This article critiques Arjun Jayadev et al's 'Patterns of Wealth Disparities in India during the Liberalisation Era', detailing problems with the data used as also the means of deflation of the stock of physical and financial assets held by households.
DISCUSSIONnovember 24, 2007 Economic & Political Weekly78Inconsistencies in Patterns of Wealth DisparitiesK G K Subba RaoIn‘Patterns of Wealth Disparities in India during the Liberalisation Era’ [EPW, September 22, 2007] by Arjun Jayadev et al have taken special efforts to analyse the results of the all-India debt and investment surveys (AIDIS), based on micro-level survey data. The data on household asset/liabilities are converted per capita by dividing the respective values by the size of the households and using household level survey weights. While this itself will affect the survey data to some extent, it is further con-founded bydeflating the 2002 data by price deflators not applicable to the stock of physical and financial assets held by households.A perusal of the results re-veals some inconsistencies whencompared totheresults published by the National Sample Survey Organisation (NSSO) in their reports. Discrepancies in DataWith reference to Tables 1 and 2 of the paper, it is not clear what the aggregates mentioned against the bottom row indicate. Presum-ably these are asset values. The data indicates that the ratio of ur-ban to rural assets is less than unity, which is contrary to the indicators pre-sented in Table 1 as well as the data pre-sented by NSSO reports, according to which the urban-rural asset ratios were 1.3 and 1.6 for 1991 and 2002, respectively.The asset structure is given in Table 2. It may be seen from this table that the struc-ture of assets shown in the article of the authors is misleading, when compared to that published in the NSSO reports. Table 3 (p 79) in the article gives the percentage of households owning each of the assets. Presumably these are propor-tions reporting assets. These are different from those published in the NSSO reports. There are similar discrepancies in the statewise data presented in other tables, but these are not detailed here. They need to be examined with reference to the data published in the reports, in view of some of the discrepancies mentioned above.Estimates of Net Worth and DebtThe authors work out a measure of net worth, by deducting debt from the value of total assets. As all the assets and liabili-ties are not recorded in the survey, the picture that emerges is one of a truncated balance sheet. In particular, inventories and stocks held by the households are excluded from the purview of the survey, in view of the difficulty in getting these details from the respondents. The assets arealso evaluated at market prices, dis-tinct from the cost of acquisition normally followed in the parlance of a balance sheet. This may give rise to inflated values of the assets, particularly assets such as land and buildings, which account for a major share of the total assets. The method of data collection in the survey is also in-direct. Illustratively, the value of assets/debt is recorded by taking the stock data as on the date of survey and adjusting for transactions that have taken place be-tween July 1, 2002 and the date of the survey. These are inherent in a sample survey,when the required information is solicited from the respondents through in-terviews. It has been the experience of earlierAIDIS surveys that debt is under-re-ported by respondents. The underesti-mates of debt is a resultant of not only the number of households reporting debt but also the amounts reported. Comparison of the data on debt revealed by the surveys with that available from lending agencies indicated substantial differences in this regard. Even so, the This article critiques Arjun Jayadev et al’s ‘Patterns of Wealth Disparities in India during the Liberalisation Era’, detailing problems with the data used as also the means of deflation of the stock of physical and financial assets held by households.Table 1: Average Value of Assets(in Rs) RuralUrban1991 2,00,179 (1,07,007) 1,01,479 (1,44,330) 2002 4,56,571 (2,65,606) 2,52,720 (4,17,158) Source: Figures from Jayadev et al (2007). Those in brackets are taken from the NSSO reports 419 and 500 for the respective years. K G K Subba Rao (kgksubbarao@gmail.com) was formerly adviser, Department of Statistical Analysis and Computer Services (DESACS), Reserve Bank of India, Mumbai.Table 2: Asset Structure (in %) 1999 2002 RuralUrbanRuralUrbanLand 49.4 (64.2) 24.7 (35.5) 48.6 (63.2) 26.9(38.5)Buildings 29.5 (21.4) 30.7 (39.3) 32.7 (23.5) 33.2(37.8)Durables 10.4 (5.9) 29.9 (11.6) 9.8 (5.1) 24.2 (8.4)Source:Figures from Jayadev et al (2007). Those in brackets are fromthe NSSO reports indicated above.
