The Indian Household Savings Landscape - NCAER€¦ · · 2016-07-27The Indian Household Savings Landscape Cristian Badarinza ... rural-urban divide, ... India, since households

The Indian Household

Savings Landscape

Cristian Badarinza National University of Singapore

Vimal Balasubramaniam Saïd Business School, Oxford & NCAER

Tarun Ramadorai Saïd Business School, Oxford & NCAER

India Policy Forum July 12–13, 2016

NCAER | National Council of Applied Economic Research 11 IP Estate, New Delhi 110002

Tel: +91-11-23379861–63, www.ncaer.org

NCAER | Quality . Relevance . Impact

NCAER is celebrating its 60th Anniversary in 2016-17

http://www.ncaer.org/

The findings, interpretations, and conclusions expressed are those of the authors and do not necessarily reflect the views of the Governing Body or Management of NCAER.

i

The Indian Household Savings Landscape*

Cristian Badarinza National University of Singapore

Vimal Balasubramaniam Saïd Business School, Oxford & NCAER

Tarun Ramadorai Saïd Business School, Oxford & NCAER

India Policy Forum July 12–13, 2016

* Preliminary draft. Please do not circulate beyond its discussion at the NCAER India Policy Forum 2016, for which this paper has been prepared. Badarinza: [email protected] Balasubramaniam: [email protected] Ramadorai: [email protected]. We are grateful to Suyash Rai and Milan Vaishnav for sharing their state-level inflation dataset, and Yunqi Zhang for excellent research assistance.

Abstract Savings and economic growth are closely linked in many prominent economic models. This paper describes and attempts to explain patterns in recent data on the allocation of Indian household wealth (accumulated savings). In contrast with micro-data from other countries, Indian household wealth is dominated by the presence of non-financial assets in the household balance sheet, with a particularly high relative weight in gold. The cross-household dispersion in the share of non-financial assets on the balance sheet is significantly related to the rural-urban divide, as well as to differences in education, family composition, wealth, and age. We also uncover evidence for a substitution effect between gold and real estate holdings. Controlling for demographic variation, there is substantial residual heterogeneity across Indian states in non-financial asset holdings, which is closely related to state-level variation in historical inflation volatility, and appears to have long-lasting effects through the experience channel. JEL Classification: D14, E21, E31, E44 Keywords: Gold, Households, Inflation, Real estate, Savings

mailto:[email protected]



ii

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 1

The Indian Household Savings Landscape Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai

1. Introduction

Household savings are inextricably linked with economic growth. In models ranging from simple Harrod-Domar formulations (Harrod, 1939; Domar, 1946) to more sophisticated formulations such as Lucas (1988); Romer (1986, 1989) and Mankiw, Romer and Weil (1992), savings directly affect economic growth, when transformed into productive investments in physical and human capital (see Deaton (2012) for an excellent survey). The international evidence suggests that domestic savings are particularly important in this context, as domestic savings and domestic investment rates appear to be highly correlated (Feldstein and Horioka, 1980).1

Of course, savings can also be related to economic growth as a consequence rather than a cause – the permanent income hypothesis (see, for example, Campbell (1987)) counterfactually predicts a low savings rate for fast-growing countries like India, since households expecting higher income in the future should lower current savings rates to smooth consumption over their lifecycles.2 To explain the data, an intriguing possibility that has been raised is that high-growth countries might exhibit high savings rates for precautionary motives, for example, to hedge against expected adverse shocks to future income (see Chamon and Prasad (2010), for evidence in China of this channel).

The academic literature has extensively explored both directions of the savings-growth relationship, and recent work uses microeconomic survey data in an attempt to gain greater clarity. From a policy perspective, a better understanding of household savings and patterns in the allocation of household wealth (the aggregated stock of household savings) is an essential input to growth policy through the first channel (savings impacting growth), and will facilitate better micro-prudential regulation through the second channel (i.e., documenting how household portfolio allocation adjusts in response to household expectations about future risks).

In this paper, we provide evidence from recent data on the household savings landscape in India, with a view towards shedding light on both of these channels. More specifically, our contribution is four-fold: (1) We document and describe broad patterns in the composition of household wealth in India using the latest wave (2012) of the All India Debt and Investment Survey (AIDIS), (2) We compare the composition of wealth of Indian households with households in other advanced countries, as well as China, using detailed micro-data from household surveys in these countries; (3) We uncover interesting variation in the allocation of household wealth within India across

1 Empirically, Prasad, Rajan and Subramanian (2007) present evidence that a reduced reliance on foreign capital is associated with higher growth, which further lends credence to the reliance on domestic savings in emerging economies. 2 In contrast, most standard lifecycle models of savings such as Modigliani (1986) would predict an increasing savings rate accompanying high levels of growth using standard age-compositional arguments (i.e., households that are saving in middle-age are more highly weighted than ageing households who are dissaving in retirement), but given the lack of wide prevalence of formal retirement savings schemes in most emerging economies, this explanation is somewhat less appealing for such economies.

2 India Policy Forum 2016

demographic and household characteristics and (4) We present evidence that macroeconomic experiences (more specifically, episodes of inflation uncertainty) importantly affect household wealth allocation decisions, in particular, the significant holdings of gold.3

We focus on the microeconomic data provided in the AIDIS to document patterns of Indian household savings for two reasons. First, several committees have looked into the nature of aggregate household savings using macroeconomic data available through the National Accounts Statistics, and have raised concerns about data gaps, measurement challenges, and estimation error (Kaur, 2011; Rangarajan, 2009), suggesting that the analysis of the AIDIS will help to provide complementary, and potentially more accurate evidence on the determinants of Indian household savings. Second, to shed light on the implications of household savings and wealth allocation for household welfare, there is a need to look into household asset holdings at the micro level and document the role of demographic characteristics (e.g., age, education, and wealth) in explaining household savings and wealth allocation (see, for example, Badarinza, Campbell and Ramadorai (2016)).

To provide a broad overview, India’s aggregate savings rate is comparable to that of emerging economies such as Indonesia, Thailand, and South Korea, and substantially higher than in most developed countries.4 While the aggregate gross savings rate in the economy has been growing in the long-historical context (see Mohan and Kapur (2015), for example), over the past several decades, this rate has leveled off at about 20 percent of GDP.

Decomposing these savings into physical savings (in assets such as gold and real estate), and financial savings (invested in claims such as deposits, debt, and equity), a striking feature of the Indian data is that Indian households have increasingly favoured physical over financial savings. To provide some context, in 2011-12, nearly 70 percent of aggregate annual household savings flow into physical assets, with comparable ratios in the average household wealth allocation (accumulated savings) being much higher. This issue has been a cause for concern among policy makers for several reasons,5 not least of which is that the models in which savings predict growth connect savings with productive investments – which are more easily mapped to financial savings rather than

3 A large literature focuses on the credit or liabilities side of the Indian household balance sheet. Household indebtedness, formal sources of credit, credit constraints and alleviating credit constraints have been the subject matter of analysis in several academic and policy work (for most recent work using survey data, see Pradhan, 2013). Yet, to our best of knowledge, no work has thus far looked into the composition or allocation of wealth for Indian households. While three different pieces of work, namely, Subramanian and Jayaraj (2006); Vaidyanathan (1993) and Divatia (1976), have used the wealth information data, they have all been in the context of wealth inequality and not on allocation decisions by households in India. Interestingly, Subramanian and Jayaraj (2006) focus on wealth inequality across different categories of asset holdings, and in computing inequalities within categories do not take into account the fact that households have a choice about the fraction held in different asset types. 4 World Bank data on savings rate in economies puts India at Rank 37 out of 164 economies in 2014. See http://goo.gl/scp5tP. However, this aggregate savings rate is much lower than China, whose savings as percent of GDP stands at 50% (Rank 9). 5 In 2013, press reports highlighted that the rise in the share of physical savings seemed to have stalled. For instance, see http://goo.gl/C7tkAg, and http://goo.gl/crivp2. However, the levels of these shares are still significant.

http://goo.gl/scp5tP

http://goo.gl/C7tkAg

http://goo.gl/crivp2

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 3 to stores of value such as gold. Furthermore, for Indian households, financial instruments could provide significant benefits not provided by physical assets.

