Solving the Income Data Puzzle

The problem with Income Data in India
Income data in India has always been a contentious issue. There is a lot of intuitive discomfort that we have with the numbers, especially when you have to explain them to someone from overseas who is evaluating the potential of the Indian market with a view to investing in it. "How can any one who earns so little, afford to buy so many things, and still manage the living expenses of a family of five?", they ask, puzzled! We can definitely vouch for the fact that the income data is generated by reputed, world class organizations, using rigorously designed, huge sample size surveys that would satisfy any survey data excellence standard, anywhere in the world. So, there is no "survey science" flaw on which to hang our discomfort with the data.

Obviously something isn't adding up. For example, consider the NCAER data (2001-02, the latest they have available). They describe a lower middle tier of consumers that they call the climbers, earning between Rs.2-5 lakhs a year, 70% of whom have basic durables like TVs and refrigerators, a little less than one third of whom have entry level cars and 13% have ACs. Take their next category - the upper middle class that they call the strivers. The survey data shows them as having a annual income between Rs.5 -10 lakhs a year and on that income, one in two have cars, and others luxuries. Since the number of rich households earning Rs.10 lakhs or more a year is a mere 8 lakh households in 2001-02 (and about 11 lakh households now), they alone cannot be contributing to all the consumption increase that we are seeing from the supply side. And we are not that much of a credit driven society in any case. So the consumption data must be correct.

Consumption data, is like maternity. A certainty. Income data is like paternity. A matter of opinion. Various people have changed survey income data to suit their logic comfort levels. For example, a presentation made by one of the big consulting firms, arbitrarily moved all the income categories upwards by about 40%, claiming that "team analysis" had led them to conclude that this is the extent of understatement of income that people give in surveys, to save themselves from income tax. But these are fact free analyses which are high on conjencture.

But, income data has its pluses...
Interestingly, as we will show later in this article, all survey income data produces more or less the same results of income distribution ie what income percentile has how much of the income. There is very consistent lying (under-reporting) from the people of India when asked the question on income - no matter which agency does the survey. We are forced to conclude that the income data we are getting is what, in statistics-speak, we would call 'reliable' but not 'valid'. 'Not valid' because it does not measure what it is supposed to. But reliable because it repeatedly identically gives you the same result on whatever it is, that it is measuring.

Obviously, the survey income data is some measure of exactly under reported income or expenditure data, which is on an interval or a relative scale, where the distance or difference between Rs.500 000 and Rs.530 000 on the scale, is the same as the distance or difference between Rs.60,000/- and Rs.90,000/-. However purchasing power has to be an absolute number, which can be compared with other such numbers from around the world. So the relative scale of income distribution doesn't really help; except to make comparisons between people or between periods of time.

And GDP per capita data its minuses
And so business leaders, economists, politicians, equity researchers - in short everyone other than marketing folk who think of markets as made of people and not macro statistics - prefer to use the GDP per capita or related number as the real income number. Yes this is a reliable and hopefully a valid number. So Indian GDP per capita in 2003-04 is US $ 550, and we know exactly where this stands relatively to any other country. And with the notion of Purchasing Power Parity, at least intellectually, the concept is clear even if not intuitively or strategically!

The only trouble with per capita GDP or any such macro number is that you cannot identify people (consumers) based on their per capita GDP, and band then together, and then study each band (or per capita GDP consumer segment) in further detail. Therefore we do not know what people in each per capita GDP category own, and how this is changing over time, where they live (in terms of town class and so on). Most of all, there is a magic number that people use of GDP per capita above which, they say consumption will 'take' off. This number ranges from US $1200 to US $2000. And is used often to determine the size of the consuming class in India. However we do not know whether this magic number, no doubt empirically validated from other economies around the world, makes sense for this market because a market's potential in terms of how many can afford to consume depends on (a) income levels and (b) cost of goods. We have seen 2 wheelers and telecom take off at well below the magic number because price thresholds were discontinuously lowered but performance maintained.

Therefore we decided to get together as a team and look at all available income constructs and see how they relate to each other. The purpose of doing this is not to arrive at a single measure of affluence for all to use; but to enable a more informed choice. Further, recognizing that given the limitations of each, multiple measures will need to be used, the endeavour is to be able to establish some consonance between GDP per capita and survey data on income or expenditure.