World Bank Reprint Series: Number 195 Kyu Sik Lee Intra-urban Location of Manufacturing Employment in Colombia Reprinted with permission from Journal of Urban Economics, vol. 9 (1981), pp. 222-41. JOURNAL OF URBAN ECONOMICS 9, 222-241 (19891) SYMPOSIUM ON URBANIZATION AND DEVELOPMENT Intra-urban Location of Manufacturing Employment in Colombia' Kyu SIK LEE Urban and Regional Economics Division, Development Economics Department, The World Bank, Washington, D.C. 20433 Received August 22, 1979 This paper describes the spatial distribution of manufacturing employment in Bogota and analyzes the changes in the location patterns over the 1970-1975 period using industrial directory data. The analysis shows strong evidence of the spatial decentralization of manufacturing employment owing to a steady movement of firms outward from central areas; new large firms with more than 100 employees tend to locate near the periphery while the births of small firms tend to take place near the center. Measures of concentration indicate an increasing dispersion of manufacturing establishments during the period. 1. INTRODUCTION Changes in the patterns of intra-urban location of manufacturing em- ployment have been fairly well documented for cities in developed countries, particularly for large cities in the U.S. and U.K.,2 where the decentralization of manufacturing employment has been observed for the past several decades.3 Comparable studies for rapidly growing cities in less developed countries (LDCs), however, are rare. The location of employ- ment tends to influence residential locations and sets the basis for the overall development patterns in an urban area. As the first step in a continuing research effort, this paper presents findings for Bogota, Co- 'An earlier version of this paper was presented at the American Economic Association meetings, Chicago, Illinois, August 29-31, 1978. Support for the work reported in this paper was provided by the City Study research project (RPO 671-47) funded by the World Bank. The views and conclusions reported here are those of the author and not of the World Bank or its affiliated organizations. The author would like to thank M. Wilhelm Wagner and Yoon Joo Lee for research assistance, members of the City Study research staff at the World Bank and at Corporacion Centro Regional de Poblacion in Bogota for comments on the work presented here, with particular appreciation to Gregory K. Ingram, Jose Fernando Pineda, Roger Schmenner, Douglas Keare, Michael Beesley, and Larry Westphal. 2A comprehensive review of the literature on intra-urban manufacturing location can be found in Kemper (1973, Chap. 2); also, Struyk and James (1975) give a review of previous research related with this paper. 3For example, Hoover and Vernon (1959), Leone (1971), and Hamer (1975). 222 0094-1 190/81/020222-20$02.00/0 Copyright 0 1981 by Academic Prsss, Inc. All rights of reproduction in any form reserved. EMPLOYMENT LOCATION IN COLOMBIA 223 lombia which also show a strong decentralization of manufacturing em- ployinent. In the theoretical literature, the location behavior of a finn has been typically postulated in the neoclassical framework: given price gradients for rents and wages, transport costs of inputs and outputs, and the location of markets, a firm will tend to locate where it maximizes its profits.4 In the recent empirical literature on employment location a popular approach has been to investigate not only the growth and decline of stationary (non- relocating) firmls but also the location decision of newly established firms and relocating firms, and analyze changes in the location patterns at the margin. Sach empirical analyses require micro establishment data. Since the 1970 Economic Census, the National Statistics Department (DANE) of Co- lombia has been updating its annual industrial directory files which pro- vide information on individual establishment operations including loca- tion, production, sales, and input uses.5 These data for the 1970- 1975 period revealed a strong trend of de- centralization of manufacturing employment.6 To fully exploit the data content we did the analysis by the location tenure of firms, i.e., newly established (births), defunct (deaths), relocating (movers), and stationary (mature) firms. This amounts to decomposing changes in the stock of employment by the flows of births and deaths, relocation of jobs, and stationary growth and decline of employment.7 The location of births and the destination of movers are particularly important for understanding the location dynamics of employment and predicting the future spatial struc- ture of an urban area. 2. OVERVIEW OF MANUFACTURING EMPLOYMENT IN BOGOTA AND CALI: COMPARISONS W1TH U.S. CITIES According to a 1977 household survey,8 in both Bogota and Cali, about 37% of the population were in the labor force: more than one million people were employed in Bogota while over a third of a million worked in Cali. Manufacturing jobs accounted for 25% of employment in Bogota and 30% in Cali. This share was the largest among all sectors in Cali, while in Bogota, it was second only to the services sector which had 32% of the 4See Alonso (1967), Mills (1972), and Ingram (1977). 