Household Income, Travel Behavior, Location and Accessibility: Sketches From Two Different Developing Contexts

This paper analyzes the differences in travel behavior and location characteristics across different income groups in two cities in very different parts of the world – Chile and China. Using recent household travel surveys, we compare vehicle ownership rates, mode choices, trip rates and purposes, and travel times and distances according to high, middle, and low income terciles in Chengdu and Santiago. We also compare household location characteristics and present different measures of accessibility. The results suggest commonalities and differences and build a foundation upon which future, more detailed analytical models can be developed and more rigorous and comparable accessibility measures might be derived. INTRODUCTION Much remains to be learned about how income affects household transportation and location choices of urban residents in less developed countries. Several major analytical challenges exist, including the rapid pace of change in land development patterns and transportation characteristics (e.g., vehicle ownership) and the lack of data (because they are not collected, quickly obsolete due to rapidly changing conditions, and/or not easily collected for important segments of the population). Attempts to compare among and generalize about the so-called developing world face further challenges of data comparability (e.g., trip definition in surveys) and the enormous range of cultural, regional, economic and other factors that make for as much variation among developing countries as between the developed and developing countries. Nonetheless, comparative analyses can be useful, to look for regularities in trends, behaviors, and conditions and explore possible reference points on development trajectories. Cameron et al, for example, using aggregate city-wide data propose a generalized model to predict aggregate private motorized distances traveled for cities based on urban area (1). Schafer uses data from about 30 travel surveys in more than 10 countries to support the hypotheses of stable average travel budgets (share of time and income) (2). Hyodo et al compare trip characteristics revealed through Japanese government-sponsored household origin destination studies for 11 different developing country cities, showing the wide ranges in vehicle ownership, trip generation rates by age and gender, mode shares, trip times, etc. (3). Gakenheimer and Zegras examine basic transportation and land use characteristics from a range of cities in Africa, China, South Asia, and Latin America, showing the wide spectrum of motor vehicle ownership levels, presence of motorized two-wheelers, urban densities, non-motorized mode shares, public transportation use, etc. and noting the important apparent role of national motor vehicle industries, common trends towards public transport decline, and near-universal difficulties in land development management (4). In this paper, we make an initial comparison of two cities from two very different parts of the developing world: China and Chile. At first glance, these two countries offer somewhat distant extremes: China, the world’s most populous country, with an Eastern-Confucian tradition and existing cities whose foundations date to the pre-Christian era; Chile, a country of 16 million, largely dominated by Western culture and traditions, with most of the existing urban areas dating essentially to the arrival of the Spanish conquistadores in the late 1500s. China, despite a huge urban population (in 2000 there were about 37 cities with over 1 million persons), still has, officially, a large share (over 60%) of its citizens living in rural areas. Chile, on the other hand, despite being highly urbanized (nearly 87% of the nation living in urban areas), has just one city with over 1 million persons. Despite these differences, we can see a few common characteristics in the two countries. China, since the early 1980s economic reforms, has exploded onto the global economy, recording average annual growth rates in per capita GDP of almost 11% since 1984 (all figures in this paper, unless otherwise noted, are converted to purchasing power parity (PPP) using implied PPP currency conversion rates (5)). Chile, the first Latin American country to embrace the neo-liberal economic model, has also sustained fairly high economic growth rates over the past two decades: with an average annual growth rate in per capita GDP of approximately 6.5% (5). As of 2005, Chile ranked 37 among nations on the UN’s Human Development Index; China ranked 85 (6). In the face of these contrasts, we make an initial exploration into income group variations in travel behavior, household location, and accessibility in Chengdu, the capital of Sichuan Province, and Zegras and Srinivasan 2 Santiago, Chile’s capital city. We set out to answer some basic questions, hoping to shed some light on relative similarities and differences in these, on the surface, very different places. Do the poorer income groups have very different travel burdens in the two