Theoretically, the benefits of participating in formal financial markets are well documented (see, for example Rodrik and Rosenzweig, 2009; Campbell, 2006), and in terms of asset allocation once the participation decision has been made, financial assets provide better liquidity and diversification properties than physical assets, and can facilitate efficient household lifecycle portfolio management. While the optimal split between physical and financial assets involves complex lifecycle portfolio choice calculations, most models would struggle to explain the observed Indian household’s average allocation to physical assets as an optimal portfolio allocation. We therefore focus on explaining this puzzlingly high allocation to non-financial assets in our analysis, leaving for future work an analysis of the asset allocation patterns within the group of financial assets.6

Another way to assess Indian households’ wealth allocation patterns is to take an international comparative perspective. We find that Indian households are broadly similar to Chinese households, in that in both countries, a substantial fraction (about 90 percent) of wealth is held in real estate and other non-financial assets. We also find substantial differences between these allocation patterns and those of U.S. and European households, who invest significantly more in financial assets. Much of this variation can be ascribed to the (lack of) retirement asset holdings in emerging markets, with defined benefit and defined contribution schemes accounting for a large fraction of household asset holdings in Europe and the U.S.

We then turn to analyzing wealth allocation patterns across the demographic distribution in India. We document that there is a great deal of cross-sectional variation in household allocation decisions to real estate and gold, which is correlated with household characteristics. The allocation to physical assets decreases significantly with education, but counter-intuitively, increases with wealth. Within the physical asset share, we show that despite the fact that the share invested in gold falls with increasing wealth, the share of real estate significantly increases in wealth (but not in education). We also find substantial variation along the dimensions of whether or not households are in urban or rural areas, and by the number of children (and daughters) in the household.

Controlling for demographic characteristics, we find substantial remaining geographical (state-level) variation in the ratio of gold to total assets. This suggests that cultural factors may play an important role in the extent to which households hold gold, but we also explore the extent to which this variation in wealth allocation choices across different states stems from exposure to different economic shocks. Combining state level gold holdings ratios with state-level measures of inflation, we find that a two-cross-sectional standard deviation increase in experienced inflation volatility increases holdings of gold by 0.6 percent, a substantial impact that is comparable in magnitude to the effect of having children. This cross-sectional variation in the impact of inflation volatility on gold holdings is somewhat offset by a corresponding decrease in household allocation to real-estate, meaning that the total impact of experienced inflation volatility

6 This is currently an active area of investigation. For example, Campbell, Ramadorai and Ranish (2014) and Anagol, Balasubramaniam and Ramadorai (2015, 2016) explore Indian households’ equity portfolio allocations using detailed micro-data.


on the physical asset holdings ratio is flat. This suggests that inflation hedging motives, while important to explain the lure of gold, mainly affect the intensive margin within the set of physical assets in the household’s portfolio.

Our conclusions are twofold. First, we do find evidence of some precautionary motivations explaining household savings, namely that inflation hedging motives appear to play an important role. The important novel finding here is that inflation experiences seem to drive Indian households specifically towards investments in gold more than other physical assets. The other important addition to this finding is that these experiences appear to be long-lasting – we find evidence that inflation experiences have greater explanatory power for the allocation to gold for individuals that were young (age 25) when they had these experiences, using a methodology inspired by Malmendier and Nagel (2011). Second, we find that there is a strong effect of education on the share of the household’s allocation to financial investments, controlling for wealth and other demographic characteristics. This suggests a significant role for education, and potentially financial education in affecting household choices (see, for example, Lusardi and Mitchell (2009)). The policy prescriptions offered by these findings are self-evident.

The rest of the paper is organized as follows. Section

2 presents the data used in this paper, section

3 characterizes the household asset composition in India, section


4 explains the heterogeneity in wealth allocation, section 5 presents the relationship between macroeconomic experiences such as inflation volatility and wealth allocation choices, and section

Table 6: The Role of Inflation Experiences

This table reports estimated γ coefficients from the following specification:

𝑓𝑖,𝑘 = 𝛼 + 𝜇𝑘 + 𝛽𝑋𝑖,𝑘 + 𝛾𝛱𝑖,𝑘25 + 𝜖𝑖,𝑘,

where 𝑓𝑖,𝑘 are wealth shares of household i in state k, and 𝑋𝑖,𝑘 are dummy variables

capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. 𝛱𝑖,𝑘

25 is inflation in state k during the year in which the head

of household i was 25 years of age. We normalize the inflation level by subtracting the in-sample mean and dividing by the standard deviation. We use household-level data from the 2012 wave of the AIDIS survey and inflation data from the Labour Statistics Bureau, Government of India. *, **, *** denote statistical significance at the 10%, 5%, and 1% level, respectively.

Non-financial ratio Real estate ratio Gold ratio

Inflation experience when young -0.000 -0.002 0.003**

State fixed effects Yes Yes Yes Demographic characteristics Yes Yes Yes No. of obs. 78,486 78,486 78,486 Adjusted R 0.18 0.39 0.32

6 concludes.

2. Data

Our main data source for this study comes from the National Sample Survey (NSS) Organization’s All India Debt and Investment Survey (AIDIS) that records asset holdings as on June 2012 for households in India.7 AIDIS is a decennial survey conducted by the NSSO since 1971, with roughly a 0.01% sample of the Indian population, through a multi-stage design that is adopted in all NSS data collection exercises.8 We observe demographic information such as the gender of the head of the household, age, education level, the number of daughters, the number of children, the household sector (rural/urban), and the location of residence (state-region-district). Over and above demographic and household characteristics, on the asset side of the household balance sheet, this survey records information on land holdings, buildings, and other constructions owned, livestock and poultry, transport equipments, farm equipments, non-farm business equipments, financial assets such as shares and

7 Liabilities of the household were recorded in June 2012 and June 2013 depending on the visit during which this information was recorded. 8 A stratified multi-stage design has a first stage (FSU) which are the census villages (as of 2001 census) in the rural areas and Urban Frame Survey (UFS) blocks in urban areas (as of 2007-12 list). Further, within these FSUs, the “ultimate stage units” (USUs) are households. Should any of the FSUs be large, an additional intermediate stage of sampling using sub-blocks (hamlet-groups in case of rural areas) is used. For more details, we refer to the NSS (2012) Handbook on Survey Deisgn and Definitions available upon request from the authors or directly from the Ministry of Statistics and Programme Implementation. Appendix Figure A.1 presents the survey sample across Indian States for AIDIS, 2012.


debentures, bank deposits, insurance, pensions and other financial assets, amounts receivable for services rendered, and gold holdings. While other surveys such as the India Human Development Survey (IHDS) contain information about participation or ownership of different asset types, to the best of our knowledge, AIDIS is the only data source that provides explicit valuation of all assets held by households in great detail.

Some of the assets are valued not merely by asking respondents what the value of the asset is, but by the government records of these assets. For instance, to assess the value of land, AIDIS records land acquired prior to the survey year on a guideline basis: These are valuations obtained from Patwaris (village accountants) for rural areas and the registrar’s office (where land transactions are registered) in the urban areas. It is important to note that these valuations are in general the lower bound of the value of these asset holdings, as the market prices of land are almost always higher, and the registered prices of land transactions are often understated to avoid paying stamp duty and state government taxes. For buildings, a similar approach is adopted and floor area prices are computed using government registration records. Residential buildings exclude the value of the land on which the building is constructed and is thus not inclusive of the value of land recorded separately. For all other asset valuations, the value as stated by the respondent for the household is recorded.

For our purpose, we classify financial assets to include shares and debentures, all types of deposits, saving schemes, annuity schemes, provident fund, pension fund, NPS, other contributory funds, and payments receivable by the household. Likewise, non-financial assets include real estate assets (including land and buildings), durable assets and equipment (including livestock and poultry, transport equipment, agricultural machinery, non-farm business equipment), and finally holdings of gold/bullion.9 The survey also provides the sampling weight of each observation. The empirical measure of interest is the non-financial ratio and its sub-components, i.e., the fraction of total assets that are held in non-financial, physical form and in categories such as real estate and gold. Although this is the best source of data on the asset composition of households in India, there are some important limitations to keep in mind while interpreting the data. For example, the valuations of real estate and buildings are likely to be understated by official sources across the distribution, and not just for one or the other household.10 As reiterated in Jayadev, Motiram and Vakulabharanam (2007) and Brandolini, Cannari, D’Alessio and Faiella (2004), unless conscious efforts are made to oversample the wealthy, the extent of financialization of wealth will be misrepresented.

We construct similar measures for other countries for international comparison. We use the Chinese Household Finance Survey (CHFS, 2012), the Household, Income and Labour Dynamics in Australia Survey (HILDA, 2012), the UK Wealth and Assets

9 Although the survey collects information on gold and bullion and classify them as “financial” assets, our rationale behind this classification is also based on the liquidity in the asset market to which each asset belongs. Gold in India is physically held, and not traded frequently. 10 The extent to which asset holdings in land, buildings and gold are understated are difficult to assess for lack of alternate and better sources of information on such asset holdings. Having said that, Subramaniam and Jayaraj (2006) document that it is likely that some households (especially in the upper tail of the wealth distribution) understate their real estate holdings for fear of being reported for potential tax implications.

