciembre 20011 (3~lM52¡
The Orinoquia and Amazonia regions have had a very difterent growth performance. \\ hile the Onnoquia income grew above the national average, in part due to oil exploitations, the Amazoma's stagnated. These regions, extensión are more than halt of the country's territory but only a small part (although growing) of the colombian population lived there.
III. | Data Sources and Constmction of theVariables
The results ot the research project presented in this paper were based on a set of municipal variables covering income, economic activity, social conditions, education, health, infrastructure, crime, and geography. This set of variable are supposed to be related to the local (municipal) level of income, its rate of growth, GDP per land, population and its rate of growth. The way in which the variables were constructed is key to understand and interpret thc results.
A. Income Variables
The income proxy used in this paper is the municipal property and the municipal industry and commerce tax revenues. \\e use these variables since they reflect both wealth and level of economic activity in the municipality, which are expected to be related to local income level. Based on municipal tax revenue \vc computed per capita income by calculating the share of cach municipality in the total municipal tax revenue ot a particular department and then multiplying such share by the departmental GDP. \\e carried out that process for 1973 and 1995.
B. Geographical Variables
Thc geographical variables were computed based on different maps of soil, rivers, roads and tables of the municipal homogeneous physical zones. For instance, the computation process to obtain a soil suitability Índex, that wc will use in the econometric exercises later, was done as follows: a) digitalization of thc soil maps; b) calculation of theárea of each typc of soil for every municipality of Colombia. \\e used the classitications of soil by aptitude, agro-ecological zones and degree of soil erosión, extracted from the colombian soil maps. The Colombian Institute of Geography elaborated these maps''. c) estimation of a soil Índex for every municipality for each typc of soil classification. In order to compute the índex we ranked the soils by its production suitability and
The ñame ol rhe insrimre is Insriruro deograheo Agusrín Oxla/zi.
3X6 F Sánchez y ]. Núñez | Planeación y Desarrollo XXXI ¡ |ulio-diciembrc 2(100 (379-452)
then calculated the weighed average of such ranking . The weigh used was the share8 in each municipality of every type of soil.
By using the same process described above we computed the municipal density of rivers by Upe (primary, secondary and tertiary) and the density of roads also by type (in 1970 such types were primary, secondary, under construcción and railroads). We also used the characteristics of the so-called homogeneous physical zones extracted from the municipal maps to construct indexes of soil suitabiüty, water availability, land slope and quality of roads. The number of physical homogeneous zones in each municipality depends on the physical characteristic differences within a particular municipality (IGAC, (1998) see Appendix 1. We calculated indexes for more than 600 hundred municipalities based on more than 20,000 thousand homogeneous zones.
We also calculated the distance of each municipality both to the main colombian markets (four main cities): Bogotá, Medellín, Cali and Barranquilla and to the most important scaports: Barranquilla and Buenaventura, the first one located on the Caribbean and the second on the Pacific. The calcularions were made based on the cartographic information of each colombian municipality.
C. Population and Human Capital Variables
The variables of population, demography and human capital were extracted and computed from the 1973 Census. The population and demographic variables computed were municipal population, migración rate, and percentage of population between 0 and 6 years oíd, 7 and 17,18 and 65 and more than 65 years oíd. The human capital variables calculated were enrollment rates in primary and secondary school, collcge graduates asa percentage of the labor forcé, average years of schooling of the labor forcé and its variance, and indexes of poverty and misery (both based on unsatisfied basic needs). We computed as well from the 1973 census, as a measure of income distribution, the variance of the personal income logarithm for each municipality.
We also computed a tropical disease incidence Índex defined as the percentage of total deaths caused by tropical diseases. The N'ational Statistics Department records the number of deaths and its causes for each municipality since 1979.
D. The Segregation of New Municipalities
One of the main difficulties in the elaboración of this paper was how to deal with the creation of new municipalities since more than 120 were created during 1973-1995. For this we programmed to always
The soil types and their ranking are presented in the appendix. s The share was calculated by running the Auto Cad software on the digitalizated maps and thegrid of the colombian municipalities. F. Sánchez y |. Xuñez | Planeación y Desarrollo XXXI ' |ulio-diciembre 2(100 (379-452)
maintain the municipal political división of 1973. Thus, every new municipality created was added back to the municipality ít was segregated froni.
IV. I The Process of Municipal Population and Income Growths
A. Population Growth
Between 1973 and 1995, the annual rate of growth of the colombian population was 2.4%, but it reduced from 3.4% during 1973-1985 to 2.1% during1985-1995. However, the population rate of growth has differed across regions and towns. Thus, the rate (non-weighted) for the Colombia Andean Región reached 1.24%, for the Pacific 3.4%, for the Caribbean 3.8% and for the ()rinoquia 4.7%. Between 1973 and 1995 the rates of population growth show persistence'1 if we use the whole national sample. However, at a regional level only the rates of population growth forthe Andean Región did showsome persistence. Thus, regressing the 1985-1995 rate of population growth on the 1973-1985 yield for the whole country a coefficient 0.13 with an R-square of 0.019. The same regression for the Andean Región yields a coefficient of 0.13, for the Caribbean of 0.39 and for the Pacific of 0.025, although the coefficients of the last two regions are not significant (Table2).
Besides the weak persistence results of the rates of population growth, we find convergence in the rate of growth of population density among the municipalities of Colombia. Regressing the 1973- 1995 rate of population density growth on the initial population density yields a coefficient of- 0.0072 with an R-square of 0.043. If the same exercise is done for the 1973-1985 period, the regression yields a coefficient of-0.012 with an R-square of 0.08. However, the coefficient found for the period 1985-1995 is 0.00053 and statistically insignificant. The result implies that around the mid-1980's the convergence dynamics of the population rate strongly decreased (Table 1). Although some of the coefficients in the population regression are significant, the "unconditional density of population convergence model" explains quite a little of the dynamic of the population. This means that the forces of regional and local densification must be linked to other sources.
Persistence means rhar the región and municipaliries wirh highest population growth during 19~3-1985 have aiso the highest population growth during 1985-1995.
Mít 'lable2
TheMunicipalGrowthProcess
PersistenceofpopulationGrowth
Dependentvariable:PopulationGrowth1985-1995 TotalCountryAndeanRegiónCaribbeanRegiónPacificRegiónOrinoquianRegión
Constant 0.0185 0.0136 0.0159 0.0331 0.0362
(12.653)v'(13.862)'' -1.5411 (13.983)v (5.004)1
Populationgrowthrate1973-1985 0.1368 0.1327 0.3966 0.0253 0.0640
(4.239)' (4.5913)' (1.2516) (0.7125)y (0.6141)
No.ofobscrvations 914 575 140 146 41
R2 0.0193 0.0354 0.0112 0.035 0.0095
DensityofPopulationConvergence1973-1995
Dependentvariable:PopulationGrowth1973-1995 Country Total AndeanRegiónCaribbeanRegiónPacificRegiónOrinoquianRegión
Constant 0.0414 0.0005 0.0315 0.0686 0.0583
(9.476)" (0.1342) (2.7226)' (7.365)" (6.295)v
Populationdensitv1973 -0.0072 0.0017 -0.0032 -0.011 -0.0133
(-6.4437)'" (1.758)1 (-1.177) (-5.062)3' (-2.619)v
No.ofobscrvations 915 576 140 146 41
R2 0.0435 0.0053 0.0013 0.2065 0.2234 'Viible2 (continua/ion)
TheMunicipalGrowthProcess
DensityofPopulationConvergence1973-1985
CaribbeanRegiónPacificRegiónOrinoquianRegión Dependentvariable:PopulationGrowth1973-1985TotalCountry AndeanRegión
0.01052 0.0359 0.0912 0.0638 Constant 0.0181 (4.1486)' (3.804)' (1.500) (2.8080)' (5.800)'
-0.016 0.0005 -0.0025 -0.0036 -0.0194 Populationdensitv1973
(-1.9634)' (0.446) (-1.3445) (-1.1266) (-4.944)'
576 140 146 41 No.ofobservations 914 0.1342 0.0002 0.0020 0.0239 0.1941 R2
PercapitaIncomeConvergence
CaribbeanRegiónPacificRegiónOrinoquianRegión Dependentvariable:IncomeGrowth1973-1995TotalCountryAndeanRegión
0.2757 0.1175 0.0548 0.1375 Constant 0.1319 (3.3474)' (7.7592)' (5.1144)' (1.2082) (3.9638)'
-0.0282 -0.0147 -0.0118 -0.0099 -0.0171 Percapitaincomegrowth1973
(-3.2964)' (-7.673)' (-4.5337)' (-2.038)2 (-4.1983)'
576 139 146 41 No.ofobservations 914
0.093 0.0817 0.0297 0.1186 0.1758 R-
17Significantat90%.
2/Significantat95%.
■vSignificantat99"'n. F. Sánchez y J. Núñez | Planeación y Desarrollo XXXI I julio-diciembre 2000 (379-452)
B. Income Growth
In this section, the municipal ratc of growth is going to be explained by using a simple versión of the Swan-Solow model. According to this model the economy reaches its "steady state" at the point in which the ratc of growth of capital/labor rado (k) is equal to ¿ero. This occurs when the saving, sf'(k)/k, equals the "effective" (including technological change) depreciation Une, (n+¿) (Aghion and Howitt, 1998; Barro and Sala-i-Martin, 1995). If k
Grapb 1 Dynamics of The Growth Rate
Growth rate >0
n + 5
s . f (k)/k
To test the Solow-Swan model, authors often estímate the following equation
(1) Y; = OC0 + Pyu)+ (p*\, +E¡ where yÁs the income per capita rate of growth of country, región or municipality between 0 and /, y is output per capita at /, and A'is a vector of control variables, such as saving rate s and popularion rate n. \X e suppose in this model that the valué of the control variables is the same across municipalities. The above equation was estimated for the colombian municipalities. Thus, regressing the rate of growth of per capita income on the per capita level of1973 yields a negative and statistically significant
391 Y. Sánchez y |. Núñez | Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
coefficient of -0.014 and anR-square of 0.09 (Table 2). The results imply that although there is unconditional income per capita convergence among municipalities it explains a little of income growth1". If we examine per capita income growth by región \ve find similar results. For Ínstance, the econometric exercisefor the Andean Región yields a coefficient of-0.012, for the Caribbean Región of -0.01, for the Pacific of -0.017 and for the Orinoquia of -0.028(Table 2 and Graph 2).
Graph 2 Relationship betvveen Income Per Capita and some Explanatory Variables
Geographical Variables
Altitude Above Sea Level Soil Suitability índex
i ■+ it-
12- 12- „.• í1. • í °° * V *4 8 o° 10- 10- • .• °°°¡ %jÜ^'Í5.s¿í i. i 6 c
c c 8 8- ° ° L. B 5- 5- 0 Ó ■3 6 o 8 o 6- °
4- 4 •
2 2 2 4 6 8 10 12 0.0 0.5 1.0 1.5 2.0 2.5
Altitude Above Sea Level Soil Suitabilitv índex
Distance to Domestic markets
5.0 5.5 6.0 6.5 7.0
Distance to Domestic Markets
:" The negative coefficient of initial income in the growth regression implies that, if \vc take two municipalities with the same rates of investment and the same level of effíciency, the poorer one will grow more quicklv for a transitional period. The reason for these "transitional dynamics" is that rclatively poor municipality have lower stocks of physical and human capital. Henee the marginal product of extra capital is higher in this economy. See Temple (1999) for a discussion on growth estimates.
i 92 I-'. Sánchez y J. Núñez | Plancación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
Graph 2 (continuation) Relatíonship between Income Per Capita and some Explanatory Variables
Infrastructure Variables
Proportion of Households with Electrical Power índex of Quality and Availability of Roads
14-, , 14,
12 o ° • « ° ",H •I 'I 10.
4. 4-
0.0 0.2 0.4 0.6 0.8 1.0 0.5 1.0 1.5 2.0 2.5
Proportion of Households with Electrical Power 1973 índex of Quality and Availability of Roads
Human Capital
Migration Rate (1973) Enrollment Rate in Primary and Secondary (1973)
0.0 0.2 0.4 0.6 0.5 1.0 1.5 2.0 2.5
Migration Rate Enrollment Rate in Primary and Secondary"
39) F. Sánchez v ). Núñez | Planeacic'm y Desarrollo XXXI | Julio-diciembre 2(100 (7T9-452)
V. I Geography and the Level of Per Capita Income
A. General Considerations
In this section \ve wül examine the role of the geographical and other variables in explaining income per capita differences among eolombian municipalities. Aceording to Table 1, the differences in income per capita between municipalities and regions are significant. Thus, in 1995 the Bogotá income per capita amounted $35,427, in contrast with the Caribbean Región income per capita (average of municipalities) of S7,250(1975 pesos), the Andean Region's of $14,490 and the Pacific Region's of S7,272. \X e will attempt to answer several questions relatedto the dispersión in income per capita ot municipalities, among them ¿what explain the income per capita differences among the eolombian municipalities. ¿How much the geographical differences matter for such differences? ¿How much the other factors contribute (education, health, infrastructure, municipal governmcnt transfers, proximity to domestic markets, etc.) to the income per capita differences?
There existsa clear relationship between geography variables and income as has been pointed out by Gallup and Sachs (1998). However, this paper is an attempt to measure such relationship in the eolombian case using municipal data. A first look to the mentioned relationship is depicted in Graph 3. It is observed that the 1995 level of per capita income is positively correlated, among the geographical variables, with soil suitability Índex and negatively with distance to domestic markets. Additionally, municipal income per capita is positively correlated with human capital variables such as 1973 years of schooling of labor forcé, 1973 school cnrollment rates and 1973 migration rates and with infrastructure variables such as road Índex and 1973-electricity coverage rate.
Grapb 3 Conditional Convergence
k(0)poor k* k(0)rlch k*nch
394 I. Sánchez y J. Núñcz | Plantación v Desarrollo XXXI ¡ Julio-diciembre 2(1(1(1 (379-452)
B. Global Results
The model \ve use to cxplain the income per capita gaps among the Colombia municipalities follows Gallup and Sachs (1998) and Rappaport (1999). To capture the impact of geographv in income per capita differences of colombian municipalities wc have carried out two econometric exercises. The first onc contains the regression of municipal income per capita on only geographical variables, which are clearly exogenous to income. These variables are: altitude above sea level, rain precipitation, distance to domestic markets and seaports, soil suitability, flatness of land, availability of water and closeness to main colombian rivers. In the sccond econometric wc add variables that are considered in the economic literature linked to income levéis: infrastructure, human capital, living standards and institutional fea tures.
Table 3 presents the results of income per capita regressions. Column (1) shows that by far the most signiheant coefficient in the regression is proximity to the main domestic markets. The coefficient of this variable is negative (meaning the farther away from the market the lower the per capita income). The rest of the geographic variable coefficients have the expected sign. Thus, precipitation (cubic centimeters of rain per ycar and indicator of unhealthiness) has a negative influence on the level of per capita income. The altitude coefficientsthat were introduced ina quadratic way have also the expected signs. The lineal specification has a positivesign whereas the quadratic a negative one. Thus, low and hot lands but also high and cold lands have, ceteris paribits, lower per capita income than land in modérate tropical climates. The Índex of soil suitability, constructed by the authors, has the expected positive sign. The better the soil the higher the income. The variable Cauca and Magdalena, which takes the valué of one for the municipalities along these rivers, have a negative sign although Magdalena is never significant. This may be due to the decreasing importance of these rivers in the economic life of the country during the last decades. As a whole and according to column (1) the geographical variables explain 36% of the variance of municipal income per capita. Most of the coefficients are significant, have the expected sign and confirm for a particular country what Gallup and Sachs (1998) found for the entire world. The results valídate as wcll the strong effect of the market proximity on income. Thus, the proximity to large markets facilitates the supply of intermedíate goods, makes easier to find the right workers, and allows information exchange among firms in the same industry clustering together (Krugman, 1991).