5This data set is very similar to the Dun and Bradstreet Market Identifier (DMI) data in the U.S., which have been used in recent employment location studies for U.S. cities. See Leone (1971), Schmenner (1973 and 1981), Kemper (1973), and Struyk and, James (1975). These files, however, include only establishments with 10 or more employees. This group accounts for more than 70% of total manufacturing employment in Bogota. 6This paper examines Cali as well as Bogota, but presents details for Bogota. 7This approach was also umed by Leone, Schnenner, Struyk and James, and Cameron (1973). 8Published in DANE Statist. Bull., No. 309, April 1977. 224 KYU SIK LEE TABLE I Manufacturing Firm Size and Employment Distribution: Comparison of Bogota and Cali with U.S. cities Size Bogota Cali Washington, D.C. Boston Chicago Los Angeles (persons) % cumm. % cumm. % cumm. % currmn. % cumm. % cumni. A. Firm size distributiona Less than 20b 57.55 57.55 55.25 55.25 59.11 59.11 50.39 50.39 46.89 46.89 53.03 53.03 20-49 24.17 81.72 22.45 77.70 24.53 83.64 25.08 75.47 24.06 70.95 24.48 77.51 50-99 10.11 91.83 11.95 89.65 8.88 92.52 12.93 88.40 11.60 82.55 11.18 88.69 100-499 7.43 99.26 9.33 98.98 6.54 99.06 10.05 98.45 14.61 97.16 9.84 98.53 500 or more 0.75 100.00 1.02 100.00 0.93 100.00 1.56 100.00 2.84 100.00 1.46 100.00 B. Employment distribution' less that 20b 14.25 14.25 10.78 10.78 12.17 12.17 7.57 7.57 4.95 4.95 8.07 8.07 20-49 18.26 32.51 12.71 23.49 17.47 29.64 12.28 19.85 8.37 13.32 13.48 21.55 50-99 17.31 49.82 15.79 39.28 13.63 43.27 13.92 33.77 9.00 22.32 12.40 33.95 100-499 35.74 85.56 36.32 75.60 31.20 74.47 31.15 64.92 33.42 55.74 30.79 64.74 500 or more 14.45 100.00 24.39 100.00 25.54 100.00 35.08 100.00 44.26 100.00 35.25 100.00 C. Average finn size (persons) less than 20 9.95 10.07 9.07 9.78 9.60 9.53 20-49 30.35 29.21 31.35 31.86 31.57 34.48 50-99 64.80 68.12 67.61 70.05 70.46 69.47 100-499 193.34 200.81 210.00 201.78 207.66 195.87 500 or more 771.63 1233.00 1,203.50 1,465.40 1,416.58 1,509.07 Total 40.18 51.58 44.04 65.07 9t.80 62.62 (1,000 persons) Population' 3,453 1,057 2,862 2,754 6,978 7,041 Total employmentc 1,157 351 1,110 1,098 2,503 2,596 Mfg. employment' 291 106 67 262 782 719 Mfg. empl, share (%o) (25.16) (30.20) (6.04) (23.87) (31.24) (27.70) Sources: Figures for U.S. cities are from Gregory K. Ingram, "Reductions in Auto Use from Carpools and Improved Transit," Harvard University, October, 1976, they are derived from County Business Patterns, 1973, and data on journey to work (Table 2) of the 1970 U.S. Census. 11970 for Bogota and Cali; and 1973 for U.S. cities. All U.S. figures are for central city. bFor Bogota and Cali, this category covers 5-19; and for U.S. cities, 4-19. C1977 estimate for the Special District of Bogota and Cali; 1970 U.S. census SMSA figures for U.S. cities. city's employment. Large U.S. cities such as Boston, Chicago, and Los Angeles also have about 25-30% of employment in manufacturing. The firm size distribution in Bogota, compared in Table 1, is not much different from that of Cali except that Cali has proportionally more firms of the largest size. The firm size distribution of Bogota looks much like that of Washington, D.C., although the latter has only a small armount of manufacturing employment. Cali, in contrast, resembles Los Angeles in terms of the firm size distribution as well as its share of manufacturing employment; Cali, however, has a smaller proportion of employment in the 500 or more category than Los Angeles. Moreover, in this last category, EMPLOYMENT LOCATION IN COLOMBIA 225 TABLE 2 Birth, Death, and Relocation Rates in Bogota, Cali, and U.S. Cities Births Deaths Moversd Establishments Employment Establishments Employment Establishments Employment % of Annual % of Annual % of Annual % of Annual % of Annual % of Annual base rate base rate base rate base rate base rate base rate Cleveland' 