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 7 Survey (WAS, 2012), European Household Finance Survey (2012) for Germany, and the US Survey of Consumer Finances (SCF, 2010).11

It is important to note that different surveys cover different asset categories, and with different degrees of precision. To ensure comparability across countries and consistency with the structure of the Indian dataset, we pool asset categories with similar features.12

One notable difference between India, China, and the remaining countries concerns holdings of private retirement savings in defined-contribution accounts. In the US, UK, Australia, Canada, and Germany, such products have been robust fixtures of financial markets for decades. Most of the population relies at least partially on such private products to finance consumption in retirement and respective household surveys adequately account for this observation. On the contrary, defined-contribution savings accounts are only held by very small part of the population both in China and India. Neither the AIDIS nor the CHFS capture exact balances of defined-contribution accounts.

Finally, state-level inflation data for this study is drawn from the Labour Statistics Bureau and uses the CPI (Agricultural Workers) inflation as representative for all sampled households.13 The correlation between CPI (AW) and CPI (Industrial Workers - IW) is high at 0.92 for the period during which the data is available. Since CPI (AW) is available for a longer time-period to capture volatility experience, we use CPI (AW) for analysis.

3. Characterizing the Allocation of Household Wealth

Our focus in this paper is on characterizing households’ savings and wealth allocation to different types of assets, with a particular focus on the distinction between physical and financial assets. We note that macroeconomic and household data show different patterns of the ratio of physical to total assets (which we term the “non-financial ratio” in the remainder of this paper) for India14. Macroeconomic data estimates gross savings using a residual approach in the construction of national account statistics and uses assumptions that often generate significant inaccuracies especially when it pertains to statistics about the household sector. Furthermore, these

11 The US Survey of Consumer Finances was the first to capture detailed categorization of household balance sheets on a large scale and for representative cross-sections of the population. In this study, we report results based on the 2010 wave, which is the closest point in time to ensure comparability with the Indian micro-level data. 12 We report the exact asset definitions and groups in an online appendix to this paper. 13 We thank Suyash Rai and Milan Vaishnav for providing us with the cleaned monthly CPI (AW) series for each state by taking the average of all centres available in each state. 14 The differences in computing the non-financial ratio using micro and macro data sources are similar to the discrepancy between the per-capita consumption expenditure estimated using National Accounts Statistics and the National Sample Survey in India. For instance, see Ravallion (2003); Srinivasan (2000) and Sen (2000). However, work by Sundaram and Tendulkar (2001) minimise the discrepancy by looking into the consumption basket measured in micro surveys and document the differences with the macroeconomic data.


estimates are often made under severe data constraints.15 For instance, the total savings in cash on hand for the household sector is determined as a proportion of the total currency in circulation and is currently set at 0.93. This proportion has been used since 1985 to determine the cash on hand with the household sector. The Central Statistical Organization notes in its documentation that, “this proportion is likely to undergo change as soon as more data based on the survey results of the RBI become available.”16 These problems with macroeconomic data are well documented in various high level committees. Both the sub-group on household sector saving for the Working Group on Savings for the Twelfth Five-Year Plan (2012-13 to 2016-17), Kaur (2011), and the High Level Committee on the Estimation of Savings and Investment set up by the Ministry of Finance, Rangarajan (2009) note that there are weaknesses mainly due to data quality, data gaps and estimation problems with respect to determining the aggregate savings (for the economy as a whole and for the household sector) in India.

Figure 1: Aggregate Allocation of Household Savings

In Figure 1 we report the aggregate composition of household savings, obtained from the Reserve Bank of India. The data cover households and informal microenterprises that are excluded from the corporate sector when calculating macroeconomic aggregates.

15 Direct estimates of household saving and its composition are not available in India as it is a sector not only of households, but also, non-government non-corporate enterprises of farm business and non-farm business like sole proprietorships and partnerships, and non-profit institutions. India does not have income-expenditure surveys that normally form the basis of analysis for savings and investments, and such surveys are not conducted for all these components of the household sector in India. Household financial savings are calculated as the sum of annual increase in financial assets net of increase in financial liabilities. The financial savings of households are estimated as residuals from the flow of funds accounts, compiled by the Reserve Bank of India. The Central Statistical Organisation (CSO) estimates household investment in physical capital (using another residual method) and this is defined as physical savings. Net addition to fixed assets include: investment in fixed assets of construction and machinery, equipment, and change in stocks. The residual approach for physical investments proceed as follows: The CSO estimates total physical capital formation, and then deducts estimates of public and private corporate sector investments from the total. The remainder is considered physical savings. 16 http://goo.gl/bByvID, accessed on 07 June 2016.

http://goo.gl/bByvID


Figure 1 shows the decomposition of the aggregate household savings flow into financial and non-financial assets. It suggests that throughout the last decades, around half of annual Indian savings were consistently allocated to non-financial assets. Most recently, the fraction of non-financial assets in annual savings has risen to 70 percent, further enhancing the concern about declining financial resources available for investment in the macroeconomy.

The implications of this phenomenon are non-trivial. In every period, households allocate limited resources to different savings vehicles, some of which are non-productive and solely serve as a store of value (for example, gold). However, households’ current income and welfare, as well as the productive capacity of the economy are determined by the accumulated stock of wealth, i.e., the aggregate stock of household capital available for productive purposes. The current economic situation of Indian households, as reflected by their wealth holdings, is the product of decades of individual decisions taken within a wider macroeconomic context, and strongly influenced by social norms and personal experiences. This paper proposes a disaggregated micro-level view of household wealth that enables us to map the Indian savings landscape and to understand the underlying factors that determine household allocation decisions.

Figure 2: Allocation of Household Wealth from Micro Surveys

(a)


(b)

In Figure 2, we compare the average allocations of household assets across countries. The data sources are the All India Investment and Debt Survey (2012 wave), the Euro system Household Finance and Consumption Survey (HFCS), the US Survey of Consumer Finances (SCF, 2010 wave), the Chinese Household Finance Survey (CHFS, 2012 wave), the Australian Household, Income and Labour Dynamics Survey (HILDA, 2010 wave), and the UK Wealth and Assets Survey (WAS, 2012 wave). Financial assets include current accounts, checking, money market, savings and transaction accounts, sovereign and corporate fixed-income products, both denominated in domestic and foreign currency, publicly traded shares, employee shares and options, trust accounts, money owed to households, certificates, leases, proceeds from lawsuits (HFCS, SCF), other non-pension assets (SFS), children’s bank accounts (HILDA), children trust funds, National Savings Products and other investments (WAS). Retirement assets include all types of defined contribution plans (public, occupational, or private), which have an account balance. Unlike other countries, financial assets include retirement assets in India. In Panel (a), we compute averages across households using population weights, as indicated in each survey. Additionally, in Panel (b), we present the value-weighted averages across households.

We proceed by comparing the average allocations of household assets across different countries in Figure 2 (a). Indian and Chinese households (on average) hold roughly 76 and 82 percent of their wealth in real estate respectively. However, Indian households hold more durable assets than the Chinese households whereas the extent of financialization of the household balance sheet is greater with Chinese households than Indian ones. Furthermore, the household allocation choices are very different in India and China when compared with more advanced economies. On average, holdings of real estate account for lower fractions of total savings in countries such as the US (43.8%), and particularly Germany (36.7%).

Figure 2 (b) presents the value-weighted average allocations of household assets across countries. Value-weighted quantities are calculated by summing up the different types of assets across all surveyed households, appropriately weighted to ensure that the result is representative for the entire population. The resulting ratios of financial and non-financial assets relative to the sum total of Indian household wealth therefore reflect the aggregate composition of asset holdings for the country as a whole. They are,

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 11 however, unrepresentative of the way in which the total wealth of the country is distributed among the population. That said, the resulting pattern is quite similar to the equally weighted average allocations and suggests that Indian and Chinese households are heavily skewed towards real estate assets in their wealth composition, in contrast with the greater prevalence of financial assets in developed economies. The overall value-weighted non-financial assets ratio (the sum of real estate, durable goods, and gold holdings relative to total assets) is highest in India (92.19%), and only slightly lower in China (87.60%). At the other extreme, non-financial assets account for only 50.1% of total household wealth in Germany. Despite high rates of homeownership, which are also partially stimulated by government intervention (such as in the US), the overall role of real estate remains lower in advanced economies.