Column (3) presents the econometric results for the IGAC sample", which allows thevariables including availability of water (in a quadratic form) and proportion of fíat land in the municipality. The water coefficients have the expected sign and are highly significant, which indicates that too little water availability is linked to lower income level but so does too much. The proportion of fíat land coefficient has a positive sign, which indicates that mountainous colombian landscape, heavily concentrated in the Andean part of the country, has not been an economic advantage at all. In fact, it has made difficult transportation, agriculture, and construction of cities. According to the results in column (3) geographical variables account for 46% of the per capita income variance.
The IGAC sample exeludes Antioquia, Bogotá and Cali.
395 DeterminantsofMunicipalPer CapitaIncome(1995)
IGAC IGAC PerCapitaTaxesPerCapitaTaxes DependentVariable Total Total TotalSample Sample Sample Sample Sample TotalSample
8.5568 10.5135 13.0179 6.8585 12.1485 Constant 17.5143
(10.494)'' (6.680)" (13.877)''(7.953)" (8.764)'" (3.741)'
GeographicalVariables
-0.1238 -0.5706 -0.2072 -0.3672 Rainprecipitation -0.5892 -0,3413 (-2.554)- (-5.185)"" (-1.992)2 (-8.155)'"(-5.694)' (-5.758)'
0.3849 0.2197 0,3602 0.8099 0.5636 Altitudeaboyesealevel 0.4264 (2.992)' (2.467)JI (1.785)'- (2.683)'' (2.780)" (3.698)'
-0.0189 -0.0329 -0.0714 -0.0491 -0.0356 Altitudeabovcsealcvel'N2 -0.0391 (-3.020)v (-2.496)2 (-1.644) (-2.734)5(-2.800)"'(-3.703)''
0.3170 0.6720 0.4817 0.5934 0.4525 0.6863 Soilsuitabilit\'Índex (7.996)' (3.531)1 (3.990)< (8.1546)' (5.148)''(6.241);
-1.3290 -1.1054 -1.6105 -1.1151 -1.6452 Distancetodomesticmarkets -1.4968 (-14.202)'-(-12.719)' (-12.724)""(-10.368)''(-14.192)'" (-9.178)"
4.7933 3.6813
WateravailabilityÍndex•
(3.654)"' (2.897)' -
-2.3645 -2.0001 Wateravailabilitvindex"2 -
; (-3.995)1 (-3.510)" -
0.6987 0.5539 Proportionoffíatlands
(4.393)"' (4.298)!
-0.0303 -0.1155 -0.5077 -0.4130 -0.4771 -0.4255 CaucaRivcr (-2.013)2 (-0.146) (-0.820) (-2.213)- (-2.678)''(-1.438)
-0.0621 -0.0337 -0.1617 -0.0730 -0.0486 MagdalenaRiver -0.0991 (-0.326) (-0.490) (-0.684) (-0.267) (-1.013) (-0.505)
0.2209 0.0438 0.2673 0.1707 0.3952 0.2008 Rivers(inKilometers)
(3.557)'' (0.813) (3.942)''(2.896)1 (3.976)' (2.513)- Vable} (continuation)
DeterminantsofMunicipalPerCapitaIncome(1995)
DependentVariable Total Total IGAC IGAC Per CapitaTaxesPerCapitaTaxes
Sample Sample Sample Sample TotalSample TotalSample
InfrastructureVariables
Proportionofhouscholds - 0.8170 0.6794 1.4117 withelcctricalpowcr1973 (3.947)'' (2.789)' (7.239)'
Roaddensitv1970 - 0.0570 0.0291 0.0806 0.0738
(2.052)•''■■■ (0.818) (3.439)' (2.771)"
Roaddensitvrateofgrowth - 0.6623 0.5804 0.9073
_ (1.838)'-' (1.427) (2.576)-
Índexofqualityandavailabilitvofroads 0.2019 -
- (1.794)'-'
Human Capital
Migrationrate1973 2.6361 3.2323 2.4658
(5.919)-' - (6.091)" (6.192)"
Rnrollmentrateinpriman1 - 1.9812 1.5525 1.1716 andsecondarvschool1973 - (5.195)" (3.536)' (3.208)'
thousandof Collegegraduatesper 0.0370 0.0583 0.0357
Laborforcé1973 (2.851)" - (4.511)" - (3.285)"
Numbcroftropicaldiseasedcaths -0.0964 -0.0759 -0.1153 per1.000peoplc1979 (-2.729)" (-1.915)'-' (-3.126)" Tabk3 (amtnuiation)
DeterminantsofMunicipalPer CapitaIncome(1995)
DependentVariable Total Total IGAC IGAC PerCapitaTaxesPerCapitaTaxes
Sample Sample Sample Sample TotalSample TotalSample
InstitutionsandLivingStandards
-0.0411 lntcractionbctwccnsoilanddcgrec -0.1123 -0.1726
(-0.796) ofurbanization(1973) (-2.O55)2 (-2.745)1' -
_ - -0.0134 Proportionoflandwithcofeecrops1980 -0.0116 -0.0125
(-4.522)' (-3.963)"' (-5.606)'
-0.2723 Incomeinequality1973 0.0069 -0.1840
(0.092) (-1.855)1' (-3.843)'
0.2317 Percapitamunicipaltransfers - 0.4398 0.4484
(4.085)' (2.815)' (vearlvavcrage1973-1995) - (5.224)' -
872 Numberofobservations 873 872 625 612 897
0.6006 R2 0.3530 0.5542 0.4571 0.6313 0.3680
" Significanrat90°/,,.
21Significantat95%. v Significantat99%. h Sánchez y J. Xúñez Planeución y Desarrollo XXXI |ulio-diciembre 2111)11 (3^9-452)
The colombian historians have stressed the importance of gcography for inter-regional trade, due to the mountainous land and the difficulties of road construction. The Magdalena River was until 1930 the main transportation mean to conneet the sea (Atlantie) and the hinterland (Bogotá and the Coffee Zonc). Besides Bogotá, the other main important cities were, as mentioned, Medellín (a coffee and mining center), Cali (an agro-industrial center) and Barranquilla (the colombian oldest port in the country on the Caribbean Sea). Since the beginning of the century the domestic market was formed around these cities awayfrom the coasts. Morcovcr, the import substitution policies adopted during most ot the century contribute to direct the production of tradable goods oriented towards domestic market (Jaramillo and Cuervo (1987)). In this sense, the proximity to the domestic markets should have been, as it was round, an important determinant of the per capita income lcvcl differences among municipalities1-.
In Table 3, columns (2) and (4) we present the determinaras of the 1995 income per capita using a large set of variables. As infrastructure variables we include 1970 kilometers of roads (per square kilometers). The other infrastructure variables are the rate of growth of road density, the 1973 percentage of household with electncal power and an Índex of quality of roads (IGAC samplc). The road variable coefficients are positive and significant as well as household coverage of electrical power. The human capital variables we use, are the 1973 enrollment rate in priman' and secondary school years of schooling, the 1973 migration rate, the number of college graduates per thousand people and the 1979 number of deaths by tropical diseases. The results of the estimations show that the municipalities that had in 1973 greater enrollment rates in priman- and secondary education reached a higher income in 1995. The same is obtained for the variables migration rate and number of college graduates. The tropical disease indicator is, as expected, negatively related to income per capita.
Finally, the econometric results for the institutional and living standard variables show that the coefticient of interaction between the soil suitability Índex and the degree of urbanization has, as expected, a negative sign since quality of soils are more important for income in the rural municipalities. Income inequality is also negatively associated with income. According to Dinninger et al. (1999) and Leibovich et al. (1999) poverty and inequality restriet the accumulation of human capital in the long run, which at the same time affect income. Finally, per capita intergovernmental transfers during 1973- 1995 turned out significant in explaining the level of income of 1995. This is an important result, given the heated colombian debate on the decentralization process and its effeets, that indicates that in average transfersto municipalities are not a waste of money and that somehow have contributed to increase the municipal income lcvcl. The econometric exercises also show that our model is ven- robust not only to the samplc size1' but as well to the introduction of other variables. Similar results to the aboye are obtained if we use the per capita municipal taxes as income proxy as presented in columns (5) and (6). In the Table 1 Appendix 3 we present the same exercise with nice results using as dependent variable GDP per land área.
'- An explanation of rhc domestic market tormation could be round in Goueset (1998). '" The regressions run based on the informador! of homogeneous phvsical zones do not include Antioquia since the intormation of this Department on this matter is incomplete.
399 F. Sánchez y J. Núñez | Planeación y Desarrollo XXXI ¡ Julio-diciembre 2000 (379-452)
C. Geography and the Income of the Poor and the Rich Municipalities
In Table 4 we present aneconometric exercise14 aimed to determining if the impactof the different variables on income differs for the poor and rich municipalities(by the colombian standards). Although there are not great differences in the magnitude of the coefficients there are somc results worth to point out. Columns (1) and (2) present the rcgression estimates of per capita income for poor and rich municipalities on only geographical variables. Although the coefhcients are very similar and bear the same sign for the two sub-samples, the geographical variables explain 24% of income per capita variance of poor municipalities and 19% of the richest. The greatest difference is on the significance and magnitude of the distance to domestic market coefficient, which is larger (-1.9 for the poorest and -1.4 for the richest) and most statistically significant for the poorest. The same results are obtained using the IGAC sample(columns (5) and (6)). According to them geographical variables explain 32% of the per capita income variance of the poorest municipalities and 25% of the richest. The altitude, the availability of water and the land flatness seem to be more important for income level of the poor. The poor municipalities are small and located in rural áreas wherc the main source of income are agriculture and livestock activities, which depend quite a lot upon the availability of the mentioned natural resources and geographical characteristics. Paradoxically, soil suitability seems to be more important for the richer municipalities, which may be due to historical reasons. They are rich, in part, precisely because the oíd settlements were made on the good soils while the later onestook place on soils of lower quality.
The impact of the 1973 electrical coverage seems to be very important for the income of poor and the rich and the coefficient is quite significant for bothof them. However, the impact of both the density ofroad and its rate of growth is higher for the poor municipalities than for the rich. Among the human capital variables, the education (enrollment rates and number of college graduates) coefficients seem to be of greater magnitude for the rich municipalities when the total sample is used. However, the coefficients look pretty similar when the IGAC sample is used.
Finally, the institutional and living standards variables do not seem to perform well in the quantile regressions. It may be the case since the poor and rich municipalities may be quite homogeneous among them with respect to these variables. The econometric exercises of the determinants of per land área income for rich and poor municipalities are presented in the Table2
Appendix 3.
14 For the poor and rich estimation we used the methodology of quantile regressions. See Appendix 3 Table 2.
400 Tabk4
QuantileRegresionsofMunicipalPerCapitaIncome
TotalSample Total!>ample IGAC Dependentvariable IGAC !Sample Sample
25% poorest25% richest25% poorest25% richest25% poorest25% richest25% poorest25% richest
Constant 20.4273 14.9455 11.2505 10.3756 15.2920 11.2163 9.1375 8.0232
(14.053)"(10.906)"(9.033)" (6.752)"'(7.036)" (4.965)"(3.200)"'(2.996)"
GeographicalVariables
Rainprecipitation -0.6258 -0.4952 -0.3830 -0.2755 -0.5586 -0.5445 -0.1573 -0.1885
(-6.934)-"(-5.181)"(-6.465)"(-3.405)"'(-4.751)-"(4.213)"(-1.159) (-1.251)
Altitudeabovesealevel 0.4003 0.3886 0.3394 0.0819 0.8168 0.6999 0.5595 0.4824
(2.703)"(2.399)2'(3.643)" (0.361) (4.213)" (3.249)"(2.670)" (2.046)2'
AltitudeabovesealeveP2 -0.0357 -0.0357 -0.0309 -0.0081 -0.0704 -0.0619 -0.0473 -0.0425
(-2.444)2'(-2.360)2'(3.357)" (-0.649) (-3.904)"(-3.115)"(-2.431)2'(-1.998)2-'
SoilsuitabilityÍndex 0.6678 0.7582 0.3360 0.5942 0.5309 0.7090 0.3018 0.5063
(5.786)" (6.950)" (3.240)" (4.965)'''(3.916)" (5.582)" (1.549) (2.972)"'
Distancetodomestic -1.8408 -1.1976 -1.2062 -0.9910 -1.7882 -1.4004 -1.2024 -1.0111
matkets (-12.460)"(-9.237)'-'(-12.212)"(-7.869)"(-11.321)"(-9.506)"(-6.157)'(-5.160)
WateravailabilityÍndex 4.5810 4.7089 0.8844 2.6278
- - - (2.340)2-'(1.770)" (0.441) (1.376)
Wateravailabilityindex/N2 - -2.0893 -2.2688 -0.7104 -1.4649
- - (-2.39Ó)2'(-1.946) (-0.795) (-1.663)"
Proportionoffíatlands - - 0.6641 0.5189 0.5876 0.5223 -
- - (3.052)" (2.758)"'(2.503)2'(2.514)2'
CaucaRiver -0.1897 -0.4480 -0.4860 -0.4459 -0.1769 -0.2534 -0.3368 -0.6040
(-0.763) (-1.954)"(-3.123)" (-2.206) (-0.540) (-0.862) (-0.909) (-2.018)2'
MagdalenaRiver -0.1676 0.0398 -0.0409 0.0683 -0.0009 -0.0829 0.0183 -0.0769
(-0.851) (0.235) (-0.339) (0.464) (-0.005) (-0.431) (0.083) (-0.362)
River(ínkilometers) 0,1387 0,3428 0.1376 0.1329 0.1846 0.5187 0.1030 0.1293
(1.610) (4.461)" (2.416)2/(1.886)" (1.636) (5.227)'(0.841) (1.116) Tuble4 (cniíliiniatioii)
QuantileRegresionsofMunicipalPerCapitaIncome
TotalSample IGAC Sample IGAC Sample DependentVariable TotalSample
25% poorest25% richest25% poorest25% richest25% poorest25% richest25% poorest25% richest
InfrastructureVariables
Proporrionofhouseholds 0.9057 0.4401 0.5266 0..1200
(1.361) (1.528)) (4.354)' (1.946)'; withclccrricalpowcr1973 i
0.0696 -0.0133 Roaddcnsity19"0 O.O8H2 0.0125 '
(2.695)'(-0.476) (3.603)' (0.445) i
2.0907 -0.3439 Roaddensirvrarcot 1.015^ 0.0320
(3.221)' (0.086) (2.955)' (-0.468) growth i
-0.0163 0.3613 Índexofquality
(-0.080) : (1.772)' andavailabilitvotroads
Human Capital
2.8399 2.8228 2.6113 2.33r>3 Mun-ationrateI9"3 i :
(3.2"6)' (3.091)' (6.105)': : - (6.645)'
1.4631 1.7333 Knrollmentrateinpriman- 1.8844 2.301"
(2.020)-:(2.437)-' (5.266)' (5.295)'' andsecondarvschoo!193 !
0.0"16 O.O"23 - 0.0191 0.0628 Colichegraduatcsper 1
(3.523)-'(3.824)' Tin>usand(>flab<>rh>rce19^3 (í.-iiy C.\2~y !
-0.1611 Xumberottropicaldisease -0.4485 -0.1834 -0.2-02
(-2."85)' (".924)' ; deathsper1.000x-ople19"9 1 Yabk(continuation) 4
QuantileRegresionsofMunicipalPerCapitaIncome
Depender»Variable TotalSample TotalSample IGAC Sample IGACSample
25%poorest25% richest25% poorest25%richest25%poorest125%richest25%poorest25%richest
InstitutionsandLivingStandards
Interacrionbetweensoiland _ -0.1181 -0.1452 -0.1220 -0.1550
degreeofurbanization1973 - - (-1.900)"(-2.380)2''- - (-1.036) (-1.766)1''
Proportionoflandwith - - -0.0125 -0.0106 -0.0055 -0.0165
coteccrops1980 (-4.487)'' (-2.880)1' - (-0.867) (-2.370)2'
Incomeinequalitv1973 - 0.0757 -0.1142 -0.0127 -0.2865
- (1.126) (-1.285) - (-0.069) (-1.771)1
Percapitamcpal.transfers - 0.4679 0.5676 0.4520 0.4633
(yearlyaverage1973-1995) (5.320)' (5.853)" (2.590)' (2.693)'
Numberofobservations873 873 872 872 613 613 612 612
PseudoR2 0.2362 0.1846 0.3636 0.3461 0.3144 0.2562 0.4203 0.4055
"Signifícantat90%.
