Anindex to Measure Socio Economic Status : Asset Approach '

Objective: To develop an index to measure socio economic status using information on household assets. Methodology: A set of standardized asset item weights, which would befinally summed up to generate an asset index for each household was computed. ‘These standardized asset item weights were calculated by multiplying the factor score coefficients corresponding to each asset item by the standadized values (Z scores) ofthose asset items measured in the binary scale. Household asset data collected in the DHS survey 2000 was used in the calculations. Results: The standardized asset item weights corresponding to different asset items, quintile cut off points of the asset index that could be used to measure socio economic status were presented. Conclusion: The asset index could be viewed as a valid, reliable, and more feasible method of measuring socio economic status using only five simple questions.


Introduction
Socio economic status. is an important background variable in health research. The traditional approach for measuring socio economicstatus, such as measuring income and expenditure patterns needs a high degree of expertise, and complex survey instruments. Further, the declared information used in traditional approach is frequentlycriticised for accuracy and reliability. A relatively new methodology, the asset approach, which considers the assets (possessions) of a household as proxies of it's socio economic status is considered to be simple and free from above problems. Household assets are shown to relate quite closely. with income and expenditure. Assessing these variables is much easier, quicker, more objective and reliable than assessing income or expenditure'.
Theobjective of this article is to describe the procedure of adapting the asset approach and identifying the standardized asset item weights 'Community physician, Family Health Bureau, Colombo.
Journal ofthe College ofCommunity Physicians ofSri Lanka to suit the current Sri Lankan context. As will be described in the text, the asset item weighis identified in this study could bedirectly applied on the asset characteristics of households identified in other sample surveys to measure their socio economicstatus.