Households in advanced economies do not necessarily directly hold more financial assets than their Indian counterparts, but they do accumulate sizeable amounts of funds in retirement accounts over the course of their lifetimes. The German case illustrates the substitution effect between public and private pension systems particularly well. Since the retirement system is one mostly based on state-sponsored defined-benefit pensions, households in Germany only save small amounts in private retirement accounts and instead decide to invest larger amounts in financial assets such as sight deposits, government bonds, publicly traded shares, and mutual funds.

Table 1: All India Debt and Investment Survey, 2012: Summary Statistics

Mean Percentile

10

th 25

th 50

th 75

th 90

th 99

th

Financial Assets 52,513 0 0 2,200 19,710 95,000 8,51,250

Non-financial assets 15,16,554 53,500 1,67,000 4,70,650 12,61,950 31,78,710 1,46,00,000

Real Estate / Housing 14,00,534 0 1,25,600 4,00,000 11,35,000 29,38,350 1,40,00,000

Durable Goods 58,151 0 1,000 11,200 41,960 1,07,700 7,57,000

Gold / Bullion 57,870 0 3,500 20,000 60,000 1,50,000 5,50,000

Total assets 15,69,066 63,500 1,79,000 4,93,000 13,15,766 33,01,500 1,50,00,000

Notes: N= 107,950. All reported statistics are weighted by the population weights and are in Rupees.

We now focus on the micro data from the AIDIS survey to understand Indian wealth allocation in more detail. Table 1 presents summary statistics of the detailed asset composition held by Indian households. The equal-weighted average household owns 96 percent of all assets in physical form, 90 percent of which are in real estate and housing assets. The next few columns present the distribution of each of variables in

columns from the 10th to the 99th percentile – we note wide variation in the composition of asset holdings by Indian households. At the median, households own about Rs. 2,200 of financial assets (which include bank deposits and small savings), and about Rs. 20,000 in gold. Holdings of financial assets vary from 0 to above Rs. 10 million – even at the maximum, the share of financial assets to total assets is 0.55. On average (and across the distribution), Indian households mostly favour physical (or non-financial) assets.


Table 2: Allocation of Household Wealth in India

Table 2 presents the composition of household assets by different household charactersitics in India. Each column presents different types of assets owned by the household: Financial assets are the sum total of shares (direct and indirect ownership), debentures, bank deposits, annuities, provident fund investments (GPF, PPF, etc.), small savings schemes (Post Office, RBI Bonds), Unit-linked Insurance Products, pensions, and amount receivable from services rendered. Real Estate and Housing comprise of both land and house ownership valued at local guideline values obtained from the Registrar’s office in urban areas and the local village accountant (Patwaris) for rural areas. Durables include livestock such as cows, buffaloes, farm and non-farm equipments (including vehicle ownership), Gold/Bullion include gold both in the form of jewelry and otherwise. Columns 1, 3–5 are presented as a fraction of total assets owned by the household. Column 2 presents the non-financial asset ratio, that is, the sum of columns 3 to 5.

Financial Assets

Non-financial Assets

Real Estate Housing

Durables Gold Bullion

(1) (2)=3+4+5 (3) (4) (5)

Education

Illiterate/Below Primary school 0.03 0.97 0.82 0.06 0.10

Primary and Middle school 0.04 0.96 0.79 0.06 0.11

Secondary school 0.05 0.95 0.75 0.07 0.13

Diploma 0.08 0.92 0.72 0.07 0.13

Graduate and postgraduate 0.15 0.85 0.65 0.08 0.12

Total Assets

< 179,000 0.10 0.90 0.55 0.11 0.24

179,000 to 493,030 0.04 0.96 0.80 0.06 0.10

493,030 to 1.3mn 0.04 0.96 0.84 0.05 0.07

1.3mn to 1.48 mn 0.04 0.96 0.88 0.04 0.04

> 1.48 mn 0.03 0.97 0.93 0.03 0.02

Age

24 to 35 0.08 0.92 0.70 0.08 0.15

36 to 43 0.06 0.94 0.75 0.07 0.12

43 to 50 0.05 0.95 0.78 0.06 0.10

51 to 60 0.05 0.95 0.80 0.06 0.09

61 to 110 0.04 0.96 0.83 0.05 0.09

Region Type

Rural 0.03 0.97 0.83 0.06 0.08

Urban 0.11 0.89 0.65 0.08 0.17

Children

0 0.09 0.91 0.72 0.06 0.13

1 0.06 0.94 0.77 0.06 0.11

> 1 0.05 0.95 0.78 0.07 0.10

Daughters

0 0.07 0.93 0.74 0.06 0.12

1 0.05 0.95 0.78 0.06 0.11

> 1 0.04 0.96 0.80 0.07 0.09

India 0.06 0.94 0.77 0.07 0.11


Delving further into the household level data, we observe important differences in asset composition across household characteristics such as education, wealth levels, age, and household sector (urban and rural). Table 2 presents the composition of assets by different household characteristics. Column (1) presents the fraction of total assets in financial instruments and Columns 3, 4, and 5 present real estate (and housing), durables, and gold (and bullion) holdings. Column (2) is the sum of Columns 3, 4, and 5 and reports the overall non-financial assets ratio. There is nearly a 12 percentage point difference in the extent of financial assets held by households that are headed by those with graduate (or post-graduate) education and those who are illiterate; most of this increase in financial assets is driven by decreases in real estate and housing allocations. The fraction of gold and bullion on the household balance sheet is invariant to education level in India. Households with total wealth of less than 1,79,000 rupees hold about 24 percent of their wealth in gold, whereas those in the highest wealth grouping (>1.48 million rupees) hold about 2 percent in gold and bullion.17 Similarly, the extent of gold in the portfolio decreases from 15 percent to 9 percent as household heads grow older, and the share of real estate increases to over 80 percent. Households in urban areas (on average) hold about 11 percent of their total assets in financial instruments, whereas this figure is very low at 3 percent for households in rural areas. Similarly, the non-financial ratio is nearly 1 (0.97) on average for rural areas, whereas it is much lower (and still relatively high compared to other countries) at 0.89 for households in urban areas.

Table 3 presents the average share of wealth in each asset type across Indian states. While a (poor) state like Bihar has households with barely no financial assets, cities/union territories18 such as Chandigarh have the highest levels of financialization of the household balance sheet. Similarly, households in Tamil Nadu, on average, hold about 28 percent of assets in gold (and bullion), followed closely by Andhra Pradesh (at 22 percent) – these high gold holdings in Southern Indian states suggest that strong cultural factors may be at play in these cross-state patterns. Across states, we observe a great deal more variation within the non-financial ratio, i.e., on the intensive margin, between real estate and gold.

17 We group households into five wealth categories – the first four based on the quartile breakpoints on the wealth distribution and the fifth for the top 1% of the wealth distribution. The remainder of our analysis uses these breakpoints to evaluate patterns across the household wealth distribution. 18 Union Territories in India, unlike the states, are ruled directly by the Central Government and are federal territories that have no elected sub-national governments of their own.


Table 3: Regional Variation in Household Wealth Allocation

This table presents the average share of wealth in each asset type across Indian states (in rows) presented in the increasing order of the financial assets ratio.

Financial

Assets Real Estate

Housing Durables Gold

Bullion Non-financial

Assets

(1) (2) (3) (4) (5)=2+3+4

Bihar 0.02 0.91 0.05 0.03 0.98

Rajasthan 0.03 0.79 0.08 0.09 0.97

Uttar Pradesh 0.03 0.85 0.06 0.06 0.97

Madhya Pradesh 0.04 0.82 0.07 0.07 0.96

Chhattisgarh 0.04 0.82 0.08 0.07 0.96

Orissa 0.04 0.79 0.07 0.10 0.96

Jharkhand 0.04 0.86 0.06 0.04 0.96

Uttarakhand 0.04 0.79 0.07 0.10 0.96

Telengana 0.04 0.71 0.08 0.18 0.96

Manipur 0.04 0.84 0.06 0.05 0.96

Kerala 0.05 0.79 0.03 0.13 0.95

Gujarat 0.06 0.73 0.08 0.14 0.94

Lakshadweep 0.06 0.80 0.03 0.11 0.94

Jammu & Kashmir 0.06 0.84 0.05 0.05 0.94

Tamil Nadu 0.06 0.59 0.06 0.28 0.94

Haryana 0.06 0.81 0.07 0.06 0.94

Maharashtra 0.07 0.77 0.06 0.10 0.93

Tripura 0.07 0.76 0.07 0.10 0.93

Andhra Pradesh 0.07 0.63 0.09 0.22 0.93

West Bengal 0.07 0.81 0.05 0.07 0.93

Punjab 0.08 0.82 0.06 0.05 0.92

Meghalaya 0.08 0.81 0.08 0.03 0.92

Assam 0.08 0.76 0.09 0.07 0.92

Nagaland 0.09 0.83 0.07 0.02 0.91

Karnataka 0.09 0.67 0.07 0.16 0.91

Delhi 0.09 0.82 0.02 0.06 0.91

Goa 0.10 0.60 0.10 0.20 0.90

Himachal Pradesh 0.10 0.72 0.04 0.14 0.90

Mizoram 0.11 0.80 0.08 0.01 0.89

Pondicherry 0.12 0.57 0.06 0.26 0.88

Arunachal Pradesh 0.13 0.63 0.18 0.05 0.87

Sikkim 0.22 0.56 0.08 0.15 0.78

Chandigarh 0.22 0.57 0.10 0.10 0.78

Daman & Diu 0.23 0.48 0.05 0.24 0.77

Dadra & Nagar Haveli 0.23 0.63 0.08 0.06 0.77

Andaman & Nicobar 0.24 0.42 0.10 0.24 0.76


Table 4: Regional Variation in Household Wealth Allocation across the Wealth Distribution

This table presents the average share of wealth in each asset type for households with wealth <179,000 and >1.4 million rupees across Indian states (in rows) presented in the increasing order of the financial assets ratio for households with assets valued less than 179,000 rupees.