11Signifícantat95%.
''Signifícantat99%. F. Sánchez y |. Núñez ¡ Planeación y Desarrollo XXXI | ]ulio-diciembre 2000 (379-452)
VI. I Geography andthe Rate of GrowthofPer Capita Income
A. Some Ideas on Geography and Growth in Colombia
Although geography has been neglectcd in the economic modclsand estimations ot colombian regional and municipal growth, historians and travelers have long noted the important role of geographical factor in shaping the colombian development. John Hamilton, a Brirish colonel traveling in Colombia in the 19th century, pointed out in 1829 the heavy burden that nature and climate imposed on trade and human transportation. The trip on the Magadalena River from Barranquilla to Honda, the only way to access Bogotá, lasted more than 100 days. In suchlong period, many passengers got ill or died of malaria, yellow fever, diarrhea or cholera. The high freights made trade quite costly, impeding the import of goods and machinen to the hinterland (Hamilton, 1970). james Parsons, a sociologist who studied in depth the antioqueña colonization, suggested that the long and eftective geographical isolationism in the inner colombian mountain moldcd the defined traditionalism and peculiar cultural features of the Antioqueña peoplc (Parsons, 1997). The scarcity of indigenous labor forcé and the pracrically non-existence of fíat lands determined that the 19th century rural population of Antioquia were mainly composed by small landholders, working their own state. This circumstance, according to Parsons, prompted an early democratic tradirion of the antioqueña labor forcé in contrast with the classiest social structure in the south and wcst that counted with a more numerous indigenous population. The mentioned special features of this society, originated in the antioqueña geography determined, inpart, the early industrialization of the región.
The geography of Caribbcan Coast brought about a particular development pattern as Posada (1998) points out. Sea, rivers and swamps conditioned the localization of the main settlements in the coast, as ways of access, and rich sources of water and food. Nevertheless, Ufe was not easy. Land and towns sufferedperiodic floods that destroyed houses and crops, and frequently altered the geography of the región. Floods and high temperatures of the tropic favored the proliferation of diseases, infections and plagues and made difficult to establish permanent mining or manufacturing activities'\ The lack economic opportunities and the high disease burden in the Coast produced migration, death, and slow population growth, which generated, at the same time, a great deal of labor forcé scarcity. The latter together with the low productivity of the labor forcé, the backward technology and lack of transportation means impeded the rise ofcommercial plantations (haciendas), ümited the development of agriculture until very late in the 20th century. In contrast, the land characteristics and the factor market conditions facilitated the rise and consolidation of cattle raising.
As seen, leading historians economists and sociologists have recognized the crucial role of colombian geography in shaping the regional development patterns. According to them the main channels through which geography directly conditioned economic development were transpon costs, human health and endowment of natural resources (land suitability, water, closeness to river, etc.). It these
fSrritrler, a French engineer, carne to work for a mining company established in the Sinu Yalley around the mid 19th century. At the beginning he anei his colleagues were very enthusiasde. However, the mosquitoes and the floods made it impossible to work and sooner than later they abandoned the mine and left the country (Posada, 1998).
404 F. Sánchez y J. Núñez | Planeación y Desarrollo XXXI ¡ |ulio-diciembre 2(10(1 (379-452)
factors influence population density and creation of markets they may also have indirect cffects on growth dynamics through agglomeration economies and other feedback mcchanisms (Gallup and Sachs, 1998). Let us treat these factors ina formal simple growth model.
B. Geography and Municipal Economic Growth: Some Theory
The introduction of geographical variables in the growth equation can be interpreted as a simple modification of the so-called "conditional convergence" model. To test for "conditional convergence" we should drop the assumption that all economies have the same parameters, and therefore, the same steady state positions. The main idea is that an economy grows faster the further is from its own steady state. To ¡Ilústrate the point let us consider two economies, rich and poor, that differ on their initial capital stocks k(0)pm< k(0)nh and saving rates, s^< s^. If their savings rates were the same, then the per capita growth rates would be larger for the poor economy, and y^> y , but if it not the case and ■*■/»'< sñ* and then Yn,h >^,«,/ In tne neoclassical thesteady state valué, k*, depends on the saving rate, s, the level of the production íuncúon,f(k), and the various government policies and productivity factorsthat shift the position oithef(k) (Graph 2). Thus, the saving rate in the Solow-Swan model can be expressed as
(2) s={n + 8)k*/f(k*)
and the rate of growth of capital y as
(3) \={'i+§)W(k)/k)/(j-(k*)/k.*y 1]
Equation (3) is onsistent with yt = 0 when k=k* of the Solow-Swan model. For a given k*, the formula impües that a reduction in k, which raises the average product of capital, f(k)/k, increases yk. But a lower k is consistent with a higher yt only if the reduction is relative to the steady state valúe, k*. Thus, a poor country orregión would not be expected to grow rapidly if its steady state valué, k*, is as low as its current valué, k (Barro and Sala-i-Martin. 1995). Equation (3) suggests that we should look empirically at the relation bctween the per capita growth rate, y', and the initial income levcl, y(0), after controlling for the variables that account for differences in the steady state position, y*. The large differences among the colombian municipalities in their geographical endowments are likely to be related to different level of y*. We will see in the above econometric estimations that physical geography affeets growth through land quality, human health (tropical diseases), water availability, land slope, etc. through productivity growth. Additionally, access to markets, which facilitatcs the functioning of the factor andgoods markets, and the diffusion of technology and knowledge is clearly beneficial to productivity growth.
In this section we are going to explore at a municipal level theforces of per capita income convergence and divergence by estimating cross municipal growth regression. In theexercise we will emphasize the role of geographical variable on the growth process. Thus, we will estímate
405 I. Sánchez y ]. Núñcz | Plantación y Desarrollo XXXI ! Julio-diciembre 200(1 (379-452)
models of annual average rate of per eapita income (or municipal per capita taxes) growth during 1973-1995, conditional to the per capita income level (or municipal per capita tax level) of 1973. \\"e will test vhether growth is affected by the initial income level, (negativelv if there are convergence torces) and by geographical variables, controlling for initial level of infrastructure indicators, human capital, and living standards.
According to Table 1, the 1973 the differences in income per capita among municipalides and regions were smaller than 1995. The per capita income of Bogotá reached 52,7561, the Caribbean región SI 1,~69, the Andean Region's $9,097 and the Pacific $6,854. Thus, the income per capita differences between municipalities and regions seem to have increased during the 1970's and 198()'s. Thc average (non-weighted) per capita income of municipalities decreased for the Caribbean Región, increased a littlc (around 5"d) for the Pacific Regiónand increased significantly for the Andean Región (more than 40%). The changes in the average per capita income of the different regions suggest uneven growth rates between municipalities and regions. ;How was the processr1
C. Geography and Economic Growth: Some Results
In the third secrion ve showed there is an unconditional converge among the colombian municipalities though its explanatory power of the per capita income growth process is very small. In Table 5 columns (1) and (2) \ve present regressions of the per capita income growth rate between 1973 and 1995 on the 1973-incomc level and only geographical variables. The results obtained quite interesting and lead us to conclude that geographical variables, in the colombian case, play an important role in explaining per capita income growth.
The variable level of initial per capita income level has, as expected, negative sign, which conhrms that there are convergence torces of income per capita among the colombian municipalities. The econometric results show the distance to the markets was a key determinant on the per capita income growth rate during 19~3-1995. Its coefficient has the expected sign and is highly signifícant. The municipalities closer to the domestic markets grew more than those farther"1. These results also imply that the economic activity could have been concentrating around these markets and suggest that there have been important development of economies of scale around the colombian domestic markets that also reinforces growtharound these markets. According to Krugman (1991) geographic concentration of production is "clear evidence of the pervasive influence of some kind of increasing returns... (p. 5) and these increasing proximatc industries promote innovation and growth". In fact, colombian domestic markets are highly diversified in their production structure. Thus, the municipalities cióse to the domestic markets and under the influence of great cides have been able to exploit the knowledge spillovers coming from them and havegrown faster (Jacobs (1969)Gleaser, (1992)).
:" The distance to domestic markets was veighted by the population of rhe ciry since ít is not the same to be cióse to Bogotá than to Barranquilla. The weights used were 0.52 for Bogotá, 0.19 for Medellín, 0.18 for Cali and 0.11 for Barranquilla. These wcights correspond to the share of each cuy in the total population of the four.
41)6 Tab/e5
DeterminantsofMunicipalPerCapitaIncomeGrowth1973-1995
DependentVariable TotalSampleTotalSample IGACSampleIGAC SamplePer CapitaTaxesPerCapitaTaxes
(1) (2) (3) (4) (5) (6)
Constanr 0.4951 0.4560 0.3267 0.2917 0.3080 0.3126
(11.236)' (8.603)' (4.892)' (4.088)'■ (6.225)' (6.074)'
Per(Zapitaincome1973 -0.0216 -0.0281 -0.0214 -0.0281
(-14.725)'(-14.798)''(-11.393)"(-10.270)v
PerCapitatax1973 - -0.0166 -0.0273
(-6.936)' (-11.851)'
GeographicalVariables
Rainprecipitation -0.0138 -0.0092 -0.0173 -0.0093 -0.0080 -0.0022
(-4.823)' (-3.584)' (-4.594)"'(-2.600)1 (-2.808)1 (-0.869)
AltitudcabovcseaIcvcl 0.0085 0.0110 0.0243 0.0212 0.0085 0.0080
(1.885)' (2.039)- (4.145)'" (2.737)' (1.527) (1.586)
AltitudeabovesealcvelA2 -0.0006 -0.0009 -0.0020 -0.0017 -0.0007 -0.0007
(-1.540) (-1.847)' (-3.746)'' (-2.633)-' (-1.518) (-1.466)
SoilsuitabilitvÍndex 0.0169 0.0125 0.0175 0.0140 0.0204 0.0115
(4.975)' (3.077)'' (4.492)'' (3.003)'' (5.872)' (2.875)'
Distancetodomesticmarkets -0.0567 -0.0498 -0.0577 -0.0475 -0.0457 -0.0464
(-14.166)"' (-11.054)''(-12.566)' (-9.437)' (-9.0912)' (-9.882)'
CaucaRivcr -0.0141 -0.0116 -0.0222 -0.0183 -0.0061 -0.0035
(-1.953)' (-1.715)'-(-2.367)-- (-2.004)2 (-0.690) (-0.5738)
MagdalenaRiver -0.0051 -0.0013 -0.0040 -0.0001 -0.0067 -0.0034
(-0.948) (-0.254) (-0.673) (-0.031) (-1.161) (-0.644)
Rivcr(inkilometers) 0.0085 0.0048 0.0107 0.0067 0.0051 0.0007
(3.663)' (2.083)J (3.458)' (2.225)-- (2.079)' (0.305) Tt/b/e5 (continmition)
DeterminantsofMunicipalPerCapitaIncomeGrowth1973-1995
GeographicalVariables
IGAC SamplePerCapitaTaxesPerCapitaTaxes DependentVariable TotalSampleTotalSampleIGAC Sample
(4) (5) (6) (1) (2) (3)
0.2369 0.1954 - - WateravailabilityÍndex - -
(3.702)' (3.624)''
-0.1139 -0.1021 WateravailabilitvirulcxA2 -
(-3.996)'' (-4.211)'
0.0190 0.0189 Proportionoffíatlands -
(3.087)" (3.324)' -
InfrastructureVariables
0.0130 0.0376 Proportionof households - 0.0215
(4.788)' (2.770)3' (1.447) - withelectricalpower1973 -
- -8.64H-05 - 0.0009 Roaddensitv1970 0.0004
(-0.233) (1.192) (0.589) - -
0.0085 - Índexofqualitvandavailabilityofroads
(1.736)
Human Capital
0.0920 0.1305 0.0953 Migrationrate1973 - (6.335)" ; (5.777)" (6.042)' -
0.0637 0.0673 - Knrollmentrateinpriman1 - 0.0811
(3.6488)'' (4.336)" andsecondaryschool1973 (5.254)"
0.0016 0.0012 College graduatesperthousandof 0.0011
; (3.248)" (3.892)''' Laborforcé1973 (3.309)-'"'
-0.0046 -0.0028 - Numberoftropicaldiseased caths -0.0041 -
(-1,654)' (-2,948)' per1.000people 1979 (-2,671)' Table5 (continuation)
DeterminantsofMunicipalPerCapitaIncomeGrowth1973-1995
InstitutionsandLivingStandards
DependentVariable TotalSample Sample Total IGACSampleIGACSamplePerCapitaTaxesPerCapitaTaxes
(2) (3) (1) (4) (5) (6)
Intcractionbetweensoilanddegrcc -0.0030 -0.0058 -0.0017 ofurbanization1973 (-1.299) (-2.119)2' - - (-0.724)
Proportionoflandwithcofeecrops1980 - -0.0004 - -0.0003 _ -0.0005
(-3.927)" - (-2.335)2 (-5.113)"
lncomcincquality1973 -0.0025 _ -0.0093 -0.0124
(-0.808) - (-2.302)" (-4.103)"
Numberofobservations 873 872 613 612 873 872
R2 0.3502 0.4860 0.3955 0.5256 0.1799 0.3844
''Significantat90%.
2 Significantat95%. v Significantat99%. !•'. Sánchez y |. Núñez | Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
The variables of natural geography also played animportant role in cxplaining per capita incomc growth during 1973-1995. Thus, rain precipitation has had a negative impact while the altitude has positive impact unril certain levcl but bcvond such level its impact on growth turns out negative. The soil suitability Índex has also a positive impact on growth. Rcgions or municipalities with good soils not only have a higher per capita income but also have grown faster during 1973-1995. The regression results indícate that the municipalities located near or on the Cauca River experienced lower economic growth. According to column (1) the set of geographical variable explain 34% of the variance of income per capita growth. The IG AC sample results of columns (3) and (4) show that both water availability and proportion of fíat lands also matter for growth1". The geographical variables and the initial level of income of that sample explain 40% of the income per capita income growth variance of colombian municipalities.
The initial level of infrastructure density has determined income per capita growth. The municipalities with higher level of kilometers of roads and railroads in 1970ls grew faster during 1973-1995''. Although the same exercises were done for railroad density the coefficients obtained (not reported) werc not significant. This result was, however, expected since railroad development stopped by the end oí the 1960's. According to Ramírez (1999) the kilometers of railroad tracks grew quite fast from the end of the XIX century until the end of the 1950's, moderately until the end of the 1970's and zero or negatively afterwards. Thus, the contribution of railroad to the colombian economic development and to domestic market integration oceurred before the 197()'s. The other infrastructurevariable used, the houschold coverage rate of electrical power at the initial period, has also a positive sign.
All the coefficients of1973 human capital variables are quite significant and have the expected sign. Thus, had the number of college graduates in 1973 (per thousand of labor forcé) been higher in one person the average rate of growth during 1973-1995 would have been 0.08'/» higher. Similar results are obtained for the enrollment ratein priman and secondary education in 1973. The municipalities with higher enrollment in 1973 rates grew faster during 1973-1995. So didthe municipalities with higher 1973 migration rate. Although the mechanism it is so is not quite clear, recent studies for Colombia (Leibovich, 1996) have shown that the migrants reach higher level of income than the people with the same socio-economic characteristics alrcady living in the migration pole.In contrast, municipalities with higher ratcs of tropical disease deaths grew lower during 1973-1995. The incidence on growth ot this variable is quite strong. I lad this variable been 1 % lower income per capita would have grown at arate 1-% higher.