Methodology
The asset approach uses an asset index (a composite score) to measure the socio economic status, The asset index was computed by summing up a numberofstandardizedassetitem weights which reflect the presence or absence of each asset item in a particular household. This index was considered as a proxy reflecting the socio economic status of that household.
Development of the asset index required following actions. 1) Identifying a representative data base which contains information on the distribution of assets, 2) Determining the factor score coefficients (weights) corresponding to different assets, 3) Calculating standardized asset item scores, 4) Identification of cut off points for the classification of socio economic status.
1) Identifying a representative data base with asset information: Estimation of standardized asset item weights corresponding to different asset variables required asset information from a larger representative sample. The DHS survey 2000 * collected data on the possession oflarge number of asset items in 8765 randomly selected households. This data base was examined for the consistency of data on household possessions and material used for the construction ofthe house. The asset items for which complete information was available were considered for inclusion in developing the asset index. The items considered included; consumer items ranging from radio, televisions, and car, dwelling characteristics such as flooring materials, type of drinking water, toilet facilities etc...Appendix 1).The materials used to build wall, floor and roof were found to be highly correlated with eachother. Hence,ofthese, only floor material which has shown the highest variation across the sample was selected 2) Determining the factor score coefficients corresponding to different assets using principal componentanalysis: Thefirst step in identifying a standardized asset item weights for a variable was to estimate the factor score coefficient corresponding to that Volume 9 , 2004 32 variable, These factor score coefficients were obtained through principal component analysis ofall selected asset variables of the DHS 2000 data base. Theprincipal componentanalysis was conducted as follows. Each variable was coded in binary scale (""1' to the presence and "0 "to the absence). It should be noted that for the questions with multiple responses (he questions referring to type of latrine, type offlooring and sources of drinking water) each response was considered as a separate variable. For example data on typeof latrine was originally collected as water seal, pourflush,pit latrine, bucket latrine, other types and nolatrine. During recoding each response was considered as separate variables with binary out come('I' to the presence and "O "to the absence). For example having water seal latrine or not, havingpitlatrine or not, etc... SPSS 11 software wasusedto identify the factor score coefficients corresponding to each asset, The factor score coefficient (1" factor) corresponding to a particular variable was considered as the relative weight corresponding to that variable. These factor score coefficients were used in the calculation of standardized assetitem weights.
3) Calculating the standardized asset item weights Eachassetitem wasassigned standardized asset item weights depending on its presence or absence. These standardized asset item weights were calculated by multiplying the factor score coefficient of that variable by the standardized values (Z scores) ofthe variable measured in the binary scale using following formula'. Talue of asset unweighted mean " raw asset variable sset Variable factor score Unweighted standard deviation of \ coefficient asset variable 4) Identification of cut off points and measuring economic status applying asset weights Thepurposeof asset approach is to compile an asset indexthatrepresents socio economic level ofa particular individual. Theasset index is the sum ofall standardized asset item weights for an individual household. This asset index is considered to reflect the socio economic level of the household, Above described procedure was applied on the data obtained from DHS survey 2000 and an asset index was generated for each household included in the survey. Then quintile cut off points that separate the different socio economic strata wereidentified using those asset indices ( Table 3).
Journal ofthe College of Community Physicians ofSri Lanka Results 1)Standardized assetitem weights Table 1 presents the mean, standard deviation, factor score coefficient related to a particular asset variable and the two standardized asset item weights comesponding to the variable depending on its presence or absence.
For example, if a particular household is supplied with electricity the corresponding asset score would be 0.1291 where as if the household is not supplied with electricity the corresponding asset score will be -0.2392 ( Table 1). As described earlier in the case of variables such as water supply andtype of latrine, where more than one response was considered, each response was assigned separate standardized asset item weight. For example, the question inquiring the type of water source could have the answer, "Yes" to one more of 9 responses while for others the answer would be "No". Each of these answers has to be assigned a separate standardized asset item weight. For example, a particular household obtains water from an unprotected well only, then this household would be assigned the weights as described in Table 2. In this manner the household would be assigned the standardized asset. weights correspondingto all the asset items.
2) Quintile cut off points ofthe asset index Table 3 presents the quintile cut off points that separate the differentsocio economicstrata.
As the standardized asset item weights and cut off points were obtained using a nationally representative sample, the same standardized asset item weights and cut offs could be applied on the data obtained from other samples. The asset. indexes that are computed using standardized asset item weights could be used to classify the individuals' in to different socio economic quintiles relative to the national cut off levels.
Validationof the Asset Index Thevalidity of the assetindex as a proxy to the socio economic status was assessed through two approaches. 1) Assessment of the internal coherence * of the asset index by reviewing the degree of separation of the poorest through second, middle and fourth to richest strata in terms of the average ownership of individual assets by households. 2) Evaluation of the constructs; a) the asset index as a proxy to socio economic status should correlate with the factors  known to associate with socio economicstatus, Thefactors, so correlated, included the risk of having a child with low birth weight, the risk of havinga stunted child in a household, and risk ofdying before the age of 1 years (IMR) b) distributionpattern of households from different socio economic quintiles across geographical strata should confirm to the patterns known by other means. If the asset index is internally coherent, the individual items used in computing it should produce a clean separation of the poorest to the richest socio economic levels in terms of the asset ownerships. Stating this fact in a different manner; the percentage of households with a particular asset should showa clear differences or gradients across the socio economicquintiles. Table 4 presents the compares the average asset 'ownership across socio economic quintiles.
Journal ofthe College ofCommunity Physicians ofSri Lanka  Table 4 shows a gradual concentration of different items from one end of the socio economic spectrum to the other. The concentrations seemed to conform to the socio economic standards. For examplein the case of sources of water supply, sources, such as protected well with in the premises or tap line with in the premises seemed to concentrate towards the richest quintile where as the protected wellout side premises and those who take water from rivers tanks, and other sources were concentrated towards the poorest quintile.
2) Association of the asset index with the factors such as law birth weight, underweight, and child mortality The DHSsurvey 2000 data was used to identify the associations of asset indices of individual households with having a low birth weight baby, stunted child and an infant dying before completing one year to a mother in the household. Binary logistic regression procedure was carried out considering these three conditions as dependant variables and socio economic status measured by asset index as the independent variable. Educational level of the mother was also included in the models to control for the potential confounding effects. Regression analysis showed that all these three conditions were significantly associated with the socio-economic status. Figure 1 depicts the distribution ofrisks (Odd ratios) of having above conditions across the socio economic quintiles. Table 5 present the regression coefficients, odd ratios, and significances of these associations.
Risk of experiencing any of these conditions show a declining trend from the poorest to the richest quintile    The distribution shows that the largest proportion of richest households is seen in the urban areas where as the largest proportion of poorest households are seenin the estate sector. This pattern confirmsthe distribution of socio economic standards identified in the other sources suchas socio economic survey *, and the vulnerability map°.
Finally, it can be concluded that the asset approachis a valid, easier and feasible method to measure socio economicstatus using only five simple questions (Appendix 1). However, it should be noted that the standardized asset item weights are relative to the current socio economic status of the country and therefore should be updated with time.