Financial Assets

Real Estate Housing

Gold Bullion

Non-financial Assets

< 179,000 > 1.4mn < 179,000 > 1.4mn < 179,000 > 1.4mn < 179,000 > 1.4mn Bihar 0.02 0.02 0.86 0.95 0.04 0.01 0.98 0.98 Lakshadweep 0.03 0.06 0.00 0.86 0.78 0.06 0.97 0.94 Orissa 0.03 0.09 0.75 0.80 0.13 0.05 0.97 0.91 Manipur 0.04 0.06 0.80 0.84 0.09 0.04 0.96 0.94 Mizoram 0.04 0.15 0.77 0.79 0.03 0.01 0.96 0.85 Jharkhand 0.05 0.08 0.76 0.86 0.12 0.02 0.95 0.92 Madhya Pradesh 0.06 0.03 0.66 0.90 0.17 0.02 0.94 0.97 Tripura 0.06 0.06 0.74 0.86 0.13 0.04 0.94 0.94 Rajasthan 0.06 0.03 0.55 0.86 0.26 0.05 0.94 0.97 Chhattisgarh 0.07 0.04 0.62 0.90 0.17 0.03 0.93 0.96 Uttarakhand 0.08 0.03 0.43 0.91 0.28 0.03 0.92 0.97 Uttar Pradesh 0.08 0.03 0.62 0.91 0.18 0.02 0.92 0.97 Telengana 0.09 0.03 0.45 0.85 0.37 0.06 0.91 0.97 Goa 0.09 0.10 0.08 0.69 0.61 0.10 0.91 0.90 Meghalaya 0.09 0.06 0.75 0.88 0.05 0.02 0.91 0.94 West Bengal 0.10 0.09 0.70 0.85 0.12 0.03 0.90 0.91 Tamil Nadu 0.12 0.04 0.29 0.81 0.49 0.11 0.88 0.96 Gujarat 0.12 0.03 0.34 0.88 0.40 0.04 0.88 0.97 Andhra Pradesh 0.12 0.04 0.40 0.82 0.35 0.08 0.88 0.96 Maharashtra 0.12 0.04 0.47 0.91 0.29 0.02 0.88 0.96 Assam 0.15 0.06 0.57 0.87 0.13 0.03 0.85 0.94 Arunachal Pradesh 0.16 0.10 0.55 0.75 0.07 0.02 0.84 0.90 Karnataka 0.21 0.06 0.29 0.83 0.39 0.05 0.79 0.94 Kerala 0.21 0.03 0.09 0.88 0.58 0.07 0.79 0.97 Nagaland 0.23 0.09 0.22 0.82 0.16 0.02 0.77 0.91 Pondicherry 0.25 0.04 0.01 0.87 0.63 0.07 0.75 0.96 Jammu & Kashmir 0.28 0.04 0.31 0.90 0.28 0.02 0.72 0.96 Haryana 0.30 0.03 0.29 0.90 0.20 0.03 0.70 0.97 Punjab 0.31 0.03 0.45 0.90 0.12 0.02 0.69 0.97 Andaman & Nicobar 0.35 0.14 0.10 0.71 0.40 0.09 0.65 0.86 Himachal Pradesh 0.37 0.04 0.05 0.89 0.49 0.03 0.63 0.96 Sikkim 0.46 0.08 0.09 0.82 0.37 0.03 0.54 0.92 Chandigarh 0.50 0.05 0.12 0.87 0.13 0.04 0.50 0.95 Daman & Diu 0.50 0.03 0.01 0.91 0.44 0.03 0.50 0.97 Delhi 0.57 0.02 0.00 0.93 0.41 0.01 0.43 0.98 Dadra & Nagar Haveli 0.63 0.03 0.04 0.92 0.16 0.01 0.37 0.97 India 0.10 0.04 0.56 0.88 0.24 0.04 0.90 0.96


Even along the wealth distribution (Table 4), the variation in asset composition across states is striking: While households with less than 1,79,000 rupees (bottom quartile) hold nearly half their wealth in gold in Tamil Nadu, only about 11 percent of the wealth for those with assets over 1.4 million rupees are held in gold. The increase in real-estate holdings as a function of wealth is also striking: While poor states such as Bihar have very similar asset composition across the wealth distribution, households in richer states like Gujarat, Tamil Nadu, and Maharashtra, exhibit a great deal more variation. The variation across states and wealth bins in the fraction allocated to financial, real estate, and gold assets naturally also contributes to the variation observed in Table 1. In the next sections, we explain the heterogeneity in wealth allocation more formally in a multiple regression setup, that we also use to study the role of inflation uncertainty and asset composition more closely.


4. Explaining Heterogeneity in Wealth Allocation

Table 5: Explaining Heterogeneity in Wealth Allocation across Indian Households

This table reports estimated β coefficients from the following estimated specification:

𝑓𝑖,𝑘 = 𝛼 + 𝜇𝑘 + 𝛽𝑋𝑖,𝑘 + 𝜖𝑖,𝑘


capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. We use household-level data from the 2012 wave of the Indian NSS survey and inflation data from the Labour Statistics Bureau, Government of India. *, **, *** denote statistical significance at the 10%, 5%, and 1% level, respectively.

Non-

financial ratio

Real estate ratio

Gold ratio

Education Illiterate/below Primary School

- - -

Primary and Middle School -0.013*** -0.039*** 0.021*** Secondary School -0.026*** -0.078*** 0.038*** Diploma -0.056*** -0.128*** 0.050*** Graduate and Postgraduate -0.109*** -0.188*** 0.042*** Total Assets < 179,000 - - - 179,000 to 493,030 0.054*** 0.239*** -0.134*** 493,030 to 1.3mn 0.070*** 0.317*** -0.184*** 1.3mn to 1.48 mn 0.104*** 0.432*** -0.244*** > 1.48 mn 0.149*** 0.537*** -0.280*** Age 24 to 35 years - - - 36 to 43 years 0.003 0.009 -0.003 43 to 50 years 0.001 0.008 -0.006 51 to 60 years 0.001 0.012 -0.012*** > 61 years 0.015*** 0.031*** -0.009* Region Type Rural - - - Urban -0.064*** -0.144*** 0.069*** Children 0 - - - 1 0.019*** -0.021*** 0.024*** 2 0.027*** -0.004 0.015* Daughters 0 - - - 1 -0.001 -0.013*** 0.011*** 2 -0.002 -0.021*** 0.015*** Constant term 0.877*** 0.593*** 0.191*** State FE Yes Yes Yes No. of observations 107,950 107,950 107,950

Adjusted 𝑹𝟐 0.19 0.40 0.32

The framework of interest is given by a cross-sectional regression at the household level, across all Indian states. We estimate the following empirical specification:



where, fi,k are wealth shares of different asset types of household i in state k, μk are the

state fixed-effects and Xi,k are dummy variables capturing wealth and age quintiles,

education groups, rural and urban residency, number of children and number of daughters. Table 5 presents the regression results where fi,k is the non-financial ratio,

real estate ratio and gold ratio across three adjacent columns.

The non-financial ratio is a declining function of education. Relative to the group of households with illiterate heads (or ones with below primary school education), having earned a diploma is associated with a 5.6 percentage points lower non-financial assets share. For graduate and postgraduate household heads, the difference is even more substantial, reaching 10.9 percentage points. This is primarily driven by the extent of land holdings declining by 12.8 (diploma holders) and 18.8 percentage points (graduate and postgraduate holders), respectively. To the contrary, investments in gold are a slightly increasing, although non-monotonic function of education, even after controlling for total asset holdings.