The living standard and institutional variables have also impact on the economic growth of municipalities. According to Table 5, the negative sign of the interaction between soil and degree of urbanization implies that the importanceof soil quality on growth decreases, as the municipality becomes more urban. Prcsence of coffee crops has also a negative effect on growth. Although the coefficient of the latter variable is small, has a negative sign and is quite significant. This may be the result of the decreasing importance of coffee asa source of national income, the loss of its comparativo
'" Water availabilitv matters until certain point as indicated by the positive and significant coefficient ot this variable in its linear specification and the negative and also significant coefficient ot its quadratic specitication. lN We use this year because the existing maps of roads and railroads are from 19^0. '" We use thevalúes of the variables at the initial year to avoid endogeneity problems. See H'dderke et al. (1998; for a discussion on the topic.
410 1. Sánchez y J. Núñez | Planeación y Desarrollo XXXI |ulio-diciembre 20(10 (379-452)
advantage to other production scctors and its lower prices compared with those prcvailing in the 1970's (Ocampo, 1987). The 1973 income inequality (measured as the variance of the logarithm of income at a municipal level) has had negativc on municipal growth. The effect of initial income distribution on per capita income growth is negative in concordance with results obtained by recent cross-country studies (Denninger and Squire, (1998)) .The transmission mechanism from distribution to lower growth takcs place through credit constraint channcl. A more unequal distribution of assets would imply that, for a given level of per capita income, a greatcr number of people are credit constraint, which restricts education and human capital accumulation and lower aggregate growth. Columns (5) and (6) contain the per capita income regression using other income proxies and municipal taxes. The results are quite similar to the obtained in columns (1) and (2) in the same table. The determinants of the income per land rate of growth are shown in Table 3 and 4 off Appendix 3.
VII. I Geography and Population
This section of the document presents the some of the determinants of population density for the colombian municipalities2". The population density refleetsin part the settlements decisions made in the past and the historical demographic trends. The settlements decisions should have been influenced by the quality of soil, the proximity to rivers, seaports and main roads,the quality of the climates and the like. At the same time, once a settlement takes place, it will determine the future economic growth of the región through the creation of economy of agglomeration, and markets of goods and factor. Additionally, changes in the production structure, demographic trends and access to social services may influence the trends in population density. No more than thirty years ago most the population of Colombia lived under poverty (measured by unsatisfied basic needs) and the majority was located in rural áreas (more than 70%). By 1995 the percentage of population under poverty decreased to less than 30% and lived mostly in urban áreas (70%) with greater access to publie and social services.
Table 6 contains the determinants of population density of colombian municipalities for 1973 and 1995. According to the results, the colombian regions with high incidence of heavy precipitation Índices are the less populated, as expected. The altitude has a quadratic specification in the equation, which means that ceterisparilms the most density-populated áreas are those with mild (not too cold not too hot) weather. The sign of the coefficient for soil suitability Índex indicates that colombian population settlements were also determined by the quality of soil and erosión Índices. However, at least for 1973, the municipalities located nearor along the colombian main rivers (Magdalena and Cauca) did not have higher population density. In contrast, municipalities with coffee crops were in 1973, ceteris parilms, more densely populated. Finally, the population density for 1973 was determined by the availability of transportation means (density measured in 1949) specially by the railroads.
Column (2) in Table 6 contains the determinants of 1995 density population. The geographical, the infrastructure and the publie and social services variables matter for population distribution. Among the geographical variables the rain precipitation temperature and the distance to seaports have a negative impact on population density. In contrast, altitude and suitability of soils have a positive effect.
Jaramillo (1998) and Flórez (1998) have studied the structure of colombian population and the migration pattern.
411 Tableó
Geography,MunicipalPopulationDensityand1973-1995PopulationGrowth
PopulationGrowth 1973Population1995Population1995PopulationPopulationGrowth DependentVariable 1973Population
1.2184 3.6500 14.7055 12.7027 0.9099 Constant 16.1558 1 (26.108)' (7.177)" (14.0847)' (13.043)''(11.929)' (8.845)''
GeographicalVariables
-0.0070 -0.0179 Populationdensity1973 - -
(-4.843)"' (-3.436)'
-0.4133 -0.2009 0.0020 -0.0008 Rainprecipitation -0.6821 -0.5403 (0.8019) (-0.417) (-8.378)' (-7.032)''(-4.805)' (-2.974)'
-0.0071 -0.6049 -0.5480 -0.3071 -0.0060 Altitudeabovcsealcvel -0.8161 (-1.700) (-2.397)- (-5.738)" (-4.796)' (-3.981)' (-3.180)''
0.0007 0.0790 0.0683 0.0328 0.0007 AltitudeaboyesealevclA2 0.0836 (1.491) (1.869)' (6.213)" (6.566)'' (5.141)"' (3.369)"''
-0.0005 0.4231 0.7910 0.2508 0.0160 Soilsuitabilitvindex 0.0607 (3.988)' (-0.116) (5.983)' (4.656)' (8.085)s' (2.748)"
0.0217 -2.1851 -0.6037 0.0105 Distancetoseaports -2.7398 (2.990)' (-11.056)'"'(-7.147)''(-2.451)2' (1.507)
0.1635 0.0347 0.0148 0.0004 Distancetodomesticmarkets -0.9777 0.1702
(4.655)' (0.054) (-8.046)' (1.343) (1.209) (0.310)
-0.5145 -0.3597 -0.1331 Krosioníndex -0.6507
- (-4.743)' (-4.186)""(-2.440)2 (-1.122)
0.0014 0.1448 0.1621 0.0014 0.0045 CaucaRivcr - (0.355) (0.763) (0.657) (0.007) (1.047)
0.0893 0.1667 0.0017 0.0051 MagdalenaRivcr 0.1833 (0.648) (1.249) (0.577) (1.642) (0.325)
-0.3624 -0.3478 Rivcr(inkilometers) -
(-3.966)' (-5.219)' - - Table6 (continuation)
Geography,MunicipalPopulationDensityand1973-1995PopulationGrowth
InfrastructureVariables
DependentVariable 1973Population 1973Population1995Population1995PopulationPopulationGrowth PopulationGrowth
Proportionofhouscholds 1.5070 0.0398
withclcctricalpowcr1973 : (6.069)v (2.686)'■
Densitvofroads1949 0.0136 -
(1.248)
Railroaddcnsirv1970 0.0745 -
Roaddensity1970 (8.080)v 0.2687 0.0070
- ; (9.604)-v (1.745)'-
RoaddensitvRafeofgrowth - 1.9134 0.0569
(5.872)' (1.454)
Human Capital
Enrollmentrateinpriman-and 0.7944 0.0062
sccondarvschool1973 - - - (1.977)2-' (0.477)
Collcgegraduatesperthousand 0.0532 0.0009 oflaborforcé1973 (2.962)' (1.881)1 Tab/e6 (continuation)
Geography,MunicipalPopulationDensityand1973-1995PopulationGrowth
InstitutionsandLivingStandards
1995PopulationPopulationGrowth PopulationGtowth DependentVariable 1973Population1973Population1995Population
-0.1486 0.0024 Interactionbetweensoiland
(-2.767)v (0.720) dcgrecofurbanization(1973) -
0.0187 0.0002 Proportionoflandwithcofcc 0.0187 -
(5.106)' (5.266)' - (1.972)2' crops1980 -
-0.8620 -0.0417 PerCapitaMunicipalTransfcrs
(-8.317)-' (-1.630) (vearlvaverage1973-95) - - -
869 898 873 Numberofobservations 894 894 893
0.6245 0.078 0.269 R- 0.2575 0.4471 0.3086
1 Significantat90"
2 Significantat95"/
' Significantat99'/ I;. Sánchez y ). Núñcz I Plancación y Desarrollo XXXI Julio-diciembre 2(KK) (379-452)
AH the infrastructure variables have a positive effect on population density. In fact, the 1970 road densin, rate of growth of road density and 1973 electrical power coverage have the largest t-statistics in the regrcssion. The 1973 enrollment rates in priman- and secondary school also have an important impact on the 1995-population densin as well as the number of college graduares (per thousand people).
Table 6 also presents the geographical determinants of population growth during 1973-1995. The coefficient of population density in 1973 has a negative sign, which indicares that there is "convergence" in density population among the colombian municipalities. The other results show that the population of municipalities away from the main domestic market, away from the seaports and with better soil suitability, more density of roads in 1970, greater household coverage of electrical power and higher density of coffeelands have grown faster. The altitude of the municipality seems to have a quadratic effect on population growth. Too high altitude discourages population growth but so does too little.
The quantitativc results obtained from the population regressions imply that geographical factors affect íts localization and growth, which, at the same time, have indirect effeets on economic growth through agglomeration, creation of markets and related mechanisms.
VIII. I The Sources of Income perCapita and Economic Growth Differences
A. Exercises of Income Per Capita Decomposition
In this section we will explore, by using the methodology decomposition based on the coefficients obtained, the impact of differentvariables on the income per capita differences among municipalities. In order to carry out the exercise presented in tables 7 and 8, we took the per capita income average difference between the Andean Región (without Bogotá) and the other colombian regions. In order to calcúlate how much each variable contributes to the difference in income per capita we multiply the regression coefficient of the variable by the difference bctween the average of the variable in the Andean Región and the average of the variable in the other región. Thus, if Y, is the average income per capita of región A and Y* the average income capita of región K, the difference of average income can be calculated as follows
(4) YA-YK = i where p, {X tA — X ¡K ) is the contribution of variable / to the difference in per capita income betweer región A and región K.
415 Tal)le7
IncomePerCapitaDifferencesbyRegión(TotalSample)
Amazonio CaribbeanPacificDrinoquianAmazonio Zaribbean PacificOrinoquian AndeanRegiónvs. (%) (%) f/o) (%)
-0.1758 100.00 100.00 100.00 100.00 Difcrcnce 0 7523 0.8642 -0.6331
13.16% -468.26 Geography 0.4934 0.4874 -0.0833 0.8232 65.59 56.40
-111.83 -12.12 7.98 -36.79 Rainprecipitation -0.0912 0.0690 0.2329 0.1966 -273.32 0.5035 0.4805 168.19 23.83 -79.53 Altitudeabovethesealevel 1.2653 0.2059
-0.5680 -162.36 -19.46 95.58 323.09 Altitudeabovethesealevel"2 -1.2214 -0.1682 -0.6051
-0.0305 -25.57 -5.32 33.58 17.35 Soilsuitabilityíndex -0.1924-0.0460-0.2126
0.9797 98.70 52.99 -24.86 -557.28 Distancetodomesticmarkets 0.7425 0.4579 0.1574
-0.0087 0.89 1.40 1.37 4.95 CaucaRiver 0.0067 0.0121 -0.0087
-0.0017 0.47 -0.20 0.27 0.97 MagdalenaRiver 0.0035 -0.0017 -0.0017
-0.0416-0.1490 -0.2247 -2.61 -4.81 23.53 127.82 River(inkilomctcrs) -0.0196
0.47 -11.47 -9.61 Infrastructure -0.01380.0041 0.0726 0.0169 -1.83
0.0608 0.1556 -1.94 3.46 -9.60 -88.51 Roaddcnsity1970 -0.0146 0.0299
-0.0055 -0.0024 2.83 0.32 0.87 1.37 Growthrateofroaddensitv 0.0213 0.0028
Proportionofhouscholds
-0.1363 -2.72 -3.31 -2.73 77.53 withelectncalpower1973 -0.0205-0.0286 0.0173
10.66 71.36 268.15 Human Capital 0.0030 0.0921 -0.4518 -0.4714 0.40
Iinrollmentrateinprimary
0.0356 -0.2198 10.10 3.90 -5.62 125.03 andsecondarvschool1973 0.0760 0.0337
170.88 0.0373 -0.5422 -0.3004 4.96 -0.64 85.64 Migrationrate1973 -0.0055
College graduaresperfhousand -0.0765
-7.03 0.84 5.09 43.52 oflaborforcé1973 -0.0529 0.0073 -0.0322 Table7 (continuation)
IncomePerCapitaDifferencesbyRegión(TotalSample)
CaribbeanPacificOrinoqutan AndeanRegiónvs. AmazonicCaribbeanPacificOrinoquianAmazonic
Numberoftropicaldiseasedeaths
per1.000people1979 -0.0574 0.0566 0.0870 0.1253 -7.63 6.55 -13.74 -71.27
InstitutionsandLivingStandards 0.1379 0.1062 0.0183 -0.0872 18.33 12.29 -2.89 49.60
Interactionbetweensoilanddegree
ofurbanization1973 0.1069 0.0259 0.1043 0.1148 14.21 3.00 -16.47 -65.30
Proportionoflandcofee withcrops1980 -0.0687-0.0137-0.0614 -0.0713 -9.13 -1.59 9.70 40.56
Incomeinequality1973 0.0005 -0.0013 0.0011 0.0004 0.07 -0.15 -0.17 -0.23
Percapitamunicipaltransfers
(yearly1973-1995) 0.0992 0.0953 -0.0257 -0.1311 13.19 11.03 4.06 74.57
Totalexplainedbytheregression 0.6205 0.6898 -0.4442 0.2815 82.48 79.82 70.16 -160.13
Residualunexplainedbytheregression 0.1318 0.1744 -0.1889 -0.4573 17.52 20.18 29.84 260.13
Note:AndeanRegióndoesn'tincludeBogotá. 'lable8
IncomePerCapitaDifferencesbyRegión(IGACSample)
CaribbeanPaciflcOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic AndeanRegiónvs.
100.00 100.00 100.00 Difercnce 0.8429 0.8539 -0.6351 -0.2295 100.00
-234.90 Geography 0.6149 0.5769 -0.0647 0.5391 72.95 67.56 10.19
-4.64 1.29 -28.97 -88.89 Rainprecipitation -0.0.3910.0110 0.1840 0.2040
0.7599 0.8753 235.44 -13.09 -119.65 -381.39 Altitudeabovethclevel sea 1.9845 -0.1118
-217.84 16.33 139.29 439.43 Altitudeabovethesealevel"2 -1.8362 0.1394 -0.8846 -1.0085
-0.1776 -0.1869 -21.07 -9.26 36.91 81.44 SoilsuitabilityÍndex -0.0791 -0.2344
90.39 60.91 -5.37 -383.70 Distancctodomesticmarkets 0.7619 0.5201 0.0341 0.8806
-0.0032 1.35 2.48 0.50 1.39 CaucaRiyer 0.0114 0.0212 -0.0032
-0.0037 -0.0037 0.97 -0.43 0.58 1.61 MagdalenaRiver 0.0082 -0.0037
-0.3460 -2.92 -1.29 20.55 150.76 River(inkilometers) -0.0246 -0.0110 -0.1305
-19.62 -7.65 138.12 420.83 WateravailabilitvÍndex -0.1654-0.0653-0.8772 -0.9658
1.2540 22.80 12.51 -180.11 -546.41 Wateravailabilitvindex^2 0.1922 0.1068 1.1439
-0.1607 -11.91 5.77 8.35 70.02 Proportionoffíatlands -0.1004 0.0493 -0.0530
-0.0249 -7.54 -1.53 Infrastructure -0.0656 0.0479 0.0035 -2.95 -7.68
-0.61 -0.59 -6.53 -5.32 Roaddcnsity1970 -0.0051-0.0050 0.0415 0.0122
-0.0170 0.0427 1.67 0.84 2.68 -18.61 Growthrateotroaddcnsity 0.0141 0.0072
Propornonothouseholds
-0.0799 -3.07 -6.52 -5.46 34.81 withelectricalpower1973 -0.0259 -0.0557 0.0347
0.0285 -0.95 -1.42 1.78 -12.42 Índexofqualitvandavailabilitvofroads -0.0080-0.0121-0.0113 Table8 (continuation)
IncomePerCapitaDifferencesbyRegión(IGACSample)
CaribbeanPacifícOrinoquianAmazonio AndeanRegiónvs. CaribbeanPacifícOrinoquianAmazonic
e/o) (%) (%) e/o)
Human Capital -0.0146 0.0799 -0.5564 -0.6234 -1.73 9.36 87.61 271.63
Enrollmentratcinprimarv
andsecondaryschool1973 0.0699 0.0165 0.1293 -0.0056 8.29 1.93 -20.36 2.44
Migrationrate1973 0.0337 -0.0124-0.6817 -0.6321 4.00 -1.45 107.34 275.42
Collegcgraduatesperthousand
oflaborforcé1973 -0.0797 0.0069 -0.0595 -0.0552 -9.46 0.81 9.37 24.05
Numberoftropicaldiseasedeaths 0.0689 per1.000people1979 -0.0385 0.0555 0.0695 -4.57 8.07 -8.74 -30.28
InstitutionsandLivingStandards 0.1807 0.1546 0.0881 0.1382 21.44 18.11 -13.87 -60.22
lnteractionbetweensoilanddegree ofurbanization1973 0.1728 0.0714 0.1674 0.1846 20.50 8.36 -26.36 -80.44
Proportionoflandwithcofeecrops1980 -0.0682-0.0094-0.0544 -0.0708 -8.09 -1.10 8.57 30.85
Incomeincqualitv1973 -0.0228 0.0043 -0.0468 0.0053 -2.70 0.50 7.37 -2.31
Percapitamunicipaltransfers
1973-1995) (yearly 0.0989 0.0883 0.0219 0.0191 11.73 10.34 -3.45 -8.32
TotalExplainedbytheRegression 0.7561 0.7458 -0.4851 0.0574 89.70 87.34 76.38 -25.01
ResidualUnexplainedbytheRegression 0.0868 0.1081 -0.1500 -0.2869 10.30 12.66 23.62 125.01 F. Sánchez y |. Xúñez | Planeación y Desarrollo XXXI | Julio-diciembre 2(10(1 (379-452)
Tables 7 and 8 present thc rcsults of the decomposition of thc average per capita income between thc Andean Regiónand the rest of the rcgions. Thc income per capita difference betwcen the Andean Regiónand thc Caribbean one is 0.75 (in logs). The geographical variables account for 0.62 (82%) of such differencc although thc magnitudc and sign of the contribution of each one varíes quite a lot. However, thc variable that contributes thc most (in favor of the Andean Región) to the per capita incomc differencc is distance to domestic markets. Infrastructure variables favor thc Caribbean Región but only because of 1973 stock variable (density of roads and coverage of electrical powcr). In contrast, the growth of road density has favored the Andean Región.