DISCUSSIONEconomic & Political Weekly november 24, 200779negative values referred to in the article make it an interesting read. When debt ex-ceeds the value of total assets for a house-hold, it is an indication of a heavy debt burden beyond the means of assets. This will be more so in case of households in the lower strata of assets. As per the AIDIS results, the debt-asset ratio is the highest in the lower asset groups in the range of 20-30 per cent for both rural and urban households, and it tapered down with the increase in the size of asset holdings. It would be more informative if estimates of households reporting such debt beyond their assets, together with the per cent of such debt in each of the lower decile groups, is worked out using the micro level data and building up the relevant estimates.This disaggregated picture will reflect the dead weight indebtedness, which needs atten-tion from the policy point of view. Another point referred to in the paper is that items like gold are not recorded in the survey. Needless to mention that it is ex-tremely difficult to get such sensitive infor-mation from the respondents in a house-hold enquiry. Even in the national accounts data, gold is not treated as a financial asset of the household sector. It is also men-tioned that land is underreported. It is extremely difficult to get a true picture in a household survey. Even so, thedistributions or the structural pat-terns obtained from survey results will provide a rough idea of the status of as-sets and liabilities of the households. In-cidentally,it may be mentioned that the NSSO report 491 on household ownership holdings in India gives a distribution of households by size of land possessed, apart from various other details, as part of the AIDIS 2002-03.Double DeflationIn the study under reference, the assets/liabilities for the year 2002 are deflated by consumption-based indicators, viz, the consumer price index(CPI) for agricul-tural workers in case of rural households andCPI for industrial workers for urban households. The use of these indices for the deflation of the stock of physical and financial assets is questionable. In particu-lar, physical assets constitute a predomi-nant share in the stock of assets owned by the households, more so in the case of rural households, where the share of land itself accounted for more than 60 per cent of the total value of assets. Admittedly, there is no deflator, which can be used for bringing the data for the two bench-mark years to a common platform, as the asset items are different and need differ-ent deflators for the purpose. Certainly the above deflators are not appropriate for measuring the growth of various asset items overtheperiod. The problem is still con-founded if the same deflators are used at the state level. As such, the growth rates worked out cannot give a true picture. In short, the household survey data sets are contaminated by double deflation used by the authors – deflating the data by the household size and also using the price deflators not connected to the variables under reference. The Gini coefficients of concentration, as given in the study are presented in Table3. It may be observed that the coefficients are almost invariant for rural and urban households over the two benchmark years. Thus, there is no change in the inequali-ties of asset holdings though the concen-tration of asset holdings is more in the higher strata of households. Suffice it to say that there is no reduction in inequalities for rural and urban house-holds in the period under reference. One argument levelled by the authors is that the concentration of wealth is understated, because the sample design does not over-sample the rich. Does it mean that the households in the higher strata will have to be taken with certainty? If details of the sample design are recapitulated, sev-enstrata are formed in each selected vil-lage/urban block on thetwincriteriaof indebtedness and land possessed in case of rural households and indebtedness and monthly per capita expenditure in case of urban households; twohouseholds are selected from each stratum.The number of households in the top strata will be comparatively less, and the sampling fraction will be relatively high in such cases, compared to that in the other strata.Table 3: Gini Coefficient in Urban and Rural Areas 19912002 Rural 0.61 0.61Urban 0.700.69Source: Jayadev et al.
DISCUSSIONnovember 24, 2007 Economic & Political Weekly78Inconsistencies in Patterns of Wealth DisparitiesK G K Subba RaoIn‘Patterns of Wealth Disparities in India during the Liberalisation Era’ [EPW, September 22, 2007] by Arjun Jayadev et al have taken special efforts to analyse the results of the all-India debt and investment surveys (AIDIS), based on micro-level survey data. The data on household asset/liabilities are converted per capita by dividing the respective values by the size of the households and using household level survey weights. While this itself will affect the survey data to some extent, it is further con-founded bydeflating the 2002 data by price deflators not applicable to the stock of physical and financial assets held by households.A perusal of the results re-veals some inconsistencies whencompared totheresults published by the National Sample Survey Organisation (NSSO) in their reports. Discrepancies in DataWith reference to Tables 1 and 2 of the paper, it is not clear what the aggregates mentioned against the bottom row indicate. Presum-ably these are asset values. The data indicates that the ratio of ur-ban to rural assets is less than unity, which is contrary to the indicators pre-sented in Table 1 as well as the data pre-sented by NSSO reports, according to which the urban-rural asset ratios were 1.3 and 1.6 for 1991 and 2002, respectively.The asset structure is given in Table 2. It may be seen from this table that the struc-ture of assets shown in the article of the authors is misleading, when compared to that published in the NSSO reports. Table 3 (p 79) in the article gives the percentage of households owning each of the assets. Presumably these are propor-tions reporting assets. These are different from those published in the NSSO reports. There are similar discrepancies in the statewise data presented in other tables, but these are not detailed here. They need to be examined with reference to the data published in the reports, in view of some of the discrepancies mentioned above.Estimates of Net Worth and DebtThe authors work out a measure of net worth, by deducting debt from the value of total assets. As all the assets and liabili-ties are not recorded in the survey, the picture that emerges is one of a