9.97 3.22 2.59 0.86 14.07 4.49 7.75 2.52 13.83 4.41 5.77 1.89 Minneapolis-a 12.29 3.94 6.17 2.02 18.00 5.67 11.25 3.62 15.93 5.05 8.28 2.69 St. Paul Boston' 6.10 1.99 1.30 0.43 13.40 4.28 8.00 2.60 9.80 3.17 4.70 1.54 Phoenix0 24.40 7.55 12.10 3.88 20.20 6.32 5.30 1.74 8.90 2.88 4.70 1.54 New Yorkb 10.21 4.98 3.95 1.96 7.56 3.71 3.55 1.76 11.45 5.57 1.24 0.62 Bogotac 52.38 8.79 31.96 5.70 27.01 4.90 12.61 2.40 19.12 3.56 16.59 3.12 Calic 43.13 7.44 24.48 4 48 26.88 4.88 11.27 2.16 18.33 3.42 10.40 2.00 'From R. Struyk and F. James, "Intrametropolitan Employment Location," Lexington Books, Lexing- ton, Mass., 1975; covered 1965-1968 period (1965 was the base year). bFrom R. Leone, "Location of Manufacturing Activity in the New York Metropolitan Area," Yale University, 1971; covered 1967-1969 period (1967 was the base year). 'The period covered was 1970-1975; 1970 was used as the base year. The base year figures can be seen in Lee (1978). dIn the case of Bogota, the figures include establishments which moved at least to another seccion changing DANE's six-digit zone code; in the case of Cali the figures include establishments which moved at least to another barrio changing the first four digits of DANE's zone code. Including the moves within the same seccion, the annual relocation rate of establishments was 5.12% for Bogota and 4.28% for Cali. the average firm size in Bogota is only about half of those in large U.S. cities. In terms of employment dynamics, Table 2 indicates that Colombian cities are somewhat different from U.S. cities. At an annual rate of 8.8%, the birth rate of firms9 in Bogota is higher than that of all five U.S. cities. Phoenix, with a birth rate of 7.6%, comes closest to Bogota's and is slightly higher than Cali's. The average death rate in Bogota, as well as in Cali, is comparable to that of Cleveland, Minneapolis, and Boston. It should be noted that the birth rate is greater than the death rate for Bogota, Call, Phoenix, and New York;10 the opposite is true in Boston, Minneapolis, and Cleveland. 9This may include other than genuine births; for example, the birth category may include those moving into the area. I0This was true for New York in the late sixties. The situation may have been reversed by now. Both studies for U.S. cities, cited in Table 2, covered a two- to three-year interval in the late sixties. A study of such a short period may reflect only what happened in that particular portion of business cycle. In this respect, our data set is somewhat more attractive because it covers a six-year period. 226 KYU SIK LEE In all cases, the activity rates in terms of establishments are greater than the rates based on employment. This indicates that the marginal firms are small ones which have a greater propensity to move, to start up and close down business. During 1970-1975, nearly 60% of firm births and deaths were in the 10-24 person category in Bogota as well as in Cali; moreover, the median age of establishments, as of 1970, was 4.5 years in both cities. Small, young firms therefore seem to contribute significantly to the changes in employment distribution in Colombian cities. 3. SPATIAL DISTRIBUTION OF EMPLOYMENT: EVIDENCE OF DECENTRALIZATION DANE's zone system divides Bogota into 38 comunas. To analyze changes in employment location patterns, we devised two zone systems by aggregating comunas into six "rings," and eight "radial sectors" (see Fig. 1). Tabulations of employment using the six rings, reported in Table 3, provide strong evidence of decentralization of manufacturing employment in Bogota: Ring 1 experienced an absolute decline in employment; moving out from the central business district (CBD), we observe an accelerating employment growth wvhich reaches an annual rate of 16% in Ring 5. The system of eight radial sectors follows the patterns of land use specialization in Bogota. The "industrial corridor," which bisects the city into the north and the south, is Radial Sectors IV and V. The residential north is Sector VIII, and the residential south is Sector II. The residual southwest area is labeled Sector III; the residual northwest is further TABLE 3 Distribution of Manufacturing Employmenta by Ring, Bogota, 1970-1975 1970 1975 Annual average Ring Persons % Persons % growth rate (%7) l(CBD) 4,538 5.60 4,102 3.47 - 2.00 2 11,767 14.53 14,898 12.59 4.83 3 34,351 42.42 47,858 40.44 6.86 4 18,112 22.37 25,958 21.94 7.46 5 11,548 14.26 24,047 20.32 15.80 6 391 0.48 729 0.62 13.27 n.a. 266 0.33 741 0.63 - Total 80,973 100.00 118,333 100.00 7.88 Data source: DANE Industrial Directory Files. aEstablishments with 10 or more employees. EMPLOYMENT LOCATION IN COLOMBIA 227 Comuno boundaries 6 Ring boundories ondnumbers ---VI Sector boundories ond numbers 6-8 Analysis Zone (ring, sector) BASED ONv 38 COMIUNAS 6-8 KILOMETERS 0 1 2 MitESa 2 3 FIG. 1. The zone system: Bogota. (This map has been prepared by the World Bank's staff, The denominations used and the boundaries shown on this map do not imply, on the part of the World Bank and its affiliates, any judgment on the legal status of any territory or any endorsement or acceptance of such boundaries.) 