I luman capital variables favor as well thc Caribbean Región mainly because of its greater initial stock of college graduatcs. However, the contribution of school cnrollment ratcs favorsthe Andean región. It is worth to point out, however, that the ovcrall contribution of human capital to average income per capita differences between the Andean región and the Caribbean is quite small (-0.038 or 5% of thc difference) (Table 7). The contribution of institutions and living standards variablesto the income per capita differences between the Andean región and the Caribbean is modérate. Thus, the interaction between soil suitability and urbanization, the 1973 proportion of household under poverty and the 1973-1995 average per capita municipal transfers favor thc Andean Región while the proportion of land with coffee crops hurt Andean Región income. As a whole these variables account for 0.16 (in log) of the income per capita differencc or about 22% of it. The decomposition exercisc cxplains 97% of thc per capita income differences between the Andean R egion and the Caribbean, 75% with Pacific Región and 103% of the difference with the Orinoco Región.
In Table 8 we present the results of the income per capita difference decomposition based on the IGAC sample (that exeludes Antioquia, Bogotá and Cali). The estimates are similar to the ones obtained using thc whole sample, which is an indication of their robustness. The decomposition exercise performs even better with thc IGAC sample. In fact, its explain 100% of the difference in income per capita between thc Andean Región and the Caribbean one, 80" o, 120% and 117% of the difference with the Pacific, Orinoco and Amazonas regions, respectively. Appendixes 4 and 5 contain decomposition excrcises for GDP by área with similar results21.
B. Exercises of Income Per Capita Growth Decomposition
Tables 9 and 10 contain decomposition exercises of per capita income growth. The non-weighted growth difference between the Andean and Caribbean regions is 4.89 percentage points. The geographical variables account for 3.02 percentage points or 61.2%) of the differencc. The distance to the domestic markets explains most of the differencc in per capita income growth rates between the Andean and Caribbean regions. The same variable explains as well most of the growth difference between the Andean and Pacific regions. This is a novel and important result, which may indícate that agglomeration economies and large market externalities are the leading forces pulling average income growth of the colombian regions. Other geographical variables that account for growth differences are rain precipitation and soil suitability, although in much lesser proportion than distance to domestic markets.
:: It is worth to point out that although the decomposition exercisc show that the difterences in regional per capita income are mostly explained by distance to domestic markets this does not imply that global income ditferences are also explained by this variable. In fact, if the sample is divided not in regions but in poor and rich municipalities the variable distance to domestic markets loses part of its relevance.
420 Tabk9
IncomePer CapitaGrowthDifferencesbyRegión(Totalsample)
CaribbeanPacific Orinoquian AndeanRegiónvs. AmazonicCaribbeanPacificOrinoquianAmazonic
Diference 0.0489 0.0228 0.0161 0.0453 100.00 100.00 100.00 100.00
0.0096 Pcrcapitaincome1973 -0.0097 0.028 0.0336
Geography 0.0302 0.0220 0.0017 0.0409 61.76 96.61 10.73 90.35
Rainprecipitation -0.0024 0.0018 0.0063 0.0053 -4.91 7.89 39.13 11.70
abovethesealevel Altitude 0.0388 0.0063 0.0154 0.0147 79.35 27.63 95.65 32.45
abovethelevel sea^2 Altitude -0.0343 -0.0047 -0.0170 -0.0159 -70.14 -20.61 -105.59 -35.10
SoilsuitabilityÍndex -0.0050 -0.0011 -0.0055 -0.0007 -10.22 -4.82 -34.16 -1.55
Distancetodomesticmarkets 0.0334 0.0206 0.0070 0.0441 68.30 90.35 43.48 97.35
(.aucaRiver 0.0001 0.0003 -0.0002 -0.0002 0.20 1.32 -1.24 -0.44
MagdalenaRiver 0.0001 -0.0000722-0.0000722-0.0000722 0.20 -0.32 -0.45 -0.16
River(inkilometers) -0.0005 -0.0011 -0.0042 -0.0063 -1.02 -4.82 -26.09 -13.91
Infrastructure -0.0006 -0.0005 0.0009 -0.0022 -1.23 -2.19 5.59 -4.86
Roaddensity1970 -0.0001 0.0002 0.0005 0.0013 -0.20 0.88 3.11 2.87
Proportionofhouseholds withelectricalpower1973 -0.0005 -0.0007 0.0004 -0.0035 -1.02 -3.07 2.48 -7.73
Human Capital 0.0004 0.0038 -0.0148 -0.0165 0.82 16.67 -91.93 -36.42
Hnrollmentrateinpriman1 andsecondan1school1973 0.0031 0.0013 0.0014 -0.0090 6.34 5.70 8.70 -19.87
Migrationrate1973 0.0013 -0.0001 -0.0189 -0.0104 2.66 -0.44 -117.39 -22.96
Collegegraduatesperthousand -0.0010 oflaborforcé1973 -0.0016 0.0002 -0.0024 -3.27 0.88 -6.21 -5.30 Table9 (contimiation)
IncomePerCapitaGrowthDifferencesbyRegión(Totalsample)
CaribbeanPacifícOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic AndeanRegiónvs. (%) (%) (%) (%)
Numbcroftropicaldiseasedeaths
22.98 11.70 per1.000people1979 -0.0024 0.0024 0.0037 0.0053 -4.91 10.53
InstitutionsandLivingStandards 0.0003 0.0008 0.0002 0.0005 0.61 3.51 1.24 1.10
lntcractionberwccnsoilanddegrec
3.07 17.39 6.84 ofurbanization1973 0.0029 0.0007 0.0028 0.0031 5.93
-13.66 -5.52 Proportionoflandcofee withcrops1980 -0.0024 -0.0004-0.0022 -0.0025 -4.91 -1.75
-0.22 -0.0002 2.19 -2.48 Incomeinequality1973 0.0005 -0.0004 -0.0001 -0.41
50.17 TotalExplainedby theRegression 0.0303 0.0261 -0.0120 0.0227 61.96 114.60 -74.36
49.83 ResidualUnexplainedbytheRegression0.0186 -0.0033 0.0281 0.0226 38.04 -14.60 174.36
Note:AndeanRegióndoesn'tineludeBogotá. Table10
IncomePerCapitaGrowthDifferencesbyRegión(IGACSample)
CaribbeanPacificOrinoquianAmazonicCaribbeanPacific AndeanRegiónvs. OrinoquianAmazonic (%) (%) (%)
Difcrence 0.0440 0.0247 -0.0115 0.0438 100.00 100.00 100.00 100.00
Per(Zapitaincome1973 0.0041 -0.0082 0.0106 0.0335 -
Geography 0.0308 0.0253 0.0047 0.0331 70.06 102.38 -40.76 75.54
Rainprecipitation -0.0017 0.0005 0.0083 0.0092 -3.86 2.02 -72.17 21.00
0.0286 Altitudcabovethcsealevcl 0.0747 -0.0042 0.0329 169.77 -17.00 -248.70 75.11
Altitudeabovcthcsealcvcl^2 -0.0665 0.0050 -0.0320 -0.0365 -151.14 20.24 278.26 -83.33
-0.0055 -0.0073 SoilsuitabilitvÍndex -0.0024 -0.0058 -12.50 -9.72 63.48 -13.24
Distancetodomesticmarkets 0.0325 0.0222 0.0014 0.0375 73.86 89.88 -12.17 85.62
CaucaRivcr 0.0004 0.0009 -0.0001 -0.0001 0.91 3.64 0.87 -0.23
MagdalenaRiver 2.74K-05-1.23K-O5-1.23K-05-1.23H-05 0.06 -0.05 0.11 -0.03
River(inkilometers) -0.0008 -0.0003 -0.0043 -0.0114 -1.82 -1.21 37.39 -26.03
-0.0034 WateravailabilitvÍndex -0.0087 -0.0464 -0.0511 -19.77 -13.77 403.48 -116.67
Wateravailabilin1index^2 0.0098 0.0054 0.0583 0.0639 22.27 21.86 -506.96 145.89
Proportionoffíatlands -0.0034 0.0016 -0.0018 -0.0055 -7.73 6.48 15.65 -12.56
Infrastructure -0.0008 -0.0015 -0.0001 -0.0004 -1.72 -5.90 0.87 -0.91
Roaddcnsity1970 4.42K-05 4.36I--05-0.0003 -0.0001 0.10 0.18 2.61 -0.23
Proportionofhouseholds withelectriealpowcr1973 -0.0005 -0.0010 0.0006 -0.0015 -1.14 -4.05 -5.22 -3.42
índexofqualirvandavailabilin1 -0.0005 ofroads -0.0003 -0.0004 0.0012 -0.68 -2.02 3.48 2.74 Tal)le10(continuation)
IncomePerCapitaGrowthDifferencesbyRegión(IGACSample)
CaribbeanPaciflcOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic AndeanRegiónvs. (%) (%) (%) (%)
Human Capital 0.0007 0.0028 -0.0215 -0.0246 1.59 11.34 186.96 -56.16
Enrollmentrateinpriman-
-48.70 -0.46 andsccondaryschool1973 0.0030 0.0007 0.0056 -0.0002 6.82 2.83
-2.02 239.13 -57.99 Migrationratc1973 0.0013 -0.0005 -0.0275 -0.0254 2.95
Collegegraduatesperthousand
-3.42 oflaborforcé1973 -0.0022 0.0001 -0.0016 -0.0015 -5.00 0.40 13.91
Numberoftropicaldiseasedeaths
5.71 per1.000pcople1979 -0.0014 0.0025 0.0020 0.0025 -3.18 10.12 -17.39
-16.52 InstitutionsandLivingStandards 0.0028 0.0024 0.0019 0.0045 6.36 9.72 10.27
Interactionbetweensoilanddegree
14.16 ofurbanization1973 0.0058 0.0024 0.0057 0.0062 13.18 9.72 -49.57
Proportionoflandwithcofeecrops1980 -0.0019 -0.0002-0.0015 -0.0019 -4.32 -0.81 13.04 -4.34
0.81 20.00 0.46 Incomeinequaliry1973 -0.00110.0002 -0.0023 0.0002 -2.50
TotalExplainedbytheRegression 0.0336 0.0290 -0.0150 0.0126 76.30 117.54 130.54 28.74
ResidualUnexplaiaedbytheRegression0.0104 -0.0043 0.0035 0.0312 23.70 -17.54 -30.54 71.26 I\ Sánchez y J. Núñez | Planeación y Desarrollo XXXI ¡ Julio-diciembre 2000 (379-452)
The effect of the infrastructure on regional growth differences is small as observed intable 9. Enrollment ratesin primar}' and secondary education, and incidence of tropical diseases account for a littlc more of the regional differences in growth. Enrollment rates and tropical diseases explain, respectively, 5.7% and 10.7% of the growth differences between the Andean and Pacific regions. Similar results are obtained by using the IGAC sample as presented in table 10. A decomposition of the per land área GDP growth with the total and IGAC samples is found in appendixes 6 and 7.
C. Effects of Geography on Income Inequality between Municipalities
Table 9 contains the effect of the different geographical variables on per capita income inequality between municipalities. In order to carry out the exercise we calculated the original income per capita Gini coefficient of the colombian municipalitics using as a weighting variable the municipal population. Then, in order to compute the effect of each variable on the Gini we estimated thc income per capita under the assumption that such variable is equal to the average in every municipality and by using the new income estimation we calculated a new Gini coefficient. Thus, the estimated income is estimated as follows
(5)
where 1"A is the new municipal income per capita, Y; is the original income per capita and -P,(-X,t -xM) is expression that subtracts from income the municipal valué of X and adds to it the average valué of such variable. Then, a new Gini coefficient is calculated based on >'v Table 11 shows the result of the exercise. The original Gini coefficient is equal to 0.606 and after accounting by the geographical variables it is obtained a Gini of 0.609. Although, that is the compound effect of all geographical variables, some of these variables havean important impact on municipal inequality. For instance, if it is assumed that either distances to domestic markets or altitudes are equal for all municipalities the Gini coefficient reaches 0.63 in the first case and 0.647 in the second.
The infrastructure variables have a greater impact on Gini than geography does, although such impact depends basically upon the 1973 electrical power household coverage. Thus, had such variable been equal (in 1973) the 1995 income per capita Gini coefficient would have been 0.482 (Table 11). The human capital variables have the strongest impact on Gini (Table 11). Had the 1973 valué of such variable been equal for all municipalities the 1995 Gini would have been just 0.388. Both 1973 enrollment rates and 1973 college graduates (per thousand people) have a large impact on Gini. Finally, the institutional and living standard variables as a whole have a modérate impact on municipal Gini coefficient. Nevertheless, it is considerable the effect of 1973 poverty rate. Had it been equal (in 1973) in all municipalities the 1995 Gini coefficient would have been 0.514. The same exercise was carried out using the IGAC sample and the results obtained were quite similar when using the whole sample (Table 11).