Compared to households with less than Rs. 1,79,000 in total assets, we observe a very large increase in the non-financial ratio at the highest ends of the wealth distribution. While the overall increase in non-financial ratio is noteworthy, the rise in the fraction of real estate (land and buildings) owned by the rich households (nearly 53 percentage point increase) is offset by the decline in the role of gold (a drop of 28 percentage points relative to households with less than 1,79,000). To the extent that the wealth distribution is representative of rich households in India, this suggests that the financialization of the asset side of the household balance sheet is meagre, even among the rich households. Urban households hold 6.4 percentage points fewer non-financial assets than their rural counterparts. This decrease is primarily due to reduction in the extent of real estate holdings. However, urban households notably also carry nearly 7 percentage points more gold on their balance sheets. One possible explanation for the decline in real estate holdings for urban households is the relative costs of these assets may be higher; another is that there is a greater prevalence of rental contracts in such locations, precluding the need to own a housing asset.

The life-cycle properties of the non-financial ratio are nearly absent. At different segments of the age distribution, there is no significant difference in the non-financial ratio or the real-estate ratio, except for the highest end of the age distribution where households tend to reduce their gold holdings (by a small margin of 1 percentage point) and increase land and housing holdings (by 3.1 percentage points). Put differently, the real-estate holdings may be a higher fraction on the household balance sheet as one grows older, as more liquid assets are drawn down towards retirement (and gold is often given to daughters of the household when they get married). As a function of the number of children in the household, gold holdings increase relative to households with no children and then decrease marginally. However, as a function of the number of daughters of the household, these holdings also increase, but only marginally. Relative real estate shares fall as the number of daughters in the household increases. This suggests that the motivation to hold gold may not only be due to social norms, and other factors may also be at play in determining the propensity of households to accumulate gold.


Figure 3: Residual Variation of Asset Ratios across Indian States

This figure reports estimated μk coefficients from the following specification:

𝑓𝑖,𝑘 = 𝛼 + 𝜇𝑘 + 𝛽𝑋𝑖,𝑘 + 𝜖𝑖,𝑘,

where fi,k are wealth shares of household i in state k, and Xi,k are dummy variables

capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. The indicator variable for the state of Jammu & Kashmir is omitted from the regressions, which implies that this state is taken as a reference. We use household-level data from the 2012 wave of the Indian NSS survey.

After controlling for household characteristics, the extent of variation across states remains large. Figure 3 suggest that the maximum (average) real estate ratio unexplained by household characteristics is in the state of Bihar and the least in the Andaman and Nicobar Island. Similarly, the maximum unexplained gold ratio is in the state of Tamil Nadu, and least in the state of Mizoram. This variation is important as it suggests that there might be region specific variation in factors that determine household asset allocation choice that can be studied.


5. Inflation Uncertainty and Wealth Allocation

Undoubtedly, cultural factors explain part of the regional variation in patterns of savings allocation. However, motivated by a growing literature on the role of experience in determining household economic decisions (and in particular inflation experience, see Malmendier and Nagel (2011)), we check how the extent of inflation uncertainty experienced by households in different regions in India affects their allocation decisions. The principal question here is whether, in the absence of financial markets that provide for a good hedge to inflation, households believe that non-financial assets, especially gold, provide a hedge against inflation eroding household savings.

To identify this effect, we use the fact that India doesn’t have a unified market for goods and the markets are deeply segmented, meaning that a regional demand or supply shock is almost always not arbitraged. This means that we can use the experienced inflation, which differs across Indian states, to attempt to explain the state-level variation in physical asset holdings.

Figure 4: Relationship between Inflation Uncertainty and Wealth Allocation in

India

This figure reports the correlation between household wealth shares and inflation developments across Indian states. We obtain state-specific wealth shares as fixed effects 𝜇𝑘 from the following estimated specification:

𝑓𝑖,𝑘 = 𝛼 + 𝜇𝑘 + 𝛽𝑋𝑖,𝑘 + 𝜖𝑖,𝑘,


capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. The sample contains 15 states for which we are able to match household-level information from the NSS survey and historical regional consumer price indexes: Maharashtra, Andhra Pradesh, Karnataka, Rajasthan, Gujarat, Uttar Pradesh, West Bengal, Bihar, Madhya Pradesh, Orissa, Tamil Nadu, Assam, Kerala, Punjab, and Jammu & Kashmir. We use household-level data from the 2012 wave of the Indian AIDIS survey. Inflation volatility is calculated as the in-sample standard deviation of the consumer price index between 2003 and 2012. Inflation data is sourced from the Labour Statistics Bureau, Government of India.

(a) Inflation and Non-financial Ratio (b) Inflation and Real Estate Assets Ratio


(c) Inflation and Gold Ratio

How do inflation and inflation uncertainty affect wealth allocations in India? We study this by reporting the correlation between household wealth shares and inflation developments across states. We obtain the state-specific wealth share as fixed effects (μk) from Section


4, after having controlled for an array of household-level characteristics. For 15 of the Indian states, state-wise inflation experience can be computed on the basis of reported inflation data collected from centres across India that is also used to compute the national inflation series. We proxy inflation uncertainty using the in-sample standard deviation of the CPI (AW) between 2003 and 2012.19 Figure 4 presents the relationship between inflation experience and the non-financial ratio and its sub-components.

While the relationship between inflation and non-financial ratio is flat, the relationship between inflation volatility and the real estate ratio, and inflation volatility and the gold ratio are strongly (and statistically significantly) decreasing and increasing respectively. Within the gamut of non-financial assets that households could choose from, these correlations suggest that there is a substitution effect between real estate and gold that can be explained by the extent of inflation uncertainty in different regions of India.20

Our explanation of the result is that if financial assets are not easily available/accessible or if demand-side frictions cause investors to eschew them, investors seek alternative savings vehicles.21 Since very few households invest internationally, and there are regulatory restrictions to how much money can be invested abroad, the set of viable alternatives is restricted to the space of non-financial, physical assets.22 In the space of non-financial assets, gold may be seen as an inflation hedge.

At the same time, the relationship between inflation uncertainty and real estate suggests that people with high experienced inflation volatility avoid real estate. Several potential channels of explanation exist. Most plausibly, real estate is a lower liquidity asset when compared with gold, and if liquidity needs are correlated with inflation volatility, gold better serves the purpose than real estate. Gold as a non-financial asset also has additional properties that are not provided for by real estate, such as a high collateral value and physical verifiability (as opposed to the great many challenges inherent in property verification in India).

Another possibility here has to do with the financing mechanism for the real estate investment. Since we have not considered the liabilities side of the balance sheet, it may be that the prevalence of adjustable rate mortgages makes mortgage-financed real estate a more risky proposition in an environment of high inflation volatility, especially if households face a current income affordability constraint (see Campbell, Cocco et al. (2003)).

For a deeper understanding of these phenomena, and in the spirit of Malmendier and Nagel (2011), we run a regression where we explicitly consider the role of experienced inflation at the age of 25 (Table 6). The magnitude of the estimated effect is

19 Appendix Figure A.2. presents the inflation variation across states and over time. 20 This effect is economically and statistically significant even after controlling for variation in State GDP. See Appendix Figure A.3. 21 Most recently, an RBI appointed committee on financial inclusion concluded that “despite improved financial access, usage remains low, underscoring the need to better leverage technology to facilitate usage.” (Mohanty, 2015). 22 Caballero and Krishnamurthy (2009) show that these have real macroeconomic consequences in terms of the global imbalances in savings.

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 23 large: Two standard-deviations higher inflation leads to a 0.6% increase in the non-financial ratio. In terms of benchmarking, the size of the effect amounts to about half the contribution of having a daughter in the household. Table 6 also shows that the substitution effect between real estate and gold ratio is robust, suggesting that inflation remains important in shaping household decisions on the extensive margin, beyond other more deep-seated household preferences that are related to social norms in India.

Table 6: The Role of Inflation Experiences

This table reports estimated γ coefficients from the following specification:

𝑓𝑖,𝑘 = 𝛼 + 𝜇𝑘 + 𝛽𝑋𝑖,𝑘 + 𝛾𝛱𝑖,𝑘25 + 𝜖𝑖,𝑘,


capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. 𝛱𝑖,𝑘

25 is inflation in state k during the year in which the head

of household i was 25 years of age. We normalize the inflation level by subtracting the in-sample mean and dividing by the standard deviation. We use household-level data from the 2012 wave of the AIDIS survey and inflation data from the Labour Statistics Bureau, Government of India. *, **, *** denote statistical significance at the 10%, 5%, and 1% level, respectively.