228 KYU SIK LEE divided into Sector VI, containing the airport, and Sector VII, which includes the "commercial corridor" of Chapinero areas and the residential northwest. As Table 4 shows, manufacturing employment growth was lowest in two residential sectors (II and VIII). In addition, employment growth in Sectors IV and V ("industrial corridor") was slightly lower than the overall growth rate. These sectors had lower than average employment growth, thus reducing their relative shares of manufacturing employment during the period. The "residual" sectors (III, VI, and VII), in contrast, increased their relative shares of manufacturing employment. Evidernce for the de- centralization of employment from the industrial corridor to the non- residential sectors is also supported by the extremely high employmers. growth in Sector III (157o) and Sector VII (12%o). 4. COMPOSITION OF CHANGES IN SPATIAL DISTRIBUTION OF EMPLOYMENT: THE INCUBATOR HYPOTHESIS To study changes in employment levels over time and space, we created a master file with the following categories of firms: (1) mature finns-those that appeared in all six annual directories with the same address; (2) births -those that appeared for the first time in any year during 1971-1975 and TABLE 4 Distribution of Manufacturing Employment' by Radial Sector, Bogota, 1970-1975 1970 1975 Annual average Radial sector Persons % Persons % growth rate (%O) I CBD 4,538 5.60 4,102 3.47 - 2.00 II South 2,451 3.03 3,218 2.72 5.60 III Southwest 6,255 7.72 12,741 10.77 15.29 IV, V Industrial corridor 57,833 71.42 84,362 71.29 7.84 VI Airport 3,532 4.36 5,333 4.51 8.59 VII Northwest 1,960 2.42 3,493 2.95 12.25 VIII North 4,138 5.11 4,343 3.67 0.97 n.a. 266 0.33 741 0.63 - Total 80,973 100.00 118,333 100.00 7.88 Data source: DANE Industrial Directory Files. 'Establishments with 10 or more employees. EMPLOYMENT LOCATION IN COLOMBIA 229 kept the same address; (3) deaths-those that dissappeared from the directory during 1971-1975 after having kept the same address; and (4) movers-those that relocated within Bogota during 1971-1975, including births and deaths that changed the address during the period. As indicated earlier, changes in spatial patterns of employment can best be understood by examining the location behavior of establishments by their location tenure. In Table 5, we find that the 4.7% growth of employ- ment in mature firms was much lower than the overall growth rate of 7.9% TABLE 5 Composition of Changes in Manufacturing Employment' by Ring, Bogota, 1970-1975 Matureb Birthsb Deathsb Movers 1970 1975 At Origin At Destination Ring Persons % Persons % Persons % Persons % Persons % Persons % I(CBD) 2.364 3.95 2,828 3.76 1,011 4.16 581 6.24 1,659 14.80 433 3.22 2 7,136 11.92 8,211 10.93 4,310 17.73 1,701 18.28 2,726 24.32 1,772 13.19 3 26,291 43.93 32,807 43.67 8,393 34.52 3,663 39.36 4,689 41.82 4,535 33.77 4 14,591 24.38 17,701 23.56 4,554 18.73 1,611 17.31 1,697 15.14 3,273 24.37 5 9,149 15.29 13,035 17.35 5,332 21.93 1,677 18.02 358 3.19 3,351 24.95 6 119 0.20 29 0.04 487 2.00 44 0.47 82 0.73 66 0.49 n.a. 191 0.32 519 0.69 224 0.92 29 0.31 - - - - Total 59,841 100.00 75,130 100.00 24,311' 100.00 9,306d 100.00 11,211 100.00 13,430 100.00 Summary Statistics Annual stationary Annual Annual Origin/destination Ring growth rate (%) birth rate (9%)' death rate (%)' ratio of movers l(CBD) 3.65 4.10 2.44 3.83 2 2.85 6.44 2.74 1.54 3 4.53 4.47 2.05 1.03 4 3.94 4.59 1.72 0.52 5 7.34 7.89 2.75 0.11 6 - 17.56 2.16 1.24 Total 4.66 5.39 2.20 0.83 Data source: DANE Industrial Directory Files. aExcluded those establishments with less than 10 employees, that appeared only for one year in the directory. The accounting identity between (I) the change in stock of employment in Table 3 and (2) the sum of flows and stationary growth in Table 5 does not hold mainly because of the omission of 135 firms (out of 2,629) which could not be classified into the four location tenure categories. See Lee (1978) for details. bExcluding movers. 