425 !■'. Sánchez v J. Núñez | Plantación y Desarrollo XXXI ¡ Julio-diciembre 2000 (379-452)
Tab/eH Effects of DifferentVariables on Municipal Income Inequality
Total Sample IGAC Sample
Gini % Change Gini % Change
Initial Gini 60.62 - 60.11 -
Recalculated after acounting by:
Geography 59.20 -2.34 59.43 -1.12
Rain precipitation 59.25 -2.25 59.87 -0.39
Altitude above the sea level 61.46 1.39 60.75 1.0"
Solí suitabilitv index 59.95 -1.10 59.49 -1.03
Distance to domestic markets 60.90 0.4" 64.22 6.85
Cauca River 60.2" -0.5" 60.28 0.29
Magdalena River 60.55 -0.11 60.20 0.15
River (in kilometers) 61.43 1.34 58J0 -2.35
Water availabilitv index 61.08 1.62
Proportion of fíat lands 59.06 -l."4
Infraestructura 47.55 -21.56 52.57 -12.53
Road density 19T0 60.30 -0.53 60.11 0.02
growth rate of road densitv 60.78 0.26 60.19 0.14
Proportion of households with electrical
power 1973 47.8" -21.03 53.15 -11.57 índex of quality and availabilitv of roads 59.68 -0."l
Human capital 34.13 -43.70 39.50 -34.28
F.nrollment rate in priman- and secondary
School 1973 49.63 -18.12 53.51 -10.97
Migration rate 1973 60.07 -0.90 55.47 -7."1
College graduates per thousand of labor forcé 43.42 -28.37 48.73 -18.93
Number of tropical disease deathsper 1.000
people 1979 61.11 0.81 61.38 2.12
Institutions and Living Standards 68.23 12.56 69.51 15.64
Interaction between soil and degree
of urbanization 1973 65.77 8.50 65.79 9.46
Proportion of land with cofee crops 1980 59.97 -1.0" 59.13 -1.62
Income inequaliry 1973 60.59 -0.05 60.67 0.93
per capita municipal transfers
(vearlv average 1973-1995) 64.10 5.74 64.53 7.36
426 F Sánchez y J. Núñez I Plantación y Desarrollo XXXI I Julio-diciembre 2000 (379-452)
The results of municipal income inequality exercises may have the following implication: the división of the country by the traditional regions could hide the real sources of income differences. A global analysis based for instance in the Gini coefficicnt, shows that education and infrastructure explain the most of the municipalper capita income differences. However, a regional analysis shows that the differences come mainly from distance to markets. This implies that the use of traditional regions may not be the most adequate approach to the study of growth and development, at least in the colombian case.
IX. I Conclusions
The objective of this document was to determine the relationship between the geographical variables and per capita income, per capita income growth, population density and population growth of colombian municipalities. In order to carry out econometric estimations at a municipal level we constructed a set of geographical variables based on soils, climates and roads maps. Additionally, we extracted some other geographical variables from the physical homogeneous zone statistics of the colombian Institute of Geography (IGAC).
We found that geography weighs both for the level of municipalper capita income and for its growth. Thus, geography explains between 36% and 47% of the municipal income per capita variance and between 35% and 40%) of the municipal income per capita growth variance. It was established that, among the geographical variables, distance to domestic markets and soil suitability have the largest influences both on per capita income and on its growth. Another important finding was that geographical variables seem to be more significant more for the poor municipalities than for the rich ones. Thus, in the poor municipalities, geography explains between 25% and 32% of per capita income variance and between 24%) and 27% of per capita income growth variance. In contrast, in the rich municipalities, geography matters less. It explains between 18% and 25% of per capita income variance and between 16% and 17% of its rate of growth variance. Thus, geography affects income level and its growth through the productivity of land, the availability of natural resources such water and rivers, the presence of tropical diseases, and agglomeration. However, geography influences the fate of a región but that is not theend. There is an important role for human action either through public policy or prívate intervcntion. Education, infrastructure and better institutions can boost economic regional economic growth, help poor regions to overeóme the poverty trap of low income and low economic growth.
Geography also matters for population density. It was found that such variable is determined by rain precipitation, altitude and soil suitability. At the same time, the more densely populated municipalities are cióse to domestic markets and awayfrom the seaports. Infrastructure availability and access to social services also influence population density.
427 I\ Sanche/ y ). Núñez | Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
In the last section of the paper wc carried out some decomposition exercises to determine the role of geography in explaining incomc per capita differences between colombian regions. The exercises show that geography accounts for more than 70% of the per capita income differences between the Andean and the Caribbean regions. A similar percentage is found with respect to the rest of colombian regions. The importance of human capital and infrastructure variables was found to be much less in explaining incomc differences. Finally, \ve measured the impactof geography on municipal income inequality. The exercises found that, asa whole, geography is neutral for municipal income inequality. Howcver, distances to seaports and rain precipitation haveby themselves a modérate effect on such variable. For municipal inequality it seems to matter much more differences in infrastructure and human capital variables.
42,H F. Sánchez y J. Núñcv | Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452)
Appendix 1
Example of a Tablewith Information on Homogeneous Physical Zones Instituto Geográfico Agustín Codazzi 1998
Seccional: Mapa desubzonas Municipio: Tausa Cundinamarca código. 25 Homogeneas-físicas Código: 793
No. Área Clima V.P. Pend. Aguas Vías Usos Puntos (Has) Investigados
1 93 F.H 49 a Suficientes 3 9-6 1 2 2.780 F.H 55 c Escasas 1—2 9-6 2-3-4-5-6-15-24-29-2A
3 450 F.H 55 c Escasas 3 12 12B
4 115 F.H 55 c Escasas ! 4 9-6 17
5 598 M.F.H 49 c Flscasas 4 9-6 30
6 526 F.S. 44 c Escasas 1 9-6 20-25 7 78 F.S. 44 c Escasas ! 9-8 23
8 2.008 F.S. 38 d Escasas 1 -3 9 - 6 19-26-27 10 2.026 M.F.H 38 d Escasas 4 9-6 11-11A-32-33-36-37 11 291 M.F.H 38 ! d Escasas ; 4 7/9 13-31
12 488 F.H 30 e Escasas 4 12 12
13 489 F.H 30 e Escasas 4 9-6 16-10A
14 112 F.H 30 e Escasas 4 12/9 10
15 160 F.H 30 e Sin 6 7/9 14
16 260 F.S. 30 ! e2p Sin : 5 9-6 43
17 460 F.S. 17 i fr Sin 6 7/9 9
18 119 F.H 23 fr Sin 1 9-6 21
19 367 M.F.H 17 fr Escasas ¡ 4 9-6 34
20 2.274 M.F.H 17 fr Sin 6 7/9 8-3-47-35-39
21 386 M.F.H 17 fr Sin 6 12 12A
22 790 M.F.H 17 fr Sin 6 12/9 --12C
23 218 M.F.H 6 fr Sin 7 8 1 IB
24 1.888 M.F.H 6 cr Sin : 8 8 12D-34B
25 934 M.R. 6 Sin 7 8 38
26 398 M.F.H 6 e Sin 6 8 40
27 652 M.F" 6 Sin ! 7 8 22-28 Aguas 830
Total 19.881
Source: Instituto Geográfico Agustín Codazzi.
429 F. Sánchez y J. Nuñez Planeación y Desarrollo XXXI ! Julio-diciembre 20011 (379-452)
Appendix 2
The aim of quantile regression is to estimate the median of the dependent variable. However, this method could also be used to estimate any quantile of the dependent variable. This method is very similar to ordinan- least squares, but its objective is to estimate the median of the dependent variable. Quantile regression finds a line through the data that minimizes the sum of the absolute residuals rather than the sum of the squares ot the residuals.
The method is very useful when the data have heteroscedaskticity. The quantik regression methodology estimates the variance-covariance matrix of the coefficients using the method of Koenker and Basset (1982) and Rogers (1993). This methodology turns the problem of finding a minimum absolute deviation into a linear programming problem. Thus, it performes an iteration process until it rinds the coefficients that minimize the residuals. The covariance matnx is:
cov(P) = R21R,R2Í where , R¡ = X WWX, R2 = X'X and IC is a diagonal matrix with elements IT equals to:
qifrM^) if r>o V-q)ifrM (0) if r 4)0 F. Sánchez y J. Nüñez , Planeación y Desarrollo XXXI Julio-diciembre 200(1 (379-432) Appendix 3 Table 1 Determinants of Municipal GDP Per Land Área Dependent Variable: Total Sample Total Sample IGAC Sample IGAC Sample GDP per Área (1) (2) (3) (4) Constant 31.2H6 23.3983 23.1622 18.8403 (14.690)' (9.491)' (8.458)'' Geographical Variables Rain precipitation -0.9847 -0.5369 -1.0257 -0.4970 (-8.098)' (-6.2803)' (-7.079)'' (-4.406)3' Altitude above sea level -0.0522 0.1269 0.4067 0.4337 (-0.222) (0.957) (1.785)1' (2.614)' Altitude above sea leve.1^2 0.0100 -0.0112 -0.0271 -0.0390 (0.462) (-0.867) (-1.303) (-2.539)1' Soil suitability Índex 1.4384 0.7021 1.2119 0.7670 (10.332)' (5.652)' (8.099)' (5.764)'' Distance to domestic markets -1.8431 -1.2898 -1.5605 -1.1450 (-10.967)' (-10.287)'' (-8.952)' (-".889)v Water availabilitv Índex - 6.4698 3.3947 (2.587)" (1.926)' Water availabilitv indcxA2 -2.8728 -1.8544 : : (-2.578)-; (-2.3Ó7)2 Proportion of fíat lands - 0.4892 0.4191 (2.037)'' (2.431)2 Cauca River -0.0278 -0.3354 -0.1952 -0.3695 (-0.082) (-1.584) (-0.530) (-1.374) Magdalena River -0.1633 0.0480 -0.3044 -0.0872 (-0.675) (0.307) (-1.310) (-0.522) Rivcr (in kilometers) -0.2060 -0.1767 -0.0268 -0.1135 (-1.816)'' (-2.430)J (-0.223) (-1.257) 431 F. Sánchez y ). Núñez | Planeación y Desarrollo XXXI | Julio-diciembre 2000 (379-452) Table 1 (continuation) Determinants of Municipal GDP Per Land Afea Dependent Variable: Total Sample Total Sample IGAC Sample IGAC Sample GDP per Atea (1) (2) (3) (4) Infrastructure Variables Proportion of households 2.5174 - 2.2508 with electrical power 1973 (10.107)"' (8.276)' Road density 1970 - 0.3281 0.2201 (10.646)' (5.649)"' Road densitv rate of growth - 2.6573 1.6575 (6.448)'-' (3.266)' índex of quality and availability - of roads Human Capital Migration rate 1973 1.1248 1.4280 (2.246)2-' (2.351)2' Enrollment rate in priman- 2.5344 2.6828 and secondary school 1973 - (5.729)'-' - (5.209)' 0.1114 College graduates per thousand of - 0.0916 - labor forcé 1973 (8.281)" p.016)""' Number of tropical disease deaths -0.0741 -0.0330 per 1.000 people 1979 (-1.806)" (-0.724) Institutions and Living Standards Interaction between soil and degree -0.2214 - -0.3228 of urbanization (1973) (-3.194)'- (-4.454)'' 0.0181 Proportion of land with cofee - 0.0106 - crops 1980 (2.666)'-' (3.597)" Income inequality 1973 0.0259 -0.1259 (0.273) (-0.9679) Per capita municipal transfers -0.4683 -0.3021 (vearly average 1973-95) (-4.488)'- (-2.421)2' Number of Observations 873 872 613 612 R2 0.3422 0.6955 0.3548 0.6875 Significant at 90%. 2 Significant at 95%. ' Significant at 99%. 