Quality and credibility problems of govt data
The leaked Consumer Expenditure Survey (CES) has once again highlighted the importance of statistical data. In this context, I would like to highlight the perennial issues associated with the consumer price index (CPI) data that is used as a policy input by the Reserve Bank of India (RBI) for inflation targeting.
In January 2015, the Central Statistics Office (CSO) revised the base year of CPI to 2012 (from 2010=100). The weighting pattern of the revised series was based on the 2011-12 CES of the National Sample Survey Office (NSSO). The weight for food and beverages in CPI was taken as 45.86 per cent. The RBI’S monetary policy stance is guided by the weights derived from the CES. However, herein lies the problem.
There is significant difference between the share of food and beverages (30 per cent) in the Private Final Consumption Expenditure (PFCE), published by the National Account Statistics (NAS), and the weights derived from the CES. Subsequently, the episodic spikes in CPI inflation, which is largely contributed by rise in food and vegetable prices, makes the RBI decision difficult. Interestingly, if we reduce the weightage of food in CPI to 30 per cent, the headline CPI declines to 3.53 per cent from the current 4.62 per cent.
The NAS and CES estimates have difference in coverage, estimation methods and databases, and, therefore, the two should be different. For example, with respect to coverage, consistency with accounting conventions requires that the PFCE by NAS be subject to a wider coverage than the one drawn from the NSS. The NSS is a directly observed standalone estimate relating to a given survey period (usually a year), while the PFCE from the NAS is an indirect, residual macro-level estimate of aggregate PFCE derived from GDP estimates. The residual nature of the PFCE arises from its use of the commodity-flow method at the disaggregated level of a commodity or service. This method employs an ex-post aggregate commodityflow balance in which economy-wide domestic production is equated to its various uses. Also, since the CES is a household consumption survey and the PFCE relates to both consumers and institutions, the CPI rightly uses the CES data.
The problem, however, is with the CES data itself. For example, the National Horticulture Board (NHB) is the primary source of production and price data for fruit not covered by the Ministry of Agriculture (MOA). The NHB compiles such data from state horticulture board (SHB). It has, however, been noticed in the past that there is a sizeable divergence between the figures the SHBS supply to MOA and those to NHB.
The problems with the NSSO survey, now a part of the CSO, is that it has failed to keep pace with the changing times. One of the common arguments against changing the questionnaire (that contains 500 questions and includes non-food consumption) for understanding the consumption pattern of individuals is that any change in survey questions will impart a structural break in time series, making it hard for researchers to compare results over time.
However, the point that is often missed is this: Though data used in such surveys has opened avenues of research that could not have been pursued otherwise, the quality of research depends on the extent and reliability of the data as well as on the validity of the restrictions upon which the statistical methods have been built. Otherwise, such data hurts the credibility of institutions, the government and even undermines the policy decisions of the RBI.
A comparison of the government data and the latest NSSO survey ( see table) shows that there are pockets where the data is largely comparable and pockets where it is not (say, LPG connections and toilets ). If the sample questions are simple and aligned with the changing times, the survey results can be closer to reality.
Most importantly, the leaked NSSO report shows the inflation adjusted monthly per capita consumption expenditure has declined by 3.7 per cent. During the same period, according to the NAS expenditure data, there was a growth in monthly per capita consumption expenditure of 43 per cent. This apart, today’s official statistics are way more complicated than before. The NSSO needs to change the way it designs survey questions. It needs to understand the shifts in the Indian economy, policy, and those brought about by the use of smartphones and social media. The Ministry of Statistics and Programme Implementation needs to embrace economics and train people in household finance.
We also need to develop several alternative ways of measuring the same thing. It is surprising why big data has still not been used. If the survey is capturing the ground reality, then lot of these alternative sources of data should converge. We should also develop mechanism to do preliminary analysis of data under survey for a proof of concept to check if the results are catching the ground realities.
Finally, there is another twist. The RBI in its 2015 technical advisory committee minutes said: “With the vendors engaged in e-commerce offering low prices, retail inflation may be lower than what the headline number suggests”. India’s top two web retailers Flipkart and Amazon clocked combined sales of ~31,000 crore during the 15-day festival period in October. Our results indicate that only by considering the discount prices the CPI inflation could be 20 bps lower than the actual. Clearly, the time has come to align the Indian statistical system with behavioural economics.