Search interesting materials

Monday, July 23, 2018

Discrepancies in the measurement of household saving

by Radhika Pandey and Renuka Sane.

Data on household financial saving is key to understanding how households save, and the flow of capital from households to firms. In India, households are seen to invest largely in physical assets, causing considerable concern for financial sector policy. This has generated debate on how to improve access to finance and get more households to participate in financial markets.

Accurate and precise measurement is the foundation for any work on financial saving. In the context of data on savings, a number of committees have recommended improvements in the quality of estimates. The issue, however, gets less attention than it deserves. In this article, we study the data on one of the key components of savings: the household financial saving, and highlight some instances of discrepancies between the two data sources on this.

Data on household financial saving in India

There are two sources of data for annual household financial saving in India.

  1. CSO: Data on annual household saving are published by the CSO in its end-January release titled 'First Revised Estimates (FRE) of National Income, Consumption Expenditure, Saving and Capital Formation', and revised in the subsequent annual releases. The Gross Financial Savings of Household Sector (at Current Prices) comprises of the following broad instruments: (a) Currency (b) Deposits (c) Shares and debentures (d) Claims on government - which include all the small savings schemes (e) Insurance funds (f) Provident and pension funds.

  2. RBI: Information on financial assets and liabilities of the household sector are available as part of the Flow of Funds (FoF) Accounts published by the RBI annually. FoF accounts map instrument-wise financial flows across five major institutional sectors of the Indian economy on a `from whom to whom basis'. These institutional sectors comprise (a) financial corporations (b) non-financial corporations (c) general government (d) household sector and (e) the rest of the world. RBI has been publishing the 'Flow of Funds' accounts since 1964. This table is part of the Handbook of Statistics on the Indian economy. The RBI estimates are released five months ahead of the CSO release.

    The data on changes in financial assets/liabilities of the household sector (at current prices) comprises of (a) Currency (b) Bank deposits (c) Non-banking deposits (d) Life insurance fund (e) Provident and pension fund (f) Claims on government (g) Shares and debentures (h) Units of UTI (i) (Net) Trade debt.

The data headings under the CSO and the RBI largely map to each other. The CSO provides us with one heading on deposits, while the RBI breaks it into bank and non-bank deposits. The UTI mutual fund is counted under Shares and debentures in both, while the "Units of UTI" heading in the RBI data pertain to Administrator of the Specified Undertaking of the UTI since 2005-06. Trade debt (net) is shown as part of deposits in the CSO scheme of financial instruments.

In theory, there should be similarities between the two. The CSO has been publishing the new series of national accounts with base year 2011-12 since 2015. In line with this new series of national accounts, the RBI also compiled the FoF accounts starting from 2011-12. In both the data-sets, the economy is divided into the five sectors mentioned above. Despite alignment of the sectors, there are discrepancies in the findings from the RBI FoF data and the CSO data.

Discrepancy in the CSO and RBI estimates

We first present the total household financial saving across the last three years between the two data sources in Table 1. The aggregate gross household financial saving for the year 2016-17 from the CSO is Rs 14,048.47 billion, while the RBI reports it to be Rs 18,204.68 billion. This amounts to a difference of more than Rs 4,000 billion for the year 2016-17. The differences in previous years are lower.

Table 1: Gross financial savings of the household sector (Rs.billion) (Base year 2011-12)
Gross financial savings CSO RBI
2014-15 12,572.47 12,826.33
2015-16 15,207.27 15,142.06
2016-17 14,048.47 18,204.68

We next analyse where the discrepancy for the year 2016-17 is coming from. Table 2 presents the instrument-wise share in total financial saving from the two sources in 2016-17.

Table 2: Share of instrument in financial saving 2016-17 (Base year 2011-12)
Share of instrument (%) RBI CSO
Currency -17.4 -22.5
Bank deposits 60.1 62.2
Non bank deposits 1.9 1.8
Insurance 24.2 24.9
Provident and pension funds 16.2 21.5
Claims on government 4.6 4.5
Shares and debentures 10.0 2.6
Net trade debt 0.2 0.3

The biggest source of discrepancy is seen in the share of shares and debentures in total household financial savings. According to the CSO numbers, their share is a meager 2.6% while the RBI numbers suggest that the share of `shares and debentures' is 10%. There are differences in the share of provident and pension funds. The CSO numbers report the share to be 21.5% while according to the RBI figures, provident and pension funds constitute 16% of the aggregate household financial saving. This is surprising because the CSO document explaining the changes in methodology in the new base year series shows that their key data source for estimating household financial savings in shares and debentures is the RBI.