'At birth. dln 1970. 'Annual average rate based on 1970 manufacturing employment (Table 3). 230 KYU SIK LEE (Table 4).11 While mature firms in Ring 5 grew at a rate close to the overall average, those in the CBD and Ring 2 grew more slowly than those in the other rings. The lowest growth of mature firms in Ring 2 with fairly high employment concentration may indicate the existence of capacity con- straint. In addition, we observe that nearly 60% more jobs were created by newly founded firms than by mature ones. In all rings, the number of Jobs created by births was more than the number resulting from the growth of mature firms; in Ring 2, the former was four times larger than the latter. It is striking that Ring 2 has the lowest growth of mature firms but is also a high birth area; its annual birth rate is higher than the industry average. Moreover, many firms start business in Ring 2 and then move out of this area. This may be evidence for the modified case"2 of the incubator hypothesis, which states that small, new firms tend to locate in centralized areas that provide essential services such as production space and financial services. Based on the criteria similar to those used by Struyk and James (1975), the incubator hypothesis is supported in Ring 2 as follows: the area's percentage share of employment by births (18%o) is greater than its share of base year employment (15% in Table 3). This is also true for Rings 5 and 6, but these areas are disqualified from being an "incubation area" because their new firms are not small. Ring 5 is impressive in terms of attracting new firms and movers; the annual birth rate was highest (except for Ring 6 which had a small employment base), and the inflow of jobs into Ring 5 was also more than ten times the outflow. It should be noted that the percentage distribution of births across the rings was very similar to that of deaths. To test the incubator hypothesis more specifically, we have defined 28 analysis zones by intersecting six rings with eight radial sectors (see Fig. 1); 13 we then determined as an incubation area those with the following characteristics: (1) the area has a high concentration of births; (2) the area's percentage share of employment by births is greater than the corresponding share of base year employment; and (3) the average firm size of births is small (i.e., about 25 persons). The areas shown in Table 6 had a high concentration of births; it is interesting to find that in the CBD and.the industrial corridor in Rings 3 and 4 the share of employment by birth; was smaller than the share of base year employment. Ort the other "Even though one might expect that the 1974 recession affected the growth of mature firms, this would not have resulted in a difference of such a magnitude in the employment growth rate. "2Hoover and Vernon (1959) were concerned with the incubation phenomenon in the city center. Struyk and James, however, "modified" the hypothesis by extending it to other centralized areas and traditional manufacturing districts. "3An analysis zone is identified by a two-digit number: the first digit refers to the ring and the second refers to the sector. EMPLOYMENT LOCATION IN COLOMBIA 231 TABLE 6 Selected Analysis Zones and "Incubation" Areas in Bogota Employment share (%o) Ratio Average firm size (persons) Analysis zone (AZ) 1970 Base (1) Births (2) (2)/(1) 1970 Base Births CBD (AZ 11) 5.60 4.16* 0,74 32.0 37.4 Ind. corridor (AZ 24,25) 11.11 15.07* 1.36* 36.0 27.5* Northwesta (AZ27,37,47) 2.16 5.94* 2.64* 21.3 20.3* Ind. corridor (AZ 34,35) 37.80 26.47* 0.70 61.3 32.2 Ind. corridor (AZ 44,45,54) 22.50 15.66* 0.70 94.4 50.8 Southwestb (AZ 53) 3.20 5.85* 1.83* 99.6 94.8 Ind. corridorc (AZ 55) 0.01 4.24* 424.00* 11.0 114.4 SubtoL-l 82.38 77.39 0.93 55.9 35.5 Total (Bogota) 100.00 100.00 1.00 55.1 34.1 ANote: * indicates passing a criterion to be an "incubation area." Data sources: DANE Industrial Directory files. aThis zone contains Chapinero commercial area. bIncludes Bosa. cIncludes Fontibon. hand, in the industrial corridor in Ring 5 and in Ring 2, the employment share by births was greater than the corresponding share of base year employment. In Ring 5, however, the average firm size of births was about 100 persons, indicating that the birth of large firms indeed occurs near the periphery. Only the industrial corridor in Ring 2 and the northwest (Chapinero) area therefore pass all three criteria for an "incubation area." To examine the direction of relocating firms, Table 7 considers