truncated balance sheet. In particular, inventories and stocks held by the households are excluded from the purview of the survey, in view of the difficulty in getting these details from the respondents. The assets arealso evaluated at market prices, dis-tinct from the cost of acquisition normally followed in the parlance of a balance sheet. This may give rise to inflated values of the assets, particularly assets such as land and buildings, which account for a major share of the total assets. The method of data collection in the survey is also in-direct. Illustratively, the value of assets/debt is recorded by taking the stock data as on the date of survey and adjusting for transactions that have taken place be-tween July 1, 2002 and the date of the survey. These are inherent in a sample survey,when the required information is solicited from the respondents through in-terviews. It has been the experience of earlierAIDIS surveys that debt is under-re-ported by respondents. The underesti-mates of debt is a resultant of not only the number of households reporting debt but also the amounts reported. Comparison of the data on debt revealed by the surveys with that available from lending agencies indicated substantial differences in this regard. Even so, the This article critiques Arjun Jayadev et al’s ‘Patterns of Wealth Disparities in India during the Liberalisation Era’, detailing problems with the data used as also the means of deflation of the stock of physical and financial assets held by households.Table 1: Average Value of Assets(in Rs) RuralUrban1991 2,00,179 (1,07,007) 1,01,479 (1,44,330) 2002 4,56,571 (2,65,606) 2,52,720 (4,17,158) Source: Figures from Jayadev et al (2007). Those in brackets are taken from the NSSO reports 419 and 500 for the respective years. K G K Subba Rao (kgksubbarao@gmail.com) was formerly adviser, Department of Statistical Analysis and Computer Services (DESACS), Reserve Bank of India, Mumbai.Table 2: Asset Structure (in %) 1999 2002 RuralUrbanRuralUrbanLand 49.4 (64.2) 24.7 (35.5) 48.6 (63.2) 26.9(38.5)Buildings 29.5 (21.4) 30.7 (39.3) 32.7 (23.5) 33.2(37.8)Durables 10.4 (5.9) 29.9 (11.6) 9.8 (5.1) 24.2 (8.4)Source:Figures from Jayadev et al (2007). Those in brackets are fromthe NSSO reports indicated above.
DISCUSSIONEconomic & Political Weekly november 24, 200779negative values referred to in the article make it an interesting read. When debt ex-ceeds the value of total assets for a house-hold, it is an indication of a heavy debt burden beyond the means of assets. This will be more so in case of households in the lower strata of assets. As per the AIDIS results, the debt-asset ratio is the highest in the lower asset groups in the range of 20-30 per cent for both rural and urban households, and it tapered down with the increase in the size of asset holdings. It would be more informative if estimates of households reporting such debt beyond their assets, together with the per cent of such debt in each of the lower decile groups, is worked out using the micro level data and building up the relevant estimates.This disaggregated picture will reflect the dead weight indebtedness, which needs atten-tion from the policy point of view. Another point referred to in the paper is that items like gold are not recorded in the survey. Needless to mention that it is ex-tremely difficult to get such sensitive infor-mation from the respondents in a house-hold enquiry. Even in the national accounts data, gold is not treated as a financial asset of the household sector. It is also men-tioned that land is underreported. It is extremely difficult to get a true picture in a household survey. Even so, thedistributions or the structural pat-terns obtained from survey results will provide a rough idea of the status of as-sets and liabilities of the households. In-cidentally,it may be mentioned that the NSSO report 491 on household ownership holdings in India gives a distribution of households by size of land possessed, apart from various other details, as part of the AIDIS 2002-03.Double DeflationIn the study under reference, the assets/liabilities for the year 2002 are deflated by consumption-based indicators, viz, the consumer price index(CPI) for agricul-tural workers in case of rural households andCPI for industrial workers for urban households. The use of these indices for the deflation of the stock of physical and financial assets is questionable. In particu-lar, physical assets constitute a predomi-nant share in the stock of assets owned by the households, more so in the case of rural households, where the share of land itself accounted for more than 60 per cent of the total value of assets. Admittedly, there is no deflator, which can be used for bringing the data for the two bench-mark years to a common platform, as the asset items are different and need differ-ent deflators for the purpose. Certainly the above deflators are not appropriate for measuring the growth of various asset items overtheperiod. The problem is still con-founded if the same deflators are used at the state level. As such, the growth rates worked out cannot give a true picture. In short, the household survey data sets are contaminated by double deflation used by the authors – deflating the data by the household size and also using the price deflators not connected to the variables under reference. The Gini coefficients of concentration, as given in the study are presented in Table3. It may be observed that the coefficients are almost invariant for rural and urban households over the two benchmark years. Thus, there is no change in the inequali-ties of asset holdings though the concen-tration of asset holdings is more in the higher strata of households. Suffice it to say that there is no reduction in inequalities for rural and urban house-holds in the period under reference. One argument levelled by the authors is that the concentration of wealth is understated, because the sample design does not over-sample the rich. Does it mean that the households in the higher strata will have to be taken with certainty? If details of the sample design are recapitulated, sev-enstrata are formed in each selected vil-lage/urban block on thetwincriteriaof indebtedness and land possessed in case of rural households and indebtedness and monthly per capita expenditure in case of urban households; twohouseholds are selected from each stratum.The number of households in the top strata will be comparatively less, and the sampling fraction will be relatively high in such cases, compared to that in the other strata.Table 3: Gini Coefficient in Urban and Rural Areas 19912002 Rural 0.61 0.61Urban 0.700.69Source: Jayadev et al.