432 Tabk2 QuantileRegressionsofGDP PerLandÁrea DependentVariable TotalSample TotalSample IGACSSample IGACSample 25%poorest25% richest25% poorest 25% richest25% poorest 25% richest25% poorest 25% richest Constant 31.2073 283463 21.6468 23.9899 24.2121 21.8126 19.9390 16.8363 (21075)"'(10612)" (17.121)v (10.877)" (9529)'-'(4993)" (5.627)" (5.623)" GeographicalVariables Rainprecipitation -0.8059 -0.9590 -0.5427 -0.5952 -0.7879 -0.9666 -0.4357 -0.4835 (-7.821)"(-4.877)-"(-7.789)" (-5.292)"(-5.180)3-'(-3.052)"(-2.375)2-1(-2.963)" Altitudeabovesealevcl 0.2101 -0.1765 0.4008 -0.0807 0.4693 -0.0656 0.4439 0.4150 (1.323) (-0.563) (3.577)" (-0.472) (2.020)" (-0.127) (1.545) (1.807)" Altitudeabovesealevel/x2-0.0150 0.0251 -0.0376 0.0044 -0.0350 0.0153 -0.0430 -0.0354 (-0.967) (0.842) (-3.380)3'(0.270) (-1.640) (0.330) (-1.633) (-1.689)" SoilsuitabilityÍndex 1.3707 1.6440 0.5757 0.7033 1.2627 1.4576 0.6035 0.9805 (11.306)'-'(7.074)" (5.310)" (4.386)" (8.083)" (4.442)"(2.382)2'(5.358)" Distancetodomestic -1.9250 -4.3498 -1.5104 -1.1272 -1.7107 -1.3055 -0.8989 -1.5097 Markets (-13.186)"(-4.981)"(-14.873)1' (-6.356)"(-9.645)"'(-3.610)"(-4.190)"(-7.270)'-' WateravailabilityÍndex - - - 6.3558 5.9675 3.4252 4.7818 - - - (2.579)" (1.437) (1.230) (2.278)2/ Wateravailabilityindex/N2 - -2.8220 -2.5619 -1.9599 -2.7147 - - - (-2.554)-'(-1.367) (-1.567) (-2.908)" Proportionoffíatlands - - - - 0.2819 0.6090 0.5787 0.3652 - - - (1.102) (1.330) (1.956)" (1.549) CaucaRiver -0.1440 0.2517 -0.3195 -0.1470 -0.3159 0.3475 -0.2248 -0.4033 (-0.527) (0.533) (-1.783)" (-0.544) (-0.828) (0.483) (-0.534) (-1.260) MagdalenaRiver -0.3360 0.1078 -0.0884 0.0675 -0.3819 -0.0444 -0.2311 0.0133 (-1.688)"(0.305) (-0.676) (0.341) (-1.658)"(-0.099) (-0.795) (0.060) River(inkilometers) -0.4411 -0.1471 -0.1067 -0.2302 -0.3036 0.0986 -0.1694 -0.2057 (-5.603)"(-0.928) (-1.888)"(-2.422)-'(-2.596)"(0.415) (-1.119) (-1.619) Tab/e2 (continnation) QuantileRegressionsofGDP PerLandÁrea IGAC Sample IGAC Sample DependentVariable TotalSample TotalSample 25% poorest25% richest 25%poorest25% richest25% poorest25% richest25% poorest25% richest InfrastructureVariables 2.0709 2.7089 1.4372 2.3703 Proportionothouscholds - (6.220)' (9.032)''(8.593)'' (2.832)' witheléctrica!power1973 0.1139 0.0051 0.4173 0.2629 - Roaddcnsity1970 - (3.643)' (0.167) (17.165)'(6.276)' 0.1529 3.3873 2.0428 1.9599 Roaddensityrate (0.190) (3.627)'' (2.337)2 otgrowth (10.294)' 0.6932 0.8641 - - índexofquality - - (2.542)^(4.172)1 andavailabilitvofroads Human Capital 1.5232 0.6813 1.1260 1.2756 Migrationrate1973 (1.527) (1.729)' (1.184) (1.843)' 2.3824 2.3770 2.4113 2.5580 Knrollmentrareinpritnary (3.346)' ; : (4.320)3 - - (3.611)' andsecondaryschool1973 (6.059)' 0.1367 0.1199 0.1173 Collcgegraduatesper 0.1064 (10.608)' (3.887)' (5.550)' Thousandoflabortorce (9.264)' 1973 -0.2361 -0.0841 -0.1380 Numberoftropicaldisease -0.5600 (-9.188)'(-3.444)' (-0.636) (•2.102)- Deathsper1.000people - 1979 Table2 (confmuation) QuantileRegressionsofGDP PerLandAxea DependentVatiable TotalSample TotalSample IGAC Sample IGAC Sample I"'"" 25%poorest25% richest25%poorest25%richest¡25%poorest25%richest:25%poorest25%richest InstitutionsandLivingStandards Intcractionbctwccnsoil;ind - -0.1248 -0.2910 -0.2625 -0.4466 DegreeofLirbanizarion1973 - (-1.903)'(-3.436)' - (-1.791)"(-4.838)" Proportionofland 0.0126 0.0094 0.0231 0.0148 with cofcccrops1980 (3.471)' (1.875)' (2.704)' (2.109)- Incomcinequalitv1973 - 0.1295 -0.0213 -0.0278 0.0281 (-0.187) - (1.465) (-0.120) (0.166) Percapitamunicipal -0.2406 -0.2919 -0.4885 -0.1877 Transfcrs(vearlvaverage (-2.650)' (-2.O86)2 (-2.192)'(-1.022) 1973-1995) Numberofobservations873 873 872 872 613 613 612 612 R: 0.2430 0.1751 0.4368 0.4794 0.2348 0.2019 0.4448 0.4972 1 Significantat90%. "!Significantat95%. * Significantat99%. F Sánchez y |. Núñez \ Plantación y Desarrollo XXXI | Julio-diciembre 2000 (379-452) Table 3 Determinants of Municipal GDP Per Land Área Growth Dependent Variable Total Sample Total Sample IGAC Sample IGAC Sample (1) (2) (3) (4) Constant 0.1218 0.3065 -0.0083 0.20T3 (2.063)2 (5.625)'' (-0.101) (2.725)'; GDP per land área 1973 -0.0037 -0.0141 -0.0038 -0.0151 (-3.268)' (-11.487)' (-2.569)' (-9.486)' Geographical Variables Rain precipitation -0.0058 -0.0066 -0.0171 -0.0103 (-1.698)'' (-2.ni)2 (-3.831)' (-2.532)2 Altitude above sea level -0.0021 -0.0029 0.0138 0.0097 (-0.412) (-0.641) (2.019)- (1.625) Altitude above sea levéis 0.0004 0.0003 -0.0007 -0.0007 (0.925) (0.855) (-1.175) (-1.259) Soil suitabilitv Índex 0.0194 0.0155 0.0247 0.0210 (4.752)" (3.536)' (5.360)' (4.296):v Distance to domestic markets -0.0323 -0.0321 -0.0284 -0.0275 (-6.621)'' (-7.307)' (-5.233)' (-5.412)' Cauca River -0.0077 -0.0069 -0.0176 -0.01 71 (-0.912) (-0.930) (-1.598) (-1.766)'' Magdalena River -0.0085 -0.0065 -0.0017 -0.0044 (-1.332) (-1.186) (-0.247) (-0.741) River (in kilometers) 0.0149 0.0029 0.0149 0.0020 (5.278)' (1.145) (4.078)' (0.627) Water availability Índex 0.2320 0.1542 (3.090)' (2.425)2' Water availabilitv index^2 - - -0.1037 -0.0822 (-3.101)' (-2.910)' Proportion of fíat lands 0.0087 0.0126 - (1.218) (2.015)-' Infrastructure Variables Proportion of Households - 0.0724 0.0685 with electrical power 1973 (8.245)' _ (6.907)' Road Density 1970 - 0.0027 - -0.0001 (2.915)' (-0.329) Índex of Qualitv and Availabilitv - - 0.0149 of Roads (2.584)' 4 36 F. Sánchez y J. Núñcz | Planeación y Desarrollo XXXI i julio-diciembre 21.100 (379-452) Tab/e 3 (continuation) Determinants of Municipal GDP Per Land Área Growth Dependent Variable Total Sample Total Sample IGAC Sample IGAC Sample (1) (2) (3) (4) Human Capital Migration rate 1973 - 0.0865 - 0.0789 (4.969)" (3.584)" Enrollment rate in priman- 0.0719 - 0.0696 and secondan7 school 1973 (4.604)" (3.686)" Collegegraduates per thousand of - 0.0016 0.0018 Labor Forcé 1973 (4.010)" (3.125)" Number of tropical disease dcaths - -0.0012 - 0.0030 per 1.000 people 1979 (-0.384) (1.809)1"' Institutions and Living Standards Interacoon between soil and degree - -0.0075 - -0.0105 of urbanization (1973) (-3.161)" (-4.026)" Proporción of land vith cofee -0.0001 - 4.01K-05 crops 1980 (-1.248) (0.214) Income inequality 1973 - -0.0043 -0.0131 (-1.298) (-2.798)" Number of observations 873 872 613 612 R2 0.1209 0.3739 0.1997 0.4320 1 Significant at 90%. 2 Significant at 95%. " Significant at 99%. 437 Tabk4 QuantilRegressionsofMunicipalGDP PerLandÁreaGrowth IGAC Sample IGAC Sample DependentVariable TotalSample TotalSample 25% poorest25% richest 25% poorest25% richest25% poorest25% richest25% poorest25% richest (4) (5) (6) (7) (8) (1) (2) (3) 0.3319 -0.0630 0.0783 0.0481 0.3254 Constant 0.0842 0.2012 0.1913 (-0.634) (0.825) (0.655) (4.688)' (1.325) (2.377)- (2.884)! (4.283)'■ -0.0161 -0.0111 -0.0138 -0.0031 -0.0021 -0.0144 GDP perlandárea1973 -0.0031 -0.0031 (-1.737)' (-1.041) (-8.444)'(-10.304)' (-2.680)'(-1.746)1(-7.519)' (-7.688)'' GeographicalVariables -0.0079 -0.0054 -0.0235 -0.0058 -0.0133 Rainprecipitation -0.0028 -0.0060 -0.0062 (-0.966) (-3.938)'(-1.472) (-3.538)' (-0.759) (-1.227) (-1.665)' (-1.923)'■' -0.0109 0.0150 0.0165 0.0093 0.0079 Altitudeabovesealevel 0.0023 -0.0129 0.0025 (1.178)' (1.504) (1.392) (0.412) (-1.692)' (0.435) (-1.766)'(1.795)'' -0.0008 -0.0011 -0.0005 -0.0005 \lt-V1 1 i-i'n1i-n-*>!'"'9 -0.0000 0.0011 Altitudeabovesealevelz 0.0001 0.0011 (-1.323) (-1.001) (-1.081) (0.269) (1.564) (-0.050) (1.923)'(-1.082) 0.0153 0.0159 0.0184 0.0322 0.0170 0.0237 SoilsuitabilitvÍndex 0.0161 0.0245 (3.042)'(6.022)r (3.499)' (4.328)' (2.509)- (2.708)-'(3.031)'(5.374)'- -0.0328 -0.0270 -0.0396 -0.0270 -0.0298 -0.0220 Distancerodomestic -0.0369 -0.0325 (-5.644)'(-4.009)'(-5.873)'(-4.618)' Markets (-6.554)'(-4.778)-'(-6.082)' (-4.209)' 0.2423 0.0819 0.2517 0.0584 - WateravailabilityÍndex - - (2.537)- (1.185) (4.424)' (1.328) -0.1037 -0.0299 -0.1202 -0.0343 Wateravailabilitvindex"2 (-2.395)2(-0.959) (-4.754)'(-1.779)' 0.0238 -0.0004 0.0200 0.0028 Proportionoffíatlands - (-0.051) (3.108)' (0.550) - (2.376)' QuantilRegressionsofMunicipalGDP PerLandÁreaGrowth DependentVariable TotalSample TotalSample IGACSample IGACSample 25%poorest25% richest25% poorest25% richest25% poorest25% richest25%poorest25% richest (1) (2) (3) (4) (5) (6) (7) (8) GeographicalVariables River(¡nkilometers) 0.0144 0.0124 0.0067 0.0024 0.0126 0.0174 0.0039 -0.0027 (4.352)"'(3.035)' (1.946)1 (0.653) (2.550)2'(3.879)' (1.126) (-0.965) CaucaRiver 0.0068 0.0076 -0.0054 -0.0025 -0.0058 0.0058 -0.0186 -0.0190 (0.674) (0.666) (-0.589) (-0.259) (-0.395) (0.448) (-2.063)2(-2.327)2- MagdalenaRiver -0.0075 -0.0037 -0.0059 0.0003 0.0018 0.0020 -0.0023 0.0029 (-0.995) (-0.435) (-0.878) (0.052) (0.194) (0.242) (-0.388) (0.564) InfrastructureVariables Proportionofhouseholds i 0.0665 0.0512 | 0.0638 0.0579 i withelectrícalpower1973 (5.513)' (4.288)'' (5.472)""'(6.842)"' Roaddensirv1970 0.0028 0.0018 -0.0000 -0.0005 (2.478)2 (1.414) - (-0.100) (-1.168) índexqualitv of and _ 0.0190 0.0215 availabilityofroads - - (3.090)-v(4.783)'' Human Capital Migrationrate1973 0.0941 0.1255 0.0762 0.1201 (4.185)' (5.412)' (3.106)'(6.942)' Enrollmentratepriman1 in - - 0.0719 0.0731 - 0.0865 0.0601 andsecondan'school1973 - (3.505)3'(3.427)'' (4.263)'(3.525)'" Collcge graduatesper 0.0016 0.0019 0.0028 0.0023 thousandoflabor (4.229)' - (3.345)'' - (4.173)' (5.089)' I'orce1973 Numberoftropicaldisease 0.0039 -0.0010 - - 0.0056 0.0081 deaths1.000 perpeople (1.177) (-0.335) (2.460)2 (3.582)' 1979 'i'abk4 (continuation) QuandlRegressionsofMunicipalGDP PerLandÁreaGrowth DependentVariable TotalSample TotalSample IGAC Sample IGAC Sample 25%poorest25% richest25% poorest25% richest25% poorest25% richest25% poorest25% richest (6) (7) (8) (1) (2) (3) (4) (5) InstitutionsandLivingStandards -0.0124 -0.0103 Interactionbetwccnsoiland -0.0099 -0.0061 - _ (-5.O95)4 Degreeofurbanization1973 (-2.824)J/(-2.013)-' (-3.884)1' 0.0002 -0.0000 Proportionofland -0.0001 -0.0001 - withcofeecrops1980 (-0.997) (-0.929) (1.601) (-0.203) - -0.0053 -0.0163 lncomeinequality1973 -0.0025 -0.0078 (-0.660) (-1.687)" (-1.004) (-4.196)" 612 Numberofobservations873 873 872 872 613 613 612 R2 0.0885 0.0549 0.3321 0.2274 0.1195 0.1169 0.2574 0.2806 "'Significantat90%. 2 Significantat95%. " Significantat99%. Appendix4 DecompositionofGDPPerLandÁreaRegión by (TotalSample) AndeanRegiónvs. CaribbeanPacifícOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic Diference 0.5633 0.9365 0.9258 1.5361 100.00 100.00 100.00 100.00 Geography 0.4929 0.6464 0.3610 1.6120 87.50 69.02 38.99 104.94 Rainprecipitation -0.1435 0.1085 0.3663 0.3093 -25.47 11.59 39.57 20.14 Altitudeabovcthesealeve] 0.4460 0.0725 0.1774 0.1693 79.18 7.74 19.16 11.02 Altitudeabovcthclcvel seaA2 -0.4162-0.0573-0.2062 -0.1935 -73.89 -6.12 -22.27 -12.60 -7.15 -33.47 SoilsuitabilityÍndex -0.2804-0.0670-0.3099 -0.0444 -49.78 -2.89 Distancetodomesticmarkets 0.8664 0.5344 0.1837 1.1432 153.81 57.06 19.84 74.42 CaucaRiver 0.0054 0.0098 -0.0071 -0.0071 0.96 1.05 -0.77 -0.46 MagdalenaRiver -0.00500.0025 0.0025 0.0025 -0.89 0.27 0.27 0.16 River(inkilometers) 0.0202 0.0430 0.1543 0.2327 3.59 4.59 16.67 15.15 Infrastructure -0.0619 0.0954 0.3814 0.4662 -10.99 10.19 41.20 30.35 Roaddensity1970 -0.0841 0.1722 0.3503 0.8960 -14.93 18.39 37.84 58.33 Growthrateofroaddensitv 0.0855 0.0114 -0.0223 -0.0097 15.18 1.22 -2.41 -0.63 Proportionofhouseholdswitheléctrica! power1973 -0.0633-0.0882 0.0534 -0.4201 -11.24 -9.42 5.77 -27.35 DecompositionofGDP PerLandÁreabyRegión(TotalSample) (coiitiniiíition) OrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic AndeanRegiónvs. CaribbeanPacific (%) (•/.) (%) <%) -11.02 10.95 -21.46 -32.73 Human Capital -0.0621 0.1025 -0.1987 -0.5027 linrollmentratcinpriman- 0.0456 -0.2812 17.26 4.60 4.93 -18.31 andSecondarySchool1973 0.0972 0.0431 0.0159 -0.0023-0.2313 -0.1282 2.82 -0.25 -24.98 -8.35 Migrationratc1973 College graduatesthousand per -0.0799 -0.1896 -23.27 1.94 -8.63 -12.34 ofLaborforcé1973 -0.13110.0182 Numberoftropicaldiseasedeaths 0.0669 0.0963 -7.83 4.64 7.23 6.27 per1.000people1979 -0.0441 0.0435 0.4333 30.21 -4.59 31.75 28.21 InstitutionsandLivingStandards 0.1702 -0.0430 0.2939 lntcractionbetweenSoilanddegrec 0.2264 37.42 5.46 2223 14.74 ofurbanizaron1973 0.2108 0.0511 0.2058 11.18 1.35 6.08 4.26 Proportionoflandwifh cofcccrops1980 0.0630 0.0126 0.0563 0.0654 0.12 -0.0052 0.0044 0.0019 0.36 -0.56 0.48 Incomcinequality1973 0.0020 Percapitamunicipaltransferí 9.09 -0.1015 0.0274 0.1396 -18.75 -10.84 2.96 (yearly1973-1995) -0.1056 2.0088 85.56 90.47 130.77 TotalExplainedbytheRegression 0.5391 0.8013 0.8376 95.70 0.0882 4.30 14.44 9.53 -30.77 ResidualunexplainedbytheRegression0.0242 0.1352 -0.4727 Note:AndeanRegióndocsn'tincludeBogotá. Appendix5 DecompositionofGDPPerLandÁreaRegión by (IGACSample) AndeanRegiónvs. Caribbean PacificOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic (%) f/o) f/.) (%) Diference 0.8429 0.8539 -0.6351 -0.2295 100.00 100.00 100.00 100.00 Geography 0.5135 0.5540 0.3008 1.3924 60.92 64.88 -47.36 -606.71 Rainprecipitation -0.0789 0.0223 0.3714 0.4117 -9.36 2.61 -58.48 -179.39 -97.86 Altitudeabovethelevel sea 1.6223 -0.0914 0.6215 0.7159 192.47 -10.70 -311.94 Altitudeabovethesealevel^2 -1.5539 0.1180 -0.7487 -0.8535 184.35 13.82 117.89 371.90 índex Soilsuitabilitv -0.2537 -0.1131-0.3348 -0.2670 -30.10 -13.25 52.72 