Against the official savings level of 35 per cent, CMIE survey estimates this to be 40 per cent.
The official statistical machinery of India does not make an estimate of the income of households. And there are no plans to make a beginning on this front either. Given that India is among the fastest-growing economies in the world today, the failure of the official statistical machinery to provide reliable information about this important source of growth is appalling.
Though we debate poverty estimates endlessly, we measure poverty by using expenditure data of households, and not income data. Thus, a household is defined as poor if it is found to be short on spending, not earning.
Personal disposable income, as estimated by the official statistical machinery, includes income of non-profit institutions serving households (NPISH) such as temples, mosques, gurdwaras and other similar organisations. There is no estimate of their share, hence no information about the income of households.
The National Council of Applied Economic Research (NCAER) has made occasional efforts to estimate household income. These estimates led to the first quantification of the size of the great Indian middle class. NCAER’s estimates not only sparked the much-needed debate on Indian households, it also ushered in several corporate investment and marketing strategies in the 1990’s.
We have learnt a lot from the pioneering efforts of NCAER and the National Sample Survey Organisation (NSSO) which estimates household expenditures among other things.
At the Centre for Monitoring Indian Economy (CMIE), we felt the need of having an integrated estimate of household income, expenditure, savings and ownership of assets. This integrated effort would automatically address the fear that households do not reveal income data correctly. If households did indeed underestimate income then, in the integrated effort, this would lead to a proportionately low estimate of savings (which can get exaggerated because savings have a low base compared to income). Or, households would have to report low expenses in proportion to the low income responses.
The best way to test the hypothesis that households do not reveal their true income was to actually conduct a large survey, and compare the estimates with official statistics. For example, if the savings rate of the households turns out to be significantly lower than that in the official estimates, there would be a prima facie case to believe that households do not reveal their incomes correctly.
When CMIE launched the survey three years ago, we believed that it was difficult for a household to lie or systematically understate its income. It may do so casually or randomly in the case of an isolated question on income. But, in the face of detailed questions on the sources of income and other financial details of expenses and savings, it would either not respond or provide reasonably accurate information.
We had to make extra efforts to deal with a higher non-response in the richer areas of large cities; but, finally we did get responses and the results negate the myth that income cannot be estimated through a sample survey.
We subjected the results to two validations. First, we compared the estimates with official statistics. For example, we checked the expenditure data with NSSO estimates and the savings rate data with official savings rate. We found that the non-food expenditure in urban areas was higher than what the NSSO had estimated. Rural per capita expenses were estimated to be 2.5 per cent higher and urban per capita expenses were 7.5 per cent higher than the NSSO estimates. We also found that households saved over 40 per cent of their incomes, while the official estimates for the households savings were closer to 35 per cent of their estimated income along with NPISHs.
Since both expenditure and savings were clearly higher than the official estimates, and since these were a part of an integrated survey, it is clear that the households were not underestimating their incomes in the surveys conducted by CMIE. Our scrutiny of the data often worried about whether households were overstating their incomes. Efforts to specifically capture transfer incomes and non-regular incomes helped households provide income more accurately than we could have managed had we not asked these questions separately. As a result, Kerala and Bihar show a much higher income than the official statistics of State Domestic Product (SDP) show.
The second validation works on a different plane. It checks for the consistency of the responses of the households in repeated surveys. CMIE surveys are conducted on a panel of 140,000 households, every quarter. The panel was created through a random selection process initially. And then, the survey was conducted on this panel repeatedly every quarter for the past eight quarters. Results of the first four quarters were trials during which we weeded out the systematically inconsistent or unreliable observations, and replaced them with appropriate alternative households. The inconsistencies were mostly because of execution problems. By the fifth quarter, the panel was robust and machinery was established.
| WHAT WE EARN, AND SPEND | ||||
| Month | (Household data in Rs million) | Savings rate (%)* | ||
| Income | Expenses | Savings | ||
| Jan ’09 | 24,58,840 | 13,60,176 | 1,098,664 | 44.7 |
| Feb ’09 | 21,50,580 | 11,59,771 | 9,90,809 | 46.1 |
| Mar ’09 | 24,33,538 | 12,25,482 | 12,08,057 | 49.6 |
| Apr ’09 | 24,42,402 | 12,64,392 | 11,78,011 | 48.2 |
| May ’09 | 21,82,183 | 13,38,560 | 8,43,623 | 38.7 |
| Jun ’09 | 21,67,970 | 12,87,524 | 8,80,447 | 40.6 |
| Jul ’09 | 21,65,071 | 13,62,200 | 8,02,871 | 37.1 |
| Aug ’09 | 21,34,030 | 12,12,762 | 9,21,268 | 43.2 |
| Sep ’09 | 26,26,904 | 14,65,770 | 11,61,134 | 44.2 |
| Oct ’09 | 23,96,450 | 14,02,644 | 9,93,807 | 41.5 |
| Nov ’09 | 24,44,590 | 12,50,473 | 11,94,116 | 48.8 |
| Dec ’09 | 26,19,471 | 13,44,234 | 12,75,238 | 48.7 |
| Total | 2,82,22,031 | 1,56,73,986 | 1,25,48,045 | 44.5 |
| Source: Consumer Pyramids, CMIE ( * % of household income) | ||||
At the end of eight integrated surveys over a large panel, we have reasonably reliable estimates of income, expenditure, savings and ownership of assets. The survey seeks estimates of monthly income and expenses. As a result, we now have a monthly series of the monthly income, expenses and, therefore, the implicit savings of households of India. Results are presented in the table. Clearly, households are not shy of revealing their incomes and they have no reason to understate their expenses.
A caveat is in place here. Official statistics make estimates of rent and include this in the expenses, although most houses in India make no expenses on rent. Almost all of rural India and most of urban India live in their own houses. The unnecessary sophistry of imputing rent as an expenditure also confounds the computation of the consumer price index. More on this some other time. For now, we should celebrate the high savings of Indians, and seek ways to deploy them productively for future growth and employment.
Consumer Pyramids (the service based on the survey) also confirms the bulge in the population pyramid in the age group of 15-20. In five years, this population will hit the labour markets. Now, juxtapose this with the capex boom that I have described in several earlier columns in this space and you will see the promise that India holds for the future. Labour is in the pipeline and companies are creating new capacities that will employ them. Incomes are bound to rise and Indians are not shy of flaunting it anymore.
The author is managing director and chief executive officer, Centre for Monitoring Indian Economy