Non-financial ratio Real estate ratio Gold ratio

Inflation experience when young -0.000 -0.002 0.003**

State fixed effects Yes Yes Yes Demographic characteristics Yes Yes Yes No. of obs. 78,486 78,486 78,486

Adjusted R2

0.18 0.39 0.32

6. Conclusion

This paper describes the allocation of Indian household wealth using international comparative data, in combination with the most recent wave of the AIDIS (2012) household survey. We document the dominant role of non-financial assets in the household balance sheet, particularly in terms of gold holdings, land and buildings (residential real estate). The share of non-financial assets varies significantly between rural and urban areas, by household characteristics (such as the education of the household head and the number of children), and by wealth. We exploit the state-level history of inflation developments and highlight the role of personal experiences in shaping the reliance on non-financial assets. We also document a strong substitution effect between gold and real estate in the household balance sheet along the wealth distribution, and show that it is related to macroeconomic experiences in terms of inflation uncertainty. Controlling for variation in local demographic structures, we find substantial residual heterogeneity across Indian states, a part of which could be attributed to the perceived role of gold as an inflation hedge.

These observations raise important policy concerns about inflation uncertainty, and financial access. The role of sound macroeconomic policies with a strong inflation target motive is important. Creating the policy environment that incentivizes new financial savings instruments to alleviate the dependence on physical savings such as gold needs to be further stimulated. Finally, education appears to increase the allocation to financial assets substantially. Further investment in education, and in particular


financial education, would appear to generate gains on this margin as well as the more obvious ones.

References

Anagol, Santosh, Vimal Balasubramaniam, and Tarun Ramadorai, 2015. "The Effects of Experience on Investor Behavior: Evidence from India's IPO Lotteries." Available at SSRN 2568748 .

Anagol, Santosh, Vimal Balasubramaniam, and Tarun Ramadorai, 2015. "Endowment Effects in the Field: Evidence from India's IPO Lotteries." Available at SSRN .

Badarinza, Cristian, John Y. Campbell, and Tarun Ramadorai, 2016. "International Comparative Household Finance." Annual Review of Economics, 8.1.

Brandolini, Andrea, et al, 2004. "Household Wealth Distribution in Italy in the 1990s, Temi di discussione (Economic working papers) 530, Bank of Italy." Economic Research Department .

Caballero, Ricardo J., and Arvind Krishnamurthy, 2009. "Global Imbalances and Financial Fragility." The American economic review, 99(2): 584-588.

Campbell, John Y, 1987. "Does Saving Anticipate Declining Labor Income? An Alternative Test of the Permanent Income Hypothesis." Econometrica: Journal of the Econometric Society, 55: 1249-1273.

Campbell, John Y. "Household finance." The Journal of Finance, 61.4 (2006): 1553-1604.

Campbell, John Y., and Joao F. Cocco. "Household Risk Management and Optimal Mortgage Choice." Quarterly Journal of Economics, 118.4 (2003).

Campbell, John Y., Tarun Ramadorai, and Benjamin Ranish, 2014. “Getting better or feeling better? How equity investors respond to investment experience.” No. w20000. National Bureau of Economic Research.

Chamon, Marcos D., and Eswar S. Prasad. "Why are saving rates of urban households in China rising?." American Economic Journal: Macroeconomics, 2.1 (2010): 93-130.

Deaton, Angus, 2012. "Franco Modigliani and the life-cycle theory of consumption."PSL Quarterly Review, .

Divatia, V. V. "Inequalities in Asset Distribution of Rural Households." Reserve Bank of India Staff Occasional Papers 1.1 (1976).

Domar, Evsey D. "Capital expansion, rate of growth, and employment."Econometrica, Journal of the Econometric Society, (1946): 137-147.

Feldstein, Martin, and Charles Horioka. "Domestic Savings and International Capital Flows." The Economic Journal, (1980).

Harrod, Roy F. "An essay in dynamic theory." The economic journal, 49.193 (1939): 14-33.

Jayadev, Arjun, Sripad Motiram, and Vamsi Vakulabharanam, 2007. "Patterns of wealth disparities in India during the liberalisation era." Economic and Political Weekl,y, 32: 3853-3863.

Cristian Badarinza, Vimal Balasubramaniam, Tarun Ramadorai 25 Kaur, Balbir, 2011, “Report of Sub-Committee on Household Sector Savings.” Planning Commission.

Lucas, Robert E, 188. “On the mechanics of economic development,” Journal of monetary economics, 22 (1): 3–42.

Lusardi, Annamaria and Olivia S Mitchell, 2009. “How ordinary consumers make complex economic decisions: Financial literacy and retirement readiness,” Technical Report, National Bureau of Economic Research.

Malmendier, Ulrike and Stefan Nagel, 2011.“Depression Babies: Do Macro Economic Experiences Affect Risk Taking?” Quarterly Journal of Economics, 126: 373–416.

Mankiw, Gregory N, David Romer, and David N Weil, 1992. “A contribution to the empirics of economic growth,” Quarterly Journal of Economics, 107 (2): 407–437.

Modigliani, Franco, 1986. “Life cycle, individual thrift, and the wealth of nations,” The American Economic Review, 76 (3): 297–313.

Mohan, Rakesh and Muneesh Kapur, 2015“Pressing the Indian Growth Accelerator: Policy Imperatives”.

Mohanty, Deepak, 2015. “Report of the Committee on Medium-term Path on Financial Inclusion,” Reserve Bank of India.

Pradhan, Narayan Chandra, 2013.“Persistence of informal credit in rural India: Evidence from all-India debt and investment survey and beyond,” WPS (DEPR), 5, 2013.

Prasad, Eswar S, Raghuram G Rajan, and Arvind Subramanian, 2007. “Foreign capital and economic growth,” Technical Report, National Bureau of Economic Research.

Rangarajan, C, 2009. “Report of the High Level Committee on Savings and Investment,” Ministry of Finance, Government of India.

Ravallion, Martin, 2003. “Measuring aggregate welfare in developing countries: How well do national accounts and surveys agree?,” Review of Economics and Statistics, 85 (3): 645–652.

Rodrik, Dani and Mark Rosenzweig, 2009. “Access to Finance Chapter 2, Handbook of Development Economics,” Volume 5.

Romer, Paul M, 1986. “Increasing returns and long-run growth,” The journal of political economy, pp. 1002–1037.

Romer, Paul M, 1989. “Human Capital and Growth: Theory and Evidence,” NBER Working Paper Series, p. 3173.

Sen, Abhijit, 2000. “Estimates of consumer expenditure and its distribution: Statistical priorities after NSS 55th round,” Economic and Political weekly, pp. 4499–4518.

Srinivasan, TN, 2000. “Growth, poverty reduction and inequality,” in “European Annual World Bank Conference on Development Economics”.

Subramanian, S and D Jayaraj, 2006. The distribution of household wealth in India number 2006/116, Research Paper, UNU-WIDER, United Nations University (UNU).

Sundaram, K and Suresh D Tendulkar, “Recent debates on data base for measurement of poverty in India: some fresh evidence,” in “Workshop on Poverty organised by the Indian Planning Commission and the World Bank (January 11-12)” Citeseer.


Vaidyanathan, A, 1993. “Asset Holdings and Consumption of Rural Households in India: A Study of Spatial and Temporal Variations,” Agricultural Development Policy: Adjustments and Reorientation.


APPENDIX

Figure A1: Survey Sample

This figure presents the survey sample across the States of India in the All India Investment and Debt Survey, 2012. Figure A2: Inflation Developments across Indian States

This figure reports the evolution of consumer price inflation for the following set of states: Andhra Pradesh, Assam, Bihar, Gujarat, Jammu and Kashmir, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh and West Bengal. The data source is the Labour Statistics Bureau and uses the CPI (Agricultural Workers) inflation as representative for all sampled households.