cities? Do common mode choices and trip purpose shares emerge by income groups? How does relative location influence accessibility? Do the two cities have more in common than first glimpse might suggest and, if so, what might this imply for policies, planning approaches, and planning technologies? We aim to answer these questions in broad-brush strokes, laying the foundation for future, more detailed comparative analyses. The Contexts of Chengdu, China and Santiago, Chile China and Chengdu China finds itself in a state of still initial, yet rapid urbanization and urban expansion, with strong peripheral growth and concurrent intensification of core employment centers (7). Despite liberalization of the land market, the state maintains a powerful role, particularly as municipalities now have the abilities and fiscal incentives to transform land to urban uses. Many cities have embarked on large-scale urban renovation projects. Investments in transportation infrastructures have been astounding; Shanghai, for example, made over US$10 billion in transportation infrastructure investments during the five-year period 1991-1996, including in bridges, tunnels, an inner ring road and a subway line (8). Gakenheimer and Yang (9) identify three intrinsically-linked growth forces influencing Chinese cities – national urbanization, increased income, motorization – which filter through four regulatory forces – land use reforms, housing commoditization, municipal finance restructuring, and changes in land development standards. According to their analysis, the resulting patterns of metropolitan development across the nation include CBD-focused, polycentric, “garden city”-type suburbs, and completely “new” cities. Generalizing the influence of these development patterns on travel behavior in the “Chinese city” is difficult. Some downtown “gentrification” is taking place, as in the replacement of Beijing’s hutongs with high-rise apartments. High income households are also tending towards the suburban fringe, in luxury villa-type housing estates, with private motor vehicle access clearly playing an important role. For lower income groups, residential relocation with urban restructuring – and the concomitant disintegration of the communist-economy “work unit” – is also a fact of life, with important travel consequences. For example, an analysis of residential relocation in Beijing and its effects on commuting time, suggests that “reluctant” re-locators (more likely to be lower income groups forced to move due to urban “renewal”) suffer increases in commute travel time, on average 11 minutes longer than those who choose to relocate (10). Across Chinese cities, there is an apparent decline in public transport share of passenger trips, as the total passengers carried from 1993 to 1997 remained constant or declined, despite an almost doubling in public transit vehicles (11). Between 1980 and 2000, average nationwide motor vehicle growth was about 11.3% a year, with the motorcycle fleet increasing at nearly three times that rate (12). Chengdu, the capital of Sichuan province, was built in 316 BC and has a 2,300-year continuous history. The broader municipality of Chengdu has a land area of 12,300 km with 10.13 million people (in 2000), accounting for 11.8% of Sichuan province’s population (13). The urban population was about 3.36 million. Population densities within the first ring road exceed 360 persons per hectare and within the second ring road are about 130 persons per hectare. The city is typical of the so-called “2 tier” cities in China and unusual only in that it is in the relatively poor south western part of the country making it likely to attract, along with the other major city in the province (Chongqing), large numbers of rural immigrants. The average annual growth rate in GDP was 14.6% in the 1980s and 21.0% from 1990 to the present (13). Steady growth in formal population has accompanied economic growth, with an average 1.08% annual growth rate since 1980 (13); the PPP-adjusted GDP of Chengdu is approximately US$35 billion, or US$11,000 per person, with continued strong growth expected (14). Chile and Santiago Chile, already a highly urbanized country, no longer has the same rural-to-urban migration forces as China does today. Like China, however, Chile’s urban areas are experiencing the major forces that derive Zegras and Srinivasan 3 from income growth – e.g., motorization, changing residential preferences – and subsequent urban expansion. Investments in transportation infrastructure have also been massive in recent years; in the Santiago region alone over 200 kilometers of private sector concessioned highways are under construction or in advanced planning stages, signifying investments of nearly $2 billion (15); another nearly $1.5 billion in public investments has been put into expanding the Metro system (urban heavy rail) over the past few years. Relevant land development drivers, in Santiago and elsewhere in Chile, include continuous demand for low income housing, large-scale real estate mega-projects, and government finance interventions including through some urban