six subareas, defined by the CBD, three segments of the industrial corridor, the north (Sectors VI, VII, and VIII) and the south (Sectors II and III).14 For these six subareas, a matrix of origin and destination was constructed in terms of the number of relocating establishments (Table 7). Nearly 40% of the movers relocated within their same subareas; 37 firms (13%o) moved within analysis zones (AZ) 34 and 35 alone. Only one-third of these 37 firms had more than 50 employees while about a half of the firms that moved from Ring 3 to Rings 4 and 5 in the industrial corridor were large size firms; long-distance movers tend to b& large size firms.'5 As expected, the CBD experienced a net loss of firms: 30 moved out of the CBD while only 4 moved into that area. Rings 2 and 3 in the industrial corridor and the north also lost more establishments than they gained, while the reverse happened in the south and in Rings 4 and 5, which attracted the largest number of movers: 52 establishments moved into that "4See Fig. 1. 15This is consistent with similar findings on Chicago by Moses and Wi'liamson (1967). 232 KYU SIK LEE TABLE 7 Origin and Destination of Movers,', Bogota (Number of Firms) Destination AZ44, AZ24, AZ34, 45, 54, Origin CBDb 25 35 55 North South Total CBD 12 7 10 3 6 4 42 (2)17 (1) (3) (1) (1) (1) (9) AZ24, 25 2 11 18 5 3 6 45 (0) (0) (1) (2) (1) (1) (5) AZ34, 35 0 '2 37 26 6 7 88 (4) (12) (13) (1) (4) (34) AZ44, 45, 0 2 6 10 4 2 24 54, 55 (1) (2) (4) (2) (2) (1 1) North 2 2 7 13 29 4 57 (1) (0) (2) (5) (2) (1) (lo) South 0 3 4 5 3 10 25 (1) (1) (1) (1) (3) (7) Total 16 37 82 62 51 33 281 (3) (7) (21) (26) (7) (11) (75) Data source: DANE Industrial Directory files. 'Establishments with 10 or more employees. bFor definition of zones, see Fig 1. AZ refers to analysis zone. cThe number of establishments with 50 or more employees at destination. area while only 14 moved out. Among the six subareas, a relatively large number of firms relocated from the CBD to Rings 2 and 3 in the industrial corridor, from Ring 2 to Ring 3, from Ring 3 to Rings 4 and 5, and also from the north to Rings 4 and 5 of the industrial corridor; only a few firims relocated from the CBD to as far as Rings 4 and 5. The fact that a large number of firms are moving out of Ring 2 is consistent with the incubator hypothesis tested above. 5. SPATIAL CONCENTRATION OF INDUSTRIES: A TEST OF AGGLOMERATION ECONOMIES The analyses presented above revealed changes in employment location for all manufacturing. Such patterns will, however, vary across industries since the location decisions of individual firms will depend partly on the characteristics of their parent industries. After describing the spatial con- centration of industries in Bogota, we will then examine to what extent the concentration of industries in particular locations might influence the location decisions of firms. The hypothesis is that in the presence of EMPLOYMENT LOCATION IN COLOMBIA 233 TABLE 8 Distribution of Manufacturing Employmenta by Ring and Industry, Bogota, 1970, 1975 (%o) Food Beverages Textile Apparel Furniture Printing (31I)b (313) (321) (322) (332) (342) Ring 1970 1975 1970 1975 1970 1975 1970 1975 1970 1975 1970 1975 1 5.0 2.9 - 0.5 2.7 1.1 18.5 9.5 3.1 2.0 28.0 23.6 2 15.2 15.5 30.9 4.3 15.4 11.0 17.2 23.2 19.6 5.1 15.4 10.7 3 62.9 53.8 54.2 52.9 46.1 44.0 49.6 46.7 36.8 48.6 29.6 38.7 4 5.0 9.2 12.7 16.7 25.8 25.9 3.3 12.5 32.3 31.7 22.6 20.2 5 11.8 18.1 2.2 25.6 9.2 17.2 11.4 8.1 8.2 12.6 4.4 5.9 6 - 0.5 - - 0.8 0.3 - - - - 1.1 n.a. 0.2 - - - - 0.7 - 0.1- - - - Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Non-electric Transport Other chemical Plastic Fabr. metal machine Elec. mach. equipment (352) (356) (381) (382) (383) (384) Ring 1970 1975 1970 1975 1970 1975 1970 1975 1970 1975 1970 1975 1 0.3 0.2 1.1 0.4 8.2 0.1 2.7 1.2 4.4 0.6 1.2 0.1 2 17.5 14.6 4.3 5.3 12.5 13.2 5.8 7.4 16.5 11.0 7.3 8.3 3 38.8 34.5 26.6 30.8 33.9 34.5 42.6 36.9 33.2 48.0 39.7 28.5 4 17.7 19.3 13.4 12.1 32.3 33.2 28.1 25.4 34.4 27.8 28.8 29.5 5 25.8 31.1 48.1 40.3 13.2 18.5 20.8 28.7 11.5 12.6 22.9 33.5 6 - 0.3 0.3 - - 0.5 - 0.4 - - - - n.a. - - 6.2 11.