The first source of data considered for the purposes of this study is the Encuesta Condiciones de Vida (ECV) database. This database stems from a large and comprehensive survey conducted on Ecuador in 1995 that forms part of the World Bank’s Living Standard Measurement Surveys (LSMS) project that started in 1980. The survey was administered to a sample of about 5 800 nationally representative households (this study makes use of a reduced sample of 5 630 households because data were missing for some variables). It collected data on all dimensions of household well-being and socio-economic characteristics including highly disaggregated data on household consumption expenditures. The survey design incorporated both clustering and stratification on the basis of the country’s three main agroclimatic zones and rural-urban breakdown. It also oversampled Ecuador’s two main cities: Quito and Guayaquil.
According to the ECV sampling design, the sample employed in this study is representative of the main agroclimatic zones. The sample size is too small to allow an estimation of the incidence of poverty at the level of provinces, counties and parroquias (municipalities). On the basis of this survey, if traditional mapping methods of spatial distribution of poverty were applied, the only working level should be the main regions in Ecuador (first level). However, by using data from another source, the INFOPLAN database, and aggregating the two databases at the common level of the county, it was possible to map the spatial distribution of poverty at county level.
INFOPLAN is an atlas that collects about 104 variables from the “Census of population and households” (INEC) conducted in Ecuador in 1990. It provides a wide variety of information on the demographic, socio-economic and geographical characteristics of the areas, and the data are available for many geographical area levels (from regions to parroquias). However, it does not contain income or consumption expenditure information for each household. Although the 1995 ECV data were collected five years after the census, the 1990-95 period was one of relatively slow growth and low inflation in Ecuador, so it is reasonable to assume that there was relatively little change.
Furthermore, the household living standards in the available counties (Table 6 in Chapter 5) are not georeferenced: the location of the respondent household was identified only by the county of residence and the type (rural or urban) of living area. This problem was overcome by locating each family randomly (assuming a uniform distribution) in the county but taking into account the type of living area. Moreover, in order to understand the relationship between poverty and environment, the study also considered some environmental variables at the county level, e.g. cereal production, amount of arable land, and the distance of the households from the main roads (data provided by FAO/SDRN GIS).
Finally, all the data from the three sources of information (ECV, INFOPLAN and FAO) were arranged in a GIS for managing the spatial dimension. The FAO and INFOPLAN data were already organized as GIS data and could be overlaid by merging the information contained in the layers. From the ECV data, a point coverage was created taking into account the geographical constraints.
|
Table 1. Estimated percentage of poor people in Ecuador: a comparison between two methods |
|||
| Area |
Method |
||
|
Lanjouw et al. (1994) |
Lanjouw et al. (1995) |
Bigman et al. (1995) |
|
|
Costa |
40* |
54 |
46 |
|
Urban |
35 |
43 |
33 |
|
Rural |
52 |
75 |
75 |
|
Sierra |
42* |
58 |
42 |
|
Urban |
33 |
42 |
17 |
|
Rural |
53 |
78 |
69 |
|
Oriente |
60* |
65 |
69 |
|
Urban |
25 |
47 |
36 |
|
Rural |
67 |
70 |
75 |
|
Ecuador |
41* |
56 |
45 |
|
Urban |
34* |
42 |
26 |
|
Rural |
53* |
76 |
72 |
* Outputs of the computation.
Table 1 reports the outcomes from the comparison between the two methods presented above using the above data. The results for the Lanjouw et al. method were taken directly from the authors’ papers (Hentschel and Lanjouw, 1996; Hentschel et al., 2000), while the outcomes from the Bigman et al. method were computed using ad hoc programs (Annex C) with the 1995 data.
In particular, Table 1 shows:
The percentage of poor individuals increases from 1994 to 1995 under the Lanjouw et al. method.
In some rural areas, the two methods produce approximately analogous results for 1995 data. However, there is a large discrepancy in the percentage of poor in urban areas.
The differences between rural and urban areas increase for the Sierra region.