revitalization subsidies (see (16)). Santiago, located in the fertile valley of the Mapocho and Maipu rivers, was founded by the Spanish conquistador, Pedro de Valdivia, in 1541. Greater Santiago, the contiguous metropolitan area, covers approximately 800-900 square kilometers; with a gross population density of roughly 65 persons per hectare and a net (of roadways, open space) density of 85 persons per hectare. The city is in the region (Chile is divided into 13 administrative regions) known as Region Metropolitana (RM), and Greater Santiago accounts for nearly 95% of the RM’s population. With roughly 5.5 million persons today, Greater Santiago accounts for one-third the nation’s population and is almost 7 times larger than Chile’s next largest city. Over the period 1991-2001, population grew by approximately 1.7% per year, average household income by approximately 6.5% per year, and the auto fleet by 6% per year; by 2001 Greater Santiago had nearly 800,000 private motor vehicles (17). The PPP-adjusted GDP of the RM in 2004 was approximately $74 billion (18), implying a per capita GDP on the order of $11,400. The Two Cities in Brief Comparing the two cities, we can see that Chengdu is a smaller city, with much higher city-wide densities, although in the latter case the figures are affected by how the areal boundaries are defined. Based on the regional GDP figures, as converted to PPP, the two cities display roughly comparable per capita GDP levels. This is a surprising result, as we expected Chengdu to have lower levels – the comparisons may be influenced by different national accounting procedures, inaccuracies in the PPP conversion rates, among other factors. Chengdu, much like the rest of China, continues to post strong economic growth figures; Chile’s and Santiago’s will likely be lower, though still strong and sustained. Given continuous rural-urban migration in China (whether formally allowed by the government or not), Chengdu may quickly reach the size of Santiago; what can we learn and what might these cities’ learn from a comparative analysis of transportation, socioeconomic, and land use characteristics? INDICATORS OF TRAVEL BEHAVIOR This section reviews basic indicators of travel behavior across the two cities, using information in common from the two available datasets. As we are especially interested in showing the differences in travel outcomes by income categories, we categorized the households from the two cities into income terciles, producing the following (again, in PPP): • Lowest: Chengdu, below US$8,100; Santiago, below US$10,000. • Middle: Chengdu, between US$8,101-US$13,500; Santiago, between US$10,001-US$20,000. • High: Above the Middle tercile cut-off; for Chengdu, the highest recorded income in the survey was over $US $101,000); in Santiago, highest income reported income was US$633,333. By these figures, we can see that Santiago indeed seems to be a higher income city than Chengdu; the differences get wider with increasing income terciles (e.g., Santiago’s broader middle tercile), suggesting higher income inequalities in Santiago. Other data support this conclusion: Chile has a national Gini coefficient of 0.57 (6); the estimate for Santiago (calculated from the OD survey), 0.51, indicates slightly less income inequality in the city. In China, the national Gini coefficient is 0.45, while the urban Gini is 0.35 (19). For comparison, Denmark, with one of the lowest levels of income inequality among the developed countries has a Gini-coefficient of 0.25; the value for the US 0.41 (6). Zegras and Srinivasan 4 Data Conventional travel demand models that account for land use-transport interactions require travel survey and land use data, which are rarely available to the local planning agencies in China. In this context, the China Project at Harvard University, in collaboration with the Research Center for Contemporary China (RCCC) at Beijing University collected travel behavior and location characteristics data for 1001 households in Chengdu. The survey used a spatial sampling technique that overcomes the inability of traditional, household list-based area samples to reach rural migrants to the city (the so-called floating population). A spatial grid was created in the sample space, to create units small enough to be enumerated quickly and cheaply using GPS receivers that can identify small Primary Sampling Units (PSU) with considerable precision. Surveyors then enumerate the households residing within the boundaries of the PSU. Once listed, each household is interviewed. The dataset included a daily activity survey of an adult in each household. The daily activity survey included 2290 trips, which recorded travel time, cost, modes used, and alternative modes available to the trip maker. Location surveys were also conducted for each of the 1001 households and survey respondents recorded travel times to regional locations like the city center, railroad stations, shopping malls, as well as local amenities like parks, markets, playgrounds, banks and post offices. The surveyor then verified the location characteristics. Although a Japanese government (JICA)-sponsored travel survey for Chengdu was carried out in 2000 (3), those data have not been made available for public use. The JICA survey did not include the floating population. Santiago has a good tradition of quality transportation data gathering and modeling, with household origin destination surveys carried out in 1977, 1991, and 2001. The 1991 survey included over 31,000 households (3% of the city’s households at the time) interviewed during the work week of the “normal” (i.e., not summer) season. The 2001 survey demonstrates advances in the state-of-the-art. While a smaller sample was used, all days of the week were included (including weekends) as was the summer season. 