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Data source: DANE Industrial Directory Files. aEstablishments with 10 or more employees. bSIC code. agglomeration economies, the location patterns of new firms will follow those of the parent industries. The degree of employment concentration varies across industries in Bogota. Of the 29 three-digit industries, about 80% of the total employ- ment was concentrated in the largest 12 industries.'6 As shown in Table 8, all industries except for the plastic industry had their highest concentration of employment in Ring 3 during the 1970-1975 period. as was the case for the total manufacturing. In six of these industries, the employment share in Ring 3 was higher than the industry average, but in most areas its decline during the period resulted in a greater dispersion of employment. In all twelve industries, there was a substantial degree of decentralization of employment. 16Each of these 12 industries had at least 3% of the total employment. 234 KYU SIK LEE The durable industries shown in Table 8 maintained a large share of employment in Ring 4 as well as in Ring 3 during the period. In the case of the plastic industry, the employment share was highest in Ring 5. It should also be noted that, as in large U.S. cities, employment in the printing industry in Bogota is also concentrated in the CBD. Although there w.as a shift of employment in this industry to Ring 3, 24% of the printing industry's jobs were located in Ring 1 in 1975. The above description is with respect to the distance from the city center. In order to measure the extent of industrial concentration around a centroid of industries, we have computed the "standard distance," 17 using the following statistic: XE,(x-Ej)2 + (yi dj= Xi ji2 E.Eij where E Eijxl Eijyi - i - I i i Eij= the number of employment in the jth industry in the ith area; (x,,y,) = the center of the ith area in terms of x,y coordinates; and (xj,Tj) = the location of centroid of thejth industry. This formula measures a standardized distance, weighted by the number of jobs (or establishments), from subareas where jobs are located to the centroid of jobs in a particular industry. The value of standard distance will show the degree of concentration relative to the industry's centroid. The standard distances are estimated for all manufacturing establish- ments for the six years from 1970 through 1975. Table 9 shows that the standard distances increased over time as average firm size increased, indicating the general tendency of greater industrial dispersion. We also find that the standard distances of small firms were consistently shorter than those of large finns. The differences in the orders of magnitudes, however, are too small to suggest any definite tendency for small firms to cluster in the presence of agglomeration economies."8 Although we also computed estimates of standard distances at the three-digit level, they showed no consistent differences between small and large firms. Moreover, 17Isard (1960) had attributed its origin to Bachi (1957). (The expression of this measure appearing in Isard (1960), however, has an error.) The form of the squared standard distance is actually the same as that of the moment of inertia. "8Without having internal economies of scale, small firms need to share facilities and service inputs among themselves; therefore, they are expected to cluster together or locate near the center of the parent industry. EMPLOYMENT LOCATION IN COLOMBIA 235 TABLE 9 Standard Distances of Establishmentsa by Firm Size, Bogota, 1970-1975 Standard distance (kln) Average Proportion of Less than 25 or more firm sizeb small firms' 25 persons persons All firmsb (persons) (So) 1970 3.54 3.69 3.66 55.1 49.0 1971 3.65 3.70 3.72 57.3 47.3 1972 3.54 3.85 3.76 62.1 44.4 1973 3.83 4.02 3.97 60.8 45.4 1974 3.94 4.02 4.02 63.6 43.7 1975 4.01 4.04 4.06 64.7 42.7 Data source: DANE Industrial Directory Files. al0 or more employees. bAll firms with 10 or more employees. CLess than 25 category. the values of standard distances for small firms tended to increase during the five-year period. This again confirms the general tendency for in- dustries to disperse over time. Table 10 reports the values of standard distances for the 12 leading industries in Bogota. First, we observe that the values of standard dis- tances increased between 1970 and 1975 for all industries except textile, plastic, and electric machinery. These results support the general trend of increasing industrial dispersion, while having revealed some industries with increasing concentration.'9 In Table 10, the location patterns of mature firms are compared with those of new (births) and defunct (deaths) firms. The standard distances of mature firms do not vary much across industries and their values are smaller than those of births for all industries reported except beverages and plastic. This implies that new firms locate differently than mature firms,20 i.e., they tend