05

00

01

000

01

50

00

Utta

r Pra

desh

Mah

aras

htra

Wes

t Ben

gal

Tam

il Nad

u

Mad

hya

Prade

sh

Bih

ar

Rajas

than

Karna

taka

Guj

arat

Keral

a

Oris

sa

And

hra

Prade

sh

Ass

am

Teleng

ana

Pun

jab

Tripur

a

Man

ipur

Jhar

khan

d

Jam

mu

& K

ashm

ir

Har

yana

Chh

attis

garh

Him

acha

l Pra

desh

Meg

halaya

Mizor

am

Nag

alan

d

Aru

nach

al P

rade

sh

Utta

rakh

and

Sikkim

Pondi

cher

ry

And

aman

& N

icob

arGoa

Cha

ndig

arh

Laks

hadw

eep

Dad

ra &

Nag

ar H

aveli

Dam

an &

Diu

Delhi

State−wise Survey Sample

1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015-30

-20

-10

0

10

20

30

40

50

60

Pe

rcen

t


Figure A3: Relationship between Inflation Uncertainty & Wealth Allocation in India

This figure reports the correlation between household wealth shares and inflation developments across Indian states, controlling for the State Gross Domestic Product. We obtain state-specific wealth shares as fixed effects μk from the following

estimated specification:


where fi,k are wealth shares of household i in state k, and Xi,k are dummy variables

capturing wealth and age quintiles, education groups, rural vs. urban residence, number of children and number of daughters. The sample contains 15 states for which we are able to match household-level information from the NSS survey and historical regional consumer price indexes: Maharashtra, Andhra Pradesh, Karnataka, Rajasthan, Gujarat, Uttar Pradesh, West Bengal, Bihar, Madhya Pradesh, Orissa, Tamil Nadu, Assam, Kerala, Punjab, and Jammu & Kashmir. We use household-level data from the 2012 wave of the Indian AIDIS survey. Inflation volatility is calculated as the in-sample standard deviation of the consumer price index between 2003 and 2012. Inflation data is sourced from the Labour Statistics Bureau, Government of India.

(a) Inflation and Non-financial Ratio (b) Inflation and Real Estate Assets Ratio

(c) Inflation and Gold Ratio


1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

-10

0

10

20

30

Pe

rce

nt

Gold price appreciation (inflation-adjusted) Deposit rate (inflation-adjusted)

Figure A4: Gold Prices, Deposit Rates and Inflation

This figure reports the evolution of gold prices, term deposits rate, and the consumer price inflation. Gold prices and term deposit rates are both obtained from the Reserve Bank of India. Inflation data is from the Labour Statistics Bureau and uses the CPI (Agricultural Workers) inflation as representative for all sampled households.

(a) Nominal

(b) Real

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

-10

-5

0

5

10

15

20

25

30

Pe

rce

nt

Gold price appreciation Consumer price inflation Deposit rate


A1. Calculation of Asset Categories

European Household Finance and Consumption Survey (Germany) Deposits and transaction accounts

• Sight accounts: All deposits usually at a bank, savings bank, credit institution, mutual bank, insurance company, against which the account holder is permitted to make daily withdrawals (from the bank counter or ATM machines) and make transfers for the purpose of making payments to third persons or others, or transfers to other accounts.

• Savings accounts: All money deposits against which the account holder is not permitted to make transfers for the purpose of making payments to third persons or others.

• Time deposits: Money deposits usually at a bank, savings bank, credit institution, mutual bank, that cannot be withdrawn for a certain "term" or period of time. When the term is over it can be withdrawn or it can be held for another term.

Mutual funds

• Investments in mutual funds, money market mutual funds or hedge funds. • Types of mutual funds: Predominantly investing in equity, bonds, money market

instruments, or real estate. Bonds

• Corporate or government bonds, bills or notes. • All financial assets which are bearer instruments, are usually negotiable and

traded on secondary markets and do not grant the holder any ownership rights to the institutional unit issuing them.

• Types of issuer: State or other general government, banks or other financial intermediaries, non-financial corporations.

Directly held stocks

• Publicly traded shares that are listed on a stock exchange or other form of secondary market.

• Excluding mutual fund shares and excluding other equity in private businesses. Retirement assets and life insurance

• Voluntary pension plans (defined-contribution accounts). • Access to these plans is not necessarily linked to an employment relationship. • Whole life insurance policies, which accumulate a cash value that the policyholder

can redeem or borrow against. Other financial assets

• Managed accounts with a bank or investment company, which makes most of the day-to-day decisions or consult more closely with the account owner. Including trust accounts. Excluding pension and insurance accounts.

• Money owed to the household. • Options, futures, index certificates, precious metals, oil and gas leases, future

proceeds from a lawsuit or estate that is being settled, royalties. Main residence


• Current estimated price of household’s main residence (for owner-occupied residential real estate).

Other real estate

• Other properties, including farm land and buildings, industrial real estate, garages, shops, offices, hotels.

Vehicles, valuables and other assets

• Cars, other vehicles, motorbikes, trucks, vans, planes, boats, and works of art, antiques, jewellery.

Private businesses

• Investment in business equity that is not publicly traded. • Self-employment and non self-employment private businesses.

Survey of Consumer Finances (US) Deposits and transaction accounts

• Transaction accounts: checking, savings, money market or call accounts. • Certificates of deposit. Mutual funds

• Directly held pooled investment funds held by household, including stock mutual funds, tax free bond mutual funds, government bond mutual funds, and combination and other mutual funds, such as hedge funds.

Bonds

• Savings bonds. • Nontaxable bonds, mortgage bonds, government bonds, and ’other’ bonds, such as

coporate or foreign bonds. Directly held stocks

• Publicly traded stock, directly held by household. Retirement assets and life insurance

• Quasi-liquid retirement assets, including IRAs, Keoghs, thrift-type accounts, and future and current account-type pensions.

• Cash value of whole life insurance. Other financial assets

• Managed accounts, including trusts, annuities and managed investment accounts in which the household has equity interest.

• Money owed to the household. • Future proceed from lawsuits, royalties, futures, non-public stock, deferred

compensation, oil, gas, mineral investments. Main residence

• Primary residence of household, excluding the part of a farm or ranch used in a farming or ranching business.

Other real estate


• Land contracts/notes owed to the household and properties other than the principal residence, including 1-4 family residences, time shares, and vacations homes.


• All types of vehicles, cars, trucks, SUVs, motorcycles, boats, airplanes. • Non-residential investment in time shares and vacation homes net of mortgages

and other loans. • Precious metals, antiques, furniture, art objects, paintings, equipment, musical

instruments and other miscellaneous non-financial items owned by household. Private businesses

• Market value of active businesses calculated as net equity if businesses were sold immediately, including loans from the household to the businesses and the value of personal assets used as collateral for business loans.

Wealth and Assets Survey (UK) Deposits and transaction accounts

• Current accounts. • Savings accounts. • Individual Savings Accounts (ISAs). Mutual funds

• Unit investment trusts. Bonds

• Fixed term investment bonds. • UK and foreign bonds and gilts. Directly held stocks

• Shares by publicly traded UK companies. • Overseas shares. • Employee shares and options. Retirement assets and life insurance

• Occupational defined contribution pensions. • Additional Voluntary Contributions (AVCs). • Personal pensions. • Pensions from former spouse or partner. Other financial assets

• Other investments, informal financial assets, children trust funds, children savings accounts, National Savings Products.

Main residence

• Property value of main household residence. Other real estate


• Other houses, buy-to-let houses, buildings, UK and overseas land and other property.


• Household goods and collectables, household contents in main property, second homes, buy-to-let property and overseas property.

• Cars, vans, motorbikes, personalized number plates. All India Debt and Investment Survey

1. Financial Assets: There are three categories of assets that are classified as financial from the AIDIS data:

a. Shares and Debentures These include direct holding of shares and bonds in unlisted firms, listed firms, direct and indirect holdings (through Mutual Funds), and equity holdings in cooperative societies.

b. Account Receivables Amounts receivable from governments, employer/trader, other households and a residual category called “any other”. This also includes any unsecured loans provided to others, “in kind loans” such as payment receivable by doctors for services rendered from their clients will be recorded here. Other receivables, the residual category, includes bonus, profits, lottery prizes, etc. that have not yet been received by the individual.

c. Other Financial Assets Deposits, insurance premium paid (cumulative) till date, provident fund (GPF, PPF, CPF, NPS), ULIPs, government deposits and small savings schemes, bank deposits, non-banking companies deposits, micro-finance institutions, self-help groups, and annuity schemes.

2. Real Estate Assets Value of land and property owned by the household. These

include both residential properties and also land used for self-employment purposes (shops, agricultural land etc.).

3. Durables and Equipment These include transport equipment, agricultural

machinery and implements owned by the household, and non-farm business equipment such as handloom, semi-automatic and power looms, ginning, pressing and baling equipment, mills, casting, melting and welding equipment, photocopying machine, printing press, computer, medical equipment and so on. These only serve as a aiding tool, and is not exhaustive as a list.

4. Livestock Total value of livestock (chicken, ducks etc.) owned by the household

(cattle, buffalo, breeding cows) and other animals such as elephants, and horses. 5. Bullion and Gold Although this is recorded at the very end of “financial assets” of

the household, we treat it as a separate category of “non-financial” assets of the household.

Documents

The Indian Household Savings Landscape - NCAER€¦ · · 2016-07-27The Indian Household Savings Landscape Cristian Badarinza ... rural-urban divide, ... India, since households