Discrepancy in estimates depending on base year

A question that is often posed is how the share of particular instruments has been changing across time. This is especially important if the government has taken special policy initiatives to promote a specific saving instrument, and wishes to evaluate the policy impact. One such example is the category of provident and pension funds, wherein the National Pension System (NPS) has been given consistent tax breaks over the years.

Table 3 presents the share of pension and provident funds in total financial saving according to different sources. The first column comes from the RBI, Changes in Financial Assets and Liabilities of the Household Sector (RBI) at Current Prices released on September 15, 2017. The second column comes from the CSO, Changes in Financial Assets and Liabilities of the Household Sector at Current Prices : Base Year 2004-05. The CSO series for the base year 2004-05 stops at 2012-13. The third column is CSO, Gross Financial Savings of Household Sector at Current Prices: Base Year 2011-12. It would be fair to expect that for the common years, the series with different base years present comparable estimates. The CSO's document on changes in methodology in the new base year series suggest that the percentage discrepancy in household financial savings between the old and new base year series is 1.8%.

Table 3: Share of pension and provident funds in financial saving
Year RBI (2004-05) CSO (2004-05)CSO (2011-12)
2011-12 10.26 10.32 10.26
2012-13 14.71 10.99 14.71
2013-14 14.93 14.93
2014-15 14.71 15.18
2015-16 18.28 19.18
2016-17 16.26 21.50

There is a huge discrepancy in how the estimates change given the base year. For example, while the RBI data and the CSO data for base year 2011-12 suggest that the share of provident and pension funds in total saving for the year 2012-13 was 14.7%, the CSO's estimates for base year 2004-05 place this at 10.9%. In another year, however we see discrepancy between the RBI data and the CSO data for base year 2011-12. The RBI data shows a decline in the share of pension and provident funds from 18% in 2015-16 to 16% in 2016-17 while the CSO data shows an increase in the share of provident and pension funds from 19% in 2015-16 to 21.5% in 2016-17.


In the past, concerns have been raised on the quality of savings data and on the wide discrepancies visible in the RBI Flow of Funds Accounts and the CSO's data. In fact, the RBI in its August 2016 bulletin has tried to align its methodology with the CSO new base year series. However, despite their efforts, key problems in measurement remain. If there are reasons for the discrepancy, they remain inaccessible in the public domain to researchers. This is a serious concern as any analysis on household saving cannot proceed without accurate, precise and consistent data.


Since the publishing of the article, we learned that the RBI overwrites its provisional estimates (released five months ahead of the CSO's release) once the CSO releases its data on household savings. This suggests that ultimately there is only one source of data on household savings, the CSO.


Radhika Pandey and Renuka Sane are researchers at the National Institute of Public Finance and Policy.


  1. The methodology to compare instrument share (table 2) to infer or explain the reasons for difference in aggregate savings (table 1) appears to be flawed.

    The level difference (table 1) and instrument distribution (table 2) are independent by design and need not explain the inequality in each other.

    For example, even if there was equality in table 1 (say rb=cso=18 k billion), it is possible to have non-identical distribution in table 2. In others, RBI and CSO may be have same definition and recording of investment but may differ in classification by instruments.

    Secondly, even if there is inequality (as shown by author) in table 1, it is possible to have perfectly identical distribution in table 2.

    In summary, there do not appear to have a consistency link between table 1 and table 2 and hence any inference could be misleading

  2. The difference of Rs 4,000 billion is attributed to differences in the components of household financial savings. They differ in absolute levels and also in terms of their shares. We want to say that the difference is attributable to differences in components but the component which shows the starkest difference is shares and debentures. We agree in principle, that the distribution can differ even if the aggregate numbers are the same. But that may not be applicable in this case when data sources are same, the classification is same. The difference is difficult to understand.

    The RBI's document suggests that they have aligned their classification of sectors with the CSO's methodology. So the two do not differ in classification by instruments. Moreover the source of data is also identical. The data for shares and debentures is sourced from the RBI.

    1. shares and debenture (in RBI data) equals 1.8K billion only. Hence, in no way they can explain the difference of 4k billion.

      It is hard to claim that the source of data for RBI & CSO is identical. If the data sets were identical, at least the total should be equal.

      If one atleast believes that the data sets are identical the error could be simply due to some non-serious (like typing error) issues done by the reporting analyst.


Please note: Comments are moderated. Only civilised conversation is permitted on this blog. Criticism is perfectly okay; uncivilised language is not. We delete any comment which is spam, has personal attacks against anyone, or uses foul language. We delete any comment which does not contribute to the intellectual discussion about the blog article in question.

LaTeX mathematics works. This means that if you want to say $10 you have to say \$10.