to locate farther from the centroid of a given industry. This observation holds for the births of small establishments as well as large ones in most industries. It is possible that new firms are unable to compete against existing establishments for the limited space in the centralized location, or that additional space may not be available. Moreover, small new firms gain the advantage of agglomeration economies near the CBD, i.e., in the "incubation" area which does not coincide with the centroids of industries. Our findings in the previous section support 19Among those iDdustries not reported in Table 10, the value of standard distance declined also for leather, industrial chemical, and "other manufacturing" industries. 20Similar results were obtained by Cameron (1973) for the Clydeside conurbation in Scotland, but his measure was a simple average distance from the core of the central city. 236 KYU SIK LEE TABLE 10 Standard Distances of Establishmentsa by Firm Type for Selected Industries, Bogota, 1970-1975 (km) Births All SIC Industry Mature Small Large Allb Deaths 1970 1975 Non- durable 311 Food 3.04 4.78 5,66 5.18 3.97 3.22 3.97 313 Beverages 2.83 2.53 - 2.53 0.86 3.25 3.32 321 Textile 3.88 4.28 3.60 4.26 4.74 4.09 3.96 322 Apparel 2.96 3.74 3.74 3.81 3.89 3.31 3.58 352 Other chemical 3.36 4.91 5.78 5.37 4.73 3.33 4.25 Durable 332 Furniture 3.79 4.08 3.25 3.93 4.12 4.02 4.29 342 Printing 2.94 2.67 5.80 4.19 2.00 2.74 3.50 356 Plastic 3.48 2.89 3.63 3.25 5.97 3.98 3.01 381 Fabricated Metal 3.54 3.66 5.61 4.41 2.41 3.29 3.72 382 Non-electric Machinery 2.65 3.83 3.35 3.66 3.19 2.85 3.36 383 Electrical Machinery 2.67 2.62 2.54 2.65 3.55 2.94 2.83 384 Transportation Equipment 3.26 3.94 4.94 4.35 2.35 3.10 4.16 Data source: DANE Industrial Directory Files. aEstablishments with 10 or more employees. bSmall establishments have less than 25 employees. this observation: Table 6 showed that the incidence of births relative to base year employment share was low in the centralized industrial districts (Rings 3 and 4 in the industrial corridor) where the concentration of manufacturing employment was highest. The plastic, the beverages, and the electric machinery industries are the only cases where the standard dit,ance of births was shorter than that of mature firms. Of the industries not reported in Table 10, this was also true for the leather, paper, instruments, and "other manufacturing" industries.21 This can be interpreted as evidence for the presence of agglomeration economies in the sense that new firms tend to follow the location patterns 2tThese four industries together, however, had only 6% of total manufacturing employment in 1975. EMPLOYMENT LOCATION IN COLOMBIA 237 of the stock of mature firms.22 It is striking that the value of the standard distance of deaths was the largest for the plastic industry; this may indicate that the incidence of deaths was high for those firms that were located far away from the industry's cluster of firms. As mentioned earlier, the standard distance of the textile industry decreased over time as did those of the plastic and the electric machinery industries. The standard distance of all firms in an industry will decrease whenever the standard distance of births is shorter than that of the stock of mature firms; i.e., the degree of concentration increases. It is then apparent that we observe the prevalence of agglomeration economies in the textile industry as well as the seven industries mentioned above, namely, the plastic products, beverages, electric machinery, leather, paper, instruments, and "other manufacturing" industries. These eight industries together, however, accounted for only 32% of the total manufacturing employment in 1975. The standard distance as a measure of the degree of concentration does not show the possibility of multi-centers of employment. In order to discriminate such a possibility from the single-centeredness of employ- ment, an "index of contiguity" has been constructed as follows. First, a measure of proximity (P) of the jth industry's employment to the sth subarea is defined in terms of employment (E) and distance (d) between subareas s and t: Pj, Ej j2 j= 1, 2,.. 12; s t = 1, 2,... 38, d#s where 38 comunas are used as subareas. Thus, the value of Pj, will be high when a large amount of the jth industry's employment is located in subareas near the sth area. The contiguity index for the jth industry (Cj) is then defined as the correlation coefficient between Pis and Ej, across all 38 subareas, namely, Cj = C(Pjs, Ejs), where O < Cj