12,000 households were surveyed during the “normal season” and 3,000 during the summer time (in total, 1% of Greater Santiago’s households), including 59,763 persons and 153,413 trips. The 2001 survey included all trips in the public space (irrespective of distance), by all household members (regardless of age), 13 trip purposes, and 28 different modes or mode combination. Trip origins and destinations were geo-coded at nearest street corner (sometimes census block). Socio-demographic data include individual educational level, job status, household income, number of motor vehicles, etc. Household information is geo-coded at the census block centroid (nearly 50,000 blocks) (for more detail, see (20)). In recent years Chilean authorities and academics have invested considerable effort to develop a land use model that can be integrated with local travel modeling capabilities (see 21). To “feed” the model, authorities initiated an effort to compile land use data from national tax records and business and land use permits (as reported to Municipal governments); the data for 2001 include information (e.g., type of use, floor space constructed, parcel size) for roughly 1.3 million residential units and 400,000 nonresidential land uses (17 different land uses, total), geo-coded at the street address or sometimes the census block. Data were not available for rapidly suburbanizing parts of the city, which were however, included in the origin-destination survey. To make the comparisons between the two cities consistent, the Santiago trip characteristics presented below include trips made during a normal work week by persons over 15 years old (reducing the sample size to 9,038 households; 26,484 persons; 76,465 trips). For the Santiago data, the averages presented below use the factors prepared for government authorities to correct for sample bias, adjusting the survey to reflect socioeconomic and demographic information from the 2002 Census. Such correction factors were not available for the Chengdu survey because the Census housing lists exclude migrants. In 2000, the population in Chengdu over 15 included 8.6% students, 68.0% employed and 23.4% others (16) in 2001 comparable figures for Santiago are 17% students, 50% employed, and 34% others. The HarvardRCCC sample indicated that only about 2% of the trips were by those over the age of 65; in Santiago 5% of all trips were by 65+. Zegras and Srinivasan 5 Vehicle Ownership and Mode Choice Unsurprisingly, higher income households have higher car ownership rates in both cities (Table 1). Generally, Santiago displays much higher motorization rates, as reflected in the share of households in each income category that has at least one automobile. Yet, the data reveal some interesting insights. One, when motorized two wheelers are added to the household motorization rate, for low and middle income households in Chengdu the level of motor vehicle availability in the two cities becomes more comparable (0.16, low income and 0.25 for middle income, compared to 0.18 and 0.40 in Santiago) – consistent with other analyses showing that motorized two-wheelers in Asia “level” the motorization rates relative to the non-two wheeler developing country “cultures” (4). Even with this adjustment, the highest income tercile in Chengdu has fairly low vehicle ownership rates – still below that for Santiago’s middle tercile. This difference may result from higher household incomes in Santiago’s middle and high income terciles. We see another interesting result when comparing motorization rates for households with at least one driver’s license: households across the income categories display almost identical motorization rates. A surprising result is that all income categories in Santiago had higher bicycle ownership rates. In fact, in Santiago household bicycle ownership increases with income while in Chengdu it declines – in Santiago bicycles appear to be a “superior good” while in Chengdu they are an “inferior good.” This likely reflects more recreational bicycle ownership in Santiago, which is confirmed by the mode share data discussed below. Motorcycles and motorized twoand three-wheelers show very low household ownership rates in Santiago compared to Chengdu, even though in Chengdu motorcycle use is banned within the inner ring (electric bicycles and tricycles are exempt from this ban). TABLE 1 Household rates of vehicle ownership by income