116.34 Distancetodomestic markets 0.7663 0.5231 0.0343 0.8858 90.91 61.26 -5.40 -385.97 CaucaRiver 0.0104 0.0194 -0.0029 -0.0029 1.23 2.27 0.46 1.26 -0.0026 MagdalenaRiver 0.0059 -0.0026 -0.0026 0.70 -0.30 0.41 1.13 River(inkilometcrs) 0.0206 0.0092 0.1091 0.2893 2.44 1.08 -17.18 -126.06 WateravailabilityÍndex -0.1455-0.0575-0.7718 -0.8497 -17.26 -6.73 121.52 370.24 WateravailabilitvindexA2 0.1773 0.0985 1.0556 1.1572 21.03 11.54 -166.21 -504.23 Proportionoffíatlands -0.0573 0.0281 -0.0303 -0.0918 -6.80 3.29 4.77 40.00 Infrastructure -0.1135 -0.24060.2752 0.0155 -13.47 -28.18 -43.33 -6.75 Roaddensitv1970 -0.0301 -0.0297 0.2432 0.0714 -3.57 -3.48 -38.29 -31.11 -41.61 Growthrateofroaddensitv 0.0314 0.0161 -0.0380 0.0955 3.73 1.89 5.98 Proportionofhouseholds withclectricalpower1973 -0.0844-0.1812 0.1128 -0.2596 -10.01 -21.22 -17.76 113.12 -0.0458 índexofqualin-andavailabilitvofroads -0.0304 -0.0428 0.1082 -3.61 -5.36 6.74 -47.15 DecompositionofGDP PerLandÁreabyRegión(IGACSample) (continuation) AmazonioCaribbeanPacificOrinoquianAmazonio AndeanRegiónvs. "JaribbeanPacificOrinoquian (%) (%) (%) e/o) 9.01 18.25 126.01 Human Capital -0.05710.0769 -0.1159 -0.2892 -6.77■ F.nrollmcntratc¡npriman' -0.0087 12.73 2.97 -31.27 3.79 andsecondaryschool1973 0.1073 0.0254 0.1986 -0.2180 1.38 -0.50 37.02 94.99 Migrationrate1973 0.0116 -0.0043-0.2351 Collegc graduatesperthousand -0.1055 -0.1137 -18.06 1.55 17.90 45.97 ofLaborforcc1973 -0.1522 0.0132 Numbcroftropicaldiscascdcaths -0.0238 -2.82 4.99 -5.40 -18.74 per1.000people1979 0.0426 0.0343 0.0430 8.24 -49.91 -171.02 InstitutionsandLivingStandards 0.2951 0.0704 0.3170 0.3925 35.01 Interactionbetweensoilanddegree 35.80 14.60 -46.04 -140.48 ofurbanization1973 0.3018 0.1247 0.2924 0.3224 0.0821 9.40 1.28 -9.94 -35.77 Proportionoflamí\vithcofeccrops1980 0.0792 0.0109 0.0631 -0.0219 0.0025 -1.27 0.23 3.45 -1.09 lncomeinequality1973 -0.0107 0.0020 Percapitamunicipaltransfers -0.0145 -8.92 -7.87 2.61 6.32 (yearly1973-1995) -0.0752-0.0672-0.0166 -122.36 -658.47 TotalExplainedbytheRegression 0.6380 0.4607 0.7771 1.5112 75.69 53.95 12.66 23.62 125.01 ResidualunexplainedbytheRegression0.0868 0.1081 -0.1500 -0.2869 10.30 U DecompositionofGDPPerLandÁreaGrowthbyRegión(TotalSample) AndeanRegiónvs. CaribbeanPacificOrinoquianAmazonicCaribbeanPacificOrinoquianAmazonic Difcrcnce 0.0308 0.0026 -0.0174 -0.0195 100.00 100.00 100.00 100.00 GDP perlandarca1973 0.0016 -0.0121 -0.018 -0.027 Geography 0.0168 0.0126 0.0019 0.0301 54.55 484.62 -10.92 -154.36 Rainprecipitation -0.0018 0.0013 0.0046 0.0039 -5.84 50.00 -26.44 -20.00 -0.0107 Altitudeabovethcsealcvel ■0.0017-0.0042 -0.0040 -34.74 -65.38 24.14 20.51 Altitudeabovethcsealevel^2 0.0134 0.0018 0.0066 0.0062 43.51 69.23 -37.93 -31.79 SoilsuitabilitvÍndex -0.0064-0.0015-0.0071 -0.0010 -20.78 -57.69 40.80 5.13 Distancetodomesticmarkets 0.0218 0.0134 0.0046 0.0288 70.78 515.38 -26.44 -147.69 CaucaRiver 0.0001 0.0002 -0.0001 -0.0001 0.32 7.69 0.57 0.51 MagdalenaRivcr 0.0007 -0.0003-0.0003 -0.0003 2.27 -11.54 1.72 1.54 River(inkilometers) -0.0003-0.0006-0.0022 -0.0034 -0.97 -23.08 12.64 17.44 Infrastructure -0.0024-0.00110.0043 -0.0046 -7.79 -42.31 -24.71 23.59 Roaddensity1970 -0.0006 0.0014 0.0028 0.0073 -1.95 53.85 -16.09 -37.44 Proportionofhouseholdsuithelectrical power1973 -0.0018-0.00250.0015 -0.0119 -5.84 -96.15 -8.62 61.03 DecompositionofGDP PerLandÁreaGrowthbyRegión(TotalSample) (continuation) CaribbeanPacificOrinoquianAmazonic AndcanRegiónvs. CaribbeanPacificOrinoquianAmazonio (%) (%) (%) (%) 18.25 126.01 Human Capital -0.05710.0769 -0.1159 -0.2892 -6.77 9.01 Unrollmcntrateinprimary 8.77 46.15 -6.90 41.03 andsecondarvschool1973 0.0027 0.0012 0.0012 -0.0080 3.57 -3.85 98.28 48.72 Migraciónrate1973 0.0011 -0.0001 -0.0171 -0.0095 Collegegraduatesperthousand 16.92 -0.0033 -7.47 11.54 8.05 oflaborforcc1973 -0.00230.0003 -0.0014 NumbcrofTropicalDiscascDcaths -26.92 6.32 8.72 per1.000pcoplc1979 0.0007 -0.0007-0.0011 -0.0017 2.27 -34.48 -35.90 InstitutionsandLivingStandards 0.0064 0.0025 0.0060 0.0070 20.78 96.15 InteractionBervveenSoiland Degree 25.00 69.23 -43.68 -42.56 ofurbani/ation1973 0.0077 0.0018 0.0076 0.0083 -0.0010-0.0002-0.0009 -0.0010 -3.25 -7.69 5.17 5.13 ProportionofI.andwithCofec('rops1980 -0.0007 -0.0003 -0.97 54.62 4.02 1.54 lncomcIncqualitv1973 -0.00030.0009 35.63 -51.28 TotalExplainedby theRegression 0.0230 0.0147 -0.0062 0.0100 74.68 565.38 64.37 151.28 ResidualuncxplainedbytheRegressior0.0078 -0.0121-0.0112 -0.0295 25.32 -465.38 Note:AndeanRegióndoesn'tincludeBogotá. DecompositionofGDPPerLandÁreaGrowthbyRegión(IGACSample) AndeanRegiónvs. Caribbean PacificOrinoquianAmazonioCaribbeanPacificOrinoquianAmazonic <%) <%) (%) (%) Difercncc 0.0260 -0.0273 -0.0264 0.0050 100.00 100.00 100.00 100.00 GDP perlandarca1973 -0.0013 -0.0063-0.0202 -0.0127 _.. Geography 0.0165 0.0131 0.0072 0.0269 63.46 -47.99 -27.27 538.00 Rainprecipitation -0.0020 0.0005 0.0095 0.0105 7.33 -1.89 190.00 1.05 altitudcabovethcseaIcvcl 0.035" -0.0020 0.0157 0.0136 -130."7 7.58 272.00 1.57 altirudeabovethcsealeve!^2 -0.0272 0.0020 -0.0131 -0.0149 99.63 -7.58 -262.00 -1.49 Soilsuitabilityíndex -0.0080-0.0036-0.0106 -0.0085 29.30 13.64 -212.00 -0.85 Distancetoclomesricmarkcts 0.0188 0.0128 0.0008 0.0217 68.86 -48.48 16.00 2.17 ("aucaRivcr 0.0004 0.0008 -0.0001 -0.0001 -1.47 -3.03 -2.00 -0.01 MagdalenaRivcr 0.0003 -0.0001 -0.0001 -0.0001 -1.10 0.38 -2.00 -0.01 River(inkilometers) -0.0003-0.0001 -0.0018 -0.0049 1.10 0.38 -36.00 -0.49 WateravailabilitvÍndex -0.0070-0.0027 -0.0374 -0.0411 25.64 10.23 -748.00 -4.11 Wateravailabilitvindcx/s2 0.0080 0.0044 0.0476 0.0522 -29.30 -16.67 952.00 5.22 Proportionoffíatlands -0.00220.0011 -0.0012 -0.0036 8.06 -4.17 -24.00 -0.36 Infrastructure -0.0032-0.00650.0047 -0.0055 ll^"124.62 94.00 -0.55 Ro;uldensitv1970 -0.0002-0.0002 0.0020 0.0005 0.73 0.76 40.00 0.05 Proportionofhouseholdswithelectrical powcr1973 -0.0025-0.00550.0034 -0.0079 9.16 20.83 68.00 -0.79 índexofqiialitvandavailabilitvofroads -0.0005 -0.0008-0.0007 0.0019 1.83 3.03 -14.00 0.19 DecompositionofGDP PerLandÁreaGrowthbyRegión(IGACSample) (conlinnafion) OrinoquianAmazonic PaciflcOrinoquianAmazonicCaribbeanPaciflc AndeanRegiónvs. Caribbean (%) (%) (%) (%) -0.0209 -10.26 7.95 -320.00 -2.09 Human Capital 0.0028 -0.0021-0.0160 linrollmcntrateinprimary 12.00 -0.02 0.0007 0.0056 -0.0002 -10.99 -2.65 1 andsecondarvschool1973 0.0030 -1.63 -0.0003-0.0175 -0.0163 -2.93 1.14 -350.00 Migrationratc1973 0.0008 Collegegraduatesperthonsand -0.0019 -0.0017 9.16 -0.76 -38.00 -0.17 oflaborforcé1973 -0.00250.0002 Numberoftropicaldiseascdeaths -0.0027 -0.0022 0.0027 5.49 10.23 -44.00 ■0.27 per1.000people1979 0.0015 0.0117 -33.33 -17.58 140.00 1.17 InstitutionsandLivingStandards 0.0091 0.0046 0.0070 lntcractionbetweenSol]anddegree 0.0101 0.0112 -38.46 -16.29 202.00 1.12 ofurbanization1973 0.0105 0.0043 0.0002 -0.^3 -0.15 4.00 0.02 ProportionofLandwithcofeecrops1980 0.0002 3.99E-O5 0.0002 0.0003 0.03 0.0003 -0.0033 5.86 1.14 -66.00 Incomcinecjualitv1973 -0.0016 0.0122 -92.31 -34.62 58.00 1.22 TotalExplainedby theRegression 0.0252 0.0091 0.0029 -0.2869 -317.95 -409.47 -3000.00 -28.69 ResidualunexplainedbytheRegression0.0868 0.1081 -0.1500 Note:AndeanRegióndoesn'tincludeBogotá. I-'. Sánchez y |. Nurk-z | Planeacion y Desarrollo XXXI , |ulio-cliciembre 20(11) (379-452) Bibliography Aghion, P and Howitt, P. (1998) "Fndogenous Growth Theory", MIT Press. Birchenall, }. and Murcia, G. (1997), "Convergencia regional: Una revisión del caso colombiano", Desarrollo y Sociedad, No.40, september. Bushnell, D. (1996) "Colombia, una Nación a pesar de sí misma", Santa Fe de Bogotá, Fxiitorial Planeta. Barro, R and Sala-i-Martin (1995) "Fxonomic Growth", McGraw-Hill. Campos, N and Nugent, |. (1998) "Instituciones y crecimiento: ;Puedc el capital humano ser un vínculo?". Revista de la Cvpcil, Cepal. Cárdenas, M. et al. (1993) "Convergencia, crecimiento y migraciones interdepartamentales: Colombia 1950-85", Coyuntura liconómica, Vol.XXIII, No.l, april. Denninger and Squire (1999) "Inequality and Growth: New FJvidence", }oumalofDevelopment Ticonomics, februarv. Diamond, |. (1999) "Guns, Germs and Steel", New York, W.VÍ'. Norton Company. Fcddcrke, | and Klitgaard,R. (1998) "Fxonomic Growth and Social Indicators: An Kxploratory Analysis", h.conomic Development and Cultural Change, april. Fernández. C. (1999) "Aglomeration and Trade: The colombian Case", Santa Fede Bogotá, Banco de la República. Flórez, C.F^. (1998) "Transformaciones socio-demográficas en Colombia en el siglo XX", Cede, Universidad de los Andes, Mimeo. Gallup, J. and Sachs, |. (1998) "Geography and Fxonomic Development", HIID. Gleaser, F^., Kallal, H., Scheinkman, |. and Shleifer, A. (1992) "Growth of Cities", journalof Political ¡iconomy, Vol.100, No.6. Gouesset, X. (1998) "Bogotá: el nacimientode una metrópoli", Tercer Mundo. Hamilton, | (1970) "Del Magdalena a Bogotá", I 'ia/eros extranjeros en Colombia, Cali, Siglo XXI Santa Fede Bogotá, Carvajal. Instituto Geográfico Agustín Codazzi -IGAC- (1998) "Proyecto áreas biofísicas homogéneas", Documento de Trabajo, Santa Fede Bogotá. jaramillo, S. y Cuervo, F. (1987) "La configuración del espacio regional en Colombia", Serie F^studios CFXJF 1, Santa Fede Bogotá, Universidad de los Andes. 44 9 F. Sanche/ v |. Núñez | Pkineación y Desarrollo XXXI | Julio-diciembre 20(10 (379-452) laramillo, S. (1998) "Migraciones e interacción regional en Colombia, 1973-1993", Revista Territorios, Cider-Uniandes, june. Jacobs, J. (1969) "The Kconomy of Cities", New York, Yintage. Krugman, P. (1991) "Geography and Trade", M1T Press. Krugman, P. (1998) "\\ hat's New about the New Kconomic Geography?", Oxford Rtriewof l.conomic Polio; Vol. 14, No.2. Landes, D. (1999) "The Wealth and Poverty of Nations", New York, W'.W". Norton Company. Leibovich, J. and Núfiez, J. (1999) "Los activos y recursos de la población pobre en Colombia", Documento CKDK, march. Leibovich, J (1996) "lü desempeño económico de los migrantes", Rerista Plantnaón r Desarrollo, Yol.XXYIl,No.4,december. Lora, H. (1999) "Cómo llega Colombia al siglo XXI", Washington, ínter Amcrican Development Bank. Meisel, A. (1993) "¿Polarización o convergencia? A propósito de Cárdenas, Pontón y Trujillo", Coyuntura liconámka, Vol.XXIII, No.2, ¡une. Meisel, A. and Bonnet, |. (1999) "¿Por qué perdió la Costa el siglo XX?", Documentos de trabajo sobre economía regional, Cartagena, Banco de la República. Montengro, S. (1996), "I'.l papel de las regiones para la estabilidad macroeconómica de Colombia", Desarrollo y Sociedad, No.38, september. Montenegro, S. y Suárez, G. (1998) "Determinantes del crecimiento regional en Colombia", Santa Fe de Bogotá, Facultad de Rconomía, Universidad de los Andes, Mimeo. Ocampo, J.A. (1987) "Café y macroeconomía, 1950-1987", Coyuntura hconówica, Yol.XXI, No.2, l'edesarrollo. Parsons, J (1997), "La colonización antioqueña en el occidente colombiano", Santa Fede Bogotá,, Banco de la República- VX Ancora Fiditores. Posada, I i. (1998) "'Kl caribe colombiano: una historia regional", Santa Fe de Bogotá, Banco de la República- Hl Ancora F.ditores. Póvcda Ramos, G. (1998) "Vapores fluviales en Colombia", Santa Fede Bogotá, Tercer Mundo I-xlitores-Colciencias. Ramírez, M.T. (1999) "The impact of Transportation Infrastructure on the Colombian Ixonomy", Borradores de heonomia, No. 124, Banco de la República. 4 so 1;. Sánchez y J. Núñcx ! Planeación y Desarrollo XXXI : Julio-diciembre 2(100 (379-452) Rappaport,J. (1999) "Local liconomic Growth", H1ID. Rocha, R. y Vivas, A. (1998) "Crecimiento regional en Colombia: ¿Persiste la desigualdad?", Revistade heonomía del Rosario, No.l, january. Sánchez Torres, F. (1999) "Pobreza, descentralización y acceso a los servicios sociales", Coyuntura Social, Fedesarrollo, may. Soto, A.(1998) "Crecimiento y convergencia departamental: Una aproximación panel al caso colombiano, 1960-1995", Santa Fede Bogotá, Tesis de Magister en Kconomía, Universidad de los Andes. Temple, J. (1999) "The New Growth Hvidence", journalof MconomkLiterature, Vol.XXXVIII, No.l, March. 45 i